Debian Codegraph @

extra/gperftools/gperftools-2.15/ChangeLog

raw
1 commit 365060c4213a48adb27f63d5dfad41b3dfbdd62e
2 Author: Aliaksey Kandratsenka <[email protected]>
3 Date: Fri Jan 5 12:23:30 2024 -0500
4
5 bump version to 2.15
6
7 M CMakeLists.txt
8 M NEWS
9 M configure.ac
10 M src/windows/config.h
11 M src/windows/gperftools/tcmalloc.h
12
13 commit 5abc5721300463b7f05a34b4c95cf45664103b50
14 Author: Aliaksey Kandratsenka <[email protected]>
15 Date: Thu Jan 4 19:33:18 2024 -0500
16
17 move dynamic_annotations.cc under libsysinfo.la
18
19 See github issue #1474 for immediate reason.
20
21 Note, this entire idea of number of convenience libraries is likely
22 simply artifact of Google's codebase past. We don't really need this
23 complexity. But I am holding big reorganization of this for after API
24 and ABI work. For now, simply moving dynamic_annotations.cc into
25 libsysinfo fixes things. Most of the code links both anyways. So lets
26 just do it.
27
28 M Makefile.am
29
30 commit d5c1edf4f5fe901f40d96f4a9d4914f651742c88
31 Author: Aliaksey Kandratsenka <[email protected]>
32 Date: Wed Jan 3 16:06:39 2024 -0500
33
34 [cmake] explicitly link more libraries for threading
35
36 See discussion on https://github.com/gperftools/gperftools/pull/1473
37
38 M CMakeLists.txt
39
40 commit 9430287d15a5c3dd7de9c1b205c6c034c82f35a1
41 Author: Aliaksey Kandratsenka <[email protected]>
42 Date: Mon Jan 1 12:38:26 2024 -0500
43
44 few editorial fixes to NEWS for 2.14 release
45
46 M NEWS
47
48 commit dcf01310551c91d83aeb9bcc8edcb4072d697bd7
49 Author: Aliaksey Kandratsenka <[email protected]>
50 Date: Mon Jan 1 01:02:20 2024 -0500
51
52 correct NEWS entry about QNX (it supports heap profiler)
53
54 M NEWS
55
56 commit e49b36f7d678443859f977bb9a2a3bada1f96ffc
57 Author: Aliaksey Kandratsenka <[email protected]>
58 Date: Sun Dec 31 22:30:20 2023 -0500
59
60 bump version to 2.14
61
62 M CMakeLists.txt
63 M NEWS
64 M configure.ac
65 M src/windows/config.h
66 M src/windows/gperftools/tcmalloc.h
67
68 commit f17d54df5c41a9880c87e623bbf354bc773f440c
69 Author: Aliaksey Kandratsenka <[email protected]>
70 Date: Sun Dec 31 23:12:09 2023 -0500
71
72 add recently added getpc-inl.h to a set of headers to package
73
74 M Makefile.am
75
76 commit fae6421668dd640afc4c7fb08134e53c240b4235
77 Author: Aliaksey Kandratsenka <[email protected]>
78 Date: Sun Dec 31 23:16:00 2023 -0500
79
80 avoid unused variable warning for TCMallocImplementation space
81
82 M src/tcmalloc.cc
83
84 commit 1ba9f605798a4040c3de306f641899dc53cfa5ec
85 Author: Aliaksey Kandratsenka <[email protected]>
86 Date: Sun Dec 31 21:59:53 2023 -0500
87
88 construct malloc extension instances in static spaces
89
90 So that malloc hooks can call into
91 MallocExtension::instance()->GetAllocatedSize() and avoid recursion
92 from new MallocExtension call inside that call.
93
94 This was proposed by github user poljak181 at issue #1472
95
96 M src/malloc_extension.cc
97 M src/tcmalloc.cc
98
99 commit a0880d78f3bb996790075b986c0fa3c81624aacf
100 Author: Aliaksey Kandratsenka <[email protected]>
101 Date: Sun Dec 31 21:33:30 2023 -0500
102
103 Revert "[osx] implement native c++ allocation operators on osx"
104
105 This reverts commit 6683e4f6eeebc52c0704500ef9f461454434578c.
106
107 M src/libc_override_osx.h
108
109 commit 6299ae9869db5084bcab917627497da397386216
110 Author: Aliaksey Kandratsenka <[email protected]>
111 Date: Sun Dec 31 21:23:25 2023 -0500
112
113 test PREFER_LIBGCC_UNWINDER via defined()
114
115 Our newly updated cmake bits now define it to empty, when it is
116 enabled, so we want to test it right. Autoconf stuff behaves the same
117 as before.
118
119 M src/stacktrace.cc
120
121 commit bbc03edc4ddbcce216bd5fa490fd9383bea26a84
122 Author: Mateusz Jakub Fila <[email protected]>
123 Date: Mon Dec 25 02:10:00 2023 +0100
124
125 added missing aggressive decommit option in cmake
126
127 M CMakeLists.txt
128
129 commit 9c2028ae5073b688da89b4866af8739a951f2fe3
130 Author: Mateusz Jakub Fila <[email protected]>
131 Date: Mon Dec 25 02:04:33 2023 +0100
132
133 added missing large alloc report option in cmake
134
135 M CMakeLists.txt
136
137 commit eeb25dd2a50d63363adc6fe07e0eeea625f9b6ea
138 Author: Mateusz Jakub Fila <[email protected]>
139 Date: Mon Dec 25 01:46:40 2023 +0100
140
141 added missing finding asm/ptrace.h in cmake
142
143 M CMakeLists.txt
144
145 commit 4e9fd729d2e7a5907fe118ff57b5218c27d5e824
146 Author: Mateusz Jakub Fila <[email protected]>
147 Date: Mon Dec 25 01:38:09 2023 +0100
148
149 added missing finding `HAVE___ATTRIBUTE__ALIGNED_FN` in cmake
150
151 M CMakeLists.txt
152
153 commit ad4a6ff8de1c04bf5b55dc87057ad84aeeb2836f
154 Author: Mateusz Jakub Fila <[email protected]>
155 Date: Sun Dec 24 02:38:59 2023 +0100
156
157 added missing gcc unwinder preference option in cmake
158
159 M CMakeLists.txt
160 M cmake/config.h.in
161
162 commit 8987d08f792c1c694849a6e779735bfcf50ac8a4
163 Author: Mateusz Jakub Fila <[email protected]>
164 Date: Sun Dec 24 01:51:00 2023 +0100
165
166 fixed setting pagesize and pageshift in cmake
167
168 M CMakeLists.txt
169 M cmake/config.h.in
170
171 commit e9a2d3c46f58581b9c4198b11fc55ac88d9cc6f1
172 Author: barracuda156 <[email protected]>
173 Date: Sat Dec 16 09:53:35 2023 +0800
174
175 mmap_hook.cc: use MAP_ANON when MAP_ANONYMOUS is not defined
176
177 M src/mmap_hook.cc
178
179 commit 85048430ac587190b83973c86ceeb16c72bce9f2
180 Author: Aliaksey Kandratsenka <[email protected]>
181 Date: Thu Dec 7 15:01:27 2023 -0500
182
183 consolidate do_mallinfo{,2}
184
185 We had 2 nearly identical implementations. Thankfully C++ templates
186 facility lets us produce 2 different runtime functions (for different
187 type widths) without duplicating source.
188
189 Amend github issue #1414
190
191 M src/debugallocation.cc
192 M src/tcmalloc.cc
193
194 commit b8e75ae6fe7659ac98d488de3c138f7c9dcacd46
195 Author: Mateusz Jakub Fila <[email protected]>
196 Date: Thu Dec 7 14:10:51 2023 +0100
197
198 Add mallinfo2 function
199
200 M CMakeLists.txt
201 M Makefile.am
202 M cmake/config.h.in
203 M cmake/tcmalloc.h.in
204 M configure.ac
205 M src/debugallocation.cc
206 M src/gperftools/tcmalloc.h.in
207 M src/libc_override_aix.h
208 M src/libc_override_gcc_and_weak.h
209 M src/libc_override_redefine.h
210 M src/tcmalloc.cc
211 M src/windows/config.h
212
213 commit a9b734e3fa697aba64ec4ef3ef228ee532f23430
214 Author: Aliaksey Kandratsenka <[email protected]>
215 Date: Sun Nov 26 23:43:24 2023 -0500
216
217 perform ucontext->pc variants testing in compile-time
218
219 As part of cpu profiler we're extracting current PC (program counter)
220 of out signal's ucontext. Different OS and hardware combinations have
221 different ways for that. We had a list of variants that we tested at
222 compile time and populated PC_FROM_UCONTEXT macro into config.h. It
223 caused duplication and occasional mismatches between our autoconf and
224 cmake bits.
225
226 So this commit changes testing to be compile-time. We remove
227 complexity from build system and add some to C++ source.
228
229 We use SFINAE to find which of those variants compile (and we silently
230 assume that 'compiles' implies 'works'; this is what config-time
231 testing did too). Occasionally we'll face situations where several
232 variants compile. And we couldn't handle this case in pure C++. So we
233 have a small Ruby program that generates chain of inheritance among
234 SFINAE-specialized class templates. This handles prioritization among
235 variants.
236
237 List of ucontext->pc extraction variants is mostly same. We dropped
238 super-obsolete (circa Linux kernel 2.0) arm variant. And NetBSD case
239 is now improved. We now use their nice architecture-independent macro
240 instead of x86-specific access.
241
242 M CMakeLists.txt
243 D cmake/PCFromUContext.cmake
244 M cmake/config.h.in
245 M configure.ac
246 D m4/pc_from_ucontext.m4
247 A src/gen_getpc.rb
248 A src/getpc-inl.h
249 M src/getpc.h
250 M src/stacktrace_generic_fp-inl.h
251 M src/tests/getpc_test.cc
252 M src/windows/config.h
253
254 commit 8edeea4e838573b689d621425044d4efc6ed7f54
255 Author: Sergey Fedorov <[email protected]>
256 Date: Sat Nov 25 15:37:06 2023 +0800
257
258 DefineTargetVariables.cmake: fix for macOS
259
260 M cmake/DefineTargetVariables.cmake
261
262 commit 68db54545ee0be0967ae5ed0e54358e8fef04b9e
263 Author: Sergey Fedorov <[email protected]>
264 Date: Sat Nov 25 15:44:16 2023 +0800
265
266 Minor fix-ups for PowerPC defines
267
268 M src/base/basictypes.h
269 M src/stacktrace_powerpc-inl.h
270
271 commit c815c760dab64a977c2c6f5d3a2a2a4bc9d10189
272 Author: Sergey Fedorov <[email protected]>
273 Date: Sat Nov 25 15:39:35 2023 +0800
274
275 CMakeLists.txt: disable libunwind for ppc
276
277 M CMakeLists.txt
278
279 commit 0ae8fe965081740547619d3d5b7f1fca1af4e0ce
280 Author: Sergey Fedorov <[email protected]>
281 Date: Sat Nov 25 15:27:20 2023 +0800
282
283 PCFromUContext.cmake: fix macOS uc_mcontext
284
285 M cmake/PCFromUContext.cmake
286
287 commit 36fa5ee9ef55103d9edc0a206270f487f2539465
288 Author: Aliaksey Kandratsenka <[email protected]>
289 Date: Mon Nov 6 13:19:17 2023 -0500
290
291 patch libtool.m4 for compiler_lib_search_path fix for qnx
292
293 Amend github issue #1429
294
295 M autogen.sh
296 A m4/libtool.patch
297
298 commit 000af9a1646ebc8c2b8120b60a2946d9718a3072
299 Author: Julian Schroeder <[email protected]>
300 Date: Mon Oct 30 11:43:42 2023 -0600
301
302 [stacktrace_generic_fp] clear aarch64 pointer auth bits
303
304 AARCH64 >= armv8.3-a supports pointer authentication. If this
305 feature is
306 enabled it modifies the previously unused upper address bits in
307 apointer.
308 The affected bits need to be cleared in order for stacktrace to work.
309
310 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
311 [[email protected]: added succinct subject line]
312
313 M src/stacktrace_generic_fp-inl.h
314
315 commit d1a0cbe1bfd8bfaa6223fb9913fa872a760e536e
316 Author: Aliaksey Kandratsenka <[email protected]>
317 Date: Mon Oct 30 19:47:02 2023 -0400
318
319 [qnx] handle broken cfree and mallopt
320
321 M configure.ac
322 M src/libc_override_gcc_and_weak.h
323
324 commit 717bf724a5468fe7125d6eab0c292ca8ec33340b
325 Author: Xiang.Lin <[email protected]>
326 Date: Thu Sep 21 15:42:33 2023 +0800
327
328 Add heap profiler support for QNX
329
330 M benchmark/unwind_bench.cc
331 M configure.ac
332 M src/base/basictypes.h
333 M src/base/sysinfo.cc
334
335 commit adf24f9962c104f463966d0580540818fa723c44
336 Author: Aliaksey Kandratsenka <[email protected]>
337 Date: Fri Oct 27 20:53:21 2023 -0400
338
339 stacktrace_unittest: add simple way to skip ucontext testing
340
341 M src/tests/stacktrace_unittest.cc
342
343 commit 4d1a9e92266f2109d9f535ad65434924765a2456
344 Author: Aliaksey Kandratsenka <[email protected]>
345 Date: Mon Oct 23 10:14:16 2023 -0400
346
347 stacktrace_unittest: test all stacktrace capturing methods
348
349 M CMakeLists.txt
350 M Makefile.am
351 M src/stacktrace.cc
352 M src/tests/stacktrace_unittest.cc
353
354 commit 4dc313870c7a91050e0ac12e411c52c68e761ba7
355 Author: Aliaksey Kandratsenka <[email protected]>
356 Date: Tue Oct 24 16:21:15 2023 -0400
357
358 add cmake warning that our support is incomplete
359
360 M CMakeLists.txt
361
362 commit 96f4f075252a946e9e538fe104621e720f359105
363 Author: Aliaksey Kandratsenka <[email protected]>
364 Date: Tue Oct 24 15:49:03 2023 -0400
365
366 avoid unused variable warning in stacktrace_libunwind
367
368 M src/stacktrace_libunwind-inl.h
369
370 commit db4eacc5d9faeef5e4d0277a45d23664aafdf81a
371 Author: Aliaksey Kandratsenka <[email protected]>
372 Date: Mon Oct 23 11:46:02 2023 -0400
373
374 avoid runtime initialization of null stacktrace implementation
375
376 As we recently found out, initializing static struct fields or
377 variables with lambdas, sets up runtime initialization instead of
378 static initialization as we assumed. So lets avoid this too for null
379 stacktrace implementation.
380
381 M src/stacktrace.cc
382
383 commit 5ba86d37a3a88060c93e350209a9365ef3afaeb3
384 Author: Yikai Zhao <[email protected]>
385 Date: Fri Oct 13 13:04:35 2023 +0800
386
387 update stacktrace_unittest to test overflow issue
388
389 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
390 [[email protected]: squashed log update commit here]
391
392 M src/tests/stacktrace_unittest.cc
393
394 commit dad9e8ceb943c108cf0687d6de273dfec3970df6
395 Author: Yikai Zhao <[email protected]>
396 Date: Wed Oct 11 19:43:28 2023 +0800
397
398 Fix result overflow in generic_fp stacktrace
399
400 In the 'with ucontext' case, the `skip_count` would be reset to 0, and
401 `max_depth` should not be modified. Otherwise the result array
402 would overflow.
403
404 M src/stacktrace_generic_fp-inl.h
405
406 commit c48d4f1407d3b496eab531f6f0e5184049fe3d9e
407 Author: Romain Geissler <[email protected]>
408 Date: Tue Oct 17 17:48:33 2023 +0000
409
410 Avoid initilizing CheckAddress with a lambda, so that it also works
411 with gcc 6.
412
413 M src/check_address-inl.h
414
415 commit d48bf6b3adad65756ed79b843b85409aa34459f8
416 Author: Aliaksey Kandratsenka <[email protected]>
417 Date: Wed Sep 27 21:37:30 2023 -0400
418
419 force inline do_mmap_with_hooks
420
421 Otherwise mmap calling to do_mmap_with_hooks might tail-call (instead
422 of inlining), which will then break GetCallerStackTrace
423 facility (since only mmap is placed into special malloc_hook section).
424
425 This unbreaks heap checker on gcc 5, but is in general right thing to
426 do.
427
428 M src/mmap_hook.cc
429
430 commit 9a123db7b441072a1892e0bab6a101763743aa1c
431 Author: Aliaksey Kandratsenka <[email protected]>
432 Date: Wed Sep 27 21:35:29 2023 -0400
433
434 work around tuple construction miscompilation on gcc 5
435
436 This fixes github issue #1432
437
438 M src/base/sysinfo.cc
439
440 commit d152d76cd14c9e439aa3d9c8a720b7601a66df97
441 Author: Yikai Zhao <[email protected]>
442 Date: Mon Sep 25 19:04:23 2023 +0800
443
444 generic_fp stacktrace: check frame size threshold for initial frame
445
446 M src/stacktrace_generic_fp-inl.h
447
448 commit 64d64bee40811ae19cb119414bdd5e078094d8a7
449 Author: Lennox Ho <[email protected]>
450 Date: Thu Sep 21 10:31:29 2023 +0800
451
452 Fix C++ aligned allocation detected in the CMake build
453
454 M CMakeLists.txt
455
456 commit 589d416977977f32fde82cd623a7a876ab1d397c
457 Author: Lennox Ho <[email protected]>
458 Date: Tue Sep 19 14:53:16 2023 +0800
459
460 Add a more performant SpinLockDelay implementation for Windows based
461 on WaitOnAddress and friends
462
463 M CMakeLists.txt
464 M Makefile.am
465 M src/base/spinlock_win32-inl.h
466
467 commit fd0fabe1831be3a3344048436f0a4ec89bf96466
468 Author: Lennox Ho <[email protected]>
469 Date: Tue Sep 19 14:47:46 2023 +0800
470
471 Fix Windows CMake build by linking page_heap_test and mmap_hook_test
472 to the entire library
473
474 M CMakeLists.txt
475
476 commit 17f23e8d1ef3832663c019e2420307c7068a5543
477 Author: Lennox Ho <[email protected]>
478 Date: Mon Sep 18 06:37:58 2023 +0800
479
480 Add the ability to disable TCMalloc replacement on Windows via
481 environment variable TCMALLOC_DISABLE_REPLACEMENT=1
482
483 M src/windows/patch_functions.cc
484
485 commit df006e880e1ca6fa3a3ccc3a44a65a9e8bec298c
486 Author: Lennox Ho <[email protected]>
487 Date: Sun Sep 17 07:59:40 2023 +0800
488
489 Also expose SetMemoryReleaseRate and GetMemoryReleaseRate as C shims
490
491 M src/gperftools/malloc_extension_c.h
492 M src/malloc_extension.cc
493
494 commit d3602c0672daebe2edd533fda968a1f36233d706
495 Author: Lennox Ho <[email protected]>
496 Date: Sun Sep 17 07:49:38 2023 +0800
497
498 Account for Windows while performing implicit TLS detection in
499 CMakeLists.txt
500
501 M CMakeLists.txt
502
503 commit dffb4a2f284700b814e2c3aa1f9263fd14e836f8
504 Author: Aliaksey Kandratsenka <[email protected]>
505 Date: Mon Sep 11 16:11:27 2023 -0400
506
507 bump version to 2.13
508
509 M CMakeLists.txt
510 M NEWS
511 M configure.ac
512 M src/windows/config.h
513 M src/windows/gperftools/tcmalloc.h
514
515 commit 4ec8c9dbb25e2844f98f7acc848aa3b190349d70
516 Author: Aliaksey Kandratsenka <[email protected]>
517 Date: Sun Sep 10 22:18:51 2023 -0400
518
519 reduce set of nallocx size testing points
520
521 Testing every 7th size is a bit slow on slower machines. No need to be
522 as thorough. We now bump by about 1/128th each step which is still
523 more steps than size classes we have.
524
525 M src/tests/tcmalloc_unittest.cc
526
527 commit e4e7ba93a03e469e1d11153bf724e422509ebab7
528 Author: Aliaksey Kandratsenka <[email protected]>
529 Date: Sun Sep 10 21:06:02 2023 -0400
530
531 unbreak unnecessary dependency on 64-bit atomics
532
533 This unbreaks builds on 32-bit arms and mipsen.
534
535 M src/tests/profile-handler_unittest.cc
536
537 commit 523b72f754246d4c3736c82ba7fa0576b672497e
538 Author: Aliaksey Kandratsenka <[email protected]>
539 Date: Sun Sep 10 18:12:45 2023 -0400
540
541 make sampling_debug_test actually test debug malloc
542
543 We do shell wrapper for actual test run, so we can inspect output of
544 pprof. But when we set up sampling_debug_test.sh we simply copied
545 regular sampling_test.sh, which ran same non-debug test binary. Now we
546 sed-replace contents of shell program when copying, so we test right
547 binary.
548
549 Another thing we fix here is our (still hardcoded) test output path is
550 now different between sampling{,_debug}_test.sh. So this fixes main
551 cause of flakiness of our unit tests.
552
553 M Makefile.am
554
555 commit 2748dd56803d61b8aee8fbb8bbdf9bc6d2913529
556 Author: Aliaksey Kandratsenka <[email protected]>
557 Date: Fri Sep 8 19:47:17 2023 -0400
558
559 unbreak address access "probing" for generic_fp backtracing
560
561 We used msync to verify that address is readable. But msync gives
562 false positives for PROT_NONE mappings. And we recently got bug report
563 from user hitting this exact condition.
564
565 For correct access check, we steal idea from Abseil and do sigprocmask
566 with address used as new signal mask and with invalid HOW
567 argument. This works in today's Linux kernels and is among fastest
568 methods available. But is brittle w.r.t. possible kernel changes. So
569 we supply fallback method that does 2 syscalls.
570
571 For non-Linux systems we implement usual "write to pipe" trick. Which
572 also has decent performance, but requires occasional pipe draining and
573 uses fds which could occasionally be damaged by some forking codes.
574
575 We also finally cover all new code with unit test.
576
577 Fixes github issue #1426
578
579 M .gitignore
580 M CMakeLists.txt
581 M Makefile.am
582 A src/check_address-inl.h
583 M src/stacktrace_generic_fp-inl.h
584 A src/tests/check_address_test.cc
585
586 commit 7ad1dc7693b7b04199681df474122248634a9808
587 Author: Ivan Dlugos <[email protected]>
588 Date: Mon Sep 4 14:40:27 2023 +0200
589
590 fix: cmake config.h defines declaration
591
592 M .gitignore
593 M cmake/config.h.in
594 M src/base/logging.h
595 M src/base/sysinfo.cc
596 M src/getpc.h
597 M src/malloc_hook.cc
598 M src/profile-handler.cc
599 M src/profiler.cc
600 M src/stacktrace_generic_fp-inl.h
601 M src/stacktrace_powerpc-linux-inl.h
602
603 commit f7172839a1a3ca99309ab322d8ea085967ecaf4d
604 Author: Aliaksey Kandratsenka <[email protected]>
605 Date: Fri Sep 8 14:22:46 2023 -0400
606
607 turn tcmalloc::TrivialOnce into POD
608
609 As we see in github issue #1428, msvc arranges full "init on first
610 use" initialization for local static usage of TrivialOnce even if that
611 initialization is completely empty. Fair game, even if stupid.
612
613 POD with no initialization should be safely zero-initialized with no
614 games or tricks from the compilers.
615
616 We could have and perhaps at some point should do constexpr for
617 TrivialOnce and SpinLock (abseil has been liberated from
618 LinkerInitialized for perphaps well over decade now, including their
619 fork of SpinLock, of course). But C++ legalese rules are complex
620 enough and bugs happened in past, so I don't want to be in the tough
621 business of interpreting standard. So at least for now we keep
622 things simple.
623
624 M src/base/spinlock.h
625 M src/debugallocation.cc
626 M src/heap-checker.cc
627 M src/malloc_hook.cc
628 M src/profile-handler.cc
629
630 commit 539ed9ca40f4d3806e7a2b79b5aa02b9a45083d3
631 Author: Aliaksey Kandratsenka <[email protected]>
632 Date: Thu Aug 24 15:03:47 2023 -0400
633
634 bump version 2.12
635
636 M CMakeLists.txt
637 M NEWS
638 M configure.ac
639 M src/windows/config.h
640 M src/windows/gperftools/tcmalloc.h
641
642 commit 8d634c1f56d282b55ad414a0d8b3bba5cad9bb93
643 Author: Aliaksey Kandratsenka <[email protected]>
644 Date: Thu Aug 24 14:06:25 2023 -0400
645
646 don't build mmap_hook when --enable-minimal is given to configure
647
648 Refers to github issue #1418
649
650 M Makefile.am
651
652 commit b4ad04982d4710f7dccae6c4d348df224cc94dad
653 Author: Brett T. Warden <[email protected]>
654 Date: Tue Aug 15 16:21:47 2023 -0700
655
656 Set Description field in generated pkg-config files (instead of
657 Summary)
658
659 Fixes #1416
660
661 M Makefile.am
662
663 commit 0a3ca5b43d7358c901d78cefef593ad8a7a96ab3
664 Author: Aliaksey Kandratsenka <[email protected]>
665 Date: Mon Aug 14 22:47:56 2023 -0400
666
667 bump version to 2.11
668
669 M CMakeLists.txt
670 M NEWS
671 M configure.ac
672 M src/windows/config.h
673 M src/windows/gperftools/tcmalloc.h
674
675 commit 83fccceffaa3e526d2612b34400c772745e8441c
676 Author: Aliaksey Kandratsenka <[email protected]>
677 Date: Mon Aug 14 10:12:11 2023 -0400
678
679 bump README freshness a bit
680
681 M README
682
683 commit c41eb9e8b594a3a6ccd973f38edb80554c1e31e1
684 Author: Ken Raffenetti <[email protected]>
685 Date: Fri Aug 11 10:23:37 2023 -0500
686
687 Add MPICH HPC environment detection
688
689 Default MPICH builds use the Hydra process manager (mpiexec)
690 which sets
691 PMI_RANK in the application environment. Update GetUniquePathFromEnv()
692 test accordingly.
693
694 Signed-off-by: Ken Raffenetti <[email protected]>
695
696 M src/base/sysinfo.cc
697 M src/tests/unique_path_unittest.cc
698
699 commit 1d2654f3a0e5da3f51da5245edb24489efd51773
700 Author: Aliaksey Kandratsenka <[email protected]>
701 Date: Thu Aug 10 00:22:45 2023 -0400
702
703 heap-checker: unbreak PTRACE_GETREGS detection on older Linux-es
704
705 This unbreaks RHEL6.
706
707 M src/heap-checker.cc
708
709 commit dbd10716809320c7fb89dd43d1e6b913ed1378c3
710 Author: Aliaksey Kandratsenka <[email protected]>
711 Date: Wed Aug 9 22:11:41 2023 -0400
712
713 link libprofiler with pthread
714
715 This unbreaks building on older Linux distros. We missed this at
716 46d3315ad7901099bd414740dbfbe6d27f33da9a when dropped maybe_thread
717 stuff, since libprofiler indeed uses pthread, and because on newer
718 libc-s pthread stuff is now part of regular libc.so.
719
720 I am also dropping bogus LIBPROFILER stuff referring to some rpath
721 badness. Unsure what it was, maybe way back we did libstacktrace as a
722 proper libtool library, so maybe something was needed. But it is just
723 a convenience archive this days, so we don't really need to add it
724 everywhere libprofiler.la is linked.
725
726 M Makefile.am
727
728 commit 729383b486b3e26626e1dd354c8b5098d104f19f
729 Author: Aliaksey Kandratsenka <[email protected]>
730 Date: Wed Aug 9 23:29:00 2023 -0400
731
732 make sure that ListerThread runs on properly aligned stack
733
734 Without this fix we're failing unit tests on ubuntu 18.04 and centos 7
735 and 6. It looks like clone() in old glibc-s doesn't align stack, so
736 lets handle it ourselves. How we didn't hit this much earlier (before
737 massive thread listing refactoring), I am not sure. Most likely pure
738 luck(?)
739
740 M src/base/linuxthreads.cc
741
742 commit 51c5e2bec7a53a7f1e5e44c866239fedb8ecd590
743 Author: Aliaksey Kandratsenka <[email protected]>
744 Date: Wed Aug 9 13:18:59 2023 -0400
745
746 massage latest GetUniquePathFromEnv changes
747
748 This fixes a number of minor bits (like build details) as well as
749 making overall code style similar to what we're doing elsewhere.
750
751 M .gitignore
752 M CMakeLists.txt
753 M Makefile.am
754 M src/base/sysinfo.cc
755 M src/base/sysinfo.h
756 M src/tests/unique_path_unittest.cc
757
758 commit 86450ad99fc9708ac5f1b79d5c7f089d50821a42
759 Author: Artem Polyakov <[email protected]>
760 Date: Thu Jul 13 17:43:29 2023 -0700
761
762 Add unit test for GetUniquePathFromEnv()
763
764 Signed-off-by: Artem Polyakov <[email protected]>
765
766 M Makefile.am
767 A src/tests/unique_path_unittest.cc
768
769 commit 881b754da0c9c8b5e711113aecfad5f7f8ef95f2
770 Author: Artem Y. Polyakov <[email protected]>
771 Date: Thu Apr 1 15:27:22 2021 -0700
772
773 Advanced UniquePathFromEnv generation
774
775 * Add support for known HPC environments (TODO: needs to be extended
776 with more nevironments)
777 * Added the "CPUPROFILE_USE_PID" environment variable to force
778 appending
779 PID for the non-covered environments
780 * Preserve the old way of handling the Child-Parent case
781
782 Signed-off-by: Artem Polyakov <[email protected]>
783
784 M src/base/sysinfo.cc
785 M src/base/sysinfo.h
786
787 commit 57512e9c3dec5de857c856d4e7783a69ea3cd872
788 Author: Aliaksey Kandratsenka <[email protected]>
789 Date: Sun Aug 6 19:24:17 2023 -0400
790
791 unbreak -Wthread-safety
792
793 It actually found real (but arguably minor) issue with memory region
794 map locking.
795
796 As part of that we're replacing PageHeap::DeleteAndUnlock that had
797 somewhat ambitious 'move' of SpinLockHolder, with more straightforward
798 PageHeap::PrepareAndDelete. Doesn't look like we can support move
799 thingy with thread annotations.
800
801 M src/base/spinlock.h
802 M src/heap-profile-table.cc
803 M src/memory_region_map.cc
804 M src/memory_region_map.h
805 M src/page_heap.cc
806 M src/page_heap.h
807 M src/tcmalloc.cc
808 M src/tests/page_heap_test.cc
809
810 commit 862039c1850d56c64f482909d615ffcffbfcdfc0
811 Author: Aliaksey Kandratsenka <[email protected]>
812 Date: Sun Aug 6 14:44:05 2023 -0400
813
814 don't -momit-leaf-frame-pointer when asked for full frame pointers
815
816 M Makefile.am
817
818 commit dc25c1fd4c4b9de71f91013a8c0b811164494eed
819 Author: Aliaksey Kandratsenka <[email protected]>
820 Date: Mon Jul 31 20:04:39 2023 -0400
821
822 bump version to 2.11rc
823
824 M CMakeLists.txt
825 M NEWS
826 M configure.ac
827 M src/windows/config.h
828 M src/windows/gperftools/tcmalloc.h
829
830 commit 909fa3e649c659bbb055f97cc9277080a1725ea8
831 Author: Aliaksey Kandratsenka <[email protected]>
832 Date: Mon Jul 31 16:27:33 2023 -0400
833
834 unbreak MallocExtension::GetAllocatedSize() for debug allocator
835
836 Some years back we fixed memalign vs realloc bug, but we preserved
837 'wrong' malloc_size/GetAllocatedSize implementation for debug
838 allocator.
839
840 This commit refactors old code making sure we always use right
841 data_size and it fixes GetAllocatedSize. We update our unittest
842 accordingly.
843
844 Closes #738
845
846 M src/debugallocation.cc
847 M src/tests/debugallocation_test.cc
848
849 commit a51e08b06a4318a201547b7458f6483e830d86e5
850 Author: Aliaksey Kandratsenka <[email protected]>
851 Date: Mon Jul 31 15:40:13 2023 -0400
852
853 drop obsolete TODO file
854
855 M Makefile.am
856 D TODO
857
858 commit a1c7ce779317adc3d4128daa37523b3ba935a34e
859 Author: Aliaksey Kandratsenka <[email protected]>
860 Date: Mon Jul 31 15:16:14 2023 -0400
861
862 refresh README-s
863
864 M README
865 M README_windows.txt
866
867 commit bef6592746445db220fc29a8ed9dc0123e628221
868 Author: Aliaksey Kandratsenka <[email protected]>
869 Date: Mon Jul 31 15:10:56 2023 -0400
870
871 refresh INSTALL
872
873 M INSTALL
874
875 commit e3de2e3242fc33aecc7b7f7519aac783bfd15b4f
876 Author: Aliaksey Kandratsenka <[email protected]>
877 Date: Mon Jul 31 14:28:58 2023 -0400
878
879 remove obsolete references to code.google.com
880
881 I.e. somehow we managed to still point to (very) old gperftools
882 hosting location, so lets fix it at last.
883
884 M CMakeLists.txt
885 M INSTALL
886 M configure.ac
887 M src/debugallocation.cc
888 M src/heap-checker.cc
889 M src/page_heap.cc
890 M src/stacktrace_win32-inl.h
891 M src/tests/large_heap_fragmentation_unittest.cc
892
893 commit 1ff09a680e6239675bbc03ab385872276409558b
894 Author: Aliaksey Kandratsenka <[email protected]>
895 Date: Mon Jul 31 14:28:40 2023 -0400
896
897 drop obsolete deb/rpm packaging stuff
898
899 M Makefile.am
900 D packages/deb.sh
901 D packages/deb/README
902 D packages/deb/changelog
903 D packages/deb/compat
904 D packages/deb/control
905 D packages/deb/copyright
906 D packages/deb/docs
907 D packages/deb/libgperftools-dev.dirs
908 D packages/deb/libgperftools-dev.install
909 D packages/deb/libgperftools0.dirs
910 D packages/deb/libgperftools0.install
911 D packages/deb/libgperftools0.manpages
912 D packages/deb/rules
913 D packages/rpm.sh
914 D packages/rpm/rpm.spec
915
916 commit 8b3f0d61451e891e1ca91a4887e472b80e9ff351
917 Author: Aliaksey Kandratsenka <[email protected]>
918 Date: Mon Jul 31 14:04:32 2023 -0400
919
920 undo MarkThreadTemporarilyIdle and make it same as MarkThreadIdle
921
922 As noted on github issue #880 'temporarily' thing saves us not just on
923 freeing thread cache, but also returning thread's share of thread
924 cache (max_size_) into common pool. And the later has caused trouble
925 to mongo folk who originally proposed 'temporarily' thing. They claim
926 they don't use it anymore.
927
928 And thus with no users and no clear benefit, it makes no sense for us
929 to keep this API. For API and ABI compat sake we keep it, but it is
930 now identical to regular MarkThreadIdle.
931
932 Fixes issue #880
933
934 M src/gperftools/malloc_extension.h
935 M src/tcmalloc.cc
936 M src/thread_cache.cc
937 M src/thread_cache.h
938
939 commit c3059a56be3f1f1581c06d77ef04b0b7ecc8cc6f
940 Author: Aliaksey Kandratsenka <[email protected]>
941 Date: Mon Jul 31 12:26:21 2023 -0400
942
943 dont workaround unknown problem in thread_dealloc_unittest
944
945 We had some sleep added at the end up thread dealloc unittest claiming
946 some race trouble with glibc. Which is likely years or even decades
947 irrelevant.
948
949 M src/tests/thread_dealloc_unittest.cc
950
951 commit 9b91ce917ad54fab8dd658daeed32bababd367bb
952 Author: Aliaksey Kandratsenka <[email protected]>
953 Date: Thu Jul 27 19:08:54 2023 -0400
954
955 [win32:patching] define single empty __expand replacement
956
957 This unbreaks some cases where patching complains about too short
958 functions to patch.
959
960 What happens is we first locate one of CRT-s (like ucrt or msvcrt) and
961 patch __expand there, redirecting to our implementation. Then "static"
962 __expand replacement is patched, but it is usually imported from that
963 same C runtime DLL. And through several jmp redirections we end up at
964 our own __expand from libc<1>. Patching that (and other cases) is
965 wrong, but I am unsure how to fix it properly. So we do most simple
966 workaround. I found that when it not fails is either in debug builds
967 where empty expand is not too short or when MSVC deduplicates multiple
968 identical __expand implementations into single function, or when
969 64-bit patching has to do extra trampoline thingy. And then our
970 patching code checks if we're trying to replace some function with
971 itself. So we "just" take advantage of that and get immediate issue
972 fixed, while punting on more general "duplicate" patching for later.
973
974 Update github issue #667
975
976 M src/windows/patch_functions.cc
977
978 commit b6cdd8f510d9b8af03cb2c115f87c4b46257ec75
979 Author: Aliaksey Kandratsenka <[email protected]>
980 Date: Thu Jul 27 19:23:15 2023 -0400
981
982 [mingw] dont add libstacktrace to libtcmalloc_minimal
983
984 There was this piece of makefile with indention to add stack tracing
985 functionality (for stuff like growthz, GetCallerStackTrace and
986 probably heap sampling) to work even in minimal configuration on
987 mingw.
988
989 What is odd is we fail to actually define libstacktrace.la target on
990 mingw, since libstacktrace.la requires WITH_STACK_TRACE automake
991 conditional which we don't enable on this platform. And yet somehow it
992 doesn't fail. It produces empty libstacktrace.la, so build kinda
993 works. Except at least on my machine it produces racy makefiles. So
994 lets not pretend and stop breaking our parallel builds.
995
996 M Makefile.am
997
998 commit a5cfd38884a34d374919eee22626c70cabd21a72
999 Author: Aliaksey Kandratsenka <[email protected]>
1000 Date: Tue Jul 25 15:58:02 2023 -0400
1001
1002 [win32] amend and unbreak previous NOMINMAX fix
1003
1004 M src/windows/config.h
1005
1006 commit d2c89ba53433f8f0940e8bf086436895e46436a6
1007 Author: Aliaksey Kandratsenka <[email protected]>
1008 Date: Mon Jul 24 18:57:52 2023 -0400
1009
1010 don't return raw span when sampling and stacktrace oomed
1011
1012 This is nearly impossible in practice, but still. Somehow we missed
1013 this logic that DoSampledAllocation always returns actual object, but
1014 in that condition where stacktrace_allocator failed to get us
1015 StackTrace object we ended up returning span instead of it's object.
1016
1017 M src/tcmalloc.cc
1018
1019 commit 59464404d1e900e9664a22960e75659eed04d8e8
1020 Author: Aliaksey Kandratsenka <[email protected]>
1021 Date: Thu Dec 30 14:21:00 2021 -0800
1022
1023 capture growthz backtraces outside of pageheap_lock
1024
1025 Actual growthz list is now lockless since we never delete anything
1026 from it. And we now pass special 'locking context' object down page
1027 heap allocation path, both as a documentation that it is under lock
1028 and for tracking whether we needed to grow heap and by how much. Then
1029 whenever lock is released in RAII fashion, we're able to trigger
1030 growthz recording outside of lock.
1031
1032 Closes #1159
1033
1034 M src/page_heap.cc
1035 M src/page_heap.h
1036 M src/static_vars.cc
1037 M src/static_vars.h
1038
1039 commit 0d42a48699059826b99dd5d1a1bb99c5669f089b
1040 Author: Aliaksey Kandratsenka <[email protected]>
1041 Date: Mon Dec 27 17:35:53 2021 -0800
1042
1043 move page heap locking under PageHeap
1044
1045 While there is still plenty of code that takes pageheap_lock outside
1046 of page_heap module for all kinds of reasons, at least
1047 bread-and-butter logic of allocating/deallocating larger chunks of
1048 memory is now handling page heap locking inside PageHeap itself. This
1049 gives us flexibility.
1050
1051 Update issue #1159
1052
1053 M src/base/spinlock.h
1054 M src/central_freelist.cc
1055 M src/page_heap.cc
1056 M src/page_heap.h
1057 M src/static_vars.cc
1058 M src/static_vars.h
1059 M src/tcmalloc.cc
1060 M src/tests/page_heap_test.cc
1061
1062 commit a3e1080c2e7038a4205168f1f9e086b530743272
1063 Author: Aliaksey Kandratsenka <[email protected]>
1064 Date: Mon Dec 27 21:26:18 2021 -0800
1065
1066 handle large alloc reporting locklessly
1067
1068 Which simplifies codes a bit.
1069
1070 Update issue #1159
1071
1072 M src/tcmalloc.cc
1073
1074 commit f1eb3c82c6b55d99bc657435b3fc42dae613440f
1075 Author: Aliaksey Kandratsenka <[email protected]>
1076 Date: Mon Dec 27 14:45:56 2021 -0800
1077
1078 correctly release memory when system's pagesize is >kPageSize
1079
1080 I.e. this covers case of arms that by default compile tcmalloc for 8k
1081 logical pages (assuming 4k system pages), but can actually run on
1082 systems with 64k pages.
1083
1084 Closes #1135
1085
1086 M src/common.cc
1087 M src/common.h
1088 M src/page_heap.cc
1089 M src/page_heap.h
1090 M src/static_vars.cc
1091 M src/tcmalloc.cc
1092 M src/tests/page_heap_test.cc
1093 M src/tests/tcmalloc_unittest.cc
1094
1095 commit d521e3b30e208af635e9812fe9bff635bbafcb1d
1096 Author: Aliaksey Kandratsenka <[email protected]>
1097 Date: Mon Dec 27 14:41:42 2021 -0800
1098
1099 move page heap allocations with alignment into page heap
1100
1101 M src/page_heap.cc
1102 M src/page_heap.h
1103 M src/tcmalloc.cc
1104 M src/tests/page_heap_test.cc
1105
1106 commit ad0ca2b83ba6743afcb8038788e9195bbd1514d1
1107 Author: Aliaksey Kandratsenka <[email protected]>
1108 Date: Mon Jul 24 18:07:25 2023 -0400
1109
1110 unbreak large heap fragmentation unittest
1111
1112 Smart compilers again (and lack of -fno-builtin-malloc which we
1113 dropped because of clang).
1114
1115 M src/tests/large_heap_fragmentation_unittest.cc
1116
1117 commit bd3bf9abd7b42acdec38a98ceabc659adf8a9225
1118 Author: Aliaksey Kandratsenka <[email protected]>
1119 Date: Mon Jul 24 19:40:49 2023 -0400
1120
1121 define NOMINMAX on windows
1122
1123 Otherwise we cannot use some std bits, like numeric_limits.
1124
1125 M src/windows/config.h
1126 M src/windows/port.cc
1127
1128 commit 814f340eede8c991885f10deca929031fe2531bb
1129 Author: Aliaksey Kandratsenka <[email protected]>
1130 Date: Sat Jul 22 20:45:52 2023 -0400
1131
1132 add --enable-libgcc-unwinder-by-default for configure
1133
1134 We want distros to ship gperftools with this options, now that they
1135 have _Unwind_Backtrace that is powered by dl_find_object.
1136
1137 M configure.ac
1138 M src/stacktrace.cc
1139
1140 commit f06ccc6f799fa2fe550a8bee08314393586fb472
1141 Author: Aliaksey Kandratsenka <[email protected]>
1142 Date: Sat Jul 22 14:21:47 2023 -0400
1143
1144 dont test HAVE_{STDINT,INTTYPES}_H
1145
1146 Those are fairly standard by now. We already require C++11 or later
1147 compiler.
1148
1149 M CMakeLists.txt
1150 M cmake/config.h.in
1151 M src/addressmap-inl.h
1152 M src/base/basictypes.h
1153 M src/central_freelist.h
1154 M src/common.h
1155 M src/debugallocation.cc
1156 M src/heap-profile-table.cc
1157 M src/heap-profiler.cc
1158 M src/internal_logging.h
1159 M src/malloc_extension.cc
1160 M src/malloc_hook.cc
1161 M src/memfs_malloc.cc
1162 M src/memory_region_map.cc
1163 M src/packed-cache-inl.h
1164 M src/page_heap.cc
1165 M src/page_heap.h
1166 M src/pagemap.h
1167 M src/sampler.h
1168 M src/stack_trace_table.h
1169 M src/symbolize.h
1170 M src/system-alloc.cc
1171 M src/tcmalloc.cc
1172 M src/tests/heap-checker_unittest.cc
1173 M src/tests/memalign_unittest.cc
1174 M src/tests/pagemap_unittest.cc
1175 M src/tests/profiledata_unittest.cc
1176 M src/tests/sampler_test.cc
1177 M src/tests/system-alloc_unittest.cc
1178 M src/tests/tcmalloc_unittest.cc
1179 M src/tests/testutil.cc
1180 M src/thread_cache.h
1181 M src/windows/config.h
1182 M src/windows/port.h
1183
1184 commit 86867674f05a94d9e761310bb5c3e7b4b4534634
1185 Author: Aliaksey Kandratsenka <[email protected]>
1186 Date: Sat Jul 22 14:03:49 2023 -0400
1187
1188 don't do syscall on osx
1189
1190 Their compiler barks and we only need this syscall business on
1191 FreeBSD.
1192
1193 M src/base/logging.h
1194
1195 commit bdae3615ca51b5aa8e7075396dc0c53748c78076
1196 Author: Aliaksey Kandratsenka <[email protected]>
1197 Date: Sat Jul 22 14:01:55 2023 -0400
1198
1199 osx's dyld_present is always true
1200
1201 As reported to us by their compiler's warning.
1202
1203 M src/base/basictypes.h
1204
1205 commit 91b8024aa1e5cd8e6588df4e9c7a6280c8f03b3a
1206 Author: Aliaksey Kandratsenka <[email protected]>
1207 Date: Sat Jul 22 14:00:32 2023 -0400
1208
1209 drop unused ThreadCache::DropCacheWhichMustBePresent
1210
1211 M src/thread_cache.h
1212
1213 commit 42dde1b883b29921ac89234369f88b133821690c
1214 Author: Aliaksey Kandratsenka <[email protected]>
1215 Date: Sat Jul 22 00:08:38 2023 -0400
1216
1217 heap-checker_unittest: run thread-unsafe libc calls less unsafely
1218
1219 Comment there says race is fine, bit it isn't. Crashes under
1220 musl. Lets not crash.
1221
1222 M src/tests/heap-checker_unittest.cc
1223
1224 commit 8be84e4a5c28c33bfdb7da3ee0c92d9620cd552f
1225 Author: Aliaksey Kandratsenka <[email protected]>
1226 Date: Fri Jul 21 14:18:12 2023 -0400
1227
1228 drop old mmap hooks and introduce internal & simpler mmap_hook.h
1229
1230 Previous implementation wasn't entirely safe w.r.t. 32-bit off_t
1231 systems. Specifically around mmap replacement hook. Also, API was a
1232 lot more general and broad than we actually need.
1233
1234 Sadly, old mmap hooks API was shipped with our public headers. But
1235 thankfully it appears to be unused externally (checked via github
1236 search). So we keep this old API and ABI for the sake of formal API
1237 and ABI compatibility. But this old API is now empty and always
1238 fails (some OS/hardware combinations didn't have functional
1239 implementations of those hooks anyways).
1240
1241 New API is 64-bit clean and only provides us with what we need. Namely
1242 being able to react to virtual address space mapping changes for
1243 logging, heap profiling and heap leak checker. I.e. no pre hooks or
1244 mmap-replacement hooks. We also explicitly not ship this API
1245 externally to give us freedom to change it.
1246
1247 New code is also hopefully tidier and slightly more portable. At least
1248 there are fewer arch-specific ifdef-s.
1249
1250 Another somewhat notable change is, since mmap hook isn't needed in
1251 "minimal" configuration, we now don't override system's
1252 mmap/munmap/etc functions in this configuration. No big deal, but it
1253 reduces risk of damage if we somehow mess those up. I.e. musl's mmap
1254 does few things that our mmap replacement doesn't, such as very fancy
1255 vm_lock thingy. Which doesn't look critical, but is good thing for us
1256 not to interfere with when not necessary.
1257
1258 Fixes issue #1406 and issue #1407. Lets also mention issue #1010 which
1259 is somewhat relevant.
1260
1261 M .gitignore
1262 M CMakeLists.txt
1263 M Makefile.am
1264 M src/base/low_level_alloc.cc
1265 M src/base/spinlock.h
1266 M src/heap-checker.cc
1267 M src/heap-profiler.cc
1268 M src/malloc_hook-inl.h
1269 M src/malloc_hook.cc
1270 D src/malloc_hook_mmap_freebsd.h
1271 D src/malloc_hook_mmap_linux.h
1272 M src/memory_region_map.cc
1273 M src/memory_region_map.h
1274 A src/mmap_hook.cc
1275 A src/mmap_hook.h
1276 M src/tests/malloc_hook_test.cc
1277 A src/tests/mmap_hook_test.cc
1278 M src/tests/tcmalloc_unittest.cc
1279 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj
1280 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj.filters
1281 M
1282 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj
1283 M
1284 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj.filters
1285 M vsprojects/page_heap_test/page_heap_test.vcxproj
1286 M vsprojects/page_heap_test/page_heap_test.vcxproj.filters
1287
1288 commit 6b691cc019500dc4547822b1eed2fd53adc6b43c
1289 Author: Aliaksey Kandratsenka <[email protected]>
1290 Date: Fri Jul 21 14:18:01 2023 -0400
1291
1292 drop patching virtual alloc fns on windows
1293
1294 It wasn't implemented exactly right and, most importantly, there is
1295 nothing on windows that uses those (now dead) mmap hooks.
1296
1297 M src/windows/patch_functions.cc
1298
1299 commit fecaf37ae4f172c62ce4d1f42062b690cc96fabc
1300 Author: Aliaksey Kandratsenka <[email protected]>
1301 Date: Fri Jul 21 13:51:36 2023 -0400
1302
1303 drop DevMemSysAllocator
1304
1305 It was never useful for anything.
1306
1307 M src/system-alloc.cc
1308
1309 commit e53ec241b6f085a770b3f5e558dac8f85963a4d4
1310 Author: Aliaksey Kandratsenka <[email protected]>
1311 Date: Thu Jul 20 21:53:54 2023 -0400
1312
1313 unbreak __sbrk detection on cmake
1314
1315 M CMakeLists.txt
1316
1317 commit d4f33620351f303e64d1a6dd3c4c79ad53532be6
1318 Author: Aliaksey Kandratsenka <[email protected]>
1319 Date: Fri Jul 21 14:21:37 2023 -0400
1320
1321 add missing .gitignore for safe_strerror_test.exe
1322
1323 M .gitignore
1324
1325 commit e0f4c0e6fb402e56e82a82438561f413090f0ac8
1326 Author: Aliaksey Kandratsenka <[email protected]>
1327 Date: Mon Jul 17 19:31:06 2023 +0000
1328
1329 manually inline MapObjectToSpan into ReleaseToSpans
1330
1331 It was originally separated into own function due to some (now
1332 obsolete) compiler optimization bug. We stopped worked around that bug
1333 few commits ago. But as github user plopresti rightfully pointed
1334 out (much, much thanks!), we lost `static' on that function. But since
1335 it is trivial function and used exactly once, it is best to simply
1336 inline it's 2 lines of code back instead of returning static.
1337
1338 Previous commit was: 68b442714a86060d1dd95de4c515cc7774e3751e
1339
1340 M src/central_freelist.cc
1341
1342 commit 46d3315ad7901099bd414740dbfbe6d27f33da9a
1343 Author: Aliaksey Kandratsenka <[email protected]>
1344 Date: Fri Jul 14 02:19:21 2023 -0400
1345
1346 amputate maybe_threads
1347
1348 This facility allowed us to build tcmalloc without linking in actual
1349 -lpthread. Via weak symbols we checked at runtime if pthread functions
1350 are available and if not, special single-threaded stubs were used
1351 instead. Not always brining in pthread dependency helped performance
1352 of some programs or libraries which depended at runtime on whether
1353 threads are linked or not. Most notable of those are libstdc++ which
1354 uses non-atomic refcounting on single threaded programs.
1355
1356 But such optional dependency on pthreads caused complications for
1357 nearly no benefit. One trouble was reported in github issue #1110.
1358
1359 This days glibc/libstdc++ combo actually depends on
1360 sys/single_threaded.h facility. So bringing pthread at runtime is
1361 fine. Also modern glibc ships pthread symbols inside libc anyways and
1362 libpthread is empty. I also found that for whatever reason on BSDs and
1363 osx we already pulled in proper pthreads too.
1364
1365 So we loose nothing and we get issue #1110 fixed. And we simplify
1366 everything.
1367
1368 M .gitignore
1369 M CMakeLists.txt
1370 M INSTALL
1371 M Makefile.am
1372 M configure.ac
1373 M src/heap-checker.cc
1374 M src/malloc_extension.cc
1375 M src/malloc_hook.cc
1376 D src/maybe_threads.cc
1377 D src/maybe_threads.h
1378 M src/profile-handler.cc
1379 M src/static_vars.cc
1380 M src/tests/low_level_alloc_unittest.cc
1381 D src/tests/maybe_threads_unittest.sh
1382 M src/thread_cache.cc
1383 M src/thread_cache.h
1384 M src/windows/port.h
1385
1386 commit 3cd265b05a1c951010ae352fe47bdab850d68357
1387 Author: Aliaksey Kandratsenka <[email protected]>
1388 Date: Fri Jul 14 01:56:28 2023 -0400
1389
1390 replace pthread_once usages with simple spinlock-backed flag
1391
1392 Our perftools_pthread_once implementation has some ifdefs and what
1393 not. I.e. some OSes have complications around early usage and windows
1394 is windows. But in fact we can implement trivial spinlock-supported
1395 implementation and depend on nothing.
1396
1397 Update issue #1110
1398
1399 M src/base/spinlock.h
1400 M src/debugallocation.cc
1401 M src/malloc_hook.cc
1402 M src/maybe_threads.cc
1403 M src/maybe_threads.h
1404 M src/profile-handler.cc
1405 M src/windows/port.cc
1406 M src/windows/port.h
1407
1408 commit 215b2d939aebc572d4e44b97fa2b82b992e6fe0c
1409 Author: Aliaksey Kandratsenka <[email protected]>
1410 Date: Fri Jul 14 01:00:32 2023 -0400
1411
1412 drop disabled windows-specific spinlock code
1413
1414 M src/windows/port.h
1415
1416 commit 37fd4adab461c2005b9855ad3d23945590155ae7
1417 Author: Aliaksey Kandratsenka <[email protected]>
1418 Date: Thu Jul 13 20:54:38 2023 -0400
1419
1420 use regular off_t on 64-bit linuxes
1421
1422 off64_t is legacy of the past.
1423
1424 Update github issue #1407
1425
1426 M src/malloc_hook_mmap_linux.h
1427
1428 commit 4a923a6b36995adf633fda63d28be05979666920
1429 Author: Gabriel Marin <[email protected]>
1430 Date: Thu Aug 23 18:31:16 2018 +0000
1431
1432 tcmalloc: enable large object pointer offset check
1433
1434 Original CL: https://chromiumcodereview.appspot.com/10391178
1435
1436 1. Enable large object pointer offset check in release build.
1437 Following code will now cause a check error:
1438 char* p = reinterpret_cast<char*>(malloc(kMaxSize + 1));
1439 free(p + 1);
1440
1441 2. Remove a duplicated error reporting function
1442 "DieFromBadFreePointer",
1443 can use "InvalidGetAllocatedSize".
1444
1445 Reviewed-on: https://chromium-review.googlesource.com/1184335
1446 [[email protected]] removed some unrelated formatting changes
1447 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
1448
1449 M src/internal_logging.h
1450 M src/tcmalloc.cc
1451
1452 commit 4b3fc02082a1f89268f44ea2d470f14375bd981e
1453 Author: Aliaksey Kandratsenka <[email protected]>
1454 Date: Thu Jul 13 18:56:19 2023 -0400
1455
1456 fix clang barking at debugallocator's magic2_ field
1457
1458 It barks because we access the field in a special way through address
1459 computation (see magic2_addr), so it is indeed not accessed
1460 directly. To emphasize that it comes directly after size2_ we simply
1461 converted size2 and magic2_ into explicit array size_and_magic2_.
1462
1463 M src/debugallocation.cc
1464
1465 commit c29e3059dd506f48e15de8af446d2bd89b5867a9
1466 Author: Aliaksey Kandratsenka <[email protected]>
1467 Date: Thu Jul 13 18:57:22 2023 -0400
1468
1469 mark CheckCachedSizeClass as used
1470
1471 It is only used from inside ASSERT and clang doesn't like it being
1472 declared but unused when NDEBUG is set.
1473
1474 M src/tcmalloc.cc
1475
1476 commit f7c6835facf58026f6db493433a0837c9dca5b97
1477 Author: Aliaksey Kandratsenka <[email protected]>
1478 Date: Thu Jul 13 18:21:24 2023 -0400
1479
1480 drop unused linuxthreads.cc:local_itoa function
1481
1482 M src/base/linuxthreads.cc
1483
1484 commit 5a3a324459558f793da77bc4c56eaeadedf9dca7
1485 Author: Aliaksey Kandratsenka <[email protected]>
1486 Date: Thu Jul 13 18:15:25 2023 -0400
1487
1488 try to avoid certain warnings in tests
1489
1490 Our tests do explicitly trigger certain "bad" cases. And compilers are
1491 getting increasingly smarter, so they start giving us warnings. People
1492 dislike warnings, so lets try to disable them specifically for unit
1493 tests.
1494
1495 Refers to issue #1401
1496
1497 M src/config_for_unittests.h
1498 M src/tests/debugallocation_test.cc
1499 M src/tests/profile-handler_unittest.cc
1500
1501 commit e63d2f855de247f78ec11115767a33f689fb9542
1502 Author: Aliaksey Kandratsenka <[email protected]>
1503 Date: Thu Jul 13 17:45:09 2023 -0400
1504
1505 avoid semi-bogus use-after-free warning
1506
1507 In heap checker we actually use test_str address after deletion, to
1508 verify at runtime that heap checker is indeed tracking deletions. But
1509 gcc barks. So lets hide allocation from it by calling
1510 tc_newarray/tc_deletearray.
1511
1512 Refers to #1401
1513
1514 M src/heap-checker.cc
1515
1516 commit 9d46f1e8904cfe38f45716ef65e36d611ce8fe21
1517 Author: Aliaksey Kandratsenka <[email protected]>
1518 Date: Thu Jul 13 17:50:55 2023 -0400
1519
1520 exercise ASSERTs as part of pagemap_unittest
1521
1522 This fixes couple gcc warnings (array bounds) when accessing page map
1523 radix tree.
1524
1525 Refers to #1401
1526
1527 M src/tests/pagemap_unittest.cc
1528
1529 commit d2bacb8e36b42af8e889c6c191d7842e3dd8d569
1530 Author: Aliaksey Kandratsenka <[email protected]>
1531 Date: Thu Jul 13 17:49:23 2023 -0400
1532
1533 help compiler see that CHECK_CONDITION failures are fatal
1534
1535 Some compiler warnings are caused by this lack of information to
1536 compiler.
1537
1538 Refers to #1401
1539
1540 M src/internal_logging.h
1541
1542 commit 2ffa4e0841a7eea2906be64a66f36f798559a7a0
1543 Author: Aliaksey Kandratsenka <[email protected]>
1544 Date: Thu Jul 13 16:07:17 2023 -0400
1545
1546 unbreak static initialization of initial malloc hooks
1547
1548 g++ 13 somehow considers reinterpret_cast<uintptr_t>(InitialNewHook)
1549 as non-constexpr. It then generates runtime constructor for malloc
1550 hooks, which run too late. That broke heap checker and heap profiler.
1551
1552 We fix by making array of hooks inside malloc to have proper function
1553 pointer type, which lets us avoid reinterpret_cast which makes
1554 construction compile-time again.
1555
1556 M src/malloc_hook-inl.h
1557 M src/malloc_hook.cc
1558 M src/tests/malloc_hook_test.cc
1559
1560 commit 55d559f7c9ea1715db9f0422d589efd7bea6481a
1561 Author: Aliaksey Kandratsenka <[email protected]>
1562 Date: Thu Jul 13 16:04:14 2023 -0400
1563
1564 fix flakiness of profile-handler_unittest
1565
1566 When checking if timer is enabled, we should look at
1567 it_interval.
1568
1569 M src/tests/profile-handler_unittest.cc
1570
1571 commit 5ba98e397cc67a3539c40692dfa51ca26b20519c
1572 Author: Aliaksey Kandratsenka <[email protected]>
1573 Date: Thu Jul 13 16:03:43 2023 -0400
1574
1575 when CHECK_OP fails print filename and line number
1576
1577 M src/base/logging.h
1578
1579 commit 9d781627bb79dc164c338c40b8085db10326d9dc
1580 Author: Aliaksey Kandratsenka <[email protected]>
1581 Date: Mon Jul 10 01:58:42 2023 -0400
1582
1583 avoid calling memory allocator under profile handler signal_lock_
1584
1585 If we don't avoid there is possibility of deadlock. Like this:
1586
1587 *) thread A runs malloc and happens to have some lock(s) there. It
1588 gets profiling signal.
1589
1590 *) thread B runs ProfileHandler::UnregisterCallback or similar and has
1591 both control_lock_ and signal_lock_.
1592
1593 *) thread B calls memory allocator (it used to delete token under
1594 lock) and hits the malloc lock, so waits
1595
1596 *) thread A that has malloc lock wants to get signal_lock_ and waits
1597 too.
1598
1599 Correct behavior is since we grab signal_lock_ from pretty much
1600 anywhere as part of signal handler, it then implies transitively that
1601 anything that grabs that lock cannot call into anything that grabs any
1602 other lock (e.g. including malloc). I.e. signal_lock_ "nests" under
1603 every other lock in the program.
1604
1605 So we refector the code some. When removing callbacks we simply copy
1606 entire list with right token removed. Then grab signal_lock_ and only
1607 swap copy with old callback list. Then signal_lock_ lock is released
1608 and all the memory freeing operations are performed.
1609
1610 When adding callback we utilize splice to also avoid anything "heavy"
1611 under signal_lock_.
1612
1613 And as part of that change we're now able to have UpdateTimer method
1614 only needing control_lock_ instead of signal_lock_.
1615
1616 We add somewhat elaborate unit test that was able to catch the issue
1617 without fix and now works fine.
1618
1619 Closes github issue #412
1620
1621 M src/profile-handler.cc
1622 M src/tests/profile-handler_unittest.cc
1623
1624 commit c44e44cecb64686e8157483b81b0dc29614c935d
1625 Author: Aliaksey Kandratsenka <[email protected]>
1626 Date: Mon Jul 10 00:20:13 2023 -0400
1627
1628 prevent potential corruption of errno in thread lister
1629
1630 See
1631 https://github.com/gperftools/gperftools/issues/500#issuecomment-133814659
1632 for explanations. We of course evolved quite a bit since that comment,
1633 but general issue still applies.
1634
1635 M src/base/linuxthreads.cc
1636
1637 commit 2b5743ffa60affcdc796dcf4bf6c4f5fff14c4de
1638 Author: Aliaksey Kandratsenka <[email protected]>
1639 Date: Sun Jul 9 18:05:01 2023 -0400
1640
1641 amputate tcmalloc unittest trylock counting
1642
1643 We had this logic to ensure that TesterThread succeeds at least 50% of
1644 times grabbing passed objects lock, apparently to ensure that multiple
1645 threads actually participate. But this check occasionally flakes. It
1646 is totally possible that lock owner gets preempted while holding lock
1647 and other threads accumulate gazillion of failed trylocks. So while it
1648 is nice to test this somehow, flakes are bad enough, so we drop this
1649 test.
1650
1651 M src/tests/tcmalloc_unittest.cc
1652
1653 commit 2d70ea9ad2b90b85db81e78e3c4e0b9e16e7977b
1654 Author: Chris Cambly <[email protected]>
1655 Date: Fri May 14 11:11:48 2021 -0400
1656
1657 initial batch of changes to enable AIX in 32-bit and 64-bit
1658
1659 - Some small automake changes. Add libc++ for AIX instead of libstdc++
1660 - Add the interface changes for AIX:User-defined malloc replacement
1661 - Add code to avoid use of pthreads library prior to its
1662 initialization
1663 - Some small changes to the unittest case
1664 - Update INSTALL for AIX
1665
1666 [[email protected]]: lower-case/de-trailing-dot for commit
1667 subject line
1668 [[email protected]]: rebase
1669 [[email protected]]: amputate unused AM_CONDITIONAL for AIX
1670 [[email protected]]: explicitly mention libc_override_aix.h
1671 in Makefile.am
1672
1673 M INSTALL
1674 M Makefile.am
1675 M configure.ac
1676 A src/libc_override_aix.h
1677 M src/libc_override_gcc_and_weak.h
1678 M src/maybe_threads.cc
1679 M src/stacktrace_powerpc-darwin-inl.h
1680 M src/static_vars.cc
1681 M src/tests/tcmalloc_unittest.cc
1682
1683 commit 26927d13334b7a06d3ec9810b7f68a9428c07223
1684 Author: Aliaksey Kandratsenka <[email protected]>
1685 Date: Sun Jul 9 16:11:51 2023 -0400
1686
1687 clean up unused link dependencies for malloc_extension_c_test
1688
1689 We used to explicitly link to libstdc++, libm and even libpthread, but
1690 this should be handled by libtool since those are dependencies of
1691 libtcmalloc_minimal. What also helps is we now build everything with
1692 C++ compiler, not C. So libstdc++ or (libc++) dependency doesn't need
1693 to be added at all, even if libtool for some reason fails to handle
1694 it.
1695
1696 M Makefile.am
1697
1698 commit ea47513a8f8b1b96e0c0e9d776503e2245e45efd
1699 Author: zhangyiru <[email protected]>
1700 Date: Fri Oct 15 18:21:36 2021 -0400
1701
1702 avoid memleak exceed int range when using heap-checker
1703
1704 [[email protected]] format size_t with %zu
1705 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
1706
1707 M src/heap-profile-table.cc
1708
1709 commit 4ac4d88a99c697fe24fd7b356fb57006c7aaeb57
1710 Author: tigeran <[email protected]>
1711 Date: Tue Jan 10 10:17:30 2023 +0800
1712
1713 Remove unused using declaration
1714
1715 std::sort in `heap-profiler.cc` is not in use, so remove it.
1716
1717 M src/heap-profiler.cc
1718
1719 commit 4e24f4adaa1f0ad982daadb118a5dde321827ad3
1720 Author: Aliaksey Kandratsenka <[email protected]>
1721 Date: Tue Jul 4 00:37:52 2023 -0400
1722
1723 fix overflow in c-f-l populate when span address is close to top
1724
1725 This fixes github issue #1323.
1726
1727 When populating span with linked list of objects ptr + size <= limit
1728 condition could overflow before comparing to limit. So fixed code
1729 carefully tests limit. We also produce version with
1730 __builtin_add_overflow since this is semi-hot loop and we want it to
1731 be fast.
1732
1733 M src/central_freelist.cc
1734
1735 commit f15425dc9916ea451a3915a8311c3bc248b225c0
1736 Author: Aliaksey Kandratsenka <[email protected]>
1737 Date: Mon Jul 3 17:36:47 2023 -0400
1738
1739 implement SafeStrError and use it inside strerror
1740
1741 This fixes issue #1371
1742
1743 From time to time things file inside tcmalloc guts where calling to
1744 malloc is not safe. Regular strerror does locale bits, so will
1745 occasionally open files/malloc/etc. We avoid this by using our own
1746 "safe" variant that hardcodes names of all POSIX errno constants.
1747
1748 M .gitignore
1749 M CMakeLists.txt
1750 M Makefile.am
1751 M src/debugallocation.cc
1752 M src/heap-checker.cc
1753 M src/memfs_malloc.cc
1754 A src/safe_strerror.cc
1755 A src/safe_strerror.h
1756 A src/tests/safe_strerror_test.cc
1757
1758 commit 88d0fd5a3bafb6f82e5786a9c70b4883f75882e7
1759 Author: Aliaksey Kandratsenka <[email protected]>
1760 Date: Mon Jul 3 17:29:13 2023 -0400
1761
1762 remove dead remains of arm_instruction_set_select header
1763
1764 M Makefile.am
1765
1766 commit d9cecd6e42d7503b9fe2592ab6d016d5d5832014
1767 Author: Aliaksey Kandratsenka <[email protected]>
1768 Date: Mon Jul 3 16:45:00 2023 -0400
1769
1770 print errno whenever dumping heap profile fails for some reason
1771
1772 Refers to issue #1116
1773
1774 M src/heap-profiler.cc
1775
1776 commit fe5ba4b5242c143661989a15a981079957835c44
1777 Author: Aliaksey Kandratsenka <[email protected]>
1778 Date: Mon Jul 3 16:20:26 2023 -0400
1779
1780 use sys/ptrace.h instead of linux/ptrace.h
1781
1782 As was pointed out at
1783 https://github.com/gperftools/gperftools/pull/1329 there is no need
1784 for us to depend on linux-headers thingy. Libc headers should be
1785 enough.
1786
1787 M src/heap-checker.cc
1788
1789 commit 55e798623fabbdaa67d68959008ba9fcd17ad024
1790 Author: Aliaksey Kandratsenka <[email protected]>
1791 Date: Mon Jul 3 15:31:21 2023 -0400
1792
1793 unbreak compilation with --disable-cpu-profiler
1794
1795 M src/stacktrace_generic_fp-inl.h
1796
1797 commit dd89dc7d01e20b83376e30aa0e52cb83c0a8e55e
1798 Author: Aliaksey Kandratsenka <[email protected]>
1799 Date: Mon Jul 3 15:29:56 2023 -0400
1800
1801 install compat headers and .pc files only with matching libs
1802
1803 Thix closes issue #1356
1804
1805 M Makefile.am
1806
1807 commit a63ea13b20baee6df1468c6dfaf8b339471f461b
1808 Author: Ali Saidi <[email protected]>
1809 Date: Fri Jul 30 12:19:02 2021 -0500
1810
1811 Add an arm64 implementation for SpinlockPause()
1812
1813 This patch adds an arm64 implementation of the SpinlockPause()
1814 function
1815 allowing the core to adaptively spin to acquire the lock and improving
1816 performance in multi-threaded environments where the locks can
1817 be contended.
1818
1819 From experience with other projects, we've found a single isb is
1820 the closest
1821 corollary to the rep; nop timing or x86.
1822
1823 Overall, across thirty runs, the binary_trees_shared benchmark
1824 improves 6% in
1825 average, median, and P90 runtime with this change.
1826
1827 M src/base/spinlock.cc
1828
1829 commit 3c0738ac1cbe51da438ca01ba74eb751f7080372
1830 Author: Sergey Fedorov <[email protected]>
1831 Date: Mon Jul 3 14:40:59 2023 +0800
1832
1833 libc_override_osx.h: a small fix for ppc
1834
1835 M src/libc_override_osx.h
1836
1837 commit cc4e289a8360810a86efd10fc6df6463778a905a
1838 Author: Aliaksey Kandratsenka <[email protected]>
1839 Date: Mon Jul 3 13:02:59 2023 -0400
1840
1841 drop weakening from cmake build
1842
1843 Weakening is optional and in github issue #1392 we apparently tried to
1844 weaken on windows and failed. So lets not even try.
1845
1846 M CMakeLists.txt
1847 D cmake/FindObjcopyWithWeaken.cmake
1848
1849 commit 44eb0ee83c3a485cf6af6366d06430fd162d0fa9
1850 Author: Aliaksey Kandratsenka <[email protected]>
1851 Date: Mon Jul 3 12:47:35 2023 -0400
1852
1853 bump heap profiler stats fields to 64 bit
1854
1855 People actually seen overflows there. Fixes github issue #1303
1856
1857 M src/heap-checker.cc
1858 M src/heap-profile-stats.h
1859 M src/heap-profile-table.cc
1860 M src/heap-profiler.cc
1861
1862 commit 972c12f77dfe27253d966a1bb58d10efb79c34c8
1863 Author: Aliaksey Kandratsenka <[email protected]>
1864 Date: Sun Jul 2 19:21:21 2023 -0400
1865
1866 refactor stacktrace.cc and drop x86 backtracer
1867
1868 We had plenty of old and mostly no more correct i386 cruft. Now that
1869 generic_fp backtracer covers i386 just fine, we can drop explicit x86
1870 backtracer.
1871
1872 With that we refactored and simplified stacktrace.cc mostly around
1873 picking default implementation, but also adding few more minor
1874 cleanups.
1875
1876 M CMakeLists.txt
1877 M Makefile.am
1878 M src/getpc.h
1879 M src/stacktrace.cc
1880 M src/stacktrace_impl_setup-inl.h
1881 D src/stacktrace_x86-inl.h
1882 M src/tests/stacktrace_unittest.cc
1883
1884 commit d9b178695f2a60f20fd2ebc6cc50c15e2b6a6a86
1885 Author: Aliaksey Kandratsenka <[email protected]>
1886 Date: Sun Jul 2 19:43:44 2023 -0400
1887
1888 support more OSes in generic-fp
1889
1890 We're still x86+arm+riscv only, but netbsd and freebsd work too. OSX
1891 as well.
1892
1893 M src/stacktrace.cc
1894 M src/stacktrace_generic_fp-inl.h
1895
1896 commit 4b78ffd03cb9a2c3bb7848f766afc964987a5c82
1897 Author: Aliaksey Kandratsenka <[email protected]>
1898 Date: Sun Jul 2 19:35:46 2023 -0400
1899
1900 try building with -fno-omit-frame-pointer -momit-leaf-frame-pointer
1901
1902 The idea is -momit-leaf-frame-pointer gives us performance pretty much
1903 same as fully omitting frame pointers. And having frame pointers
1904 elsewhere allows us to support cases when user's code is built with
1905 frame pointers. We also pass all tests with
1906 TCMALLOC_STACKTRACE_METHOD=generic_fp (not just libunwind or libgcc).
1907
1908 M CMakeLists.txt
1909 M Makefile.am
1910 M configure.ac
1911 M src/heap-checker.cc
1912
1913 commit e5ac219780ec940904bc0bafea4bdb59f9e0b650
1914 Author: Aliaksey Kandratsenka <[email protected]>
1915 Date: Sun Jul 2 19:31:27 2023 -0400
1916
1917 restore unwind-bench
1918
1919 We previously deleted it, since it wasn't portable enough. But
1920 unportable bits are now ifdef-ed out, so we can return it.
1921
1922 M .gitignore
1923 M Makefile.am
1924 A benchmark/unwind_bench.cc
1925
1926 commit cf046c84214cd9e21a836f840dad23a0f8b50665
1927 Author: Aliaksey Kandratsenka <[email protected]>
1928 Date: Tue Jun 27 23:30:01 2023 -0400
1929
1930 extend generic frame pointer backtracer to support x86-32 (aka x32)
1931
1932 M src/stacktrace_generic_fp-inl.h
1933
1934 commit f56c27910a768a4c2b47026a13f507a89c94f491
1935 Author: Aliaksey Kandratsenka <[email protected]>
1936 Date: Tue Jun 27 23:14:38 2023 -0400
1937
1938 extend generic frame pointer backtracer to support i386
1939
1940 M src/stacktrace.cc
1941 M src/stacktrace_generic_fp-inl.h
1942
1943 commit 58d6842576b35f4ce81d5d7dde389968924a590f
1944 Author: Aliaksey Kandratsenka <[email protected]>
1945 Date: Tue Jun 27 21:43:43 2023 -0400
1946
1947 more coverage for stacktrace_unittest
1948
1949 M src/tests/stacktrace_unittest.cc
1950
1951 commit 25698cd1b811ff8dca54a236571e2a4adba2ed84
1952 Author: Aliaksey Kandratsenka <[email protected]>
1953 Date: Tue Jun 27 16:37:59 2023 -0400
1954
1955 improve diagnostics for stacktrace_unittest
1956
1957 M Makefile.am
1958 M src/tests/stacktrace_unittest.cc
1959
1960 commit a25e7fa8b0564b17cd592a64b618ea451b990601
1961 Author: Aliaksey Kandratsenka <[email protected]>
1962 Date: Sun Jul 2 18:11:51 2023 -0400
1963
1964 cleanup cmake's config.h stuff
1965
1966 Some header defines were not cmakedefine01.
1967
1968 M cmake/config.h.in
1969
1970 commit 88d7e65cc2bb0c5e92892d41eec1e4ea9a7bd6fa
1971 Author: Aliaksey Kandratsenka <[email protected]>
1972 Date: Tue Jun 27 14:08:14 2023 -0400
1973
1974 drop unused libtool target in our Makefile.am
1975
1976 Not sure what it was for, but it is not useful today.
1977
1978 M Makefile.am
1979
1980 commit ae4aafa4689289e42bce4063720d4955b575c176
1981 Author: Aliaksey Kandratsenka <[email protected]>
1982 Date: Wed Jun 28 00:07:49 2023 -0400
1983
1984 freebsd+x86-64 pc-from-ucontext is not untested
1985
1986 M cmake/PCFromUContext.cmake
1987 M m4/pc_from_ucontext.m4
1988
1989 commit cddd759bd1b1ef63d51373afcda76b4cd1dd1a5c
1990 Author: Aliaksey Kandratsenka <[email protected]>
1991 Date: Mon Jun 26 20:33:30 2023 -0400
1992
1993 disable libgcc stack trace capturing on freebsd
1994
1995 M CMakeLists.txt
1996 M configure.ac
1997
1998 commit 4c72b14d35529e989f761050954dc6e4e6d1ca7f
1999 Author: Aliaksey Kandratsenka <[email protected]>
2000 Date: Mon Jun 26 17:37:19 2023 -0400
2001
2002 support proc maps iterator on NetBSD
2003
2004 Turns out it's procfs is available by default, and is fairly similar
2005 to Linux's. So we can just reuse our Linux codes. All tests now pass
2006 on NetBSD!
2007
2008 M src/base/sysinfo.cc
2009
2010 commit 7ffd35a54b13238b408dc773282fe64adb30aeb8
2011 Author: Aliaksey Kandratsenka <[email protected]>
2012 Date: Mon Jun 26 19:57:48 2023 -0400
2013
2014 correctly detect and link to backtrace_symbols
2015
2016 BSDs need -lexecinfo
2017
2018 M Makefile.am
2019 M configure.ac
2020
2021 commit 5b86aa0a51dfaa252c31f971cae487f2e0357955
2022 Author: Aliaksey Kandratsenka <[email protected]>
2023 Date: Mon Jun 26 18:01:58 2023 -0400
2024
2025 reimplement GetProgramInvocationName via reading /proc/self/exe
2026
2027 This enables us to support NetBSD, as it lacks
2028 program_invocation_name, but has fairly full-featured procfs.
2029
2030 M src/symbolize.cc
2031
2032 commit 50b5219635dafe08e1170956fb025bf1bf1f958b
2033 Author: Aliaksey Kandratsenka <[email protected]>
2034 Date: Mon Jun 26 14:30:05 2023 -0400
2035
2036 don't NO_INTR close
2037
2038 This is wrong. See man 2 close.
2039
2040 M src/base/logging.cc
2041 M src/base/sysinfo.cc
2042 M src/tests/profiledata_unittest.cc
2043
2044 commit 7dd1b823786f5953c65bdfb448c7b36447162ec9
2045 Author: Aliaksey Kandratsenka <[email protected]>
2046 Date: Mon Jun 26 14:28:02 2023 -0400
2047
2048 simplify project by making it C++-only
2049
2050 I.e. no need for any AC_LANG_PUSH stuff in configure. Most usefully,
2051 only CXXFLAGS needs to be set now when you need to tweak compile
2052 flags.
2053
2054 M CMakeLists.txt
2055 M Makefile.am
2056 R100 benchmark/run_benchmark.c benchmark/run_benchmark.cc
2057 M configure.ac
2058 R096 src/base/dynamic_annotations.c src/base/dynamic_annotations.cc
2059 R099 src/tests/malloc_extension_c_test.c
2060 src/tests/malloc_extension_c_test.cc
2061 M vsprojects/addressmap_unittest/addressmap_unittest.vcxproj
2062 M vsprojects/addressmap_unittest/addressmap_unittest.vcxproj.filters
2063 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj
2064 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj.filters
2065 M
2066 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj
2067 M
2068 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj.filters
2069 M vsprojects/page_heap_test/page_heap_test.vcxproj
2070 M vsprojects/page_heap_test/page_heap_test.vcxproj.filters
2071
2072 commit cf28e0356782fab54f51e4a394e30642c7371bc5
2073 Author: Aliaksey Kandratsenka <[email protected]>
2074 Date: Tue Jun 27 01:20:01 2023 -0400
2075
2076 correctly order weakening step to avoid race
2077
2078 Previously we allowed test programs to be linked at the same time as
2079 weakening is performed, rewriting the .a archives. So lets be more
2080 explicit. We weaken after all-am (which "runs" everything including
2081 libraries and programs), but before all target.
2082
2083 M Makefile.am
2084
2085 commit a39073886a8e06f441a97b6f2cfb833157dac6c6
2086 Author: Aliaksey Kandratsenka <[email protected]>
2087 Date: Mon Jun 26 13:23:20 2023 -0400
2088
2089 unbreak symbol weakening
2090
2091 It is kinda minor feature, and apparently we never had it working. But
2092 is a nice to have. Allows our users to override malloc/free/etc while
2093 still being able to link to us (for tc_malloc for example). With
2094 broken weakening we had this use-case broken for static library
2095 case. And it should now work.
2096
2097 M Makefile.am
2098 M cmake/FindObjcopyWithWeaken.cmake
2099 M configure.ac
2100
2101 commit 630dac81ea55a775044998de69b17a89fa75cacd
2102 Author: Aliaksey Kandratsenka <[email protected]>
2103 Date: Tue Jun 27 01:43:44 2023 -0400
2104
2105 implement simpler ChangeLog generation for source tarballs
2106
2107 We used ax_generate_changelog which works great. But it made our
2108 makefile require GNU make, which was causing annoyance on bsd systems.
2109
2110 D ChangeLog
2111 M Makefile.am
2112 M configure.ac
2113 D m4/ax_generate_changelog.m4
2114
2115 commit b30b984281bae0b5a921938b885476e23796f377
2116 Author: Aliaksey Kandratsenka <[email protected]>
2117 Date: Sun Jun 25 23:38:55 2023 -0400
2118
2119 [osx] make profiler unittest pass
2120
2121 somehow VerifyIdentical tests fail for some formatting details. But
2122 since it tests --raw feature of our deprecated perl pprof
2123 implementation, which we don't intend supporting anyways, we drop this
2124 test.
2125
2126 There is some details about how wc -l returns stuff and how zsh uses
2127 it to compare to 3. So we now explicitly strip whitespace and it
2128 works.
2129
2130 M src/tests/profiler_unittest.sh
2131
2132 commit 7437e0f612ff54bbf837ea3f7f9b366a60dba629
2133 Author: Aliaksey Kandratsenka <[email protected]>
2134 Date: Sun Jun 25 23:38:44 2023 -0400
2135
2136 [osx] make heap profiler unittest pass
2137
2138 Apparently awk's comparison $2 > 90 doesn't work when $2 is 100.0. I
2139 frobbed it some and it works now.
2140
2141 M src/tests/heap-profiler_unittest.sh
2142
2143 commit 6683e4f6eeebc52c0704500ef9f461454434578c
2144 Author: Aliaksey Kandratsenka <[email protected]>
2145 Date: Sun Jun 25 23:37:44 2023 -0400
2146
2147 [osx] implement native c++ allocation operators on osx
2148
2149 OSX has really screwed performance for it's memory allocator due to
2150 their malloc zones stuff. We and our competitors hook into that
2151 facility instead directly overriding allocation functions. Which has
2152 massive performance overhead. This is, apparently, so that malloc is
2153 same as allocating from default zone.
2154
2155 As a result, osx'es C++ operator new/delete performance is even worse.
2156 Because we call operator new, it calls malloc, which calls
2157 malloc_zone_malloc, which calls more stuff and eventually tc_malloc.
2158
2159 Thankfully, for C++ API there is no need for such "integration"
2160 between 'stock' memory allocator and malloc zones stuff. So we can
2161 hook those APIs directly. This speeds up malloc_bench numbers on OSX
2162 about 3x.
2163
2164 This change also unbreaks couple tests (e.g. heap profiler unittest)
2165 which depend on MallocHook_GetCallerStackTrace to return precisely
2166 stack trace up to memory allocator call.
2167
2168 Performance-wise, OSX still lacks good TLS and there is still one jump
2169 indirection even for operator new/delete API due to lack support of
2170 aliases. So Linux or BSD, or even windows would still be significantly
2171 faster on malloc performance on same hardware.
2172
2173 M src/libc_override_osx.h
2174
2175 commit 11d4253ead0f0db77aac0d6f5779416c7ab9efa2
2176 Author: Aliaksey Kandratsenka <[email protected]>
2177 Date: Sun Jun 25 23:37:10 2023 -0400
2178
2179 [osx] don't crash debugallocator on 0-sized allocations
2180
2181 OSX's malloc zones facility is calling 'usable size' before each (!)
2182 free and uses size to choose between our allocator and something
2183 else. And returning 0 breaks this choice. Which happens if we got
2184 0-sized malloc request, which our tests exercise couple times. So we
2185 don't let debug allocator on OSX to have 0 sized chunks. This unbreaks
2186 debug allocation tests on OSX.
2187
2188 M src/debugallocation.cc
2189 M src/tests/debugallocation_test.cc
2190
2191 commit b0c2ab298adf2be8d9fe23cd7f166cae875e9549
2192 Author: Aliaksey Kandratsenka <[email protected]>
2193 Date: Sun Jun 25 23:36:08 2023 -0400
2194
2195 [osx] place malloc-zone-related functions to google_malloc section
2196
2197 On OSX we hook into "system's" malloc through this zones facility, so
2198 we need to place those those interface functions into google_malloc
2199 section. This largely unbreaks MallocHook_GetCallerStackTrace on OSX.
2200
2201 M src/base/basictypes.h
2202 M src/libc_override_osx.h
2203
2204 commit 531ca4fdcaebd225b4616fe8a693d26a89019792
2205 Author: Aliaksey Kandratsenka <[email protected]>
2206 Date: Sun Jun 25 23:35:26 2023 -0400
2207
2208 [osx] unbreak LowLevelAlloc
2209
2210 For some reason clang shipped with osx miscompiles LowLevelAlloc when
2211 it tries to place Alloc methods into malloc_hooks section. But since
2212 we don't really need that placement on OSX, we can simply drop that
2213 section placement.
2214
2215 M src/base/low_level_alloc.h
2216
2217 commit e78238d94d41764dc72edb37e3d6510255d3dcd4
2218 Author: Aliaksey Kandratsenka <[email protected]>
2219 Date: Wed Jun 21 10:39:38 2023 -0400
2220
2221 reworked heap leak checker for more portability
2222
2223 In most practical terms, this expands "official" heap leak checker
2224 support to Linux/arm64 and Linux/riscv (mips-en and legacy arm are
2225 likely to work & pass tests too now).
2226
2227 The code is now explicitly Linux-only, without trying to pretend
2228 otherwise. Main goal of this change is to finally amputate
2229 linux_syscall_support.h, which we historically had trouble maintaining
2230 well. Biggest challenge was around thread listing facility which uses
2231 clone (ptrace explicitly fails between threads) and that causes
2232 difficulties around parent and child tasks sharing
2233 errno. linux_syscall_support stuff had special feature to "redirect"
2234 errno accesses. But it caused us for more trouble. We switched to
2235 regular syscalls, and errno stamping avoidance is now simply via
2236 careful programming.
2237
2238 A number of other cleanups is made (such us thread finding codes in
2239 procfs which clearly was built for some ages old and odd kernels).
2240
2241 sem_post/sem_wait synchronization was previously potentially prone to
2242 deadlock (if parent died at bad time). We now use pipe pair for this
2243 synchronization and it is fully robust.
2244
2245 M CMakeLists.txt
2246 M Makefile.am
2247 M cmake/config.h.in
2248 M configure.ac
2249 M docs/heap_checker.html
2250 D src/base/elfcore.h
2251 D src/base/linux_syscall_support.h
2252 M src/base/linuxthreads.cc
2253 M src/base/linuxthreads.h
2254 D src/base/thread_lister.c
2255 D src/base/thread_lister.h
2256 M src/gperftools/heap-checker.h
2257 M src/heap-checker.cc
2258 M src/tests/heap-checker_unittest.cc
2259 M src/windows/config.h
2260
2261 commit 21869679878a3d578adb6bc4ed8eaacb9f86ac3c
2262 Author: Aliaksey Kandratsenka <[email protected]>
2263 Date: Wed Jun 21 10:39:35 2023 -0400
2264
2265 fix heap checker unittest
2266
2267 We had shell wrapper for heap checker unittest, but it failed to deal
2268 with heap-checker-debug variant. So we now posix_spawn from .cc test
2269 instead.
2270
2271 M CMakeLists.txt
2272 M Makefile.am
2273 M src/tests/heap-checker_unittest.cc
2274 D src/tests/heap-checker_unittest.sh
2275
2276 commit 1c3f0dda1b3f4a5b6ce720604bcea1680e7cbb9a
2277 Author: Aliaksey Kandratsenka <[email protected]>
2278 Date: Wed Jun 21 10:39:31 2023 -0400
2279
2280 ensure that heap checker initialization actually calls malloc
2281
2282 I.e. all recent compilers (both gcc and clang) optimize out delete(new
2283 int) sequence to nothing. We call to tc_new and tc_delete directly to
2284 suppress this optimization.
2285
2286 M src/heap-checker-bcad.cc
2287
2288 commit b96dc99dbf322195389bbb43b58941c43c744a49
2289 Author: Aliaksey Kandratsenka <[email protected]>
2290 Date: Wed Jun 21 10:39:28 2023 -0400
2291
2292 one trivial config cleanup
2293
2294 M configure.ac
2295 M src/profile-handler.h
2296
2297 commit b7e47a77c0382544eeef50e8e004932bbcae1c0e
2298 Author: Aliaksey Kandratsenka <[email protected]>
2299 Date: Wed Jun 21 10:39:25 2023 -0400
2300
2301 simplify heap checker building default to be Linux-only
2302
2303 This also fixes cmake and freebsd where previously check for freebsd
2304 wasn't working.
2305
2306 M CMakeLists.txt
2307 M configure.ac
2308
2309 commit b58cbd2e23e582059d8d554354d30ef40a93724f
2310 Author: Aliaksey Kandratsenka <[email protected]>
2311 Date: Wed Jun 21 10:39:21 2023 -0400
2312
2313 make preamble patcher build and run (win64 only)
2314
2315 First, ml64 (amd64 version of masm) doesn't support /coff. Second, it
2316 also doesn't support (nor use) .model directive.
2317
2318 Sadly, asm is inherently amd64-only, so this entire test won't build
2319 in i386 configuration.
2320
2321 M src/windows/shortproc.asm
2322 M vsprojects/preamble_patcher_test/preamble_patcher_test.vcxproj
2323
2324 commit 54605b8a5807fe893a4b9ff8def982a264cdc6c9
2325 Author: Aliaksey Kandratsenka <[email protected]>
2326 Date: Wed Jun 21 10:39:18 2023 -0400
2327
2328 amputate old atomic ops implementation
2329
2330 M .gitignore
2331 M CMakeLists.txt
2332 M Makefile.am
2333 M cmake/config.h.in
2334 M configure.ac
2335 D src/base/arm_instruction_set_select.h
2336 D src/base/atomicops-internals-arm-generic.h
2337 D src/base/atomicops-internals-arm-v6plus.h
2338 D src/base/atomicops-internals-gcc.h
2339 D src/base/atomicops-internals-linuxppc.h
2340 D src/base/atomicops-internals-macosx.h
2341 D src/base/atomicops-internals-mips.h
2342 D src/base/atomicops-internals-windows.h
2343 D src/base/atomicops-internals-x86.cc
2344 D src/base/atomicops-internals-x86.h
2345 D src/base/atomicops.h
2346 M src/base/spinlock.h
2347 M src/base/spinlock_internal.cc
2348 M src/base/spinlock_internal.h
2349 M src/base/spinlock_posix-inl.h
2350 M src/base/vdso_support.cc
2351 M src/malloc_hook-inl.h
2352 D src/tests/atomicops_unittest.cc
2353 M src/windows/config.h
2354 M vsprojects/addressmap_unittest/addressmap_unittest.vcxproj
2355 M vsprojects/addressmap_unittest/addressmap_unittest.vcxproj.filters
2356 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj
2357 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj.filters
2358 M
2359 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj
2360 M
2361 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj.filters
2362 M vsprojects/packed-cache_test/packed-cache_test.vcxproj
2363 M vsprojects/packed-cache_test/packed-cache_test.vcxproj.filters
2364
2365 commit ea0988b02033627b03521435b689483b38600136
2366 Author: Aliaksey Kandratsenka <[email protected]>
2367 Date: Wed Jun 21 10:39:14 2023 -0400
2368
2369 port spinlocks atomic ops usage to std::atomic
2370
2371 M src/base/spinlock.cc
2372 M src/base/spinlock.h
2373 M src/base/spinlock_internal.h
2374 M src/base/spinlock_linux-inl.h
2375 M src/base/spinlock_win32-inl.h
2376
2377 commit 3494eb29d629363b84eb7fcd23f8fc1c049c403e
2378 Author: Aliaksey Kandratsenka <[email protected]>
2379 Date: Wed Jun 21 10:39:11 2023 -0400
2380
2381 switch malloc hooks to std::atomic
2382
2383 This makes our code one step closer to killing unportable custom
2384 atomics implementation. I did manually inspect generated code that
2385 fast-path is identical and slow path's changes of generated code are
2386 cosmetic.
2387
2388 M src/malloc_hook-inl.h
2389 M src/malloc_hook.cc
2390 M src/tests/malloc_hook_test.cc
2391
2392 commit 0e21f36843279581e5fddda97fb797e93c1dc210
2393 Author: Aliaksey Kandratsenka <[email protected]>
2394 Date: Wed Jun 21 10:39:07 2023 -0400
2395
2396 unbreak frame skipping in generic-fp backtrace method
2397
2398 M src/stacktrace_generic_fp-inl.h
2399 M src/tests/stacktrace_unittest.cc
2400
2401 commit 6f3cacf698c1f3aaa5f21812c83541ee4d565147
2402 Author: Aliaksey Kandratsenka <[email protected]>
2403 Date: Wed Jun 21 10:39:04 2023 -0400
2404
2405 stacktrace_unittest: test backtracing from signals more reliably
2406
2407 Somehow I originally choose to segfault to test this, and it caused us
2408 to deal with 'skipping over' null pointer access. Which ended up not
2409 so easy to do semi-portably. But easier way is to do same
2410 itimer/sigprof thing that we do in actual profiler.
2411
2412 We're still somewhat at the mercy of compiler placing code "normally",
2413 but this implementation seems more robust anyways.
2414
2415 M src/tests/stacktrace_unittest.cc
2416
2417 commit 68b442714a86060d1dd95de4c515cc7774e3751e
2418 Author: Aliaksey Kandratsenka <[email protected]>
2419 Date: Wed Jun 21 10:39:00 2023 -0400
2420
2421 stop working around obsolete compilation bug
2422
2423 This was targeting (some obsure and perhaps even Google-internal
2424 version) of now-ancient compiler. No need to keep this anymore.
2425
2426 M src/central_freelist.cc
2427
2428 commit 16889f6ddb51fdae833455ed9d000bc0951fd8d4
2429 Author: Aliaksey Kandratsenka <[email protected]>
2430 Date: Wed Jun 21 10:38:56 2023 -0400
2431
2432 fix minor clang warning
2433
2434 M src/central_freelist.cc
2435
2436 commit 246782e1d62d6c10343a955c102b48c3d507d4cb
2437 Author: Aliaksey Kandratsenka <[email protected]>
2438 Date: Wed Jun 21 10:38:52 2023 -0400
2439
2440 use more noopt in various unit tests
2441
2442 Many our unit tests require specific stack traces or that
2443 allocation/deallocation is not elimininated. As compilation gets more
2444 aggressive (such as when cross-object optimization is enabled during
2445 LTO), we break more of those cases.
2446
2447 This commit extends noopt usage that disables inlining optimization in
2448 specific cases. This fixes github issue #1358.
2449
2450 M src/tests/sampling_test.cc
2451 M src/tests/tcmalloc_large_unittest.cc
2452 M src/tests/tcmalloc_unittest.cc
2453 M src/tests/testutil.h
2454
2455 commit 6bdeec0ecb0ae49a5277897b60492d4f6d0acac4
2456 Author: Aliaksey Kandratsenka <[email protected]>
2457 Date: Wed Jun 21 10:38:44 2023 -0400
2458
2459 mark stacktrace capturing functions as noinline
2460
2461 It is required for LTO builds, specifically it fixes
2462 stacktrace_unittest. This is because API of stacktrace capturing
2463 functions implicitly assumes that those are not inlined.
2464
2465 Refers to github issue #1358.
2466
2467 M src/stacktrace.cc
2468
2469 commit 777182fbdf37f24627410389f829d3bf7d86367a
2470 Author: Aliaksey Kandratsenka <[email protected]>
2471 Date: Wed Jun 21 10:38:40 2023 -0400
2472
2473 enable ATTRIBUTE_NOINLINE for MSVC
2474
2475 M src/base/basictypes.h
2476
2477 commit 9bde8d837b8ed2a17ddd1206ae03e9b2265a85ee
2478 Author: Aliaksey Kandratsenka <[email protected]>
2479 Date: Wed Jun 21 10:38:35 2023 -0400
2480
2481 unbreak building with visual studio clang platform toolset
2482
2483 M src/windows/mini_disassembler_types.h
2484 M src/windows/patch_functions.cc
2485
2486 commit 0451d21e83483060d6c3e727ff2554692f5bc16e
2487 Author: Aliaksey Kandratsenka <[email protected]>
2488 Date: Wed Jun 21 10:38:31 2023 -0400
2489
2490 use libunwind when it actually works
2491
2492 Previously we blindly tried to use libunwind whenever header is
2493 detected. Even if actually working libunwind library is missing. This
2494 is now fixed, so we attempt to use libunwind when it actually works.
2495
2496 Somehow recent freebsd ships libunwind.h (which seems to belong to
2497 llvm's implementation), but apparently without matching .so. So
2498 then building
2499 and linking failed.
2500
2501 M CMakeLists.txt
2502 M cmake/config.h.in
2503 M configure.ac
2504 M src/stacktrace.cc
2505
2506 commit 90eff0637bf17a8df2665287fc0b523b48e5fa92
2507 Author: Aliaksey Kandratsenka <[email protected]>
2508 Date: Wed Jun 21 10:38:26 2023 -0400
2509
2510 upgrade malloc bench away from std::random_shuffle
2511
2512 C++17 dropped it, so we use C++11 and later std::shuffle
2513 instead. Fixes github issue #1378
2514
2515 M benchmark/malloc_bench.cc
2516
2517 commit b501806f2af8be4b83527cbb7bd6afa040d818d5
2518 Author: Aliaksey Kandratsenka <[email protected]>
2519 Date: Wed Jun 21 23:08:25 2023 -0400
2520
2521 fix malloc_bench on clang
2522
2523 We got bitten again by compiler optimizing out free(malloc(sz))
2524 combination. We replace calls to malloc/free with calls to global
2525 function operator new/delete. Those are apparently forbidden by
2526 standard to be optimized out.
2527
2528 M benchmark/malloc_bench.cc
2529
2530 commit ff68bcab609b1e99a7a0f3f73693b0873c4f3001
2531 Author: Aliaksey Kandratsenka <[email protected]>
2532 Date: Wed Jun 21 23:03:33 2023 -0400
2533
2534 [benchmark] detect iterations overflow
2535
2536 If for some reason benchmark function is too fast (like when it got
2537 optimized out to nothing), we'd previously hang in infinite loop. Now
2538 we'll catch this condition due to iterations count overflowing.
2539
2540 M benchmark/run_benchmark.c
2541
2542 commit e003e91b742e5e65645d80c684d14eacaecdc0f1
2543 Author: Aliaksey Kandratsenka <[email protected]>
2544 Date: Wed Jun 21 10:38:22 2023 -0400
2545
2546 drop dependency on PTHREAD_CREATE_JOINABLE
2547
2548 M CMakeLists.txt
2549 M cmake/config.h.in
2550 M src/tests/profile-handler_unittest.cc
2551 M src/windows/config.h
2552
2553 commit 8fc84c29a34a5feaa928e5027a1c0ba028a57b15
2554 Author: Aliaksey Kandratsenka <[email protected]>
2555 Date: Wed Jun 21 10:38:15 2023 -0400
2556
2557 mix thread stack address into sampler
2558
2559 This fixes test failures on freebsd (assert that sampler is seeded
2560 with non-0 value).
2561
2562 M src/thread_cache.cc
2563
2564 commit d9b14b9da1a3e0fc55286390aa2ab2490d7eee3d
2565 Author: Aliaksey Kandratsenka <[email protected]>
2566 Date: Wed Jun 21 10:38:03 2023 -0400
2567
2568 actually test malloc/free from InitLateMaybeRecursive
2569
2570 This days compilers ~always optimize out free(malloc(N)) to nothing.
2571
2572 M src/static_vars.cc
2573 M src/thread_cache.cc
2574
2575 commit a80ce7b0e1510577c21a38dfb5596ddc27c59bdf
2576 Author: Aliaksey Kandratsenka <[email protected]>
2577 Date: Wed Jun 21 10:37:58 2023 -0400
2578
2579 slow down profiler_unittest some more
2580
2581 On modern machines xoring volatile int is fast enough to occasionally
2582 make specific test case's profiling results not stable enough. So we
2583 now force atomics into the picture which is slightly slower (could be
2584 lots on some platforms and/or optimizer flags).
2585
2586 M src/tests/profiler_unittest.cc
2587
2588 commit c37f6c4c7c83027e920fa422b471203501be7a13
2589 Author: Aliaksey Kandratsenka <[email protected]>
2590 Date: Wed Jun 21 10:37:51 2023 -0400
2591
2592 fix deprecated autoconf bits
2593
2594 As part of that we also upgrade required autoconf version to 2.69
2595 which is what I see in rhel/centos 7 and ubuntu 14.04. Both are old
2596 enough. And, of course, .tar.gz releases still ship "packaged"
2597 configure,
2598 so will work on older distros.
2599
2600 This fixes issue #1335.
2601
2602 M cmake/config.h.in
2603 M configure.ac
2604 M m4/ac_have_attribute.m4
2605 M m4/acx_nanosleep.m4
2606 D m4/acx_pthread.m4
2607 A m4/ax_pthread.m4
2608 M m4/pc_from_ucontext.m4
2609 M m4/program_invocation_name.m4
2610 M src/windows/config.h
2611
2612 commit f06052d7362f5d71e8674d3cf46f761c793321e1
2613 Author: Aliaksey Kandratsenka <[email protected]>
2614 Date: Wed Jun 21 10:22:39 2023 -0400
2615
2616 synchronize ucontext-from-pc tests autoconf -> cmake
2617
2618 Refers to github issue #1398.
2619
2620 M cmake/PCFromUContext.cmake
2621
2622 commit 5e923877aaee4ec4bece60148dadcba7ea1f86e6
2623 Author: Henrik Reinstädtler <[email protected]>
2624 Date: Mon Jun 19 18:11:10 2023 +0200
2625
2626 Fix uc_mcontext for apple silicon
2627
2628 M cmake/PCFromUContext.cmake
2629
2630 commit bf8b714bf5075d0a6f2f28504b43095e2b1e11c5
2631 Author: Aliaksey Kandratsenka <[email protected]>
2632 Date: Mon May 30 21:21:21 2022 -0700
2633
2634 bump version to 2.10
2635
2636 M CMakeLists.txt
2637 M NEWS
2638 M configure.ac
2639 M src/windows/config.h
2640 M src/windows/gperftools/tcmalloc.h
2641
2642 commit 9d44463380716ebad2f8ba13ce8a737f22bb2602
2643 Author: Aliaksey Kandratsenka <[email protected]>
2644 Date: Mon May 30 20:33:18 2022 -0700
2645
2646 Remove basic support for Elbrus 2000 (e2k)
2647
2648 This reverts commit 3b1c60cc4ea737ff151977dbcd164c5e2eb859fb.
2649
2650 M cmake/PCFromUContext.cmake
2651 M m4/pc_from_ucontext.m4
2652 M src/base/basictypes.h
2653 M src/malloc_hook_mmap_linux.h
2654
2655 commit 6c99def347460adbb9f17815feb641cf5474548e
2656 Author: Aliaksey Kandratsenka <[email protected]>
2657 Date: Mon May 30 19:50:19 2022 -0700
2658
2659 avoid casting NULL to uinptr_t
2660
2661 Which somehow fails to compile on musl. Fixes issue #1338.
2662
2663 M src/heap-checker.cc
2664
2665 commit fe85bbdf4cb891a67a8e2109c1c22a33aa958c7e
2666 Author: Jingyun Hua <[email protected]>
2667 Date: Tue Feb 8 20:47:10 2022 +0800
2668
2669 Add support for LoongArch.
2670
2671 Only 64-bit is supported at the moment.
2672
2673 Signed-off-by: Jingyun Hua <[email protected]>
2674
2675 M m4/pc_from_ucontext.m4
2676 M src/base/basictypes.h
2677 M src/base/linux_syscall_support.h
2678 M src/base/linuxthreads.h
2679 M src/malloc_hook_mmap_linux.h
2680 M src/stacktrace.cc
2681 M src/tcmalloc.cc
2682
2683 commit e80652b6279a22bd4a4264e4403cd1981293e82f
2684 Author: Aliaksey Kandratsenka <[email protected]>
2685 Date: Fri Jan 14 23:08:57 2022 -0800
2686
2687 ship cmake bits with tar.gz distribution
2688
2689 Fixes issue #1321.
2690
2691 M Makefile.am
2692
2693 commit 003775aa816cb3fe01215455825245089d423a7c
2694 Author: Volodymyr Nikolaichuk <[email protected]>
2695 Date: Wed Sep 15 01:17:51 2021 +0300
2696
2697 free memory with mmap(PROT_NONE, MAP_FIXED)
2698
2699 M src/system-alloc.cc
2700
2701 commit 852fb6df031560c5e353b497222b76394190e27d
2702 Author: Natale Patriciello <[email protected]>
2703 Date: Thu Nov 18 18:16:40 2021 +0100
2704
2705 Implement free_definite_size
2706
2707 In Monterey, it seems that free_definite_size is required (despite
2708 the documentation saying that is optional). The implementation
2709 just forward the call to free.
2710
2711 Signed-off-by: Natale Patriciello <[email protected]>
2712
2713 M src/libc_override_osx.h
2714
2715 commit c25941200ef4ce39d0774c1332ff7abfbeab7035
2716 Author: stdpain <[email protected]>
2717 Date: Sat Mar 13 12:42:01 2021 +0800
2718
2719 fix cmake gperftools_enable_libunwind invalid
2720
2721 M cmake/config.h.in
2722
2723 commit 43504ab7091c0bcac9a233aa0ee0720a1549791b
2724 Author: Robert Scott <[email protected]>
2725 Date: Sat Feb 22 18:55:46 2020 +0000
2726
2727 tcmalloc page fences: add TCMALLOC_PAGE_FENCE_READABLE option
2728
2729 M src/debugallocation.cc
2730
2731 commit d57a9ea8bc92958733ac1c61a3c4f1ba2e150e96
2732 Author: Matt T. Proud <[email protected]>
2733 Date: Tue Apr 20 11:11:46 2021 +0200
2734
2735 README: replace "golang" moniker with "Go"
2736
2737 This commit replaces the "golang" moniker in the README file with the
2738 appropriate name of "Go" per https://golang.org/doc/faq#go_or_golang.
2739
2740 In this context, no disambiguation is required.
2741
2742 M README
2743
2744 commit f7c6fb6c8e99d6b1b725e5994373bcd19ffdf8fd
2745 Author: Aliaksey Kandratsenka <[email protected]>
2746 Date: Tue Mar 2 19:32:07 2021 -0800
2747
2748 bump version to 2.9.1
2749
2750 M CMakeLists.txt
2751 M NEWS
2752 M configure.ac
2753 M src/windows/config.h
2754 M src/windows/gperftools/tcmalloc.h
2755
2756 commit c2f60400a86a5211bf009941bf1ccf9b2947a7c1
2757 Author: Aliaksey Kandratsenka <[email protected]>
2758 Date: Sun Feb 28 17:52:17 2021 -0800
2759
2760 prefer backtrace() on OSX
2761
2762 It seems to work and _Unwind_Backtrace which is configured otherwise
2763 doesn't.
2764
2765 M CMakeLists.txt
2766 M cmake/DefineTargetVariables.cmake
2767 M configure.ac
2768
2769 commit a015377a54eb09fca4b893ae530187a705164df5
2770 Author: Aliaksey Kandratsenka <[email protected]>
2771 Date: Sun Feb 28 17:47:56 2021 -0800
2772
2773 Set tcmalloc heap limit prior to testing oom
2774
2775 Otherwise it can take long time to OOM on osex.
2776
2777 M src/tcmalloc.cc
2778 M src/tests/tcmalloc_unittest.cc
2779
2780 commit c939dd5531fd95e8af2181ac60e0b6e6230226c8
2781 Author: Aliaksey Kandratsenka <[email protected]>
2782 Date: Sun Feb 28 15:42:00 2021 -0800
2783
2784 correctly check sized delete hint when asserts are on
2785
2786 We previously tested wrong assumption that larger than page size size
2787 classes have addresses aligned on page size. New code is making proper
2788 check of size class.
2789
2790 Also added is unit test coverage for this previously failing
2791 condition. And we now also run "assert-ful" unittests for big tcmalloc
2792 too, not only tcmalloc_minimal configuration.
2793
2794 This fixes github issue #1254
2795
2796 M .gitignore
2797 M Makefile.am
2798 M src/tcmalloc.cc
2799 M src/tests/tcmalloc_unittest.cc
2800
2801 commit 47b5b59ca9b62ee42cf30273cce463aa9cbba8c7
2802 Author: Aliaksey Kandratsenka <[email protected]>
2803 Date: Sun Feb 21 15:16:06 2021 -0800
2804
2805 bump version to 2.9
2806
2807 M CMakeLists.txt
2808 M NEWS
2809 M configure.ac
2810 M src/windows/config.h
2811 M src/windows/gperftools/tcmalloc.h
2812
2813 commit d7cbc8c2ff4ab8dbdc411ab1f25eca459f88502e
2814 Author: Aliaksey Kandratsenka <[email protected]>
2815 Date: Sun Feb 21 14:55:15 2021 -0800
2816
2817 unbreak cmake build
2818
2819 It was trying to use third_party/valgrind.h header which I recently
2820 removed.
2821
2822 M CMakeLists.txt
2823
2824 commit be0bbdb3403685798a7c1d5acf6405d2635fba4c
2825 Author: Aliaksey Kandratsenka <[email protected]>
2826 Date: Sun Feb 21 13:48:11 2021 -0800
2827
2828 amputate various unused bits from elfcore.h
2829
2830 M src/base/elfcore.h
2831
2832 commit 42bab59f25dee53f91c0ba45e95b73754e197c20
2833 Author: Aliaksey Kandratsenka <[email protected]>
2834 Date: Sun Feb 21 13:35:43 2021 -0800
2835
2836 liberate profile handler from linux_syscall_support
2837
2838 We can get thread's tid in more portable way via libc syscall wrapper.
2839
2840 M src/profile-handler.cc
2841
2842 commit 4629511e99f85fa8ca40fcbe105e014a725059d2
2843 Author: Aliaksey Kandratsenka <[email protected]>
2844 Date: Sun Feb 21 13:32:08 2021 -0800
2845
2846 liberate spinlock futex waits from linux_syscall_support includes
2847
2848 We've been using libc syscall wrapper some time now.
2849
2850 M src/base/spinlock_linux-inl.h
2851
2852 commit 2e7094a86215118f3dcd3ad6f2c9099dc409ecc5
2853 Author: Aliaksey Kandratsenka <[email protected]>
2854 Date: Sun Feb 21 13:28:48 2021 -0800
2855
2856 liberate malloc_hook_mmap_linux.h from linux_syscall_support
2857
2858 It was not used some time now, as we're rightfully trusting libc
2859 syscall wrapper instead.
2860
2861 M src/malloc_hook_mmap_linux.h
2862
2863 commit 35301e2e59b30b7379981e60f3046124a2dfd68e
2864 Author: Aliaksey Kandratsenka <[email protected]>
2865 Date: Sun Feb 21 12:40:45 2021 -0800
2866
2867 add missing noopt wrappings around more operator new calls
2868
2869 This fixes tests passing on clang which otherwise eliminates those
2870 new/delete calls so checking for hooks being called failed.
2871
2872 M src/tests/tcmalloc_unittest.cc
2873
2874 commit fa412adfe38ffd3f545a0e10139bd20b38b688e9
2875 Author: Venkatesh Srinivas <[email protected]>
2876 Date: Wed Feb 17 17:50:43 2021 -0800
2877
2878 Fix thread-safety (annotalysis) annotations
2879
2880 tcmalloc contains some thread-safety annotations; however those
2881 annotations have not been exercised for some time, as they used
2882 macros/attributes only supported by a legacy branch of gcc.
2883
2884 Pull request #1251 converted those macros to support modern
2885 versions of clang; this CR fixes the annotations that were
2886 enabled. For the most part, this just requires re-enabling
2887 annotations on member functions that take/release locks. For the
2888 tcmalloc fork (pre-fork and post-fork) handlers, we mark the
2889 functions as exempt from this analysis, as it takes a dynamic
2890 number of locks.
2891
2892 M src/base/low_level_alloc.cc
2893 M src/base/spinlock.h
2894 M src/central_freelist.h
2895 M src/static_vars.cc
2896
2897 commit cc496aecb81ee5966c865f3723743ff02046c5ad
2898 Author: Venkatesh Srinivas <[email protected]>
2899 Date: Wed Feb 17 16:23:17 2021 -0800
2900
2901 tcmalloc: Switch thread-safety annotations to support clang
2902
2903 tcmalloc contains its own copy of thread_annotations.h, wrapper
2904 macros for static thread-safety analysis expressions. These thread-
2905 safety expressions allow asserting (at compile time) that certain
2906 locks are held or excluded or certain data is protected by specific
2907 locks; they are checked at compile-time by recent versions of clang
2908 or a gcc branch (https://gcc.gnu.org/wiki/ThreadSafetyAnnotation).
2909
2910 Convert the #if-guard and macro names from the no-longer-supported
2911 gcc branch's defines & macros to the versions supported by recent
2912 versions of clang.
2913
2914 M src/base/thread_annotations.h
2915
2916 commit 96ba58e19b7b93fcf1db6a0551c9510296ce26da
2917 Author: Aliaksey Kandratsenka <[email protected]>
2918 Date: Sun Feb 14 23:18:57 2021 -0800
2919
2920 bump version to 2.9rc
2921
2922 M CMakeLists.txt
2923 M NEWS
2924 M configure.ac
2925 M src/windows/config.h
2926 M src/windows/gperftools/tcmalloc.h
2927
2928 commit 9ce32aefa9c66698817be50e141c5a39fe3823c8
2929 Author: Aliaksey Kandratsenka <[email protected]>
2930 Date: Sun Feb 14 23:02:28 2021 -0800
2931
2932 upgrade test bot to xenial (ubuntu 16.04 LTS)
2933
2934 Older glibc seemingly doesn't have right unwind info for signal frame,
2935 so fails recently upgraded stacktrace_unittest. But since this version
2936 is not supported anymore, lets just test newer glibc.
2937
2938 M .travis.yml
2939
2940 commit 91ff3114492f368a651ee790c8311eb6fadf3e8e
2941 Author: Aliaksey Kandratsenka <[email protected]>
2942 Date: Sun Feb 14 22:36:55 2021 -0800
2943
2944 don't default to generic_fp without frame pointers
2945
2946 I.e. on x86-64 and riscv.
2947
2948 M src/stacktrace.cc
2949
2950 commit 4cf7dd0a7502849b117e15ad5eb1eb68eab4b6dc
2951 Author: Aliaksey Kandratsenka <[email protected]>
2952 Date: Sun Feb 14 22:28:34 2021 -0800
2953
2954 enable emergency_malloc on all architectures with mmap
2955
2956 It is actually needed for libgcc backtracer from time to time. And
2957 we've seen libunwind to need it too. Plus we've not heard of any
2958 problems with it. So lets just always enable it.
2959
2960 This should fix github issue #1248.
2961
2962 M configure.ac
2963
2964 commit 37087ec5360e821627155d3129f9b55a321a0001
2965 Author: Aliaksey Kandratsenka <[email protected]>
2966 Date: Sun Feb 14 21:37:20 2021 -0800
2967
2968 prefer libunwind on x86-64 even with -fno-omit-frame-pointer
2969
2970 Libunwind is mostly safe on x86-64 and most of OS (e.g. libc) is
2971 compiled without frame pointers anyways. Even when libc is built with
2972 frame pointers, various libc asm routines (e.g. memcpy) are not. And
2973 even for code built with frame pointers we actually sometimes get
2974 SIGPROF inside function prologues, before frame pointer is
2975 established. Which then causes imperfect backtraces.
2976
2977 Libunwind should handle this all (when it doesn't crash or deadlock).
2978
2979 M src/stacktrace.cc
2980
2981 commit f4aa2a435eed63fc047448635f705a9c6037bd97
2982 Author: Aliaksey Kandratsenka <[email protected]>
2983 Date: Sun Feb 7 11:59:44 2021 -0800
2984
2985 implement generic frame pointer backtracer
2986
2987 This supports frame pointer backtracing on x86-64, aarch64 and
2988 riscv-s (should work for both 32 and 64 bits).
2989
2990 Also added is detection of borked libunwind on aarch64-s. In this case
2991 frame pointer unwinder is preferred.
2992
2993 M Makefile.am
2994 M src/stacktrace.cc
2995 A src/stacktrace_generic_fp-inl.h
2996 M src/tests/stacktrace_unittest.cc
2997
2998 commit 17bab484aea43cf1a5247c823e036dfb52f5d92b
2999 Author: Aliaksey Kandratsenka <[email protected]>
3000 Date: Sat Feb 6 13:21:12 2021 -0800
3001
3002 always respect --enable-frame-pointers
3003
3004 Previously it only was respected on x86_64, but this days lots
3005 of modern
3006 ABIs are without frame pointers by default (e.g. arm64 and riscv, and
3007 even older mips).
3008
3009 M CMakeLists.txt
3010 M Makefile.am
3011 D cmake/CheckNoFPByDefault.cmake
3012 M configure.ac
3013
3014 commit 22c0eceddcf986a9b3f853b36d855acb38fa678f
3015 Author: Aliaksey Kandratsenka <[email protected]>
3016 Date: Sun Feb 14 16:29:14 2021 -0800
3017
3018 add emacs mode line annotations to remaining files
3019
3020 M src/base/arm_instruction_set_select.h
3021 M src/base/dynamic_annotations.c
3022 M src/base/dynamic_annotations.h
3023 M src/base/linuxthreads.h
3024 M src/base/thread_annotations.h
3025 M src/base/thread_lister.c
3026 M src/base/vdso_support.cc
3027 M src/base/vdso_support.h
3028 M src/gperftools/heap-profiler.h
3029 M src/gperftools/profiler.h
3030 M src/gperftools/tcmalloc.h.in
3031 M src/stacktrace_powerpc-darwin-inl.h
3032 M src/tests/stacktrace_unittest.cc
3033 M src/windows/addr2line-pdb.c
3034 M src/windows/config.h
3035 M src/windows/ia32_modrm_map.cc
3036 M src/windows/ia32_opcode_map.cc
3037 M src/windows/nm-pdb.c
3038 M src/windows/patch_functions.cc
3039 M src/windows/system-alloc.cc
3040
3041 commit b12139ddbaa3e16ba491a24ce6fec51009e70cce
3042 Author: Aliaksey Kandratsenka <[email protected]>
3043 Date: Sun Feb 14 16:13:15 2021 -0800
3044
3045 delete-trailing-whitespace on all files
3046
3047 M src/addressmap-inl.h
3048 M src/base/atomicops-internals-gcc.h
3049 M src/base/atomicops-internals-linuxppc.h
3050 M src/base/atomicops-internals-macosx.h
3051 M src/base/atomicops-internals-windows.h
3052 M src/base/atomicops-internals-x86.cc
3053 M src/base/atomicops-internals-x86.h
3054 M src/base/atomicops.h
3055 M src/base/commandlineflags.h
3056 M src/base/googleinit.h
3057 M src/base/linuxthreads.cc
3058 M src/base/logging.cc
3059 M src/base/logging.h
3060 M src/base/low_level_alloc.cc
3061 M src/base/low_level_alloc.h
3062 M src/base/simple_mutex.h
3063 M src/base/spinlock.cc
3064 M src/base/spinlock.h
3065 M src/base/spinlock_linux-inl.h
3066 M src/base/spinlock_posix-inl.h
3067 M src/base/spinlock_win32-inl.h
3068 M src/base/stl_allocator.h
3069 M src/base/sysinfo.cc
3070 M src/base/sysinfo.h
3071 M src/base/thread_annotations.h
3072 M src/base/vdso_support.cc
3073 M src/common.cc
3074 M src/config_for_unittests.h
3075 M src/getenv_safe.h
3076 M src/google/heap-profiler.h
3077 M src/google/malloc_extension.h
3078 M src/google/malloc_extension_c.h
3079 M src/google/malloc_hook.h
3080 M src/google/malloc_hook_c.h
3081 M src/google/profiler.h
3082 M src/google/stacktrace.h
3083 M src/google/tcmalloc.h
3084 M src/gperftools/heap-profiler.h
3085 M src/gperftools/malloc_extension.h
3086 M src/gperftools/malloc_extension_c.h
3087 M src/gperftools/malloc_hook.h
3088 M src/gperftools/malloc_hook_c.h
3089 M src/gperftools/profiler.h
3090 M src/gperftools/stacktrace.h
3091 M src/heap-checker-bcad.cc
3092 M src/heap-profile-stats.h
3093 M src/heap-profile-table.cc
3094 M src/heap-profile-table.h
3095 M src/heap-profiler.cc
3096 M src/internal_logging.cc
3097 M src/internal_logging.h
3098 M src/malloc_extension.cc
3099 M src/malloc_hook-inl.h
3100 M src/malloc_hook.cc
3101 M src/maybe_threads.cc
3102 M src/maybe_threads.h
3103 M src/memfs_malloc.cc
3104 M src/memory_region_map.cc
3105 M src/memory_region_map.h
3106 M src/pagemap.h
3107 M src/profile-handler.cc
3108 M src/profile-handler.h
3109 M src/raw_printer.cc
3110 M src/raw_printer.h
3111 M src/sampler.cc
3112 M src/sampler.h
3113 M src/stacktrace_powerpc-inl.h
3114 M src/symbolize.cc
3115 M src/symbolize.h
3116 M src/system-alloc.cc
3117 M src/system-alloc.h
3118 M src/tests/addressmap_unittest.cc
3119 M src/tests/atomicops_unittest.cc
3120 M src/tests/frag_unittest.cc
3121 M src/tests/heap-profiler_unittest.cc
3122 M src/tests/low_level_alloc_unittest.cc
3123 M src/tests/malloc_extension_c_test.c
3124 M src/tests/malloc_extension_test.cc
3125 M src/tests/markidle_unittest.cc
3126 M src/tests/memalign_unittest.cc
3127 M src/tests/pagemap_unittest.cc
3128 M src/tests/profiler_unittest.cc
3129 M src/tests/realloc_unittest.cc
3130 M src/tests/sampler_test.cc
3131 M src/tests/stacktrace_unittest.cc
3132 M src/tests/system-alloc_unittest.cc
3133 M src/tests/tcmalloc_large_unittest.cc
3134 M src/tests/testutil.cc
3135 M src/tests/testutil.h
3136 M src/tests/thread_dealloc_unittest.cc
3137 M src/windows/get_mangled_names.cc
3138 M src/windows/google/tcmalloc.h
3139 M src/windows/ia32_modrm_map.cc
3140 M src/windows/ia32_opcode_map.cc
3141 M src/windows/mingw.h
3142 M src/windows/mini_disassembler.cc
3143 M src/windows/mini_disassembler.h
3144 M src/windows/mini_disassembler_types.h
3145 M src/windows/nm-pdb.c
3146 M src/windows/override_functions.cc
3147 M src/windows/patch_functions.cc
3148 M src/windows/port.cc
3149 M src/windows/preamble_patcher.cc
3150 M src/windows/preamble_patcher.h
3151 M src/windows/preamble_patcher_test.cc
3152 M src/windows/preamble_patcher_with_stub.cc
3153 M src/windows/system-alloc.cc
3154
3155 commit 419c85814de497d6dfaf6f9a19d153138a7dec08
3156 Author: Aliaksey Kandratsenka <[email protected]>
3157 Date: Sun Feb 14 16:02:56 2021 -0800
3158
3159 amputate unused dynamic annotations support
3160
3161 M CMakeLists.txt
3162 M Makefile.am
3163 M cmake/config.h.in
3164 M configure.ac
3165 M src/base/dynamic_annotations.c
3166 M src/base/dynamic_annotations.h
3167 M src/base/low_level_alloc.cc
3168 M src/base/spinlock.h
3169 M src/profile-handler.cc
3170 D src/third_party/valgrind.h
3171 M src/windows/config.h
3172
3173 commit 73a72cdb6154ab7bdd49b1dde91349a30e8d7973
3174 Author: Aliaksey Kandratsenka <[email protected]>
3175 Date: Sun Feb 14 15:11:12 2021 -0800
3176
3177 don't check for snprintf
3178
3179 M CMakeLists.txt
3180 M cmake/config.h.in
3181 M src/windows/config.h
3182 M src/windows/mingw.h
3183 M src/windows/port.h
3184
3185 commit 95b52b0504a7cb2bce36d72b8bb0825833500b4c
3186 Author: Aliaksey Kandratsenka <[email protected]>
3187 Date: Sun Feb 14 15:02:57 2021 -0800
3188
3189 don't check for unused uname symbol
3190
3191 M CMakeLists.txt
3192 M cmake/config.h.in
3193 M configure.ac
3194 M src/windows/config.h
3195
3196 commit 01c2697fac41a5d6175ad0da21d9d753f0b385a1
3197 Author: Aliaksey Kandratsenka <[email protected]>
3198 Date: Sun Feb 14 14:30:29 2021 -0800
3199
3200 amputate unused SleepForMilliseconds from sysinfo.{h,cc}
3201
3202 M src/base/sysinfo.cc
3203 M src/base/sysinfo.h
3204
3205 commit ac68c97187e3729668ff5048c9bd27aae14e974d
3206 Author: Aliaksey Kandratsenka <[email protected]>
3207 Date: Sun Feb 14 14:25:49 2021 -0800
3208
3209 don't check for useless __builtin_stack_pointer
3210
3211 It doesn't seem to be supported anyways, and we're not using it too.
3212
3213 M CMakeLists.txt
3214 M cmake/config.h.in
3215 M configure.ac
3216 M src/windows/config.h
3217
3218 commit 7271bb72be08db491d3f0c60c1b0d5593cc1b5b5
3219 Author: Aliaksey Kandratsenka <[email protected]>
3220 Date: Sun Feb 14 14:22:08 2021 -0800
3221
3222 unbreak cmake check for TLS support
3223
3224 M CMakeLists.txt
3225
3226 commit 7c106ca241a344993f8011397de8004a224e094a
3227 Author: Aliaksey Kandratsenka <[email protected]>
3228 Date: Sun Feb 14 14:14:55 2021 -0800
3229
3230 don't bother checking for stl namespace and use std
3231
3232 Because there are no compilers left that don't do std namespace.
3233
3234 M CMakeLists.txt
3235 M cmake/config.h.in
3236 M configure.ac
3237 D m4/namespaces.m4
3238 D m4/stl_namespace.m4
3239 M src/heap-profiler.cc
3240 M src/malloc_extension.cc
3241 M src/tcmalloc.cc
3242 M src/windows/config.h
3243
3244 commit 0d6f32b9cef7ee044e55a746e7c76db62d23cd86
3245 Author: Aliaksey Kandratsenka <[email protected]>
3246 Date: Sun Feb 14 14:04:10 2021 -0800
3247
3248 use standard way to print size_t-sized ints
3249
3250 I.e. just use zu/zd/zx instead of finding out right size and defining
3251 PRI{u,x,d}S defines. Compilers have long caught up to this part of
3252 standard.
3253
3254 M cmake/config.h.in
3255 M configure.ac
3256 D m4/compiler_characteristics.m4
3257 M src/debugallocation.cc
3258 M src/emergency_malloc.cc
3259 M src/heap-checker.cc
3260 M src/heap-profile-table.cc
3261 M src/heap-profiler.cc
3262 M src/memory_region_map.cc
3263 M src/profiledata.cc
3264 M src/tcmalloc.cc
3265 M src/tests/markidle_unittest.cc
3266 M src/tests/tcmalloc_unittest.cc
3267 M src/windows/config.h
3268
3269 commit 0c11d35f4a2a0f5f42ee684a3f5b27cbab8abae2
3270 Author: Aliaksey Kandratsenka <[email protected]>
3271 Date: Sun Feb 14 13:39:55 2021 -0800
3272
3273 amputate checking for __int64
3274
3275 Since everything supports stdint.h this days.
3276
3277 M CMakeLists.txt
3278 M cmake/config.h.in
3279 M configure.ac
3280 M src/windows/config.h
3281 M src/windows/port.h
3282
3283 commit 92718aaaeb08d139b346633333235bf2260d0848
3284 Author: Aliaksey Kandratsenka <[email protected]>
3285 Date: Sun Feb 14 13:34:58 2021 -0800
3286
3287 amputate checking for conflict-signal.h
3288
3289 It was never found or used.
3290
3291 M CMakeLists.txt
3292 M cmake/config.h.in
3293 M src/profiler.cc
3294 M src/windows/config.h
3295
3296 commit 9bb2937261bca6c0257a8cfd62ff97f69c450acb
3297 Author: Aliaksey Kandratsenka <[email protected]>
3298 Date: Sun Feb 14 13:33:10 2021 -0800
3299
3300 amputate checking for inline keyword support
3301
3302 It is supported everywhere now.
3303
3304 M CMakeLists.txt
3305 M cmake/config.h.in
3306 M configure.ac
3307 M src/windows/config.h
3308
3309 commit d9c4c3b481e641b719d3d790987ed7d094157bf2
3310 Author: Đoàn Trần Công Danh <[email protected]>
3311 Date: Fri Feb 12 19:21:18 2021 +0700
3312
3313 profile-handler: use documented sigev_notify_thread_id in sigevent
3314
3315 sigevent(7) is documented to have sigev_notify_thread_id as its
3316 member.
3317 In glibc system, it's a macro expanded to the legacy _sigev_un._tid,
3318 _sigev_un._tid is obviously an internal implementation detail as
3319 signaled by its underscore prefix. And this macro was hidden inside
3320 linux/signal.h in older version of glibc.
3321
3322 On Linux that use musl libc, sigev_notify_thread_id is also a
3323 macro, but
3324 it's expanded to __sev_fields.sigev_notify_thread_id
3325
3326 [[email protected]: amputated broken linux/signal.h dependency]
3327 [[email protected]: see
3328 https://github.com/gperftools/gperftools/pull/1250]
3329 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
3330
3331 M src/profile-handler.cc
3332
3333 commit 43459feb33b87d3300e3b8b4917e5226678e4a0f
3334 Author: Đoàn Trần Công Danh <[email protected]>
3335 Date: Sat Feb 13 17:35:20 2021 +0700
3336
3337 configure.ac: check for features.h once
3338
3339 M configure.ac
3340
3341 commit 290b123c01f6578391b20bb31849f26b02401112
3342 Author: Venkatesh Srinivas <[email protected]>
3343 Date: Mon Feb 8 12:32:23 2021 -0800
3344
3345 atomicops: Remove Acquire_Store / Release_Load
3346
3347 gperftools' internal atomicops library included atomic
3348 Release Load and Acquire Store operations; those operations
3349 were unused and expressed ordering constraints that aren't
3350 expressible in the C++ standard memory model.
3351
3352 Remove them, to make a transition to C++11 atomics easier
3353 and to avoid confusing use of them.
3354
3355 M src/base/atomicops-internals-arm-generic.h
3356 M src/base/atomicops-internals-arm-v6plus.h
3357 M src/base/atomicops-internals-gcc.h
3358 M src/base/atomicops-internals-linuxppc.h
3359 M src/base/atomicops-internals-macosx.h
3360 M src/base/atomicops-internals-mips.h
3361 M src/base/atomicops-internals-windows.h
3362 M src/base/atomicops-internals-x86.h
3363 M src/base/atomicops.h
3364 M src/tests/atomicops_unittest.cc
3365
3366 commit 3b1c60cc4ea737ff151977dbcd164c5e2eb859fb
3367 Author: SSE4 <[email protected]>
3368 Date: Wed Jan 27 11:15:56 2021 +0300
3369
3370 Add support for Elbrus 2000 (e2k)
3371
3372 M cmake/PCFromUContext.cmake
3373 M m4/pc_from_ucontext.m4
3374 M src/base/basictypes.h
3375 M src/malloc_hook_mmap_linux.h
3376
3377 commit c5747615da78f7c2d91112f2a6b6b4da4f18540a
3378 Author: Jarno Rajahalme <[email protected]>
3379 Date: Fri Jan 29 09:54:18 2021 -0800
3380
3381 syscall: Mark x8 as clobbered
3382
3383 Mark arm64 register x8 as clobbered by syscall body inline assembly as
3384 it is being used to store the syscall number. Otherwise the compiler
3385 may try to use it for some other purpose.
3386
3387 This fix is derived from a resolution to clang Bugzilla report
3388 https://bugs.llvm.org/show_bug.cgi?id=48798. See this report for a
3389 minimal reproducer derived from the code fixed here as well as the
3390 resolution.
3391
3392 This should fix SEGFAULTs as reported in
3393 https://github.com/envoyproxy/envoy/issues/14756.
3394
3395 Fixes: #1241
3396 Signed-off-by: Jarno Rajahalme <[email protected]>
3397
3398 M src/base/linux_syscall_support.h
3399
3400 commit d8eb315fb18f6fb0d6efa923401f166343727bc6
3401 Author: Aliaksey Kandratsenka <[email protected]>
3402 Date: Sun Dec 20 23:35:08 2020 -0800
3403
3404 bump version to 2.8.1
3405
3406 M CMakeLists.txt
3407 M NEWS
3408 M configure.ac
3409 M src/windows/config.h
3410 M src/windows/gperftools/tcmalloc.h
3411
3412 commit 6ed61f8e91ffc00310b253ad2c777c0e862bd742
3413 Author: Aliaksey Kandratsenka <[email protected]>
3414 Date: Sat Dec 19 19:14:50 2020 -0800
3415
3416 add note that cmake build is preliminary
3417
3418 M CMakeLists.txt
3419
3420 commit 6bbf2ed1501e27e215bb412b9b0b930921598bca
3421 Author: okhowang(王沛文) <[email protected]>
3422 Date: Sat Sep 12 02:43:56 2020 +0800
3423
3424 Update cmake
3425
3426 M CMakeLists.txt
3427 M cmake/FindObjcopyWithWeaken.cmake
3428 D cmake/GPerftoolsAddTarget.cmake
3429 M cmake/config.h.in
3430 A cmake/pkgconfig.pc
3431 M cmake/tcmalloc.h.in
3432 M src/tests/addressmap_unittest.cc
3433 A src/windows/CMakeLists.txt
3434
3435 commit 913d3eb7d7fc1c5cc025920d5699b528c268f0fa
3436 Author: Isaac Hier <[email protected]>
3437 Date: Wed Nov 8 11:20:25 2017 -0500
3438
3439 Fix a few macros for Apple
3440
3441 M CMakeLists.txt
3442 M cmake/config.h.in
3443
3444 commit 64a73b1cb87a6661e4c3092b07bfd5032fe9a1fa
3445 Author: Isaac Hier <[email protected]>
3446 Date: Wed Nov 8 07:24:08 2017 -0500
3447
3448 Work on fixing linking errors in stacktrace
3449
3450 M CMakeLists.txt
3451 M cmake/config.h.in
3452
3453 commit b788d51eb4cf80adc7d782fa8df7175ee96d41d9
3454 Author: Isaac Hier <[email protected]>
3455 Date: Tue Nov 7 21:25:34 2017 -0500
3456
3457 Fix conditional definitions
3458
3459 M CMakeLists.txt
3460
3461 commit 495229b625d65b19f3fc98fed5b626d31a89ee03
3462 Author: Isaac Hier <[email protected]>
3463 Date: Tue Nov 7 13:26:10 2017 -0500
3464
3465 Make internal tcmalloc libs
3466
3467 M CMakeLists.txt
3468
3469 commit cca7f6f669657218a15ef5a176706660ad2c1d67
3470 Author: Isaac Hier <[email protected]>
3471 Date: Sun Nov 5 16:53:28 2017 -0500
3472
3473 More unit tests and libraries
3474
3475 M CMakeLists.txt
3476
3477 commit 11dc65c3c40618065ae48ccdb1e1548e339bff32
3478 Author: Isaac Hier <[email protected]>
3479 Date: Sun Nov 5 12:53:07 2017 -0500
3480
3481 Fix config headers, add more unit tests
3482
3483 M CMakeLists.txt
3484 M cmake/config.h.in
3485 M cmake/tcmalloc.h.in
3486
3487 commit 6078fe40d9a5b829cbbadc6ec824163d6d9eda5b
3488 Author: Isaac Hier <[email protected]>
3489 Date: Sun Nov 5 12:23:42 2017 -0500
3490
3491 Finish configure.ac conversion to CMake, start on Makefile.am
3492
3493 M CMakeLists.txt
3494 A cmake/CheckNoFPByDefault.cmake
3495 M cmake/DefineTargetVariables.cmake
3496 A cmake/FindObjcopyWithWeaken.cmake
3497 A cmake/GPerftoolsAddTarget.cmake
3498 M cmake/PCFromUContext.cmake
3499 M cmake/config.h.in
3500
3501 commit 515fb221966a2413d00184186da97719b90e3915
3502 Author: Isaac Hier <[email protected]>
3503 Date: Fri Nov 3 09:10:27 2017 -0400
3504
3505 Generate config header
3506
3507 M CMakeLists.txt
3508 D cmake/DefineArchVariables.cmake
3509 A cmake/DefineTargetVariables.cmake
3510 M cmake/PCFromUContext.cmake
3511 A cmake/config.h.in
3512 A cmake/tcmalloc.h.in
3513
3514 commit 4adb5ff74d35900dd8d0ed07959eef68f85c8886
3515 Author: Isaac Hier <[email protected]>
3516 Date: Thu Nov 2 20:07:29 2017 -0400
3517
3518 Add architecture checks
3519
3520 M CMakeLists.txt
3521 A cmake/DefineArchVariables.cmake
3522 M src/tcmalloc.h
3523
3524 commit fa9bedc82cbbc289875a07b6e431e553b1a15b77
3525 Author: Isaac Hier <[email protected]>
3526 Date: Wed Nov 1 09:52:02 2017 -0400
3527
3528 Add most of CMake build
3529
3530 M CMakeLists.txt
3531 A cmake/PCFromUContext.cmake
3532
3533 commit 9e4f72fd1faa3f406f468796b4b6908391c8019c
3534 Author: Isaac Hier <[email protected]>
3535 Date: Mon Oct 30 22:26:13 2017 -0400
3536
3537 Define options, start system checks
3538
3539 M CMakeLists.txt
3540
3541 commit a6ce98174b47213f11c84d57cfbc1e3cdd609bfb
3542 Author: Isaac Hier <[email protected]>
3543 Date: Mon Oct 30 09:35:34 2017 -0400
3544
3545 Add CMakeLists.txt
3546
3547 A CMakeLists.txt
3548
3549 commit 313495587598fb1a69b94c44a83735f6056f6049
3550 Author: Steve Langasek <[email protected]>
3551 Date: Thu Oct 8 16:49:21 2020 -0700
3552
3553 Additional porting for riscv64.
3554
3555 Adds handling for the PC register, malloc_hook_mmap, and syscalls.
3556
3557 Successfully built in Ubuntu with these changes.
3558
3559 M m4/pc_from_ucontext.m4
3560 M src/base/linux_syscall_support.h
3561 M src/malloc_hook_mmap_linux.h
3562
3563 commit f0e289bdbb2fbbb7fa71496d846988c1bac0e310
3564 Author: Steve Langasek <[email protected]>
3565 Date: Thu Oct 8 07:54:03 2020 -0700
3566
3567 Enable build on riscv64.
3568
3569 Cacheline length may be implementation-specific, but this is known
3570 to be
3571 correct on the U54.
3572
3573 M src/base/basictypes.h
3574
3575 commit 6c715b4fa183e221a29c5d7a4ea87aeba4ee02cf
3576 Author: Tim Gates <[email protected]>
3577 Date: Sat Oct 10 17:26:00 2020 +1100
3578
3579 docs: fix simple typo, defininitions -> definitions
3580
3581 There is a small typo in src/base/linux_syscall_support.h.
3582
3583 Should read `definitions` rather than `defininitions`.
3584
3585 M src/base/linux_syscall_support.h
3586
3587 commit 02d5264018cc76a36713f97329870dbf85969519
3588 Author: Aliaksey Kandratsenka <[email protected]>
3589 Date: Sat Dec 19 17:15:31 2020 -0800
3590
3591 Revert "drop page heap lock when returning memory back to kernel"
3592
3593 This reverts commit be3da70298bf3d25c7d64655922ab82dd819ec98.
3594
3595 There are reports of crashes and false-positive OOMs from this
3596 patch. Crashes under aggressive decommit mode are understood, but I
3597 have yet to get confirmations whether false-positive OOMs were seen
3598 under aggressive decommit or not. Thus lets revert for now.
3599
3600 Updates issue #1227 and issue #1204.
3601
3602 M src/page_heap.cc
3603 M src/page_heap.h
3604 M src/tests/page_heap_test.cc
3605
3606 commit 151cbf5146cd5b359a4470860d1f7d8f6a843d62
3607 Author: Grant Henke <[email protected]>
3608 Date: Thu Nov 26 20:56:46 2020 -0600
3609
3610 Add OS X arm64 program counter
3611
3612 [[email protected]: updated comment that it is tested]
3613
3614 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
3615
3616 M m4/pc_from_ucontext.m4
3617
3618 commit 140e3481d0541044cdc27f45129b47f92d5940b7
3619 Merge: bda3c82e 0fc5cabd
3620 Author: Aliaksey Kandratsenka (aka Aliaksei Kandratsenka)
3621 <[email protected]>
3622 Date: Tue Nov 17 14:30:59 2020 -0800
3623
3624 Merge pull request #1231 from PatriosTheGreat/master
3625
3626 Fix implicit-int-float-conversion warning.
3627
3628 commit 0fc5cabdfc966affb6f336848d9587ec5d904d0c
3629 Author: Levon Ter-Grigoryan <[email protected]>
3630 Date: Tue Nov 17 12:29:22 2020 +0100
3631
3632 Fix implicit-int-float-conversion warning.
3633
3634 M src/sampler.cc
3635
3636 commit bda3c82e11615ca9e7751d1f3cfb161026ee742a
3637 Author: Kerrick Staley <[email protected]>
3638 Date: Wed Jul 8 16:42:12 2020 -0700
3639
3640 Increase kMaxStackDepth to 254
3641
3642 [[email protected]: reduced value from 256 to 254]
3643
3644 M src/profiledata.h
3645
3646 commit 1d9b8bb59d5610f2c3e9faad6bbd91ceca17fb93
3647 Author: Aliaksey Kandratsenka <[email protected]>
3648 Date: Sat Jul 11 20:33:40 2020 -0700
3649
3650 don't test sbrk hook if we're on linux and don't have __sbrk
3651
3652 "mmap" hooks code for linux only hooks sbrk when __sbrk is
3653 defined. But musl doesn't offer this function (reasonably so), so
3654 sbrk hooking doesn't work there. Not big loss, but lets make sure
3655 tests don't fail.
3656
3657 Lets reference this to issue #1198
3658
3659 M src/tests/tcmalloc_unittest.cc
3660
3661 commit 180bfa10d7cb38e8b3784d60943d50e8fcef0dcb
3662 Author: Aliaksey Kandratsenka <[email protected]>
3663 Date: Mon Jul 6 02:51:43 2020 -0700
3664
3665 bumped version to 2.8
3666
3667 M Makefile.am
3668 M NEWS
3669 M configure.ac
3670 M src/windows/config.h
3671 M src/windows/gperftools/tcmalloc.h
3672
3673 commit c1bcc412ba4a94deaf8c3230519ecb4dda60557d
3674 Author: Ryan Macnak <[email protected]>
3675 Date: Fri May 1 11:50:16 2020 -0700
3676
3677 Don't try to mark esp as clobbered in linux syscall support.
3678
3679 Compare 9e5b1628737c67b4587f937164572774592978c4, issue #1076
3680
3681 M src/base/linux_syscall_support.h
3682
3683 commit 50f89afaed3d07d384fa8d86c98460ccff1e870e
3684 Author: Aliaksey Kandratsenka <[email protected]>
3685 Date: Sun Apr 19 22:21:07 2020 -0700
3686
3687 liberate gperftools tests from relying on -fno-builtin-XXX flags
3688
3689 Clang mostly ignores those anyways, so our tests needed better way to
3690 disable optimizations (clang is quite aggressive replacing new/delete
3691 pair with stack allocation).
3692
3693 M Makefile.am
3694 M README
3695 M src/tests/debugallocation_test.cc
3696 M src/tests/system-alloc_unittest.cc
3697 M src/tests/tcmalloc_large_unittest.cc
3698 M src/tests/tcmalloc_unittest.cc
3699 M src/tests/testutil.h
3700
3701 commit 98ccd0f1029c11a3df0a51e4a14b543e430a96f8
3702 Author: Aliaksey Kandratsenka <[email protected]>
3703 Date: Sun Apr 19 22:50:16 2020 -0700
3704
3705 prevent inlining in heap-checker unittest
3706
3707 The test relies on stuff "mmaper" function not being inlined, but
3708 compilers are getting smarter and latest clang actually does inline it
3709 now. Adding volatile to pointer declaration fixes it.
3710
3711 M src/tests/heap-checker_unittest.cc
3712
3713 commit e521472f1ac1ecf028e8c922ca4f8741473afd66
3714 Author: Aliaksey Kandratsenka <[email protected]>
3715 Date: Sun Jul 5 22:49:41 2020 -0700
3716
3717 fix linking of page_heap_test on windows
3718
3719 After change to release page heap lock around returning memory back to
3720 kernel, page heap test got dependency on page heap lock. Which was not
3721 available on windows since relevant symbols are not exported.
3722
3723 Proposed fix is to simply duplicate all needed .cc files in
3724 page_heap_test project instead of linking to dll. This is not perfect
3725 but gets job done, until we figure out better solution (GNU/Linux will
3726 eventually get hidden visibility and will need it).
3727
3728 This fixes github issue 1189.
3729
3730 M vsprojects/page_heap_test/page_heap_test.vcxproj
3731 M vsprojects/page_heap_test/page_heap_test.vcxproj.filters
3732
3733 commit e5f77d6485bd2f6ce43862e3e57118b1bb97d30a
3734 Author: Aliaksey Kandratsenka <[email protected]>
3735 Date: Mon Mar 23 01:21:27 2020 -0700
3736
3737 chmod -x Makefile.am gperftools.sln
3738
3739 M Makefile.am
3740 M gperftools.sln
3741
3742 commit 6b92e00cecb82e7fae6e019583865faced29c6f7
3743 Author: Aliaksey Kandratsenka <[email protected]>
3744 Date: Mon Mar 23 01:15:40 2020 -0700
3745
3746 don't assume HAVE_MMAP on mingw builds
3747
3748 Fixes github issue #1108.
3749
3750 M configure.ac
3751
3752 commit 4cddede3990a96969aac9ff13ce3c8195ec91ce7
3753 Author: Kirill Müller <[email protected]>
3754 Date: Tue Feb 25 09:09:22 2020 +0100
3755
3756 New ProfilerGetStackTrace()
3757
3758 M Makefile.am
3759 M src/gperftools/profiler.h
3760 M src/profiler.cc
3761
3762 commit db7aa547abb5abdd558587a15502584cbc825438
3763 Author: Aliaksey Kandratsenka <[email protected]>
3764 Date: Sun Mar 8 20:57:01 2020 -0700
3765
3766 bumped version to 2.8rc
3767
3768 M NEWS
3769 M configure.ac
3770 M src/windows/config.h
3771 M src/windows/gperftools/tcmalloc.h
3772
3773 commit be3da70298bf3d25c7d64655922ab82dd819ec98
3774 Author: Aliaksey Kandratsenka <[email protected]>
3775 Date: Sun Mar 8 19:32:02 2020 -0700
3776
3777 drop page heap lock when returning memory back to kernel
3778
3779 Fixes issue #754.
3780
3781 M src/page_heap.cc
3782 M src/page_heap.h
3783 M src/tests/page_heap_test.cc
3784
3785 commit 87acc2782fe49a8b57d19783f61ff8bc667db68d
3786 Author: Aliaksey Kandratsenka <[email protected]>
3787 Date: Sun Mar 8 18:22:17 2020 -0700
3788
3789 amputate span events history
3790
3791 Nobody used it in years. Abseil tcmalloc had it amputated years ago.
3792
3793 M src/central_freelist.cc
3794 M src/page_heap.cc
3795 M src/span.cc
3796 M src/span.h
3797
3798 commit e40c7f231ad89e1ee8bf37a1d6680880c519c901
3799 Author: Ilya Leoshkevich <[email protected]>
3800 Date: Thu Feb 27 13:10:29 2020 +0100
3801
3802 Fix mmap syscall on s390
3803
3804 mmap arguments must be passed in an array on s390.
3805
3806 Signed-off-by: Ilya Leoshkevich <[email protected]>
3807
3808 M src/malloc_hook_mmap_linux.h
3809
3810 commit b7607ee7d4ab35504a54588e5912a8e850686db2
3811 Author: Jon Kohler <[email protected]>
3812 Date: Sun Jan 19 22:09:08 2020 -0500
3813
3814 tcmalloc: ability to disable fallback allocator in memfs
3815
3816 Whenever tcmalloc memfs allocator runs out of huge memory, it falls
3817 back to the
3818 default system allocator, which will start allocating from normal
3819 anonymous
3820 memory pool (small pages). For use cases that exclusively want
3821 hugepage backed
3822 memory, such as using tcmalloc in combination with Intel SPDK, this is
3823 problematic.
3824
3825 This change adds an environment variable
3826 "TCMALLOC_MEMFS_DISABLE_FALLBACK"
3827 which when set to true will ensure that tcmalloc doesn't fallback
3828 to using the
3829 system allocator when the system runs out of hugepages. This is
3830 false by
3831 default, such that all other tcmalloc consumers will not see any
3832 changes.
3833
3834 For those curious, you can see part of our upstream SPDK work here:
3835 https://review.gerrithub.io/c/spdk/spdk/+/475943
3836
3837 M src/memfs_malloc.cc
3838
3839 commit 1bfcb5bc3a779b6c0c63125737f035c2afcc9e95
3840 Author: Jon Kohler <[email protected]>
3841 Date: Tue Feb 18 11:14:49 2020 -0500
3842
3843 tcmalloc: fragmentation overhead instrumentation
3844
3845 This patch adds visibility into the overhead due to fragmentation
3846 for each size
3847 class in the tcmalloc central free list, which is helpful when
3848 debugging
3849 fragmentation issues.
3850
3851 M src/tcmalloc.cc
3852
3853 commit 36bf1309de52433f0088274b347b4fc385ea056f
3854 Author: Todd Lipcon <[email protected]>
3855 Date: Mon Jun 3 10:29:06 2019 -0700
3856
3857 Fix a clang-tidy readability warning for static member access
3858
3859 This fixes the following warning:
3860
3861 src/thread_cache.cc:281:10: warning: static member accessed through
3862 instance [readability-static-accessed-through-instance]
3863 return sampler_.GetSamplePeriod();
3864 ^~~~~~~~~~~~~~~~~~~~~~~~
3865 tcmalloc::Sampler::
3866
3867 M src/thread_cache.cc
3868
3869 commit 2b2a962c2bdc3ecbdbd45873a07989c7ccb56560
3870 Author: prehistoricpenguin <[email protected]>
3871 Date: Fri Jul 12 16:25:55 2019 +0800
3872
3873 Remove executable flag for c++ files
3874
3875 M src/heap-checker.cc
3876 M src/heap-profiler.cc
3877 M src/stacktrace_instrument-inl.h
3878 M src/symbolize.cc
3879 M src/system-alloc.cc
3880 M src/tests/sampler_test.cc
3881
3882 commit 8f308afbfedeca0991ef431e3b8e0b01bd5b8a8f
3883 Author: Henrik Edin <[email protected]>
3884 Date: Tue Apr 9 15:16:55 2019 -0400
3885
3886 Increase kClassSizesMax to 128 to allow for page size of 4K
3887
3888 M src/common.h
3889
3890 commit d3fefdb69439292ec914cd8c65dd83ed1bcbdc65
3891 Author: Henrik Edin <[email protected]>
3892 Date: Tue Jan 29 13:25:40 2019 -0500
3893
3894 Allow configuring page size to 4K, 8K, 16K, 32K, 64K, 128K and 256K
3895
3896 M INSTALL
3897 M configure.ac
3898 M src/common.h
3899 M src/windows/config.h
3900
3901 commit cf2df3b000c58155d3c83bd01b603894b4559c0b
3902 Author: Laurent Stacul <[email protected]>
3903 Date: Fri Feb 7 09:37:34 2020 +0000
3904
3905 Fix the removed std::allocator::pointer member type removed in C++20
3906
3907 M src/tests/heap-checker_unittest.cc
3908
3909 commit 31024506c543132adb315b8de89d5faa85e0ff0e
3910 Author: shipujin <[email protected]>
3911 Date: Tue Aug 20 13:24:57 2019 +0800
3912
3913 Add mips64* support
3914
3915 M src/base/linux_syscall_support.h
3916
3917 commit fe62a0baab87ba3abca12f4a621532bf67c9a7d2
3918 Author: Holy Wu <[email protected]>
3919 Date: Tue Apr 9 18:33:26 2019 +0800
3920
3921 Update config.h in Windows
3922
3923 M src/windows/config.h
3924
3925 commit 8272631b5a9f9e840eb1b014f43b3ddea16ff67d
3926 Author: 风 <[email protected]>
3927 Date: Wed Apr 17 10:55:06 2019 +0800
3928
3929 Fix a long time typo
3930
3931 Thought this fix may not effect the behavior or performance, but it
3932 seems indeed a bug.
3933
3934 M src/pprof
3935
3936 commit c1d546d7b22cc503f37e9a6efa1d249be60243a3
3937 Author: Aliaksey Kandratsenka <[email protected]>
3938 Date: Sun Apr 7 15:27:17 2019 -0700
3939
3940 never test and always default HAVE_MMAP to on
3941
3942 This should alleviate vboxfs problems from issue #1008.
3943
3944 M configure.ac
3945
3946 commit fba6ce0e7aaa094fec09ef8763e86b19dc815624
3947 Author: pkubaj <[email protected]>
3948 Date: Mon Feb 4 16:10:06 2019 +0000
3949
3950 Fix build on FreeBSD
3951
3952 On FreeBSD the following syntax is used:
3953 mr REGA,REGB
3954
3955 M src/stacktrace_powerpc-darwin-inl.h
3956
3957 commit 98ac4ee9bc6ab285429fc5d97335b4c94fc0b919
3958 Author: Jakub Wilk <[email protected]>
3959 Date: Wed Apr 3 09:50:40 2019 +0200
3960
3961 Fix typos
3962
3963 M README
3964
3965 commit 9e5b1628737c67b4587f937164572774592978c4
3966 Author: Aliaksey Kandratsenka <[email protected]>
3967 Date: Sun Feb 10 12:46:18 2019 -0800
3968
3969 don't try to mark rsp as clobbered in linux syscall support
3970
3971 rsp is not actually clobbered by that code and later gccs actually
3972 (correctly) bark
3973 at it.
3974
3975 Fixed issue #1076.
3976
3977 M src/base/linux_syscall_support.h
3978
3979 commit 1e36ed705548ae2a5170bb00dfee0954b8ef76de
3980 Author: Romain Geissler <[email protected]>
3981 Date: Tue Jan 29 18:31:19 2019 +0000
3982
3983 Use initial exec TLS model for all thread local variables from
3984 thread_cache.cc
3985
3986 This avoids a deadlock when a library which is being dlopen'ed creates
3987 as part of its static constructors a thread which quickly need to call
3988 malloc. We are still in the dlopen call (so with some internal glibc
3989 mutex taken) when the thread executes code and later needs to call
3990 malloc which in term calls tls_get_addr_tail, which wait for the
3991 dlopen
3992 mutex to be unlocked. If later the dlopen'ing thread also calls malloc
3993 as part of its constructors, we are in a deadlock.
3994
3995 Fix is similar to
3996 https://github.com/gperftools/gperftools/commit/7852eeb75b9375cf52a7da01be044da6e915dd08
3997
3998 Stack of the dlopening thread:
3999 #0 0x00007fd5406ca93c in __lll_lock_wait () from
4000 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
4001 #1 0x00007fd5406c45a5 in pthread_mutex_lock () from
4002 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
4003 ... proprietary code in the stack
4004 #9 0x00007fd5074f0367 in
4005 __static_initialization_and_destruction_0 (__initialize_p=1,
4006 __priority=65535) at src/ClientImpl.cpp:15
4007 #10 0x00007fd5074f06d7 in _GLOBAL__sub_I_ClientImpl.cpp(void)
4008 () at src/ClientImpl.cpp:85
4009 #11 0x00007fd50757aa46 in __do_global_ctors_aux ()
4010 #12 0x00007fd5073e985f in _init () from ...
4011 #13 0x00007fd53bf9dec8 in ?? () from ...
4012 #14 0x00007fd54d637a5d in call_init.part () from
4013 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
4014 #15 0x00007fd54d637bab in _dl_init () from
4015 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
4016 #16 0x00007fd54d63c160 in dl_open_worker () from
4017 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
4018 #17 0x00007fd54d637944 in _dl_catch_error () from
4019 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
4020 #18 0x00007fd54d63b7d9 in _dl_open () from
4021 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
4022 #19 0x00007fd54d61f2b9 in dlopen_doit () from
4023 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libdl.so.2
4024 #20 0x00007fd54d637944 in _dl_catch_error () from
4025 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
4026 #21 0x00007fd54d61f889 in _dlerror_run () from
4027 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libdl.so.2
4028 #22 0x00007fd54d61f351 in dlopen@@GLIBC_2.2.5 () from
4029 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libdl.so.2
4030
4031 Stack of the newly created thread calling tls_get_addr_tail:
4032 #0 0x00007fd5406ca93c in __lll_lock_wait () from
4033 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
4034 #1 0x00007fd5406c4622 in pthread_mutex_lock () from
4035 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
4036 #2 0x00007fd54d63a2ed in tls_get_addr_tail () from
4037 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/ld-linux-x86-64.so.2
4038 #3 0x00007fd53fee877d in
4039 tcmalloc::ThreadCache::CreateCacheIfNecessary () at
4040 src/thread_cache.cc:344
4041 #4 0x00007fd53fecb4ab in tcmalloc::ThreadCache::GetCache ()
4042 at src/thread_cache.h:437
4043 #5 0x00007fd53fefeccb in (anonymous namespace)::do_malloc
4044 (size=56) at src/tcmalloc.cc:1354
4045 #6 tcmalloc::do_allocate_full<tcmalloc::cpp_throw_oom> (size=56)
4046 at src/tcmalloc.cc:1762
4047 #7 tcmalloc::allocate_full_cpp_throw_oom (size=56) at
4048 src/tcmalloc.cc:1776
4049 #8 0x00007fd53ff01b80 in
4050 tcmalloc::dispatch_allocate_full<tcmalloc::cpp_throw_oom>
4051 (size=56) at src/tcmalloc.cc:1785
4052 #9 malloc_fast_path<tcmalloc::cpp_throw_oom> (size=56) at
4053 src/tcmalloc.cc:1845
4054 #10 tc_new (size=56) at src/tcmalloc.cc:1980
4055 ... proprietary code in the stack
4056 #26 0x00007fd5406c1ef4 in start_thread () from
4057 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libpthread.so.0
4058 #27 0x00007fd5403ba01d in clone () from
4059 /data3/mwrep/rgeissler/core.tls/opt/1A/toolchain/x86_64-2.6.32-v2/lib64/libc.so.6
4060
4061 M src/thread_cache.cc
4062
4063 commit 8f9a873fce14337e113a3837603a11ade06da533
4064 Author: pkubaj <[email protected]>
4065 Date: Mon Feb 4 16:06:32 2019 +0000
4066
4067 Fix accessing PC on FreeBSD/powerpc and powerpc64
4068
4069 M m4/pc_from_ucontext.m4
4070
4071 commit fc00474ddc21fff618fc3f009b46590e241e425e
4072 Author: Tulio Magno Quites Machado Filho <[email protected]>
4073 Date: Fri Feb 8 10:33:09 2019 -0200
4074
4075 Include asm/ptrace.h when accessing ucontext_t
4076
4077 In some architectures (e.g. powerpc, powerpc64 and powerpc64le), the
4078 macro that specifies the position of the program counter is defined in
4079 header asm/ptrace.h.
4080
4081 When glibc bug 21457 got fixed in version 2.27, header sys/ucontext.h
4082 stopped including signal.h, which indirectly removed asm/ptrace.h too.
4083
4084 Signed-off-by: Tulio Magno Quites Machado Filho <[email protected]>
4085
4086 M m4/pc_from_ucontext.m4
4087 M src/getpc.h
4088
4089 commit 5574c87e39ee592c909cb48245c1d91e90ddaf4d
4090 Author: Laurent Stacul <[email protected]>
4091 Date: Mon Oct 15 16:56:18 2018 +0200
4092
4093 Compile time aggressive decommit option
4094
4095 [[email protected]: performed s/agressive/aggressive/g]
4096 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
4097
4098 M configure.ac
4099 M src/static_vars.cc
4100
4101 commit e9ab4c53041ac62feefbbb076d326e9a77dd1567
4102 Author: Fabrice Fontaine <[email protected]>
4103 Date: Sun Oct 7 18:12:03 2018 +0200
4104
4105 undef mmap64 function
4106
4107 mmap64 can already been defined and as a result the following error is
4108 raised:
4109
4110 In file included from src/malloc_hook.cc:686:0:
4111 src/malloc_hook_mmap_linux.h: In function ‘void* mmap(void*,
4112 size_t, int, int, int, off_t)’:
4113 src/malloc_hook_mmap_linux.h:173:18: error: redefinition of ‘void*
4114 mmap(void*, size_t, int, int, int, off_t)’
4115 extern "C" void* mmap(void *start, size_t length, int prot,
4116 int flags,
4117 ^~~~
4118 In file included from src/malloc_hook.cc:41:0:
4119 src/malloc_hook_mmap_linux.h:159:18: note: ‘void* mmap(void*,
4120 size_t, int, int, int, off_t)’ previously defined here
4121 extern "C" void* mmap64(void *start, size_t length, int prot,
4122 int flags,
4123 ^
4124 Makefile:4874: recipe for target
4125 'src/libtcmalloc_minimal_internal_la-malloc_hook.lo' failed
4126
4127 Signed-off-by: Fabrice Fontaine <[email protected]>
4128
4129 M src/malloc_hook_mmap_linux.h
4130
4131 commit 5eec9d0ae396aa90ebcb5756dfd0f3d101667490
4132 Author: Aliaksey Kandratsenka <[email protected]>
4133 Date: Sun Oct 7 08:17:04 2018 -0700
4134
4135 Drop not very portable and not very useful unwind benchmark.
4136
4137 M .gitignore
4138 M Makefile.am
4139 D benchmark/getcontext_light.cc
4140 D benchmark/unwind_bench.cc
4141
4142 commit 1561f0946fcc96d1384dae2a81f3ed111338a341
4143 Author: Fabrice Fontaine <[email protected]>
4144 Date: Sat Oct 6 20:43:19 2018 +0200
4145
4146 check for __sbrk
4147
4148 __sbrk is not defined in musl so add a check for it in configure.ac
4149 and
4150 replace the check on ifndef __UCLIBC__ by a check on ifdef HAVE__SBRK
4151 in src/malloc_hook_mmap_linux.h
4152
4153 Signed-off-by: Fabrice Fontaine <[email protected]>
4154
4155 M configure.ac
4156 M src/malloc_hook_mmap_linux.h
4157
4158 commit 1de76671d478e848f0dd210e00df43bfa463ca9f
4159 Author: Gabriel Marin <[email protected]>
4160 Date: Thu Oct 4 22:52:35 2018 +0000
4161
4162 Fix mmap region iteration while no regions are recorded.
4163
4164 If no mmap regions are recorded, iteration failed since the RegionSet
4165 (std::set) object is not initialized.
4166
4167 Original CL https://codereview.chromium.org/14769008
4168
4169 Reviewed-on: https://chromium-review.googlesource.com/c/1130807
4170
4171 M src/memory_region_map.cc
4172 M src/memory_region_map.h
4173
4174 commit acdcacc28f17d2e8db1d1bb634025f820f912d93
4175 Author: Fabrice Fontaine <[email protected]>
4176 Date: Sun Apr 29 23:48:41 2018 +0200
4177
4178 Use off64_t instead of __off64_t
4179
4180 __off64_t is a kernel internal symbol, which happens to be
4181 user-visible
4182 with glibc, but not necessarily with other C libraries such as musl.
4183
4184 Switching from __off64_t to off64_t allows the code to build properly
4185 with musl.
4186
4187 Signed-off-by: Fabrice Fontaine <[email protected]>
4188
4189 M src/malloc_hook_mmap_linux.h
4190
4191 commit 0177a2420aa68229f55bec970c59136813a50613
4192 Author: Gabriel Marin <[email protected]>
4193 Date: Fri Oct 5 21:49:50 2018 +0000
4194
4195 Return early in WriteProfile to reduce indentation
4196
4197 Small code refactoring to reduce indentation. This change was
4198 originally
4199 part of https://codereview.chromium.org/9963095.
4200
4201 Reviewed-on: https://chromium-review.googlesource.com/c/1130804
4202
4203 M src/heap-profile-table.cc
4204
4205 commit b85652bf268d061a0ad0aec90f0d3c29aa80a6b5
4206 Author: Gabriel Marin <[email protected]>
4207 Date: Wed Oct 3 18:15:16 2018 +0000
4208
4209 Add generic.total_physical_bytes property to MallocExtension
4210
4211 Original CL:
4212
4213 - https://codereview.chromium.org/1410353005
4214
4215 Add generic.total_physical_bytes property to MallocExtension
4216
4217 The actual physical memory usage of tcmalloc cannot be obtained by
4218 GetNumericProperty. This accounts for the current_allocated_bytes,
4219 fragmentation and malloc metadata, and excludes the unmapped memory
4220 regions. This helps the user to understand how much memory is
4221 actually
4222 being used for the allocations that were made.
4223
4224 Reviewed-on: https://chromium-review.googlesource.com/1130803
4225
4226 M src/gperftools/malloc_extension.h
4227 M src/tcmalloc.cc
4228
4229 commit 90df23c81fe6baf7799b5d3df37d2846ab88fc60
4230 Author: Gabriel Marin <[email protected]>
4231 Date: Fri Oct 5 22:00:33 2018 +0000
4232
4233 Make some tcmalloc constants truly const
4234
4235 Reviewed-on: https://chromium-review.googlesource.com/c/1130809
4236
4237 M src/tcmalloc.cc
4238
4239 commit 49dbe4362b431629111b85929d91fe9a46c42295
4240 Author: Aliaksey Kandratsenka <[email protected]>
4241 Date: Mon Sep 3 16:44:52 2018 -0700
4242
4243 Add comment about gperftools 2.8 not deduplicating heapz samples.
4244
4245 M src/gperftools/malloc_extension.h
4246
4247 commit 63a12a5ed3c4aca61cc46078b6cdf1d161425a69
4248 Author: Aliaksey Kandratsenka <[email protected]>
4249 Date: Mon Aug 27 20:10:09 2018 -0700
4250
4251 Drop de-duplication of heap sample (aka heapz) entries.
4252
4253 pprof can aggregate them, and it can do it way better than we
4254 can. With
4255 proper unsampling etc.
4256
4257 M src/stack_trace_table.cc
4258 M src/stack_trace_table.h
4259 M src/static_vars.cc
4260 M src/static_vars.h
4261 M src/tests/stack_trace_table_test.cc
4262
4263 commit 954f9dc0e37ff46cb0cb24edfb39dc77fd2e7d0b
4264 Author: Aliaksey Kandratsenka <[email protected]>
4265 Date: Sun Aug 26 11:37:59 2018 -0700
4266
4267 Add flag to disable installing unmaintained & deprecated pprof.
4268
4269 Everyone should be using golang pprof from github.com/google/pprof,
4270 but
4271 distros still ship our perl version and not everyone is aware of
4272 better pprof yet.
4273
4274 This is another step in completely dropping perl pprof. We still
4275 default to installing it, but hopefully we'll be able to convince
4276 distros to disable this soon.
4277
4278 We still install pprof under pprof-symbolize name because
4279 stack traces symbolization depends on it, and because golang pprof
4280 won't support this feature.
4281
4282 This is related to issue #1038.
4283
4284 M .gitignore
4285 M Makefile.am
4286 M configure.ac
4287 M src/symbolize.cc
4288
4289 commit 893bff51bcf220b724a812d340d878b5fb8ce911
4290 Author: Aliaksey Kandratsenka <[email protected]>
4291 Date: Sun Aug 26 11:35:44 2018 -0700
4292
4293 Avoid static initialization of pprof path for symbolization.
4294
4295 This is one of the things that chrome's fork fixes, but with c++11 we
4296 can do it even nicer. Proposed fix is to use c++11 local static
4297 variable to ensure that pprof path is initialized once on as-needed
4298 basis.
4299
4300 M src/symbolize.cc
4301
4302 commit 69867c523b334028c90f2acf22b5ae528fdc2356
4303 Author: Holy Wu <[email protected]>
4304 Date: Fri Aug 10 17:19:29 2018 +0800
4305
4306 Clean up MSVC projects
4307
4308 1.Remove superfluous per file settings for include directory and
4309 runtime library.
4310 2.Remove unnecessary project tcmalloc_minimal_unittest-static. We
4311 can simply build libtcmalloc_minimal as a static library and then
4312 link against the single .lib file.
4313 3.Add separate configurations of patching and overriding facility
4314 for release mode.
4315
4316 M Makefile.am
4317 M gperftools.sln
4318 M vsprojects/addr2line-pdb/addr2line-pdb.vcxproj
4319 M vsprojects/addressmap_unittest/addressmap_unittest.vcxproj
4320 M
4321 vsprojects/current_allocated_bytes_test/current_allocated_bytes_test.vcxproj
4322 M vsprojects/frag_unittest/frag_unittest.vcxproj
4323 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj
4324 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj.filters
4325 M
4326 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj
4327 M vsprojects/malloc_extension_test/malloc_extension_test.vcxproj
4328 M vsprojects/malloc_hook_test/malloc_hook_test.vcxproj
4329 M vsprojects/markidle_unittest/markidle_unittest.vcxproj
4330 M vsprojects/nm-pdb/nm-pdb.vcxproj
4331 M vsprojects/packed-cache_test/packed-cache_test.vcxproj
4332 M vsprojects/page_heap_test/page_heap_test.vcxproj
4333 M vsprojects/pagemap_unittest/pagemap_unittest.vcxproj
4334 M vsprojects/preamble_patcher_test/preamble_patcher_test.vcxproj
4335 M vsprojects/realloc_unittest/realloc_unittest.vcxproj
4336 M vsprojects/sampler_test/sampler_test.vcxproj
4337 M vsprojects/stack_trace_table_test/stack_trace_table_test.vcxproj
4338 M vsprojects/system-alloc_unittest/system-alloc_unittest.vcxproj
4339 M
4340 vsprojects/tcmalloc_minimal_large/tcmalloc_minimal_large_unittest.vcxproj
4341 M
4342 vsprojects/tcmalloc_minimal_unittest/tcmalloc_minimal_unittest.vcxproj
4343 M vsprojects/thread_dealloc_unittest/thread_dealloc_unittest.vcxproj
4344 D vsprojects/tmu-static/tmu-static.vcxproj
4345 D vsprojects/tmu-static/tmu-static.vcxproj.filters
4346
4347 commit f2bca77aed83107197abadee0652c39e2c4c73d3
4348 Author: Brian Silverman <[email protected]>
4349 Date: Thu Aug 9 16:33:39 2018 -0700
4350
4351 Fix page_heap_test flakiness
4352
4353 After 71fa9f87 (use 2-level page map for 48-bit addresses), there is a
4354 potential for allocs of additional leaves in the middle of the test.
4355 With ASLR enabled on amd64 (without TCMALLOC_SMALL_BUT_SLOW), this
4356 results in the test failing sometimes (< 1% of the time). Instead,
4357 make
4358 sure all those leaves are allocated before deciding on the baseline
4359 heap
4360 usage in the test.
4361
4362 M src/tests/page_heap_test.cc
4363
4364 commit c41688bf20186723367b560ceb539b2330951ddb
4365 Author: Aliaksey Kandratsenka <[email protected]>
4366 Date: Sun Aug 5 20:39:07 2018 -0700
4367
4368 Use standard-conforming alignof in debugallocation.cc
4369
4370 Clang was giving warning that alignof(<variable>) is extension. So
4371 we're now doing alignof of decltype.
4372
4373 M src/debugallocation.cc
4374
4375 commit 71c8cedacafe700e662c008b65f5064b23426070
4376 Author: Aliaksey Kandratsenka <[email protected]>
4377 Date: Sun Aug 5 18:54:01 2018 -0700
4378
4379 Fix incompatible aliasing warnings
4380
4381 We aliased functions with different signatures and gcc now correctly
4382 gives warning for that. Originally gcc 5 same code merging feature
4383 caused us to alias more than necessary, but I am not able to reproduce
4384 this problem anymore. So we're now aliasing only compatible functions.
4385
4386 M src/libc_override_gcc_and_weak.h
4387 M src/tcmalloc.cc
4388
4389 commit 8dd304035848e780372a5a4bb1afac2b2d20bc43
4390 Author: Aliaksey Kandratsenka <[email protected]>
4391 Date: Sun Aug 5 19:52:11 2018 -0700
4392
4393 Format and fix out of bound access in CpuProfilerSwitch
4394
4395 GCC was giving warning on snprintf and it hinted at base_profile_name
4396 and full_profile_name not being long enough.
4397
4398 Fix is to ensure base_profile_name is long enough for PATH_MAX and
4399 that full_profile_name is enough longer to fit extra chars.
4400
4401 M src/profiler.cc
4402
4403 commit 467502e70acb89f36b475bf29fe12333939fe5c7
4404 Author: Aliaksey Kandratsenka <[email protected]>
4405 Date: Sun Aug 5 17:42:04 2018 -0700
4406
4407 provide constexpr constructor for Sampler
4408
4409 Constexpr constructor lets us use Sampler in TLS variable in cleaner
4410 way.
4411
4412 M src/sampler.h
4413 M src/tests/sampler_test.cc
4414
4415 commit 1fb543cc708f22017cb39ae874a28c0addcf5a63
4416 Author: HolyWu <[email protected]>
4417 Date: Tue May 1 18:04:57 2018 +0800
4418
4419 Patch _free_dbg to make Debug mode in MSVC works
4420
4421 Like _free_base, _free_dbg is called by CRT internal functions or
4422 operator delete in Debug mode.
4423
4424 This closes #719 and closes #894.
4425
4426 [[email protected]: trivial formatting fixes]
4427 [[email protected]: build free_dbg even in release builds]
4428
4429 M src/windows/patch_functions.cc
4430
4431 commit 267f431d8071799282f9f4af10455b7b6adeee38
4432 Author: Gabriel Marin <[email protected]>
4433 Date: Thu Aug 2 17:16:25 2018 -0700
4434
4435 Use indirect system calls in the linux spinlock implementation
4436
4437 M src/base/spinlock_linux-inl.h
4438
4439 commit 73ee9b15440d72d5c4f93586ea1179c0a265980c
4440 Author: Gabriel Marin <[email protected]>
4441 Date: Mon Jul 2 09:31:10 2018 -0700
4442
4443 Use indirect system calls in the mmap malloc hooks.
4444
4445 sys_{mmap|munmap|mremap}(...) calls are replaced with
4446 syscall(SYS_{mmap|munmap|mremap}, ...).
4447
4448 M src/malloc_hook_mmap_linux.h
4449
4450 commit 3af509d4f91aa1fae3f8caeb17b158f20e9b5e59
4451 Author: Gabriel Marin <[email protected]>
4452 Date: Tue Jul 17 18:28:01 2018 -0700
4453
4454 benchmark: use angle brackets to include ucontext.h
4455
4456 Using quotes for a system header file fails a presubmit check in
4457 Chromium.
4458
4459 M benchmark/unwind_bench.cc
4460
4461 commit 0cdda6d7cc5e082bb704a9352934148e13dc0a0d
4462 Author: Gabriel Marin <[email protected]>
4463 Date: Tue Jul 17 18:36:31 2018 -0700
4464
4465 use utf-8 for special symbols
4466
4467 A few files in the windows port that haven't been touched in a
4468 long time,
4469 use non-utf8 codes for special symbols, breaking a presubmit
4470 checker in
4471 Chromium.
4472
4473 M src/windows/ia32_modrm_map.cc
4474 M src/windows/ia32_opcode_map.cc
4475 M src/windows/mini_disassembler.h
4476
4477 commit c7a0cfda880a7310a030c62546f04a2987a11ec2
4478 Author: etienneb <[email protected]>
4479 Date: Thu Apr 20 12:58:02 2017 -0700
4480
4481 Fix potential missing nul character on resolved symbol names
4482
4483 The symbol name returned by SymFromName may not contains a NUL
4484 character
4485 when the symbol name is exactly the size of the buffer. It believe
4486 this
4487 may also happen when the symbol name is too long and truncated.
4488
4489 The original code is based on:
4490 https://msdn.microsoft.com/en-us/library/windows/desktop/ms680580(v=vs.85).aspx
4491
4492 A right implementation can be found here:
4493 https://cs.chromium.org/chromium/src/base/debug/stack_trace_win.cc?l=145&rcl=f4ecb9e37e9e2d59e32b8b96f23ac4a1e33b9552
4494
4495 As described here:
4496 https://msdn.microsoft.com/en-us/library/windows/desktop/ms680686(v=vs.85).aspx
4497
4498 NameLen
4499 The length of the name, in characters, not including the
4500 null-terminating character.
4501 MaxNameLen
4502 The size of the Name buffer, in characters. If this member is 0,
4503 the Name member is not used.
4504
4505 This issue was causing the catapult symbolisation script to encode
4506 incorrect (random) characters into the symbol names.
4507 See the example in the bug.
4508
4509 Original BUG=713741
4510
4511 Review-Url: https://codereview.chromium.org/2832643004
4512 Cr-Commit-Position: refs/heads/master@{#466098}
4513
4514 BUG=724399,b:70905156
4515
4516 Change-Id: Ia6de4ddbd5c66a966b74ecec3fda5e853f36073b
4517
4518 M src/windows/addr2line-pdb.c
4519
4520 commit e42bfc8c06abd1ea1a7e92d1d6d72d1959ec27b6
4521 Author: Gabriel Marin <[email protected]>
4522 Date: Thu Jul 19 11:10:31 2018 -0700
4523
4524 tcmalloc: use relative addresses with the windows addr2line wrapper
4525
4526 Modifies the Windows addr2line wrapper to expect addresses relative to
4527 DllBase to better simulate how addr2line works with modules in Linux.
4528
4529 Windows DLLs have a concept of "default load address" which hints
4530 to the OS
4531 where to load the binary image after relocation. The dbghelp.dll
4532 symbolization library will load the module at this location in
4533 the virtual
4534 address space meaning the caller of these functions would need to
4535 be aware
4536 of the base address. This makes things unnecessarily complex in the
4537 face of
4538 ASLR and also diverges from the behavior of addr2line when used
4539 with linux-
4540 style DSOs. This CL simply adds the module base address to the
4541 incoming
4542 addresses, thereby making the input relative addresses for the
4543 module which
4544 both is easier to use and lines up better with linux's addr2line
4545 behavior.
4546
4547 These changes were made originally as part of CL
4548 https://codereview.chromium.org/2730473002.
4549
4550 BUG=724399,b:70905156
4551
4552 Change-Id: I0abe9e0c380e7e60ae29a11021bb805b31718d08
4553
4554 M src/windows/addr2line-pdb.c
4555
4556 commit d8f8d1cced71c37ff86ff577c3afe03b7cd1648f
4557 Author: Gabriel Marin <[email protected]>
4558 Date: Wed Jul 18 18:50:30 2018 -0700
4559
4560 tcmalloc: add long form flag '--exe' to specify the binary
4561
4562 This change was originally made as part of CL
4563 https://codereview.chromium.org/2730473002
4564
4565 BUG=724399,b:70905156
4566
4567 Change-Id: I33ee28eac4dab6077cf162358cf4c10b41c0f6a2
4568
4569 M src/windows/addr2line-pdb.c
4570
4571 commit 25c53aca125b54df1d5c0e64c96f074bdb92ade2
4572 Author: Gabriel Marin <[email protected]>
4573 Date: Wed Jul 18 18:39:15 2018 -0700
4574
4575 tcmalloc: fixes for the windows addr2line wrapper
4576
4577 Fixes a few small format string bugs and makes the windows library
4578 configuration #defines more friendly for chromium builds.
4579
4580 Based on original CL: https://codereview.chromium.org/2730473002
4581
4582 BUG=724399,b:70905156
4583
4584 Change-Id: If54c6cfc5c508744d2ee18f251677b2456acda72
4585
4586 M src/windows/addr2line-pdb.c
4587
4588 commit f02e28f3488122f497755de59dc822f08c6449b9
4589 Author: Gabriel Marin <[email protected]>
4590 Date: Tue Jul 31 15:35:53 2018 -0700
4591
4592 Replace builtin_expect configure test with a direct GCC compiler check
4593
4594 M configure.ac
4595 M src/base/basictypes.h
4596
4597 commit 62c4eca6e7ec9dc78333dc6b499f90eef9849378
4598 Author: obones <[email protected]>
4599 Date: Fri Jul 20 11:07:02 2018 +0200
4600
4601 Under x64, the PE loader looks for callbacks in constant sections
4602
4603 [[email protected]: deleted some trailing whitespace]
4604
4605 M src/windows/port.cc
4606
4607 commit 0b588e7490eac4ef37b59c9265f1f3f05f4bb27c
4608 Author: Brian Silverman <[email protected]>
4609 Date: Tue Jul 31 15:53:47 2018 -0700
4610
4611 Fix uninitialized memory use in sampler_test
4612
4613 Sampler's documentation states the following:
4614 C++03 requires that types stored in TLS be POD. As a result,
4615 you must
4616 initialize these members to {0, 0, false} before using this class!
4617
4618 However, the test code wasn't doing that. MemorySanitizer and
4619 UndefinedBehaviorSanitizer both failed because of it.
4620
4621 M src/tests/sampler_test.cc
4622
4623 commit 51a5613f21a34197ed655c363c54c71cf68271ba
4624 Author: Aliaksei Kandratsenka <[email protected]>
4625 Date: Sun Aug 5 15:28:27 2018 -0700
4626
4627 Upgrade MSVC projects to MSVC2015
4628
4629 M Makefile.am
4630 M gperftools.sln
4631 D vsprojects/addr2line-pdb/addr2line-pdb.vcproj
4632 A vsprojects/addr2line-pdb/addr2line-pdb.vcxproj
4633 A vsprojects/addr2line-pdb/addr2line-pdb.vcxproj.filters
4634 D vsprojects/addressmap_unittest/addressmap_unittest.vcproj
4635 A vsprojects/addressmap_unittest/addressmap_unittest.vcxproj
4636 A vsprojects/addressmap_unittest/addressmap_unittest.vcxproj.filters
4637 D
4638 vsprojects/current_allocated_bytes_test/current_allocated_bytes_test.vcproj
4639 A
4640 vsprojects/current_allocated_bytes_test/current_allocated_bytes_test.vcxproj
4641 A
4642 vsprojects/current_allocated_bytes_test/current_allocated_bytes_test.vcxproj.filters
4643 D vsprojects/frag_unittest/frag_unittest.vcproj
4644 A vsprojects/frag_unittest/frag_unittest.vcxproj
4645 A vsprojects/frag_unittest/frag_unittest.vcxproj.filters
4646 D vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
4647 A vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj
4648 A vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcxproj.filters
4649 D
4650 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
4651 A
4652 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj
4653 A
4654 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcxproj.filters
4655 D vsprojects/malloc_extension_test/malloc_extension_test.vcproj
4656 A vsprojects/malloc_extension_test/malloc_extension_test.vcxproj
4657 A
4658 vsprojects/malloc_extension_test/malloc_extension_test.vcxproj.filters
4659 D vsprojects/malloc_hook_test/malloc_hook_test.vcproj
4660 A vsprojects/malloc_hook_test/malloc_hook_test.vcxproj
4661 A vsprojects/malloc_hook_test/malloc_hook_test.vcxproj.filters
4662 D vsprojects/markidle_unittest/markidle_unittest.vcproj
4663 A vsprojects/markidle_unittest/markidle_unittest.vcxproj
4664 A vsprojects/markidle_unittest/markidle_unittest.vcxproj.filters
4665 D vsprojects/nm-pdb/nm-pdb.vcproj
4666 A vsprojects/nm-pdb/nm-pdb.vcxproj
4667 A vsprojects/nm-pdb/nm-pdb.vcxproj.filters
4668 D vsprojects/packed-cache_test/packed-cache_test.vcproj
4669 A vsprojects/packed-cache_test/packed-cache_test.vcxproj
4670 A vsprojects/packed-cache_test/packed-cache_test.vcxproj.filters
4671 D vsprojects/page_heap_test/page_heap_test.vcproj
4672 A vsprojects/page_heap_test/page_heap_test.vcxproj
4673 A vsprojects/page_heap_test/page_heap_test.vcxproj.filters
4674 D vsprojects/pagemap_unittest/pagemap_unittest.vcproj
4675 A vsprojects/pagemap_unittest/pagemap_unittest.vcxproj
4676 A vsprojects/pagemap_unittest/pagemap_unittest.vcxproj.filters
4677 D vsprojects/preamble_patcher_test/preamble_patcher_test.vcproj
4678 A vsprojects/preamble_patcher_test/preamble_patcher_test.vcxproj
4679 A
4680 vsprojects/preamble_patcher_test/preamble_patcher_test.vcxproj.filters
4681 D vsprojects/realloc_unittest/realloc_unittest.vcproj
4682 A vsprojects/realloc_unittest/realloc_unittest.vcxproj
4683 A vsprojects/realloc_unittest/realloc_unittest.vcxproj.filters
4684 D vsprojects/sampler_test/sampler_test.vcproj
4685 A vsprojects/sampler_test/sampler_test.vcxproj
4686 A vsprojects/sampler_test/sampler_test.vcxproj.filters
4687 D vsprojects/stack_trace_table_test/stack_trace_table_test.vcproj
4688 A vsprojects/stack_trace_table_test/stack_trace_table_test.vcxproj
4689 A
4690 vsprojects/stack_trace_table_test/stack_trace_table_test.vcxproj.filters
4691 D vsprojects/system-alloc_unittest/system-alloc_unittest.vcproj
4692 A vsprojects/system-alloc_unittest/system-alloc_unittest.vcxproj
4693 A
4694 vsprojects/system-alloc_unittest/system-alloc_unittest.vcxproj.filters
4695 D
4696 vsprojects/tcmalloc_minimal_large/tcmalloc_minimal_large_unittest.vcproj
4697 A
4698 vsprojects/tcmalloc_minimal_large/tcmalloc_minimal_large_unittest.vcxproj
4699 A
4700 vsprojects/tcmalloc_minimal_large/tcmalloc_minimal_large_unittest.vcxproj.filters
4701 D
4702 vsprojects/tcmalloc_minimal_unittest/tcmalloc_minimal_unittest.vcproj
4703 A
4704 vsprojects/tcmalloc_minimal_unittest/tcmalloc_minimal_unittest.vcxproj
4705 A
4706 vsprojects/tcmalloc_minimal_unittest/tcmalloc_minimal_unittest.vcxproj.filters
4707 D vsprojects/thread_dealloc_unittest/thread_dealloc_unittest.vcproj
4708 A vsprojects/thread_dealloc_unittest/thread_dealloc_unittest.vcxproj
4709 A
4710 vsprojects/thread_dealloc_unittest/thread_dealloc_unittest.vcxproj.filters
4711 D vsprojects/tmu-static/tmu-static.vcproj
4712 A vsprojects/tmu-static/tmu-static.vcxproj
4713 A vsprojects/tmu-static/tmu-static.vcxproj.filters
4714
4715 commit 44da4ce5397344342e71ea3cc2c246c2ca4aedb6
4716 Author: Aliaksey Kandratsenka <[email protected]>
4717 Date: Sun Aug 5 15:09:20 2018 -0700
4718
4719 build with c++11 or later
4720
4721 M configure.ac
4722 A m4/ax_cxx_compile_stdcxx.m4
4723
4724 commit f47a52ce85c3d8d559aaae7b7a426c359fbca225
4725 Author: HolyWu <[email protected]>
4726 Date: Mon Apr 30 15:41:50 2018 +0800
4727
4728 Make _recalloc adhere to MS's definition
4729
4730 M src/tcmalloc.cc
4731 M src/windows/override_functions.cc
4732
4733 commit fe87ffb7ea5a7c6ce4dea45222331716907ddbf4
4734 Author: Junhao Li <[email protected]>
4735 Date: Sun May 20 13:45:32 2018 -0400
4736
4737 Disable large allocation report by default
4738
4739 Fixes issue #360.
4740
4741 [[email protected]: adjusted commit message a bit]
4742 [[email protected]: adjusted configure help message]
4743 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
4744
4745 M configure.ac
4746 M src/tcmalloc.cc
4747
4748 commit 9608fa3bcf8020d35f59fbf70cd3cbe4b015b972
4749 Author: Aliaksey Kandratsenka <[email protected]>
4750 Date: Sun Apr 29 16:55:59 2018 -0700
4751
4752 bumped version to 2.7
4753
4754 M NEWS
4755 M configure.ac
4756 M src/windows/config.h
4757 M src/windows/gperftools/tcmalloc.h
4758
4759 commit db890ccfade3cf054ec330dfb4a9ddfab971367e
4760 Author: HolyWu <[email protected]>
4761 Date: Mon Apr 30 12:08:51 2018 +0800
4762
4763 Clean up src/windows/config.h
4764
4765 M src/windows/config.h
4766
4767 commit 497ea331654f7ea9e4e6405f5beaea8ebcc9d10b
4768 Author: HolyWu <[email protected]>
4769 Date: Sun Apr 15 22:47:17 2018 +0800
4770
4771 Fix WIN32_OVERRIDE_ALLOCATORS for VS2017
4772
4773 At first I try to add some functions as what Chrome does at their
4774 https://chromium.googlesource.com/chromium/src/+/master/base/allocator/allocator_shim_override_ucrt_symbols_win.h,
4775 but it still fails. So I decide to remove all heap-related objects
4776 from libucrt.lib to see what happens. At the end I find that a lot of
4777 functions in the CRT directly invoke _malloc_base instead of
4778 malloc (and the others alike), hence we need to override them as well.
4779
4780 This should close issue #716.
4781
4782 [[email protected]: added reference to ticket]
4783 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
4784
4785 M src/libc_override.h
4786 M src/tcmalloc.cc
4787 M src/windows/override_functions.cc
4788 M src/windows/port.h
4789
4790 commit ebc85cca9083241ae360e0c68e2f1d7538001fa3
4791 Author: HolyWu <[email protected]>
4792 Date: Tue Apr 17 11:40:49 2018 +0800
4793
4794 Enable aligned new/delete declarations on Windows when applicable
4795
4796 M src/windows/gperftools/tcmalloc.h
4797 M src/windows/gperftools/tcmalloc.h.in
4798
4799 commit a3badd6d2127c161f6fbad853fccd6e8064fcde7
4800 Author: Aliaksey Kandratsenka <[email protected]>
4801 Date: Sun Apr 29 16:34:54 2018 -0700
4802
4803 Really fix CheckAddressBits compilation warning
4804
4805 It looks like previous fix introduced another warning on gcc for
4806 i386. Somehow it barked on kAddressBits shift even that shift was dead
4807 code.
4808
4809 Anyways, we now avoid possibility of undefined shift even
4810 stronger. And it actually removes compile warning on all compilers I
4811 tested.
4812
4813 M src/system-alloc.cc
4814
4815 commit 7c718fe1768273f94acf2bd43cfa1f44a4b9bdcc
4816 Author: HolyWu <[email protected]>
4817 Date: Thu Apr 19 13:49:28 2018 +0800
4818
4819 Add tests for sized deallocation
4820
4821 M src/tests/tcmalloc_unittest.cc
4822
4823 commit 30e5e614a846d5f046454341f108bc4cb2b70189
4824 Author: Fabrice Fontaine <[email protected]>
4825 Date: Sun Apr 22 19:17:12 2018 +0200
4826
4827 Fix build without static libraries
4828
4829 Only add -static to malloc_bench_LDFLAGS and binary_trees_LDFLAGS if
4830 ENABLE_STATC is set otherwise build with some compilers will fail if
4831 user has decided to build only the shared version of gperftools
4832 libraries
4833
4834 Signed-off-by: Fabrice Fontaine <[email protected]>
4835
4836 M Makefile.am
4837
4838 commit 836c4f29a585e4b176738f64cf8d312171be5fe0
4839 Author: Ben Dang <[email protected]>
4840 Date: Fri Apr 13 10:54:05 2018 -0700
4841
4842 Update documentation for heap_checker.html
4843
4844 Make it clear that the static methods used require the "namespace",
4845 `HeapLeakChecker::`.
4846
4847 M docs/heap_checker.html
4848
4849 commit e47d0d1c5142a4ad1203e58db31af2f58bafcda9
4850 Author: Tulio Magno Quites Machado Filho <[email protected]>
4851 Date: Tue Apr 10 15:56:52 2018 -0300
4852
4853 powerpc: Re-enable VDSO support
4854
4855 Powerpc requires VDSO support in order to produce a stack trace.
4856 Without this, it isn't possible to complete a build.
4857 Tested on powerpc, powerpc64 and powerpc64le.
4858
4859 M src/base/vdso_support.h
4860
4861 commit 0a66dd3a6ac60e57d9c01f17309a40e2ebe06837
4862 Author: Christoph Müllner <[email protected]>
4863 Date: Sat Apr 7 09:00:09 2018 +0200
4864
4865 linux: add aarch64_ilp32 support.
4866
4867 This patch adds aarch64 (arm64) ILP32 support by using
4868 the proper syscalls for sys_fcntl(), sys_fstat(), sys_lseek()
4869 and sys_mmap().
4870
4871 Signed-off-by: Christoph Müllner <[email protected]>
4872
4873 M src/base/linux_syscall_support.h
4874
4875 commit 05dff0966362932631a2120eaeb8eebb50dd9300
4876 Author: Christoph Müllner <[email protected]>
4877 Date: Sat Apr 7 08:30:32 2018 +0200
4878
4879 Fix signature of sbrk.
4880
4881 The manpage of sbrk says that its argument is of type intptr_t.
4882 This patch fixes a compiler warning on gcc 7.3.0.
4883
4884 Signed-off-by: Christoph Müllner <[email protected]>
4885
4886 M src/malloc_hook_mmap_linux.h
4887
4888 commit 33ae0ed2ae7ce143276658be029c28ec6f2a73c9
4889 Author: Aliaksey Kandratsenka <[email protected]>
4890 Date: Mon Apr 9 20:57:29 2018 -0700
4891
4892 unbreak compilation on GNU/Linux i386
4893
4894 Recent commit to fix int overflow for implausibly huge allocation
4895 added call to std::min. Notably, first arg was old size divided by
4896 unsigned long 4. And on GNU/Linux i386 size_t is not long. So such
4897 division was promoting first arg to unsigned long while second arg was
4898 still size_t, so just unsigned. And that caused compilation to fail.
4899
4900 Fix is droping 'ul'.
4901
4902 M src/tcmalloc.cc
4903
4904 commit 977e0d45003514d088e4bb83474c985094acee47
4905 Author: Christoph Müllner <[email protected]>
4906 Date: Sat Apr 7 08:35:42 2018 +0200
4907
4908 Remove not needed header in vdso_support.cc.
4909
4910 Signed-off-by: Christoph Müllner <[email protected]>
4911
4912 M src/base/vdso_support.cc
4913
4914 commit 36bfa9a4046109efa40ccc0806c72331de9c915b
4915 Author: Mao <[email protected]>
4916 Date: Mon Mar 26 17:22:19 2018 +0800
4917
4918 Enable tcmalloc VDSO support only on x86 to reduce static initializers
4919
4920 Background context
4921 ------------------
4922 crrev.com/1466173002 switched the GN tcmalloc target from source_set
4923 -> static_library. There are good reasons for keeping tcmalloc a
4924 source_set (see "Note on static libraries" in [1]). However, in the
4925 current state source_set was exposing extra static initializers in the
4926 GN build which, are not present in the gyp build due to the linker gc
4927 sections.
4928
4929 Resolution of this CL
4930 ---------------------
4931 The fact that vdso_support.cc is GC-ed by the linker is the symptom
4932 that such code is unreachable. A search in the codebase shows that the
4933 only client is stacktrace_x86-inl.h, which depends on VDSO only when
4934 defined(__linux__) && defined(__i386__) This CL is therefore matching
4935 this condition in vdso_support.h and conditioning the #define
4936 HAVE_VDSO_SUPPORT with the same conditions.
4937
4938 [1]
4939 https://chromium.googlesource.com/chromium/src/+/master/tools/gn/docs/cookbook.md
4940
4941 References:
4942 https://bugs.chromium.org/p/chromium/issues/detail?id=559766
4943 https://bugs.chromium.org/p/chromium/issues/detail?id=564618
4944
4945 M src/base/vdso_support.h
4946
4947 commit 1cb5de6db9f30daa56a0b54652a7ff6da288c1bd
4948 Author: Mao <[email protected]>
4949 Date: Mon Mar 26 17:28:28 2018 +0800
4950
4951 Explicitly prevent int overflow
4952
4953 M src/tcmalloc.cc
4954
4955 commit 8f63f2bb983e93b57e24a05964104f1d833c575b
4956 Author: Aliaksey Kandratsenka <[email protected]>
4957 Date: Sat Mar 24 21:15:01 2018 -0700
4958
4959 Correctly detect presence of various functions in tcmalloc.h
4960
4961 This fixes test miscompilations on some clang 6.0
4962 installations. Particularly issue #946.
4963
4964 Defines tested by tcmalloc.h where totally wrong defines. configure
4965 actually puts defines of HAVE_DECL_FOOBAR kind in config.h. Which is
4966 what we're using now.
4967
4968 M src/tcmalloc.h
4969
4970 commit 736648887b0b27ec70a50f0954a614dd65c1eb12
4971 Author: Aliaksey Kandratsenka <[email protected]>
4972 Date: Sat Mar 24 15:54:48 2018 -0700
4973
4974 Don't test OOM handling of debugallocator
4975
4976 This may fix issue #969.
4977
4978 When testing OOM handling we set up sys allocator that fails memory
4979 allocation. But debugallocator itself allocates some internal metadata
4980 memory via malloc and crashes if those allocations fail. So
4981 occasionally this test failed when debugallocator's internal malloc
4982 ended up causing sys allocator.
4983
4984 So instead of failing tests from time to time, we drop it for debug
4985 allocator. It's OOM handling is already crashy anyways.
4986
4987 M src/tests/tcmalloc_unittest.cc
4988
4989 commit c4a8e00da41c43da7e78858e06f3153f09f6967d
4990 Author: Aliaksey Kandratsenka <[email protected]>
4991 Date: Sat Mar 24 19:09:15 2018 -0700
4992
4993 Fix warning about one of CheckAddressBits functions unused
4994
4995 M src/system-alloc.cc
4996
4997 commit 47c99cf492fa340671fa89afec263b8d6e859755
4998 Author: Aliaksey Kandratsenka <[email protected]>
4999 Date: Sat Mar 24 18:29:06 2018 -0700
5000
5001 unbreak printing large span stats
5002
5003 One of recent commits started passing kMaxPages to printf but not used
5004 it. Thankfully compilers gave us warning. Apparently intention was to
5005 print real value of kMaxPages, so this is what we're doing now.
5006
5007 M src/tcmalloc.cc
5008
5009 commit 34f78a2dcd8160b483d86474702785337bbeb7ac
5010 Author: Aliaksey Kandratsenka <[email protected]>
5011 Date: Sun Mar 18 16:39:06 2018 -0700
5012
5013 bumped version to 2.7rc
5014
5015 M NEWS
5016 M configure.ac
5017 M src/windows/config.h
5018 M src/windows/gperftools/tcmalloc.h
5019
5020 commit db98aac55a90d45e9abe188c8a928c3561d3f223
5021 Author: Todd Lipcon <[email protected]>
5022 Date: Sat Mar 17 09:43:34 2018 -0700
5023
5024 Add a central free list for kMaxPages-sized spans
5025
5026 Previously, the central free list with index '0' was always unused,
5027 since freelist index 'i' tracked spans of length 'i' and there are no
5028 spans of length 0. This meant that there was no freelist for spans of
5029 length 'kMaxPages'. In the default configuration, this corresponds to
5030 1MB, which is a relatively common allocation size in a lot of
5031 applications.
5032
5033 This changes the free list indexing so that index 'i' tracks spans of
5034 length 'i + 1', meaning that free list index 0 is now used and
5035 freelist[kMaxPages - 1] tracks allocations of kMaxPages size (1MB by
5036 default).
5037
5038 This also fixes the stats output to indicate '>128' for the large
5039 spans
5040 stats rather than the incorrect '>255' which must have referred to a
5041 historical value of kMaxPages.
5042
5043 No new tests are added since this code is covered by existing tests.
5044
5045 M docs/pageheap.dot
5046 M docs/pageheap.gif
5047 M docs/tcmalloc.html
5048 M src/common.h
5049 M src/page_heap.cc
5050 M src/page_heap.h
5051 M src/tcmalloc.cc
5052
5053 commit d7be9385609328a03c5cfe775473a7dc8e014fd3
5054 Author: Aliaksey Kandratsenka <[email protected]>
5055 Date: Sun Mar 4 23:29:46 2018 -0800
5056
5057 implement more robust detection of sized delete support
5058
5059 As reported in issue #954, osx clang compiler is able to optimize our
5060 previous detection away while not really having runtime support for
5061 sized delete. So this time we use AC_LINK_IFELSE and more robust code
5062 to prevent compiler from optimizing away sized delete call. This
5063 should reliably catch "bad" compilers.
5064
5065 Special thanks to Alexey Serbin for reporting the issue, suggesting a
5066 fix and verifying it.
5067
5068 Fixes issue #954.
5069
5070 M configure.ac
5071
5072 commit f1d3fe4a21e339a3fd6e4592ee7444484a7b92dc
5073 Author: Aliaksey Kandratsenka <[email protected]>
5074 Date: Sun Feb 25 13:55:53 2018 -0800
5075
5076 refactored handling of reverse span set iterator for correctness
5077
5078 I.e. no more questionable memcpy and we run iterator's destructor when
5079 we remove span from SpanSet.
5080
5081 M src/page_heap.cc
5082 M src/span.h
5083
5084 commit 59c77be0fad2a49e31d51877985e7c48f73afcea
5085 Author: Todd Lipcon <[email protected]>
5086 Date: Sun Feb 11 16:21:42 2018 -0800
5087
5088 Update docs for central page heap to reflect tree
5089
5090 M docs/pageheap.dot
5091 M docs/pageheap.gif
5092 M docs/tcmalloc.html
5093
5094 commit 06c9414ec423ffe442c047b2560555f9d5847b1d
5095 Author: Aliaksey Kandratsenka <[email protected]>
5096 Date: Sat Jun 3 15:31:06 2017 -0700
5097
5098 Implemented O(log n) searching among large spans
5099
5100 This is implemented via std::set with custom STL allocator that
5101 delegates to PageHeapAllocator. Free large spans are not linked
5102 together via linked list, but inserted into std::set. Spans also store
5103 iterators to std::set positions pointing to them. So that removing
5104 span from set is fast too.
5105
5106 Patch implemented by Aliaksey Kandratsenka and Todd Lipcon based on
5107 earlier research and experimentation by James Golick.
5108
5109 Addresses issue #535
5110
5111 [[email protected]: added Todd's fix for building on OSX]
5112 [[email protected]: removed unnecessary Span constructor]
5113 [[email protected]: added const for SpanSet comparator]
5114 [[email protected]: added operator != for STLPageHeapAllocator]
5115
5116 M src/page_heap.cc
5117 M src/page_heap.h
5118 M src/page_heap_allocator.h
5119 M src/span.h
5120 M src/tests/tcmalloc_unittest.cc
5121
5122 commit a42e44738a330783781541411392ba27df88b8b3
5123 Author: Ishan Arora <[email protected]>
5124 Date: Tue Jan 9 01:33:13 2018 +0530
5125
5126 typo in docs/tcmalloc.html
5127
5128 M docs/tcmalloc.html
5129
5130 commit 71bf09aabe93dbe390ebb54c57d3c5b3d627c142
5131 Author: Aliaksey Kandratsenka <[email protected]>
5132 Date: Sat Dec 9 09:58:31 2017 -0800
5133
5134 bumped version to 2.6.3
5135
5136 M NEWS
5137 M configure.ac
5138 M src/windows/config.h
5139 M src/windows/gperftools/tcmalloc.h
5140
5141 commit 0bccb5e658762f30e6e3abbc40dc244e8f8911c1
5142 Author: Aliaksey Kandratsenka <[email protected]>
5143 Date: Sat Dec 9 08:51:19 2017 -0800
5144
5145 fix malloc fast path for patched windows functions
5146
5147 malloc_fast_path now receives oom function instead of full allocation
5148 function and windows/patch_function.cc wasn't updated until now. It
5149 caused assertion failures as reported in issue #944.
5150
5151 M src/windows/patch_functions.cc
5152
5153 commit 8b1d13c631024ce7e80508c284c7bc5b1327e131
5154 Author: Stephan Zuercher <[email protected]>
5155 Date: Tue Dec 5 09:49:36 2017 -0800
5156
5157 configure.ac: use link check for std::align_val_t
5158
5159 M configure.ac
5160
5161 commit 36ab068baa3b07c083bd79029622aa382aeff84a
5162 Author: Stephan Zuercher <[email protected]>
5163 Date: Tue Dec 5 09:31:38 2017 -0800
5164
5165 configure.ac: better test for -faligned-new
5166
5167 XCode 9 provides only partial support for aligned new/delete when
5168 -faligned-new
5169 is specified. Require successful linking to enable aligned new/delete.
5170
5171 M configure.ac
5172
5173 commit 6a4b0799975576c2722daa75b35f6f65220d22b6
5174 Author: Aliaksey Kandratsenka <[email protected]>
5175 Date: Thu Nov 30 18:19:17 2017 +0000
5176
5177 bumped version to 2.6.2
5178
5179 M NEWS
5180 M configure.ac
5181 M src/windows/config.h
5182 M src/windows/gperftools/tcmalloc.h
5183
5184 commit 22917145186bc8bbb6d98090d788ed7c87b282b1
5185 Author: Aliaksey Kandratsenka <[email protected]>
5186 Date: Thu Nov 30 16:36:33 2017 +0000
5187
5188 implement fast-path for memalign/aligned_alloc/tc_new_aligned
5189
5190 We're taking advantage of "natural" alignedness of our size classes
5191 and instead of previous loop over size classes looking for suitably
5192 aligned size, we now directly compute right size. See align_size_up
5193 function. And that gives us ability to use our existing malloc
5194 fast-path to make memalign neat and fast in most common
5195 cases. I.e. memalign/aligned_alloc now only tail calls and thus avoids
5196 expensive prologue/epilogue and is almost as fast as regular malloc.
5197
5198 M src/common.cc
5199 M src/tcmalloc.cc
5200
5201 commit 8b9728b02325a9bee9afd3957e506c111ed40d40
5202 Author: Aliaksey Kandratsenka <[email protected]>
5203 Date: Thu Nov 30 18:12:32 2017 +0000
5204
5205 add memalign benchmark to malloc_bench
5206
5207 M benchmark/malloc_bench.cc
5208
5209 commit 79c91a9810c928d7f6d0cb168c10ce470a714aa8
5210 Author: Aliaksey Kandratsenka <[email protected]>
5211 Date: Wed Nov 29 21:15:04 2017 +0000
5212
5213 always define empty PERFTOOLS_NOTHROW
5214
5215 Because somehow clang still builds "this function will not throw" code
5216 even with noexcept. Which breaks performance of
5217 tc_malloc/tc_new_nothrow. The difference with throw() seems to be just
5218 which function is called when unexpected exception happens.
5219
5220 So we work around this sillyness by simply dropping any exception
5221 specification when compiling tcmalloc.
5222
5223 M src/tcmalloc.cc
5224
5225 commit 03da6afff57c12845260997213b6ad89e420bab3
5226 Author: Aliaksey Kandratsenka <[email protected]>
5227 Date: Wed Nov 29 20:19:07 2017 +0000
5228
5229 unbreak throw declarations on operators new/delete
5230
5231 We now clearly separate PERFTOOLS_NOTHROW (used for tc_XXX functions)
5232 and throw()/noexcept (used for operators we define).
5233
5234 The former is basically "nothrow() for our callers, nothing for
5235 us". It is roughly equivalent of what glibc declares for malloc and
5236 friends. If some exception-full C++ code calls such function it
5237 doesn't have to bother setting up exception handling around such
5238 call. Notably, it is still important for those functions to _not have
5239 throw() declarations when we're building tcmalloc. Because C++ throw()
5240 requires setting up handling of unexpected exceptions thrown from
5241 under such functions which we don't want.
5242
5243 The later is necessary to have operators new/delete definitions have
5244 "correct" exception specifications to calm down compiler
5245 warnings. Particularly older clang versions warn if new/delete aren't
5246 defined with correct exception specifications. Also this commit fixes
5247 annoying gcc 7+ warning (and gnu++14 mode) that complains about
5248 throw() being deprecated.
5249
5250 M src/base/basictypes.h
5251 M src/libc_override.h
5252 M src/libc_override_gcc_and_weak.h
5253 M src/libc_override_redefine.h
5254 M src/tests/tcmalloc_unittest.cc
5255 M src/windows/get_mangled_names.cc
5256
5257 commit 89fe59c8318b2f8cec8ce00182dd24c357252c96
5258 Author: Aliaksey Kandratsenka <[email protected]>
5259 Date: Wed Nov 29 13:37:50 2017 +0000
5260
5261 Fix OOM handling in fast-path
5262
5263 Previous fast-path malloc implementation failed to arrange proper oom
5264 handling for operator new. I.e. operator new is supposed to call new
5265 handler and throw exception, which was not arranged in fast-path case.
5266
5267 Fixed code now passes pointer for oom function to
5268 ThreadCache::FetchFromCentralCache which will call it in oom
5269 condition. Test is added to verify correct behavior.
5270
5271 I've also updated some fast-path-related comments for more accuracy.
5272
5273 M src/tcmalloc.cc
5274 M src/tests/tcmalloc_unittest.cc
5275 M src/thread_cache.cc
5276 M src/thread_cache.h
5277
5278 commit a29a0cf348e131d5b8ec26c39dabeac89cf13fcd
5279 Author: Aliaksey Kandratsenka <[email protected]>
5280 Date: Tue Nov 28 14:43:11 2017 +0000
5281
5282 delete-trailing-whitespace on thread_cache.*
5283
5284 M src/thread_cache.cc
5285 M src/thread_cache.h
5286
5287 commit e6cd69bdecde074c3c7a52ccb099eafcecf668dc
5288 Author: Aliaksey Kandratsenka <[email protected]>
5289 Date: Wed Nov 29 19:49:18 2017 +0000
5290
5291 reintroduce aliasing for aligned delete
5292
5293 Without aliasing performance is likely to be at least partially
5294 affected. There is still concern that aliasing between functions of
5295 different signatures is not 100% safe. We now explicitly list of
5296 architectures where aliasing is known to be safe.
5297
5298 M src/tcmalloc.cc
5299
5300 commit fb30c3d435c4712785a50f8a422c76ad490a12b2
5301 Author: Aliaksey Kandratsenka <[email protected]>
5302 Date: Wed Nov 29 19:10:23 2017 +0000
5303
5304 fully disable aligned new on windows for now
5305
5306 M src/windows/gperftools/tcmalloc.h
5307 M src/windows/gperftools/tcmalloc.h.in
5308
5309 commit 7efb3ecf37d88edf9cf9a43efb89b425eaf81d5e
5310 Author: Andrey Semashev <[email protected]>
5311 Date: Sat Sep 30 17:47:35 2017 +0300
5312
5313 Add support for C++17 operator new/delete for overaligned types.
5314
5315 - Add auto-detection of std::align_val_t presence to configure
5316 scripts. This
5317 indicates that the compiler supports C++17 operator new/delete
5318 overloads
5319 for overaligned types.
5320
5321 - Add auto-detection of -faligned-new compiler option that appeared
5322 in gcc 7.
5323 The option allows the compiler to generate calls to the new
5324 operators. It is
5325 needed for tests.
5326
5327 - Added overrides for the new operators. The overrides are enabled
5328 if the
5329 support for std::align_val_t has been detected. The implementation
5330 is mostly
5331 based on the infrastructure used by memalign, which had to be
5332 extended to
5333 support being used by C++ operators in addition to C functions. In
5334 particular,
5335 the debug version of the library has to distinguish memory
5336 allocated by
5337 memalign from that by operator new. The current implementation
5338 of sized
5339 overaligned delete operators do not make use of the supplied
5340 size argument
5341 except for the debug allocator because it is difficult to calculate
5342 the exact
5343 allocation size that was used to allocate memory with
5344 alignment. This can be
5345 done in the future.
5346
5347 - Removed forward declaration of std::nothrow_t. This was not
5348 portable as
5349 the standard library is not required to provide nothrow_t
5350 directly in
5351 namespace std (it could use e.g. an inline namespace within
5352 std). The <new>
5353 header needs to be included for std::align_val_t anyway.
5354
5355 - Fixed operator delete[] implementation in libc_override_redefine.h.
5356
5357 - Moved TC_ALIAS definition to the beginning of the file in
5358 tcmalloc.cc so that
5359 the macro is defined before its first use in nallocx.
5360
5361 - Added tests to verify the added operators.
5362
5363 [[email protected]: fixed couple minor warnings, and some
5364 whitespace change]
5365 [[email protected]: removed addition of TC_ALIAS in debug
5366 allocator]
5367 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
5368
5369 M Makefile.am
5370 M configure.ac
5371 M src/debugallocation.cc
5372 M src/gperftools/tcmalloc.h.in
5373 M src/libc_override_gcc_and_weak.h
5374 M src/libc_override_redefine.h
5375 M src/tcmalloc.cc
5376 M src/tests/tcmalloc_unittest.cc
5377 M src/windows/gperftools/tcmalloc.h
5378 M src/windows/gperftools/tcmalloc.h.in
5379
5380 commit 7a6e25f3b143c35ed6fff8bb2c0f8de1ab348a39
5381 Author: Andrew Morrow <[email protected]>
5382 Date: Tue Nov 7 15:57:17 2017 -0500
5383
5384 Add new statistics for the PageHeap
5385
5386 [[email protected]: addressed init order mismatch warning]
5387 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
5388
5389 M src/page_heap.cc
5390 M src/page_heap.h
5391 M src/tcmalloc.cc
5392
5393 commit 6e3a702fb9c86eb450f22b326ecbceef4b0d6604
5394 Author: Jianbo Yang <[email protected]>
5395 Date: Tue Oct 17 01:09:18 2017 +0800
5396
5397 Fix data race setting size_left_ in ThreadCache::SetMaxSize
5398
5399 This commit is to fix the data race in ThreadCache::SetMaxSize.
5400 ThreadCache::size_left_ is removed and ThreadCache::size_ is
5401 added. ThreadCache::size_left_ was introduced for optimization.
5402 It is updated in several functions of ThreadCache, including the
5403 ThreadCache::SetMaxSize. But thread A can update size_left_ of
5404 thread B via SetMaxSize without protection or synchronization.
5405
5406 There should not be data race around ThreadCache::size_, for it
5407 isn't accessed by multi threads.
5408
5409 The optimization of tail-call in tc_{malloc, new, free} is kept
5410 and no other logics are affected.
5411
5412 M src/thread_cache.cc
5413 M src/thread_cache.h
5414
5415 commit 235471f96564aa95354839bb135661e561966001
5416 Author: cs-lee <[email protected]>
5417 Date: Sun Aug 6 04:57:42 2017 -0700
5418
5419 fix memory leak in Symbolize function
5420
5421 [[email protected]: reworded commit message]
5422 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
5423
5424 M src/symbolize.cc
5425
5426 commit 47efdd60f596f7180689337cbba2dab35539ba98
5427 Author: cs-lee <[email protected]>
5428 Date: Sun Aug 6 04:51:52 2017 -0700
5429
5430 Added mising va_end() in TracePrintf function
5431
5432 Normally the va_end function does not do anything,
5433 but it should be called because some platforms need it.
5434
5435 [[email protected]: reworded commit message]
5436 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
5437
5438 M src/debugallocation.cc
5439
5440 commit 497b60ef0fa93ed41ccacf6ffd0d536789401d29
5441 Author: Vladimir <[email protected]>
5442 Date: Fri Jul 14 15:13:32 2017 +0300
5443
5444 Implemented GetProgramInvocationName on FreeBSD
5445
5446 Few lines of code was taken from
5447 /usr/src/contrib/libexecinfo/backtrace.c
5448
5449 [[email protected]: updated commit message
5450 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
5451
5452 M src/symbolize.cc
5453
5454 commit ac072a3fc768b9699b1e809a379f576aa3e48983
5455 Author: Aliaksey Kandratsenka <[email protected]>
5456 Date: Sat Sep 23 14:55:33 2017 -0700
5457
5458 Revert "Ignore current_instance heap allocation when leak sanitizer
5459 is enabled"
5460
5461 This reverts commit 70a35422b5509a456584b132ad8ce4466af323ea.
5462
5463 M Makefile.am
5464 M src/malloc_extension.cc
5465
5466 commit fb5987d57913867e3b0f20490461ea55bfabe9ee
5467 Author: Aliaksey Kandratsenka <[email protected]>
5468 Date: Sat Sep 23 14:55:20 2017 -0700
5469
5470 Revert "Ensure that lsan flags are appended on all necessary targets"
5471
5472 This reverts commit a3bf61ca81b68e7792739c451aceef00cf7d7d03.
5473
5474 M Makefile.am
5475
5476 commit 5815f02105acd5d7fd0e6ec038a9e96ccc36911f
5477 Author: Aliaksey Kandratsenka <[email protected]>
5478 Date: Sat Sep 16 21:14:23 2017 -0700
5479
5480 Use safe getenv for setting up backtrace capturing method
5481
5482 This code runs very early, so using special "early" version of getenv
5483 is reasonable. It should fix issue #912.
5484
5485 M src/stacktrace.cc
5486
5487 commit aab4277311759b32ae8c442b49edf002230ea165
5488 Author: Dorin Lazăr <[email protected]>
5489 Date: Thu Sep 21 22:11:49 2017 +0300
5490
5491 Fixed LTO warning about the mismatch between return values for
5492 ProfilingIsEnabledForAllThreads()
5493
5494 M src/heap-checker.cc
5495
5496 commit d406f2285390c402e824dd28e6992f7f890dcdf9
5497 Author: Aliaksey Kandratsenka <[email protected]>
5498 Date: Sat Aug 5 22:05:10 2017 -0700
5499
5500 implement support for C11 aligned_alloc
5501
5502 Just like glibc does, we simply alias it to memalign.
5503
5504 M Makefile.am
5505 M src/libc_override_gcc_and_weak.h
5506 M src/libc_override_redefine.h
5507
5508 commit 92a27e41a137704772479277786d035763938921
5509 Author: Piotr Sikora <[email protected]>
5510 Date: Mon Aug 21 15:06:23 2017 -0700
5511
5512 Fix build on macOS.
5513
5514 Fixes #910.
5515
5516 Signed-off-by: Piotr Sikora <[email protected]>
5517
5518 M Makefile.am
5519
5520 commit e033431e5a9e02e4d1882bb6fcd212c4fab56173
5521 Author: Khem Raj <[email protected]>
5522 Date: Sun Jul 16 19:28:17 2017 -0700
5523
5524 include fcntl.h for loff_t definition
5525
5526 Fixes
5527 linux_syscall_support.h:2641:26: error: 'loff_t' has not been declared
5528
5529 Signed-off-by: Khem Raj <[email protected]>
5530
5531 M src/base/linux_syscall_support.h
5532
5533 commit e41bc414048eb4a37743e294bed3f1c63a1fa559
5534 Author: Khem Raj <[email protected]>
5535 Date: Sat Jul 1 13:21:21 2017 -0700
5536
5537 Use ucontext_t instead of struct ucontext
5538
5539 Newer glibc has dropped the ucontext tag from exposing
5540
5541 Signed-off-by: Khem Raj <[email protected]>
5542
5543 M src/stacktrace_powerpc-linux-inl.h
5544
5545 commit bf840dec0495e17f5c8403e68e10b9d6bf05c559
5546 Author: Aliaksey Kandratsenka <[email protected]>
5547 Date: Sun Jul 9 14:55:33 2017 -0700
5548
5549 bumped version to 2.6.1
5550
5551 M NEWS
5552 M configure.ac
5553 M src/windows/config.h
5554 M src/windows/gperftools/tcmalloc.h
5555
5556 commit 2d220c7e26bd279ec87f9d1fc694ca5114b20bee
5557 Author: Romain Geissler <[email protected]>
5558 Date: Sun Jun 18 15:21:19 2017 +0200
5559
5560 Replace "throw()" by "PERFTOOLS_NOTHROW"
5561
5562 Automatically done with:
5563 sed -e 's/\<throw[[:space:]]*([[:space:]]*)/PERFTOOLS_NOTHROW/g' -i
5564 $(git grep -l 'throw[[:space:]]*([[:space:]]*)')
5565
5566 [[email protected]: updated to define empty PERFTOOLS_NOTHROW
5567 only on pre-c++11 standards]
5568
5569 M src/libc_override_gcc_and_weak.h
5570 M src/libc_override_redefine.h
5571 M src/tcmalloc.cc
5572 M src/windows/get_mangled_names.cc
5573
5574 commit c4de73c0e69b9a75b6795fdd4598234baed8496d
5575 Author: Romain Geissler <[email protected]>
5576 Date: Sun Jun 18 15:16:16 2017 +0200
5577
5578 Add PERFTOOLS_THROW where necessary (as detected by GCC).
5579
5580 M src/base/basictypes.h
5581 M src/libc_override_gcc_and_weak.h
5582 M src/tests/tcmalloc_unittest.cc
5583
5584 commit e5fbd0e24e19bf3c9a8777cfc78e6188c5557025
5585 Author: Romain Geissler <[email protected]>
5586 Date: Sun Jun 18 15:05:46 2017 +0200
5587
5588 Rename PERFTOOLS_THROW into PERFTOOLS_NOTHROW.
5589
5590 Automatically done with:
5591 sed -e 's/\<PERFTOOLS_THROW\>/PERFTOOLS_NOTHROW/g' -i $(git grep
5592 -l PERFTOOLS_THROW)
5593
5594 M src/debugallocation.cc
5595 M src/gperftools/tcmalloc.h.in
5596 M src/tcmalloc.cc
5597 M src/windows/gperftools/tcmalloc.h
5598 M src/windows/gperftools/tcmalloc.h.in
5599
5600 commit eeb7b84c20146c0e2e039ce72a2ea083a94ba80d
5601 Author: Aliaksey Kandratsenka <[email protected]>
5602 Date: Sat Jul 8 15:23:37 2017 -0700
5603
5604 Register tcmalloc atfork handler as early as possible
5605
5606 This is what other mallocs do (glibc malloc and jemalloc). The idea is
5607 malloc is usually initialized very eary. So if we register atfork
5608 handler at that time, we're likely to be first. And that makes our
5609 atfork handler a bit safer, since there is much less chance of some
5610 other library installing their "take all locks" handler first and
5611 having fork take malloc lock before library's lock and deadlocking.
5612
5613 This should address issue #904.
5614
5615 M src/static_vars.cc
5616 M src/static_vars.h
5617 M src/thread_cache.cc
5618
5619 commit 208c26caeff24fd8c1defea80cb723ec1ab4bf06
5620 Author: Aliaksey Kandratsenka <[email protected]>
5621 Date: Tue Jul 4 19:50:04 2017 -0700
5622
5623 Add initial syscall support for mips64 32-bit ABI
5624
5625 This applies patch by Adhemerval Zanella from
5626 https://github.com/gperftools/gperftools/issues/845.
5627
5628 Only malloc (i.e. tcmalloc_minimal) was tested to work so far.
5629
5630 M src/base/linux_syscall_support.h
5631 M src/malloc_hook_mmap_linux.h
5632
5633 commit a3bf61ca81b68e7792739c451aceef00cf7d7d03
5634 Author: Francis Ricci <[email protected]>
5635 Date: Wed Jul 5 12:39:41 2017 -0400
5636
5637 Ensure that lsan flags are appended on all necessary targets
5638
5639 M Makefile.am
5640
5641 commit 97646a193237af0f941e5bd3078eb72edbc2d548
5642 Author: Aliaksey Kandratsenka <[email protected]>
5643 Date: Tue Jul 4 21:02:34 2017 -0700
5644
5645 Add missing NEWS entry for recent 2.6 release
5646
5647 Somehow I managed to miss this last commit in 2.6 release. So lets add
5648 it now even if it is too late.
5649
5650 M NEWS
5651
5652 commit 4be05e43a1adab9ceea9bdaaae546fb938f0a86c
5653 Author: Aliaksey Kandratsenka <[email protected]>
5654 Date: Tue Jul 4 20:35:25 2017 -0700
5655
5656 bumped version up to 2.6
5657
5658 M NEWS
5659 M configure.ac
5660 M src/windows/config.h
5661 M src/windows/gperftools/tcmalloc.h
5662
5663 commit 70a35422b5509a456584b132ad8ce4466af323ea
5664 Author: Francis Ricci <[email protected]>
5665 Date: Thu Jun 15 11:04:33 2017 -0400
5666
5667 Ignore current_instance heap allocation when leak sanitizer is enabled
5668
5669 Without this patch, any user program that enables LeakSanitizer will
5670 see a leak from tcmalloc. Add a weak hook to __lsan_ignore_object,
5671 so that if LeakSanitizer is enabled, the allocation can be ignored.
5672
5673 M Makefile.am
5674 M src/malloc_extension.cc
5675
5676 commit 6eca6c64fabbff55c43a78537e466312524b3acd
5677 Author: Aliaksey Kandratsenka <[email protected]>
5678 Date: Sat Jul 1 18:48:58 2017 -0700
5679
5680 Revert "issue-654: [pprof] handle split text segments"
5681
5682 This reverts commit 8c3dc52fcfe02412a529769a22cbc75388a5d368.
5683
5684 People have reported issues with this so lets stay safe and use older
5685 even if less powerful code.
5686
5687 M src/pprof
5688
5689 commit a495969cb6157cc361fc03228ae12ff7c0ecac4b
5690 Author: KernelMaker <[email protected]>
5691 Date: Wed May 24 16:46:49 2017 +0800
5692
5693 update the prev_class_size in each loop, or the min_object_size of
5694 tcmalloc.thread will always be 1 when calling GetFreeListSizes
5695
5696 M src/tcmalloc.cc
5697
5698 commit 163224d8af5abdbbd8208ad6d37f13fb450af46f
5699 Author: Kim Gräsman <[email protected]>
5700 Date: Fri May 26 13:09:03 2017 +0200
5701
5702 Document HEAPPROFILESIGNAL environment variable
5703
5704 M docs/heapprofile.html
5705
5706 commit 5ac82ec5b96d24219efd4c8aec47a45466eabd00
5707 Author: Aliaksey Kandratsenka <[email protected]>
5708 Date: Mon May 29 13:07:39 2017 -0700
5709
5710 added stacktrace capturing benchmark
5711
5712 M .gitignore
5713 M Makefile.am
5714 A benchmark/getcontext_light.cc
5715 A benchmark/unwind_bench.cc
5716
5717 commit c571ae2fc9433e958f29b3c3525d34c22a9cb884
5718 Author: Aliaksey Kandratsenka <[email protected]>
5719 Date: Mon May 22 19:04:20 2017 -0700
5720
5721 2.6rc4
5722
5723 M NEWS
5724 M configure.ac
5725 M src/windows/config.h
5726 M src/windows/gperftools/tcmalloc.h
5727
5728 commit f2bae51e7e609855c26095f14ffbb84082694acb
5729 Author: Aliaksey Kandratsenka <[email protected]>
5730 Date: Mon May 22 18:58:15 2017 -0700
5731
5732 Revert "Revert "disable dynamic sized delete support by default""
5733
5734 This reverts commit b82d89cb7c8781a6028f6f5959cabdc5a273aec3.
5735
5736 Dynamic sized delete support relies on ifunc handler being able to
5737 look up environment variable. The issue is, when stuff is linked with
5738 -z now linker flags, all relocations are performed early. And sadly
5739 ifunc relocations are not treated specially. So when ifunc handler
5740 runs, it cannot rely on any dynamic relocations at all, otherwise
5741 crash is real possibility. So we cannot afford doing it until (and if)
5742 ifunc is fixed.
5743
5744 This was brought to my attention by Fedora people at
5745 https://bugzilla.redhat.com/show_bug.cgi?id=1452813
5746
5747 M configure.ac
5748
5749 commit 6426c0cc8049dd50a681dc337ac9962577d5fa14
5750 Author: Aliaksey Kandratsenka <[email protected]>
5751 Date: Sun May 21 22:55:29 2017 -0700
5752
5753 2.6rc3
5754
5755 M NEWS
5756 M configure.ac
5757 M src/windows/config.h
5758 M src/windows/gperftools/tcmalloc.h
5759
5760 commit 0c0e2fe43b5095d19470abdc3b4d83613fe37c6d
5761 Author: Aliaksey Kandratsenka <[email protected]>
5762 Date: Mon May 22 02:39:29 2017 -0700
5763
5764 enable 48-bit page map on msvc as well
5765
5766 M src/common.h
5767
5768 commit 83d6818295cb337b6201fe307e15755df9dcd47f
5769 Author: Aliaksey Kandratsenka <[email protected]>
5770 Date: Mon May 22 02:38:13 2017 -0700
5771
5772 speed up 3-level page map access
5773
5774 There is no need to have pointer indirection for root node. This also
5775 helps the case of early free of garbage pointer because we didn't
5776 check root_ pointer for NULL.
5777
5778 M src/pagemap.h
5779
5780 commit f7ff175b92df5577aef9caf3fb02f9ff37b822ca
5781 Author: Aliaksey Kandratsenka <[email protected]>
5782 Date: Sun May 21 23:28:48 2017 -0700
5783
5784 add configure-time warning on unsupported backtrace capturing
5785
5786 Both libgcc and libc's backtrace() are not really options for stack
5787 trace capturing from inside profiling signal handler. So lets warn
5788 people.
5789
5790 M configure.ac
5791
5792 commit cef582350c952bad761476d01ea64bb4087371ce
5793 Author: Aliaksey Kandratsenka <[email protected]>
5794 Date: Sun May 21 23:15:59 2017 -0700
5795
5796 align fast-path functions only if compiler supports that
5797
5798 Apparently gcc only supports __attribute__((aligned(N))) on functions
5799 only since version 4.3. So lets test it in configure script and only
5800 use when possible. We now use CACHELINE_ALIGNED_FN macro for aligning
5801 functions.
5802
5803 M configure.ac
5804 M src/base/basictypes.h
5805 M src/tcmalloc.cc
5806
5807 commit bddf862b189c4508d5212f6e0e8ea81c4dd18811
5808 Author: Aliaksey Kandratsenka <[email protected]>
5809 Date: Sun May 21 22:24:19 2017 -0700
5810
5811 actually support very early freeing of NULL
5812
5813 This was caught by unit tests on centos 5. Apparently some early
5814 thingy is trying to do vprintf which calls free(0). Which used to
5815 crash since before size class cache is initialized it'll report
5816 hit (with size class 0) for NULL pointer, so we'd miss the case of
5817 checking NULL pointer free and crash.
5818
5819 The fix is to check for IsInited in the case when thread cache is
5820 null, and if so then we escalte to free_null_or_invalid.
5821
5822 M src/tcmalloc.cc
5823 M src/thread_cache.cc
5824
5825 commit 07a124d8c16bc7d52524ceb9f50d7a65b868e129
5826 Author: Aliaksey Kandratsenka <[email protected]>
5827 Date: Sun May 21 22:22:21 2017 -0700
5828
5829 don't use arg-ful constructor attribute for early nallocx test
5830
5831 101 is not very early anyways and arg-ful constructor attribute is
5832 only supported since gcc 4.3 (and e.g. rhel 5's compiler fails to
5833 compile it). So there seems to be very little value trying to ask for
5834 priority of 101.
5835
5836 M src/tests/tcmalloc_unittest.cc
5837
5838 commit 5346b8a4def33915e6f877e308fb948f570a6120
5839 Author: Aliaksey Kandratsenka <[email protected]>
5840 Date: Sun May 21 22:18:01 2017 -0700
5841
5842 don't depend on SIZE_MAX definition in sampler.cc
5843
5844 It was reported that SIZE_MAX isn't getting defined in C++ mode when
5845 C++ standard is less than c++11. Because we still want to support
5846 non-c++11 systems (for now), lets make it simple and not depend on
5847 SIZE_MAX (original google-internal code used
5848 std::numeric_limits<ssize_t>::max, but that failed to compile on
5849 msvc).
5850
5851 Fixes issue #887 and issue #889.
5852
5853 M src/sampler.cc
5854
5855 commit 50125d8f70007fb2aa4d1c87df6a3092a3371edf
5856 Author: Aliaksey Kandratsenka <[email protected]>
5857 Date: Mon May 15 00:02:43 2017 -0700
5858
5859 2.6rc2
5860
5861 M NEWS
5862 M configure.ac
5863 M src/windows/config.h
5864 M src/windows/gperftools/tcmalloc.h
5865
5866 commit a5e8e42a47fc0175c8044d0d77bf192b03347964
5867 Author: Aliaksey Kandratsenka <[email protected]>
5868 Date: Sun May 14 23:45:08 2017 -0700
5869
5870 don't link-in libunwind if libunwind.h is missing
5871
5872 I got report that some build environments for
5873 https://github.com/lyft/envoy are having link-time issue due to
5874 linking libunwind. It was happening despite libunwind.h being present,
5875 which is clear bug as without header we won't really use libunwind.
5876
5877 M configure.ac
5878
5879 commit e92acdf98db229e8dc84993fd8895d55e6c69129
5880 Author: Rajalakshmi Srinivasaraghavan <[email protected]>
5881 Date: Wed Apr 26 06:46:43 2017 -0300
5882
5883 Fix compilation error for powerpc32
5884
5885 Fix the following compilation error for powerpc32 platform when using
5886 latest glibc.
5887 error: ‘siginfo_t’ was not declared in this scope
5888
5889 M src/stacktrace_powerpc-linux-inl.h
5890
5891 commit b48403a4b065830129e238feffe022abd93af807
5892 Author: Aliaksey Kandratsenka <[email protected]>
5893 Date: Sun Apr 16 21:45:51 2017 -0700
5894
5895 2.6rc
5896
5897 M NEWS
5898 M configure.ac
5899 M src/windows/config.h
5900 M src/windows/gperftools/tcmalloc.h
5901
5902 commit 53f15325d93fbe0ba17bb3fac3da86ffd3f0f1ad
5903 Author: Aliaksey Kandratsenka <[email protected]>
5904 Date: Sun May 14 20:24:27 2017 -0700
5905
5906 fix compilation of tcmalloc_unittest.cc on older llvm-gcc
5907
5908 M src/tests/tcmalloc_unittest.cc
5909
5910 commit b1d88662cb4a38ee47aa48076745898033526f9d
5911 Author: Aliaksey Kandratsenka <[email protected]>
5912 Date: Sun Apr 2 19:27:00 2017 -0700
5913
5914 change size class to be represented by 32 bit int
5915
5916 This moves code closer to Google-internal version and provides for
5917 slightly tighter code encoding on amd64.
5918
5919 M src/common.h
5920 M src/packed-cache-inl.h
5921 M src/page_heap.cc
5922 M src/page_heap.h
5923 M src/tcmalloc.cc
5924 M src/tests/packed-cache_test.cc
5925 M src/thread_cache.cc
5926 M src/thread_cache.h
5927
5928 commit 991f47a159f0e169883f06686f13c31688fa2bf0
5929 Author: Aliaksey Kandratsenka <[email protected]>
5930 Date: Thu Feb 2 15:58:08 2017 -0800
5931
5932 change default transfer batch back to 32
5933
5934 Some tensorflow benchmarks are seeing large regression with elevated
5935 values. So lets stick to old safe default until we understand how
5936 to make
5937 larger values work for all workloads.
5938
5939 M src/common.cc
5940
5941 commit 7bc34ad1f60be2df6ca38f4bffdba4daa9aa9a7d
5942 Author: Aliaksey Kandratsenka <[email protected]>
5943 Date: Thu Feb 2 15:38:41 2017 -0800
5944
5945 support different number of size classes at runtime
5946
5947 With TCMALLOC_TRANSFER_NUM_OBJ environment variable we can change
5948 transfer batch size. And with that comes slightly different number of
5949 size classes depending on value of transfer batch size.
5950
5951 We used to have hardcoded number of size classes, so we couldn't
5952 really support any batch size setting.
5953
5954 This commit adds support for dynamic number of size classes (runtime
5955 value returned by Static::num_size_classes()).
5956
5957 M src/central_freelist.cc
5958 M src/common.cc
5959 M src/common.h
5960 M src/page_heap.cc
5961 M src/static_vars.cc
5962 M src/static_vars.h
5963 M src/tcmalloc.cc
5964 M src/thread_cache.cc
5965 M src/thread_cache.h
5966
5967 commit 4585b78c8dae9183dbf5f124c0343a9f1244ed67
5968 Author: Aliaksey Kandratsenka <[email protected]>
5969 Date: Tue Feb 21 00:27:39 2017 -0800
5970
5971 massage allocation and deallocation fast-path for performance
5972
5973 This is significant speedup of fast-path of malloc. Large part comes
5974 from avoiding expensive function prologue/epilogue. Which is achieved
5975 by making sure that tc_{malloc,new,free} etc are small functions that
5976 do only tail-calls. We keep only critical path in those functions and
5977 tail-call to slower "full" versions when we need to deal with less
5978 common case. This helps compiler generate much tidier code.
5979
5980 Fast-path readyness check is now different too. We used to have "min
5981 size for slow path" variable, which was set to non-zero value when we
5982 know that thread cache is present and ready. We now have use
5983 thread-cache pointer not equal to NULL as readyness check.
5984
5985 There is special ThreadCache::threadlocal_data_.fast_path_heap copy of
5986 that pointer that can be temporarily nulled to disable malloc fast
5987 path. This is used to enable emergency malloc.
5988
5989 There is also slight change to tracking thread cache size. Instead of
5990 tracking total size of free list, it now tracks size headroom. This
5991 allows for slightly faster deallocation fast-path check where we're
5992 checking headroom to stay above zero. This check is a bit faster than
5993 comparing with max_size_.
5994
5995 M src/linked_list.h
5996 M src/tcmalloc.cc
5997 M src/thread_cache.cc
5998 M src/thread_cache.h
5999 M src/windows/patch_functions.cc
6000
6001 commit 5964a1d9c98ea3c178435ff01f9b06e03eeda58d
6002 Author: Aliaksey Kandratsenka <[email protected]>
6003 Date: Tue Feb 21 00:42:29 2017 -0800
6004
6005 always inline a number of hot functions
6006
6007 M src/page_heap.h
6008 M src/pagemap.h
6009 M src/tcmalloc.cc
6010 M src/thread_cache.h
6011
6012 commit e419b7b9a66c39b44115b01520fb25a5100cec83
6013 Author: Aliaksey Kandratsenka <[email protected]>
6014 Date: Mon Feb 20 21:11:14 2017 -0800
6015
6016 introduce ATTRIBUTE_ALWAYS_INLINE
6017
6018 M src/base/basictypes.h
6019 M src/tcmalloc.cc
6020
6021 commit 7d588da7ec4f315ea2d02824d7e8813b0f95171d
6022 Author: Aliaksey Kandratsenka <[email protected]>
6023 Date: Sun Dec 18 18:36:57 2016 -0800
6024
6025 synchronized Sampler implementation with Google-internal version
6026
6027 This is mostly dropping FastLog2 which was never necessary for
6028 performance, and making sampler to be called always, even if sampling
6029 is disabled (this benefits more for always-sampling case of Google
6030 fork).
6031
6032 We're also getting TryRecordAllocationFast which is not used yet, but
6033 will be as part of subsequent fast-path speedup commit.
6034
6035 M src/sampler.cc
6036 M src/sampler.h
6037 M src/static_vars.cc
6038 M src/tests/sampler_test.cc
6039 M src/thread_cache.h
6040
6041 commit 27da4ade70d45312bfdf334aa8cf0d63bf78df14
6042 Author: Aliaksey Kandratsenka <[email protected]>
6043 Date: Tue Feb 21 00:41:43 2017 -0800
6044
6045 reduce size of class_to_size_ array
6046
6047 Since 32-bit int is enough and accessing smaller array will use a bit
6048 less of cache.
6049
6050 M src/common.h
6051 M src/tcmalloc.cc
6052 M src/thread_cache.cc
6053 M src/thread_cache.h
6054
6055 commit 335f09d4e43a8413642e17e5ac374f925906c4e3
6056 Author: Aliaksey Kandratsenka <[email protected]>
6057 Date: Mon Feb 20 23:49:15 2017 -0800
6058
6059 use static location for pageheap
6060
6061 Makes it a bit faster to access, since we're dropping pointer
6062 indirection.
6063
6064 M src/static_vars.cc
6065 M src/static_vars.h
6066
6067 commit 6ff332fb517734cf01e86272fe37521b7155c995
6068 Author: Aliaksey Kandratsenka <[email protected]>
6069 Date: Mon Feb 20 23:37:33 2017 -0800
6070
6071 move size classes map earlier in SizeMap
6072
6073 Since we access them more often, having at least one of them at offset
6074 0 makes pi{c,e} code a bit smaller.
6075
6076 M src/common.h
6077
6078 commit 121b1cb32e7416ecb7b0c739588f1bd916c4913a
6079 Author: Aliaksey Kandratsenka <[email protected]>
6080 Date: Mon Feb 20 22:57:43 2017 -0800
6081
6082 slightly faster size class cache
6083
6084 Lower bits of page index are still used as index into hash
6085 table. Those lower bits are zeroed, or-ed with size class and
6086 placed into hash table. So checking is just loading value from hash
6087 table, xoring with higher bits of address and checking if resultant
6088 value is lower than 128. Notably, size class 0 is not considered
6089 "invalid" anymore.
6090
6091 M src/central_freelist.cc
6092 M src/packed-cache-inl.h
6093 M src/page_heap.cc
6094 M src/page_heap.h
6095 M src/tcmalloc.cc
6096 M src/tests/packed-cache_test.cc
6097
6098 commit b57c0bad41700238f524b28483a1c6c18c0a621f
6099 Author: Aliaksey Kandratsenka <[email protected]>
6100 Date: Sun Apr 16 23:15:16 2017 -0700
6101
6102 init tcmalloc prior to replacing system alloc
6103
6104 Currently on windows, we're depending on uninitialized tcmalloc
6105 variables to detect freeing foreign malloc's chunks. This works
6106 somewhat by chance due to 0-initialized size classes cache working as
6107 cache with no values. But this is about to change, so lets do explicit
6108 initialization.
6109
6110 M src/tcmalloc.cc
6111
6112 commit 71fa9f873065e3d7c1f4ce0581d26b6498712f00
6113 Author: Aliaksey Kandratsenka <[email protected]>
6114 Date: Mon Feb 20 21:17:48 2017 -0800
6115
6116 use 2-level page map for 48-bit addresses
6117
6118 48 bits is size of x86-64 and arm64 address spaces. So using 2 levels
6119 map for them is slightly faster. We keep 3 levels for small-but-slow
6120 configuration, since 2 levels consume a bit more memory.
6121
6122 This is partial port of Google-internal commit by Sanjay
6123 Ghemawat (same idea, different implementation).
6124
6125 M src/page_heap.h
6126 M src/pagemap.h
6127
6128 commit bad70249dd5c829b4981aecdc25953800d6745c3
6129 Author: Aliaksey Kandratsenka <[email protected]>
6130 Date: Mon Feb 20 21:12:58 2017 -0800
6131
6132 use 48-bit addresses on 64-bit arms too
6133
6134 M src/common.h
6135
6136 commit 5f12147c6dbfe2cfbdc7553521fe0110073135f0
6137 Author: Aliaksey Kandratsenka <[email protected]>
6138 Date: Mon Feb 20 21:03:58 2017 -0800
6139
6140 use hidden visibility for some key global variables
6141
6142 So that our -fPIC code is faster
6143
6144 M src/base/basictypes.h
6145 M src/static_vars.h
6146
6147 commit dfd53da578137d5962fe8d588980870bb0ab8aa9
6148 Author: Aliaksey Kandratsenka <[email protected]>
6149 Date: Sun Jan 22 21:08:51 2017 -0800
6150
6151 set ENOMEM in handle_oom
6152
6153 M src/tcmalloc.cc
6154
6155 commit 14fd551072426cef317f43d587f91004d4fdae75
6156 Author: Aliaksey Kandratsenka <[email protected]>
6157 Date: Sun Jan 8 06:02:04 2017 +0300
6158
6159 avoid O(N²) in thread cache creation code
6160
6161 M src/thread_cache.cc
6162
6163 commit 507a105e849422d5ceff4348d38aaf72371a6161
6164 Author: Aliaksey Kandratsenka <[email protected]>
6165 Date: Sat Mar 12 14:25:10 2016 -0800
6166
6167 pass original size to DoSampledAllocation
6168
6169 It makes heap profiles more accurate. Google's internal malloc
6170 is doing
6171 it as well.
6172
6173 M src/tcmalloc.cc
6174
6175 commit bb77979dea796ab743e1308af25e9259ec97f2b1
6176 Author: Aliaksey Kandratsenka <[email protected]>
6177 Date: Sat Mar 12 18:10:19 2016 -0800
6178
6179 don't declare throw() on malloc funtions since it is faster
6180
6181 Apparently throw() on functions actually asks compiler to generate
6182 code
6183 to detect unexpected exceptions. Which prevents tail calls
6184 optimization.
6185
6186 So in order to re-enable this optimization, we simply don't tell
6187 compiler about throw() at all. C++11 noexcept would be even better,
6188 but
6189 it is not universally available yet.
6190
6191 So we change to no exception specifications. Which at least for gcc &
6192 clang on Linux (and likely for all ELF platforms, if not just all)
6193 really eliminates all overhead of exceptions.
6194
6195 M src/debugallocation.cc
6196 M src/gperftools/tcmalloc.h.in
6197 M src/tcmalloc.cc
6198 M src/windows/gperftools/tcmalloc.h
6199 M src/windows/gperftools/tcmalloc.h.in
6200
6201 commit 89c74cb79ca41cd75a1f9131af4ea2ab362593ae
6202 Author: Aliaksey Kandratsenka <[email protected]>
6203 Date: Sat Mar 12 20:26:40 2016 -0800
6204
6205 handle duplicate google_malloc frames in malloc hook stack trace
6206
6207 Subsequent optimization may cause multiple malloc functions in
6208 google_malloc section to be in call stack. Particularly when fast-path
6209 malloc function calls slow-path and compiler chooses to implement such
6210 call as regular call instead of tail-call.
6211
6212 Because we need stacktrace just until first such function, once
6213 we find
6214 innermost such frame, we're simply checking if next outer frame
6215 is also
6216 google_malloc and consider it instead.
6217
6218 M src/malloc_hook.cc
6219
6220 commit 0feb1109ac957144a50d964b0ff244a6a286174a
6221 Author: Aliaksey Kandratsenka <[email protected]>
6222 Date: Sat Mar 12 21:08:57 2016 -0800
6223
6224 fix stack trace capturing in debug malloc
6225
6226 Particularly, hardcoded skip count was relying on certain behavior of
6227 compiler. Namely, that tail calls inside DebugDeallocate path are not
6228 actually implemented as tail calls.
6229
6230 New implementation is using google_malloc section as a marker
6231 of malloc
6232 boundary. But in order for this to work, we have to prevent
6233 tail-call in
6234 debugallocation's tc_XXX functions. Which is achieved by doing
6235 volatile
6236 read of static variable at the end of such functions.
6237
6238 M src/debugallocation.cc
6239
6240 commit 0506e965ee5184490fbe7acea470458562128a79
6241 Author: Aliaksey Kandratsenka <[email protected]>
6242 Date: Sun Dec 18 18:36:00 2016 -0800
6243
6244 replace LIKELY/UNLIKELY with PREDICT_{TRUE,FALSE}
6245
6246 Google-internal code is using PREDICT_TRUE/FALSE, so we should be
6247 doing it too.
6248
6249 M src/base/basictypes.h
6250 M src/common.cc
6251 M src/common.h
6252 M src/emergency_malloc.h
6253 M src/malloc_hook-inl.h
6254 M src/tcmalloc.cc
6255 M src/thread_cache.h
6256
6257 commit 59a4987054021eaf24143a27cf01252fafb5052d
6258 Author: Aliaksey Kandratsenka <[email protected]>
6259 Date: Sun Mar 13 16:04:43 2016 -0700
6260
6261 prevent inlining ATTRIBUTE_SECTION functions
6262
6263 So that their code is always executing in prescribed section.
6264
6265 M src/base/basictypes.h
6266
6267 commit ebb575b8a0153a70f43fd1c3ec33fe30f94c97d0
6268 Author: Aliaksey Kandratsenka <[email protected]>
6269 Date: Thu Feb 2 15:41:13 2017 -0800
6270
6271 Revert "enabled aggressive decommit by default"
6272
6273 This reverts commit 7da5bd014d77ddaf694054b1e3ae0a3ef92ab384.
6274
6275 Some tensorflow benchmarks are getting slower with aggressive
6276 decommit.
6277
6278 M src/static_vars.cc
6279 M src/tests/tcmalloc_unittest.sh
6280
6281 commit b82d89cb7c8781a6028f6f5959cabdc5a273aec3
6282 Author: Aliaksey Kandratsenka <[email protected]>
6283 Date: Sun Jan 22 13:15:39 2017 -0800
6284
6285 Revert "disable dynamic sized delete support by default"
6286
6287 This reverts commit 06811b3ae4c5365880898148d188ae91d20f52f2.
6288
6289 M configure.ac
6290
6291 commit fac0bb44d5661946d9839496ddf104942dd00abe
6292 Author: Aliaksey Kandratsenka <[email protected]>
6293 Date: Sun Jan 22 13:11:05 2017 -0800
6294
6295 Do not depend on memchr in commandlineflags::StringToBool
6296
6297 It lets us use it from inside ifunc handler.
6298
6299 M src/base/commandlineflags.h
6300
6301 commit 7d49f015a01e35155517332a2dfd95f16b568939
6302 Author: Aliaksey Kandratsenka <[email protected]>
6303 Date: Sun Jan 22 13:08:34 2017 -0800
6304
6305 Make GetenvBeforeMain work inside ifunc handler
6306
6307 Depending on things such us strlen, which are ifunc-ed themselves is
6308 not supported from inside ifunc handler.
6309
6310 Thus we implement strlen, memchr and memcmp in slow naive C++ just for
6311 that place.
6312
6313 M src/base/sysinfo.cc
6314
6315 commit a2550b6309153c250f67255b6271ec988cc7d0fb
6316 Author: Aliaksey Kandratsenka <[email protected]>
6317 Date: Sun Feb 26 23:46:18 2017 -0800
6318
6319 turn bench_fastpath_throughput into actual throughput benchmark
6320
6321 Previously we bumped size by 16 between iterations, but for many size
6322 classess that gave is subsequent iteration into same size
6323 class. Multiplying by prime number randomizes sizes more so speeds up
6324 this benchmark on at least modern x86.
6325
6326 M benchmark/malloc_bench.cc
6327
6328 commit b762b1a4923f769487de818175655c1683d84db5
6329 Author: Aliaksey Kandratsenka <[email protected]>
6330 Date: Sat Mar 12 14:12:02 2016 -0800
6331
6332 added sized free benchmarks to malloc_bench
6333
6334 M benchmark/malloc_bench.cc
6335
6336 commit 71ffc1cd6b3ea99c97ed6d0e16caeb9d1b20eec7
6337 Author: Aliaksey Kandratsenka <[email protected]>
6338 Date: Sat Mar 12 14:11:40 2016 -0800
6339
6340 added free lists randomization step to malloc_bench
6341
6342 M benchmark/malloc_bench.cc
6343
6344 commit 732dfeb83d726de0aea47f0e6aa2abff3e0ad680
6345 Author: Aliaksey Kandratsenka <[email protected]>
6346 Date: Sun Jan 22 13:57:31 2017 -0800
6347
6348 Run StartStopNoOptionsEmpty profiledata unittest
6349
6350 Somehow we had this test, but didn't arrange to actually run it. GCC
6351 gave warning and now we do.
6352
6353 M src/tests/profiledata_unittest.cc
6354
6355 commit cbb312fbe8022378c4635b3075a80a7827555170
6356 Author: Aliaksey Kandratsenka <[email protected]>
6357 Date: Sun Dec 18 11:08:54 2016 -0800
6358
6359 aggressive decommit: only free necessary regions and fix O(N²)
6360
6361 We used to decommit (with MADV_FREE or MADV_DONTNEED) whole combined
6362 span when freeing span in aggressive decommit mode. The issue with
6363 that is preceding or following span with which we combined span we're
6364 freeing could be freed already and fairly large. By passing all of
6365 that memory to decommit, we force kernel to scan all of those pages.
6366
6367 When mass-freeing many objects old behavior led to O(N^2) behavior
6368 since freeing single span could lead to MADV_{FREE,DONTNEED}-ing of
6369 most of the heap.
6370
6371 New implementation just does decommit of individual bits as needed.
6372
6373 While there, I'm also adding locking to calls of
6374 PageHeap::{Get,Set}AggressiveDecommit.
6375
6376 This partially (or mostly) fixes issue #839.
6377
6378 M src/page_heap.cc
6379 M src/page_heap.h
6380 M src/tcmalloc.cc
6381
6382 commit 6d98223a9035ec1c5e9a7eb0196b21021d48a41f
6383 Author: Aliaksey Kandratsenka <[email protected]>
6384 Date: Sun Feb 12 17:12:44 2017 -0800
6385
6386 don't build with -fno-exceptions
6387
6388 It looks like, in past it could produce better code. But since
6389 unwinding is totally different since almost forever now, there is no
6390 perfomance benefit of it anymore.
6391
6392 M Makefile.am
6393
6394 commit d6a1931cce303ea279b744c9b438686e4f5994e5
6395 Author: Aliaksey Kandratsenka <[email protected]>
6396 Date: Sat Mar 12 15:33:10 2016 -0800
6397
6398 fixed warning in casting heap of checker's main_thread_counter
6399
6400 I.e. compiler didn't like int to void * casting and upcasting to
6401 intptr_t first (which has same size as void *) makes it happier.
6402
6403 M src/heap-checker.cc
6404
6405 commit 5c778701d98c642ecff97d044cd1c0b8b862aba3
6406 Author: Aliaksey Kandratsenka <[email protected]>
6407 Date: Sun Feb 12 17:54:59 2017 -0800
6408
6409 added tcmalloc minimal unittest with ASSERTs checked
6410
6411 M .gitignore
6412 M Makefile.am
6413
6414 commit a9167617abbcebe9d472141c53a20ce00c8c8d75
6415 Author: Aliaksey Kandratsenka <[email protected]>
6416 Date: Sat Apr 29 19:09:47 2017 -0700
6417
6418 drop unused g_load_map variable in patch_functionc.cc
6419
6420 M src/windows/patch_functions.cc
6421
6422 commit d52e56dcb5aad0521ccc67d7e57dfcb8d41ea568
6423 Author: Aliaksey Kandratsenka <[email protected]>
6424 Date: Sat Apr 29 19:06:55 2017 -0700
6425
6426 don't compare integer to NULL
6427
6428 M src/windows/preamble_patcher.cc
6429
6430 commit bae00c0341f49ade473d24c117226552a59d3e75
6431 Author: Aliaksey Kandratsenka <[email protected]>
6432 Date: Sun Apr 23 20:33:16 2017 -0700
6433
6434 add fake_stacktrace_scope to few msvc projects
6435
6436 Otherwise it predictably fails at linking phase due to missing
6437 symbols.
6438
6439 M
6440 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
6441 M vsprojects/tmu-static/tmu-static.vcproj
6442
6443 commit 79aab4fed40f9aac03dacb32a108861eff689beb
6444 Author: Aliaksey Kandratsenka <[email protected]>
6445 Date: Sun Apr 16 22:17:56 2017 -0700
6446
6447 correctly dllexport nallocx on windows
6448
6449 M src/gperftools/nallocx.h
6450 M src/tcmalloc.cc
6451
6452 commit b010895a088703f512cb0e225c838306b2904069
6453 Author: Aliaksey Kandratsenka <[email protected]>
6454 Date: Sun Apr 30 18:32:10 2017 -0700
6455
6456 don't undef PERFTOOLS_DLL_DECL
6457
6458 This is not necessary and will help adding more headers with
6459 dll-exported functions.
6460
6461 M src/gperftools/tcmalloc.h.in
6462 M src/windows/gperftools/tcmalloc.h
6463 M src/windows/gperftools/tcmalloc.h.in
6464
6465 commit 491b1aca7edbbaed9ea6f7baaa547a35cf385130
6466 Author: Aliaksey Kandratsenka <[email protected]>
6467 Date: Sat Apr 29 16:46:21 2017 -0700
6468
6469 don't try to use pthread_atfork on windows
6470
6471 M src/maybe_threads.cc
6472 M src/windows/mingw.h
6473
6474 commit 691045b95761bedb9170d758be82f9b2f15c673a
6475 Author: Aliaksey Kandratsenka <[email protected]>
6476 Date: Sat Mar 12 14:41:28 2016 -0800
6477
6478 suppress warnings from legacy headers while building legacy headers
6479 test
6480
6481 M src/google/heap-checker.h
6482 M src/google/heap-profiler.h
6483 M src/google/malloc_extension.h
6484 M src/google/malloc_extension_c.h
6485 M src/google/malloc_hook.h
6486 M src/google/malloc_hook_c.h
6487 M src/google/profiler.h
6488 M src/google/stacktrace.h
6489 M src/google/tcmalloc.h
6490 M src/tests/simple_compat_test.cc
6491
6492 commit 22f7ceb97a63bcbc9d824a80916bb7a809fad1f2
6493 Author: Aliaksey Kandratsenka <[email protected]>
6494 Date: Sat Apr 29 15:25:06 2017 -0700
6495
6496 use unsigned for few flags in mini_disassembler_types.h
6497
6498 We're initializing them with value that doesn't fit into signed 32-bit
6499 int and somehow mingw version that I have not just warns, but actually
6500 errors on that.
6501
6502 M src/windows/mini_disassembler_types.h
6503
6504 commit 9b17a8a5ba1c8320d12c6e6df0ab54ff218b50e6
6505 Author: Mostyn Bramley-Moore <[email protected]>
6506 Date: Thu May 11 20:40:18 2017 +0200
6507
6508 remove superfluous size_t value >= 0 check
6509
6510 M src/debugallocation.cc
6511
6512 commit 86ce69d77ff5f881c5701901bf7d1ef884f9e33a
6513 Author: Eugene <[email protected]>
6514 Date: Sat Feb 4 17:57:34 2017 +0300
6515
6516 Update binary_trees.cc
6517
6518 M benchmark/binary_trees.cc
6519
6520 commit cd8586ed6c8bed550ad727ce8af83a1a6bd1ca8e
6521 Author: Ola Olsson <[email protected]>
6522 Date: Tue Mar 21 14:07:16 2017 +0100
6523
6524 Fix path names in README
6525
6526 M README
6527
6528 commit 98753aa73772bb0fae8cee668e513d5754ab7737
6529 Author: Aliaksey Kandratsenka <[email protected]>
6530 Date: Mon Feb 20 16:04:14 2017 -0800
6531
6532 test that sized deallocation really works before enabling it
6533
6534 Weirdly, some OSX compiler does allow -fsized-deallocation, yet barks
6535 on any attempt to actually use it at compile time (!). So lets detect
6536 this as well in configure and opt out as necessary.
6537
6538 Should fix issue #870.
6539
6540 M configure.ac
6541
6542 commit 5618ef78505a854bd0d2c145fd9cf7967226a20e
6543 Author: Aliaksey Kandratsenka <[email protected]>
6544 Date: Mon Feb 20 15:31:41 2017 -0800
6545
6546 Don't assume memalign exists in memalign vs nallocx test
6547
6548 OSX and windows have issues with memalign. So test against tc_memalign
6549 instead.
6550
6551 This should fix _memalign linker-time part of issue #870.
6552
6553 M src/tests/tcmalloc_unittest.cc
6554
6555 commit bf640cd740fe38f6f10faa8683e8361fee971aba
6556 Author: Aliaksey Kandratsenka <[email protected]>
6557 Date: Mon Feb 20 14:51:27 2017 -0800
6558
6559 rename sys allocator's sys_alloc symbol to tcmalloc_sys_alloc
6560
6561 Since we're not building with hidden visibility, symbols of other DSOs
6562 can interpose our symbols. And sys_alloc is just too generic name.
6563
6564 And in fact erlang runtime has sys_alloc function. Which means we're
6565 trying to write to it's first bytes as part of initializing system
6566 allocator and crash.
6567
6568 This should fix issue #843.
6569
6570 M src/system-alloc.cc
6571 M src/system-alloc.h
6572 M src/tcmalloc.cc
6573 M src/windows/system-alloc.cc
6574
6575 commit 069e3b16551186f8da0a41c185550fdfd11af35b
6576 Author: Aliaksey Kandratsenka <[email protected]>
6577 Date: Mon Feb 20 14:25:18 2017 -0800
6578
6579 build malloc_bench_shared_full only when full tcmalloc is built
6580
6581 I.e. because otherwise, when --enable-minimal is given, we're building
6582 empty libtcmalloc.la and linking it to malloc_bench_shared_full. Which
6583 has no effect at all and actually breaks builds on OSX.
6584
6585 Should fix issue #869.
6586
6587 M Makefile.am
6588
6589 commit b8f9d0d44f94177d34b069180618b7d002e85b69
6590 Author: Aliaksey Kandratsenka <[email protected]>
6591 Date: Sun Dec 18 09:35:02 2016 -0800
6592
6593 ported nallocx support from Google-internal tcmalloc
6594
6595 nallocx is extension introduced by jemalloc. It returns effective size
6596 of allocaiton without allocating anything.
6597
6598 We also support MALLOCX_LG_ALIGN flag. But all other jemalloc
6599 flags (which at the moment do nothing for nallocx anyways) are
6600 silently ignored, since there is no sensible way to return errors in
6601 this API.
6602
6603 This was originally contributed by Dmitry Vyukov with input from
6604 Andrew Hunter. But due to significant divergence of Google-internal
6605 and free-software forks of tcmalloc, significant massaging was done by
6606 me. So all bugs are mine.
6607
6608 M Makefile.am
6609 A src/gperftools/nallocx.h
6610 M src/tcmalloc.cc
6611 M src/tests/tcmalloc_unittest.cc
6612
6613 commit b0abefd93834d9d9c7ffaae2d23bd48ed8e96477
6614 Author: Andrew Morrow <[email protected]>
6615 Date: Sat Dec 17 13:57:41 2016 -0500
6616
6617 Fix a typo in the page fence flag declaration
6618
6619 M src/debugallocation.cc
6620
6621 commit 855b3800064db49af823b85a54be269923eb6f4d
6622 Author: Kirill Müller <[email protected]>
6623 Date: Tue Nov 15 09:58:11 2016 +0100
6624
6625 replace docs by doc
6626
6627 M Makefile.am
6628 M README
6629 D doc
6630 M packages/deb/docs
6631 M packages/deb/libgperftools0.manpages
6632 M src/gperftools/heap-checker.h
6633 M src/gperftools/heap-profiler.h
6634 M src/gperftools/profiler.h
6635 M src/profiledata.h
6636 M src/tcmalloc.cc
6637
6638 commit 664210ead806d700cdbe5eeaf75d7a066fdac541
6639 Author: Kirill Müller <[email protected]>
6640 Date: Tue Nov 15 09:42:12 2016 +0100
6641
6642 doc -> docs, with symlink
6643
6644 A doc
6645 R100 doc/cpuprofile-fileformat.html docs/cpuprofile-fileformat.html
6646 R100 doc/cpuprofile.html docs/cpuprofile.html
6647 R100 doc/designstyle.css docs/designstyle.css
6648 R100 doc/heap-example1.png docs/heap-example1.png
6649 R100 doc/heap_checker.html docs/heap_checker.html
6650 R100 doc/heapprofile.html docs/heapprofile.html
6651 R100 doc/index.html docs/index.html
6652 R100 doc/overview.dot docs/overview.dot
6653 R100 doc/overview.gif docs/overview.gif
6654 R100 doc/pageheap.dot docs/pageheap.dot
6655 R100 doc/pageheap.gif docs/pageheap.gif
6656 R100 doc/pprof-test-big.gif docs/pprof-test-big.gif
6657 R100 doc/pprof-test.gif docs/pprof-test.gif
6658 R100 doc/pprof-vsnprintf-big.gif docs/pprof-vsnprintf-big.gif
6659 R100 doc/pprof-vsnprintf.gif docs/pprof-vsnprintf.gif
6660 R100 doc/pprof.1 docs/pprof.1
6661 R100 doc/pprof.see_also docs/pprof.see_also
6662 R100 doc/pprof_remote_servers.html docs/pprof_remote_servers.html
6663 R100 doc/spanmap.dot docs/spanmap.dot
6664 R100 doc/spanmap.gif docs/spanmap.gif
6665 R100 doc/t-test1.times.txt docs/t-test1.times.txt
6666 R100 doc/tcmalloc-opspercpusec.vs.threads.1024.bytes.png
6667 docs/tcmalloc-opspercpusec.vs.threads.1024.bytes.png
6668 R100 doc/tcmalloc-opspercpusec.vs.threads.128.bytes.png
6669 docs/tcmalloc-opspercpusec.vs.threads.128.bytes.png
6670 R100 doc/tcmalloc-opspercpusec.vs.threads.131072.bytes.png
6671 docs/tcmalloc-opspercpusec.vs.threads.131072.bytes.png
6672 R100 doc/tcmalloc-opspercpusec.vs.threads.16384.bytes.png
6673 docs/tcmalloc-opspercpusec.vs.threads.16384.bytes.png
6674 R100 doc/tcmalloc-opspercpusec.vs.threads.2048.bytes.png
6675 docs/tcmalloc-opspercpusec.vs.threads.2048.bytes.png
6676 R100 doc/tcmalloc-opspercpusec.vs.threads.256.bytes.png
6677 docs/tcmalloc-opspercpusec.vs.threads.256.bytes.png
6678 R100 doc/tcmalloc-opspercpusec.vs.threads.32768.bytes.png
6679 docs/tcmalloc-opspercpusec.vs.threads.32768.bytes.png
6680 R100 doc/tcmalloc-opspercpusec.vs.threads.4096.bytes.png
6681 docs/tcmalloc-opspercpusec.vs.threads.4096.bytes.png
6682 R100 doc/tcmalloc-opspercpusec.vs.threads.512.bytes.png
6683 docs/tcmalloc-opspercpusec.vs.threads.512.bytes.png
6684 R100 doc/tcmalloc-opspercpusec.vs.threads.64.bytes.png
6685 docs/tcmalloc-opspercpusec.vs.threads.64.bytes.png
6686 R100 doc/tcmalloc-opspercpusec.vs.threads.65536.bytes.png
6687 docs/tcmalloc-opspercpusec.vs.threads.65536.bytes.png
6688 R100 doc/tcmalloc-opspercpusec.vs.threads.8192.bytes.png
6689 docs/tcmalloc-opspercpusec.vs.threads.8192.bytes.png
6690 R100 doc/tcmalloc-opspersec.vs.size.1.threads.png
6691 docs/tcmalloc-opspersec.vs.size.1.threads.png
6692 R100 doc/tcmalloc-opspersec.vs.size.12.threads.png
6693 docs/tcmalloc-opspersec.vs.size.12.threads.png
6694 R100 doc/tcmalloc-opspersec.vs.size.16.threads.png
6695 docs/tcmalloc-opspersec.vs.size.16.threads.png
6696 R100 doc/tcmalloc-opspersec.vs.size.2.threads.png
6697 docs/tcmalloc-opspersec.vs.size.2.threads.png
6698 R100 doc/tcmalloc-opspersec.vs.size.20.threads.png
6699 docs/tcmalloc-opspersec.vs.size.20.threads.png
6700 R100 doc/tcmalloc-opspersec.vs.size.3.threads.png
6701 docs/tcmalloc-opspersec.vs.size.3.threads.png
6702 R100 doc/tcmalloc-opspersec.vs.size.4.threads.png
6703 docs/tcmalloc-opspersec.vs.size.4.threads.png
6704 R100 doc/tcmalloc-opspersec.vs.size.5.threads.png
6705 docs/tcmalloc-opspersec.vs.size.5.threads.png
6706 R100 doc/tcmalloc-opspersec.vs.size.8.threads.png
6707 docs/tcmalloc-opspersec.vs.size.8.threads.png
6708 R100 doc/tcmalloc.html docs/tcmalloc.html
6709 R100 doc/threadheap.dot docs/threadheap.dot
6710 R100 doc/threadheap.gif docs/threadheap.gif
6711
6712 commit 75dc9a6e1470fa82b828f9687edad48f53d740b1
6713 Author: zmertens <[email protected]>
6714 Date: Fri Nov 18 15:27:32 2016 -0700
6715
6716 Fix Post(s)cript tyos
6717
6718 M doc/pprof.1
6719 M src/pprof
6720
6721 commit dde32f8bbc95312379f9f5a651799815bb6327c5
6722 Author: Francis Ricci <[email protected]>
6723 Date: Tue Oct 11 11:14:06 2016 -0700
6724
6725 Fix unaligned memory accesses in debug allocator
6726
6727 M src/base/basictypes.h
6728 M src/debugallocation.cc
6729
6730 commit 02eeed29df112728564a5dde6417fa4622b57a06
6731 Author: Ryan Macnak <[email protected]>
6732 Date: Tue Sep 27 14:03:05 2016 -0700
6733
6734 Fix redefinition of mmap on aarch64.
6735
6736 M src/base/linux_syscall_support.h
6737
6738 commit c07a15cff4b904ac45e4019f8e36eeffd3e8186c
6739 Author: Aliaksey Kandratsenka <[email protected]>
6740 Date: Sat Sep 24 18:56:22 2016 -0700
6741
6742 [windows] patch _free_base as well
6743
6744 Looks like _free_base is used as generic free/delete, which is not
6745 checked for free/delete mismatch by their debug runtime. Issue #817
6746 occurs because something that is allocated by regular allocation
6747 functions (new or malloc) is freed by _free_base. Patch it as
6748 well fixes
6749 the problem.
6750
6751 Closes #817.
6752
6753 M src/windows/patch_functions.cc
6754
6755 commit acac6af26b0ef052b39f61a59507b23e9703bdfa
6756 Author: Aliaksey Kandratsenka <[email protected]>
6757 Date: Sat Sep 24 16:19:27 2016 -0700
6758
6759 Fix finding default zone on macOS sierra
6760
6761 This is taken from jemalloc patch at
6762 https://github.com/jemalloc/jemalloc/pull/427/commits/19c9a3e828ed46f1576521c264640e60bd0cb01f
6763 by Mike Hommey (aka glandium).
6764
6765 The issue was reported together with advise to take that jemalloc
6766 fix by
6767 Koichi Shiraishi and David Ribeiro Alves at gperftools issue #827.
6768
6769 Closes issue #827.
6770
6771 M src/libc_override_osx.h
6772
6773 commit 7822b5b0b9fa7e016e1f6b46ea86f26f4691a457
6774 Author: Tulio Magno Quites Machado Filho <[email protected]>
6775 Date: Mon Aug 8 13:50:10 2016 -0300
6776
6777 Stop using glibc malloc hooks
6778
6779 glibc deprecated malloc hooks in 2011 and will be removing them soon.
6780 These hooks aren't safe and aren't necessary when the malloc API is
6781 fully exported.
6782
6783 M configure.ac
6784 M src/libc_override_glibc.h
6785
6786 commit c92f0ed0891d69888cdc7e904c82048362b2dd8d
6787 Author: Tulio Magno Quites Machado Filho <[email protected]>
6788 Date: Wed Jul 27 17:15:52 2016 -0300
6789
6790 Remove references to __malloc_initialize_hook
6791
6792 __malloc_initialize_hook has been deprecated in glibc since 2011
6793 and is
6794 being removed on glibc 2.24.
6795
6796 M src/libc_override_glibc.h
6797
6798 commit 9709eef361aa8c46d6b14c08bebead7b8185e731
6799 Merge: eb474c99 44f276e1
6800 Author: Aliaksey Kandratsenka (aka Aliaksei Kandratsenka)
6801 <[email protected]>
6802 Date: Wed Aug 24 22:33:05 2016 -0700
6803
6804 Merge pull request #821 from jtmcdole/patch-1
6805
6806 Rename TCMALLOC_DEBUG to PERFTOOLS_VERBOSE
6807
6808 commit 44f276e132eef8ba013a04b4af92ce26464e2b54
6809 Author: John McDole <[email protected]>
6810 Date: Wed Aug 24 17:47:28 2016 -0700
6811
6812 Rename TCMALLOC_DEBUG to PERFTOOLS_VERBOSE
6813
6814 M README
6815
6816 commit eb474c995eadef3edb47bc081f09693228832794
6817 Author: Aliaksey Kandratsenka <[email protected]>
6818 Date: Sun Jul 3 18:20:32 2016 -0700
6819
6820 Summary: support gcc atomic ops on clang too
6821
6822 Clang actually does support __atomic_XXX atomic ops builtins but does
6823 not advertise itselt as gcc 4.7 or later. So we now detect clang
6824 separetely.
6825
6826 We're enabling gcc atomic ops on clang >= 3.4 since this is the oldest
6827 version that I can test.
6828
6829 This should fix issue #797.
6830
6831 M src/base/atomicops.h
6832
6833 commit 7f86eab1f3e0636a79a96fd44e3bb51db0457a86
6834 Author: Kenton Varda <[email protected]>
6835 Date: Thu Jun 30 18:32:39 2016 -0700
6836
6837 Recognize .node files as shared libraries
6838
6839 node.js extensions are regular shared libraries named with the `.node`
6840 extension. With this small change I was able to get a usable heap
6841 profile out of node to debug an extension I'm working on.
6842
6843 M src/pprof
6844
6845 commit bf8eacce69df877d1cecdab8c7a6cc3f218ebcc2
6846 Author: Bryan Chan <[email protected]>
6847 Date: Wed Jun 15 00:10:45 2016 -0400
6848
6849 Add support for 31-bit s390; merge linux_syscall_support.h changes
6850 from upstream.
6851
6852 M configure.ac
6853 M m4/pc_from_ucontext.m4
6854 M src/base/basictypes.h
6855 M src/base/linux_syscall_support.h
6856 M src/base/linuxthreads.h
6857 M src/getpc.h
6858 M src/malloc_hook_mmap_linux.h
6859
6860 commit c54218069b96e5a64e81186cd14107bf6660b61d
6861 Author: savefromgoogle <[email protected]>
6862 Date: Sat May 14 17:50:34 2016 +0200
6863
6864 Update README
6865
6866 typo corrected and direct link to Google Group mailinglist added
6867
6868 M README
6869
6870 commit 06f4ce65c226e2bed432119bd830a478012ea336
6871 Author: iivlev <[email protected]>
6872 Date: Fri Apr 15 17:46:25 2016 +0300
6873
6874 Small performance tweak: avoid calling time() if we don't need it
6875
6876 M src/heap-profiler.cc
6877
6878 commit db8d4836091b93100b176e8cab4e842974234e4e
6879 Author: Aliaksey Kandratsenka <[email protected]>
6880 Date: Sat Jun 25 16:24:40 2016 -0700
6881
6882 Autogenerate ChangeLog from git on make dist
6883
6884 This fixes build breakage introduced in preceding commit for issue
6885 #796.
6886
6887 A ChangeLog
6888 M Makefile.am
6889 M configure.ac
6890 A m4/ax_generate_changelog.m4
6891
6892 commit 4a1359831938b5a71b2faf38c02003053af253d9
6893 Author: Aliaksey Kandratsenka <[email protected]>
6894 Date: Sat Jun 25 16:12:48 2016 -0700
6895
6896 renamed ChangeLog to ChangeLog.old
6897
6898 I.e. to avoid confusion. This should fix issue #796.
6899
6900 R100 ChangeLog ChangeLog.old
6901
6902 commit 7852eeb75b9375cf52a7da01be044da6e915dd08
6903 Author: Aliaksey Kandratsenka <[email protected]>
6904 Date: Sat Apr 9 13:09:18 2016 -0700
6905
6906 Use initial-exec tls for libunwind's recursion flag
6907
6908 If we don't do it, then reading variable calls to __tls_get_addr,
6909 which
6910 uses malloc on first call. initial-exec makes dynamic linker
6911 reserve tls
6912 offset for recusion flag early and thus avoid unsafe calls to malloc.
6913
6914 This fixes issue #786.
6915
6916 M src/base/basictypes.h
6917 M src/stacktrace_libunwind-inl.h
6918 M src/thread_cache.h
6919
6920 commit a07f9fe75af25ce388af3d4ff4514b42513d766f
6921 Author: Mike Gaffney <[email protected]>
6922 Date: Thu Mar 31 11:27:21 2016 -0700
6923
6924 gerftools -> gperftools in readme
6925
6926 M README
6927
6928 commit 9fd6d2687914a1f58a8ce457d6a1bd3d55ea0747
6929 Author: Aliaksey Kandratsenka <[email protected]>
6930 Date: Sun Mar 20 12:29:40 2016 -0700
6931
6932 added define to enable MADV_FREE usage on Linux
6933
6934 Building with -DTCMALLOC_USE_MADV_FREE will enable usage of
6935 MADV_FREE on
6936 Linux if glibc copy of kernel headers has MADV_FREE defined.
6937
6938 I.e. so that people can test this more easily.
6939
6940 Affects ticket #780.
6941
6942 M src/system-alloc.cc
6943
6944 commit 6f7a14f45e2e2e16fbbd250543f5758078e2f5f0
6945 Author: Mathias Stearn <[email protected]>
6946 Date: Wed Mar 16 18:22:32 2016 -0400
6947
6948 Don't use MADV_FREE on Linux
6949
6950 Addresses #780.
6951
6952 M src/system-alloc.cc
6953
6954 commit 55cf6e6281f2f19ed04a6b2a90e2dcaec08e16d0
6955 Author: Aman Gupta <[email protected]>
6956 Date: Mon Mar 14 12:20:40 2016 -0700
6957
6958 Fix symbol resolution on OSX
6959
6960 The library mapping ranges on OSX and Linux are sorted in opposite
6961 orders
6962
6963 7f71c3323000-7f71c3339000 r-xp 00000000 09:02 29099128
6964 /lib/x86_64-linux-gnu/libz.so.1.2.3.4
6965 7f71c3339000-7f71c3538000 ---p 00016000 09:02 29099128
6966 /lib/x86_64-linux-gnu/libz.so.1.2.3.4
6967 7f71c3538000-7f71c3539000 r--p 00015000 09:02 29099128
6968 /lib/x86_64-linux-gnu/libz.so.1.2.3.4
6969 7f71c3539000-7f71c353a000 rw-p 00016000 09:02 29099128
6970 /lib/x86_64-linux-gnu/libz.so.1.2.3.4
6971
6972 vs
6973
6974 108f8d000-108f95000 r-xp 00025000 00:00 0
6975 /usr/local/opt/libmemcached/lib/libmemcached.11.dylib
6976 108f8c000-108f8d000 r-xp 00024000 00:00 0
6977 /usr/local/opt/libmemcached/lib/libmemcached.11.dylib
6978 108f68000-108f8c000 r-xp 00000000 00:00 0
6979 /usr/local/opt/libmemcached/lib/libmemcached.11.dylib
6980
6981 M src/pprof
6982
6983 commit 8e858436223db7729d41304ad29a29946d47382b
6984 Author: Aliaksey Kandratsenka <[email protected]>
6985 Date: Sun Mar 13 15:11:50 2016 -0700
6986
6987 added simple .travis.yml config
6988
6989 A .travis.yml
6990
6991 commit 05e40d29c034c0d68322e589b8d6e234f562be2e
6992 Author: Brian Silverman <[email protected]>
6993 Date: Tue Mar 8 15:58:02 2016 -0500
6994
6995 Recognize modern Linux ARM
6996
6997 Commit e580d7888154fa7f95b3cef9e18f1ce69182212b fixed the macros
6998 in some
6999 of the code but not other places.
7000
7001 `make check` still fails in the same places on a Debian Jessie armhf
7002 system.
7003
7004 M src/base/elfcore.h
7005 M src/base/linuxthreads.h
7006
7007 commit 632de2975e63f89613af9ab99bc1603a4a6332aa
7008 Author: Aliaksey Kandratsenka <[email protected]>
7009 Date: Sat Mar 12 11:48:20 2016 -0800
7010
7011 bumped version up to 2.5
7012
7013 M NEWS
7014 M configure.ac
7015 M src/windows/config.h
7016 M src/windows/gperftools/tcmalloc.h
7017
7018 commit 6682016092c735feb29021acbb8c2067fdf1a209
7019 Author: Aliaksey Kandratsenka <[email protected]>
7020 Date: Sat Mar 12 10:54:05 2016 -0800
7021
7022 Unbreak profiling with CPUPROFILE_FREQUENCY=1
7023
7024 This closes ticket #777.
7025
7026 No test sadly, since it's not trivial to unittest this case. But
7027 fix with single-shot manual testing is better than nothing.
7028
7029 M src/profile-handler.cc
7030
7031 commit 6ff86ff6a783a277956e45db68c6a2b729f27000
7032 Author: Aliaksey Kandratsenka <[email protected]>
7033 Date: Sat Mar 5 16:17:28 2016 -0800
7034
7035 bumped version to 2.4.91 for 2.5rc2
7036
7037 M NEWS
7038 M configure.ac
7039 M src/windows/config.h
7040 M src/windows/gperftools/tcmalloc.h
7041
7042 commit 782165fa7f2c49d6a67c2415626a1f215cc21ac2
7043 Author: Aliaksey Kandratsenka <[email protected]>
7044 Date: Sat Mar 5 15:47:25 2016 -0800
7045
7046 build sized delete aliases even when sized-delete is disabled
7047
7048 In this case we alias to regular delete. This is helpful because if we
7049 don't override sized delete, then apps will call version in libstdc++
7050 which delegates calls to regular delete, which is slower than calling
7051 regular delete directly.
7052
7053 M configure.ac
7054 M src/libc_override_gcc_and_weak.h
7055
7056 commit 06811b3ae4c5365880898148d188ae91d20f52f2
7057 Author: Aliaksey Kandratsenka <[email protected]>
7058 Date: Sat Mar 5 15:08:50 2016 -0800
7059
7060 disable dynamic sized delete support by default
7061
7062 IFUNC relocations don't support our advanced use case (calling
7063 application function or looking up environment variable).
7064
7065 Particularly, it doesn't work on PPC and arm when tcmalloc is linked
7066 with -Wl,-z,now. See RedHat's bugzilla ticket
7067 https://bugzilla.redhat.com/show_bug.cgi?id=1312462 for more details.
7068
7069 M configure.ac
7070
7071 commit d4d99eb608f3d778fa301137a85cc4903a1dd33e
7072 Author: Aliaksey Kandratsenka <[email protected]>
7073 Date: Sat Mar 5 14:53:42 2016 -0800
7074
7075 unbreak compilation with visual studio
7076
7077 Specifically, this commit adds missing fake_stacktrace_scope.cc
7078 to msvc
7079 build and removes undef-ing of PERFTOOLS_DLL_DECL by tcmalloc.h.
7080
7081 M src/gperftools/tcmalloc.h.in
7082 M src/windows/config.h
7083 M src/windows/gperftools/tcmalloc.h
7084 M src/windows/gperftools/tcmalloc.h.in
7085 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
7086
7087 commit 126d4582c18d053452d22ce12351c5429513aabe
7088 Author: Brian Silverman <[email protected]>
7089 Date: Tue Mar 1 17:57:33 2016 -0500
7090
7091 Call function pointers with the right type
7092
7093 I think it's undefined behavior, but it's definitely weird. ubsan
7094 complains too.
7095
7096 M src/tests/tcmalloc_unittest.cc
7097
7098 commit e0fa28ef7d50158b271da90f53ab3aee5ba441d8
7099 Author: Brian Silverman <[email protected]>
7100 Date: Tue Mar 1 17:56:32 2016 -0500
7101
7102 Don't shift a type by more than its width
7103
7104 It's undefined behavior, so ubsan complains.
7105
7106 M src/tests/sampler_test.cc
7107
7108 commit a1c764d2632cc0090f31d8d56effb94a8d049b54
7109 Author: Brian Silverman <[email protected]>
7110 Date: Tue Mar 1 17:55:59 2016 -0500
7111
7112 Initialize counters in test
7113
7114 M src/tests/profile-handler_unittest.cc
7115
7116 commit 22123a37c236e26535d3f3fff7f31a5b6515d7d6
7117 Author: Brian Silverman <[email protected]>
7118 Date: Tue Mar 1 17:55:53 2016 -0500
7119
7120 Don't overflow a signed integer
7121
7122 It's undefined behavior and ubsan catches it.
7123
7124 M src/base/low_level_alloc.cc
7125
7126 commit 66e1e94f38467b5c7bbfb05e3c7267f3039a2c69
7127 Author: Aliaksey Kandratsenka <[email protected]>
7128 Date: Mon Feb 29 20:09:39 2016 -0800
7129
7130 added minimal "header section" to README
7131
7132 M README
7133
7134 commit 2804b7cfee8d49d5ab98340cbf3050c0a5f57b98
7135 Author: Aliaksey Kandratsenka <[email protected]>
7136 Date: Sun Feb 21 20:27:48 2016 -0800
7137
7138 bumped version to 2.5rc
7139
7140 M configure.ac
7141 M src/windows/config.h
7142 M src/windows/gperftools/tcmalloc.h
7143
7144 commit f47fefbfc1083e7efbada3e7e7d2859d31e151e7
7145 Author: Aliaksey Kandratsenka <[email protected]>
7146 Date: Sun Feb 21 22:38:23 2016 -0800
7147
7148 updated NEWS for 2.5rc
7149
7150 M NEWS
7151
7152 commit cef60361741e858041f4bb4d46ce00c6db89b711
7153 Author: Aliaksey Kandratsenka <[email protected]>
7154 Date: Sun Feb 21 23:03:41 2016 -0800
7155
7156 alias same malloc/free variants to their canonical versions
7157
7158 This is similar to what gcc 5 does anyways, except gcc 5 is placing
7159 jumps which adds a bit of overhead.
7160
7161 Instead of letting gcc do it, we alias using ELF symbol aliasing. All
7162 free variants (tc_delete{,array}_{,nothrow}) are aliased to
7163 tc_free. There are 3 malloc variants that differ by oom
7164 handling. tc_newarray is aliased to tc_new. And tc_newarray_nothrow is
7165 aliased to tc_new_nothrow.
7166
7167 This aliasing only happens in non-debug malloc, since debug malloc
7168 does
7169 distinguish between different variants since it needs checking for
7170 mismatch.
7171
7172 M src/tcmalloc.cc
7173
7174 commit ea8d242061460309d174d9053a224cf94e65e6fe
7175 Author: Aliaksey Kandratsenka <[email protected]>
7176 Date: Sun Feb 21 20:16:37 2016 -0800
7177
7178 Re-enable MultipleIdleNonIdlePhases test
7179
7180 Which was accidentally removed as part of MarkThreadTemporarilyIdle
7181 support.
7182
7183 M src/tests/markidle_unittest.cc
7184
7185 commit c9962f698b0edb37315445b33a06820f9bcf5df9
7186 Author: Aliaksey Kandratsenka <[email protected]>
7187 Date: Sun Feb 21 20:07:37 2016 -0800
7188
7189 added maybe_emergency_malloc.h to Makefile.am
7190
7191 Because without this reference it isn't packaged by make dist.
7192
7193 M Makefile.am
7194
7195 commit 7dd4af65365d74a5d8d30d5811c26117a9192238
7196 Author: Aliaksey Kandratsenka <[email protected]>
7197 Date: Sun Feb 21 19:26:37 2016 -0800
7198
7199 don't round up sizes for large allocation when sampling
7200
7201 This closes #723.
7202
7203 Since rounding up prior to sampling is introducing possibility of
7204 arithmetic overflow, we're just not doing it.
7205
7206 It introduces some error (up to 4k), but since we're dealing with at
7207 least 256k allocations, we're fine.
7208
7209 M src/tcmalloc.cc
7210
7211 commit 4f3410e759ec42cb307429222d690a81e3cd37b0
7212 Author: Aliaksey Kandratsenka <[email protected]>
7213 Date: Sun Feb 21 13:52:47 2016 -0800
7214
7215 enable emergency malloc by default on arm when using libunwind
7216
7217 Because by default libunwind on arm does fopen which mallocs
7218 which will
7219 deadlock unless we enable emergency malloc.
7220
7221 M configure.ac
7222
7223 commit 7f12051dbe1f402771a747c1192a66e7571d94f9
7224 Author: Aliaksey Kandratsenka <[email protected]>
7225 Date: Sun Jan 31 23:17:50 2016 -0800
7226
7227 implemented emergency malloc
7228
7229 Emergency malloc is enabled for cases when backtrace capturing
7230 needs to
7231 call malloc. In this case, we enable emergency malloc just prior to
7232 calling such code and disable it after it is done.
7233
7234 M Makefile.am
7235 M configure.ac
7236 M src/debugallocation.cc
7237 A src/emergency_malloc.cc
7238 A src/emergency_malloc.h
7239 A src/emergency_malloc_for_stacktrace.cc
7240 A src/fake_stacktrace_scope.cc
7241 M src/malloc_hook.cc
7242 A src/maybe_emergency_malloc.h
7243 M src/stacktrace.cc
7244 M src/tcmalloc.cc
7245 M src/thread_cache.h
7246
7247 commit 3ee2360250c639f02b354820bc50efc6e104b754
7248 Author: Aliaksey Kandratsenka <[email protected]>
7249 Date: Sat Feb 20 21:49:58 2016 -0800
7250
7251 replaced invalid uses of __THROW
7252
7253 We're now using it only when overriding glibc functions (such
7254 as malloc
7255 or mmap). In other cases (most importantly in public tcmalloc.h
7256 header)
7257 we're doing our own throw() to avoid possible breakage on future glibc
7258 changes.
7259
7260 M src/debugallocation.cc
7261 M src/gperftools/tcmalloc.h.in
7262 M src/libc_override_glibc.h
7263 M src/libc_override_redefine.h
7264 M src/tcmalloc.cc
7265 M src/windows/gperftools/tcmalloc.h
7266 M src/windows/gperftools/tcmalloc.h.in
7267
7268 commit 013b82abcfeaa158da2490ff86d6bff5a5434d7f
7269 Author: Aliaksey Kandratsenka <[email protected]>
7270 Date: Sat Feb 20 21:19:37 2016 -0800
7271
7272 unbreak <malloc.h> inclusion in gperftools/tcmalloc.h
7273
7274 We have shipped header which checked HAVE_XXX defines which we only
7275 defined in project-local config.h. So it could never work correctly.
7276
7277 We're now doing #include <malloc.h> just like tc_mallinfo on constant
7278 which we detect at configure time and write into header that we
7279 install.
7280
7281 M src/gperftools/tcmalloc.h.in
7282
7283 commit 19903e6f156d0ff8063880ddc5051b5dc261c1ff
7284 Author: Aliaksey Kandratsenka <[email protected]>
7285 Date: Sat Feb 20 21:18:18 2016 -0800
7286
7287 drop detection of sys/malloc.h and malloc/malloc.h
7288
7289 Which we don't really need.
7290
7291 M configure.ac
7292
7293 commit cdff090ebd7a3cef8de60109bef52251758181d7
7294 Author: Aliaksey Kandratsenka <[email protected]>
7295 Date: Sat Feb 20 20:43:21 2016 -0800
7296
7297 Fix several harmless clang warnings
7298
7299 M src/common.cc
7300 M src/internal_logging.cc
7301 M src/malloc_hook.cc
7302 M src/tests/current_allocated_bytes_test.cc
7303
7304 commit 9095ed08402620b04591c2d27ca222594f0da6db
7305 Author: Aliaksey Kandratsenka <[email protected]>
7306 Date: Sat Feb 20 20:33:27 2016 -0800
7307
7308 implemented stacktrace capturing via libgcc's C++ ABI function
7309
7310 Particularly _Unwind_Backtrace which seems to be gcc extension.
7311
7312 This is what glibc's backtrace is commonly is using.
7313
7314 Using _Unwind_Backtrace directly is better than glibc's backtrace,
7315 since
7316 it doesn't call into dlopen. While glibc does dlopen when it is
7317 built as
7318 shared library apparently to avoid link-time dependency on libgcc_s.so
7319
7320 M Makefile.am
7321 M configure.ac
7322 M src/stacktrace.cc
7323 A src/stacktrace_libgcc-inl.h
7324
7325 commit 728cbe102108be8296f2485e384392a4408ce9db
7326 Author: Aliaksey Kandratsenka <[email protected]>
7327 Date: Sat Feb 20 13:06:08 2016 -0800
7328
7329 force profiler_unittest to do 'real' work
7330
7331 'XOR loop' in profiler unittest wasn't 100% effective because
7332 it allowed
7333 compiler to avoid loading and storing to memory.
7334
7335 After marking result variable as volatile, we're now forcing
7336 compiler to
7337 read and write memory, slowing this loops down sufficiently. And
7338 profiler_unittest is now passing more consistently.
7339
7340 Closes #628
7341
7342 M src/tests/profiler_unittest.cc
7343
7344 commit fff6b4fb8814b5f25d4f97889d5003537f7dcc92
7345 Author: Aliaksey Kandratsenka <[email protected]>
7346 Date: Sun Jan 31 23:14:43 2016 -0800
7347
7348 Extend low-level allocator to support custom pages allocator
7349
7350 M src/base/low_level_alloc.cc
7351 M src/base/low_level_alloc.h
7352
7353 commit 32d992679589ea83e044d3f8263583ae960b0b16
7354 Author: Aliaksey Kandratsenka <[email protected]>
7355 Date: Sun Jan 31 22:42:20 2016 -0800
7356
7357 added malloc_bench_shared_full
7358
7359 M .gitignore
7360 M Makefile.am
7361
7362 commit 00d8fa1ef8d6650f08e00f59baa22cd31b908432
7363 Author: Aliaksey Kandratsenka <[email protected]>
7364 Date: Sat Feb 6 19:13:07 2016 -0800
7365
7366 always use real throw() on operators new/delete
7367
7368 Since non-glibc-s have no __THROW and lack of throw() on operators
7369 gives
7370 us warning.
7371
7372 M src/libc_override_gcc_and_weak.h
7373
7374 commit 08e034ad5940f24f99a24630c7b71a0de728d05b
7375 Author: Aliaksey Kandratsenka <[email protected]>
7376 Date: Sat Feb 6 16:19:54 2016 -0800
7377
7378 Detect working ifunc before enabling dynamic sized delete support
7379
7380 Particularly, on arm-linux and x86-64-debian-kfreebsd compilation
7381 fails
7382 due to lack of support for ifunc. So it is necessary to test at
7383 configure time whether ifunc is supported.
7384
7385 M configure.ac
7386
7387 commit a788f354a0cd6de3d714a58ad3e6ae5baf806c3b
7388 Author: Aliaksey Kandratsenka <[email protected]>
7389 Date: Sat Feb 6 16:01:25 2016 -0800
7390
7391 include unistd.h for getpid in thread_lister.c
7392
7393 This fixes warning produced on arm-linux.
7394
7395 M src/base/thread_lister.c
7396
7397 commit 644a6bdbdb50b26a25e4428c43556467a6e8b5cc
7398 Author: Bryan Chan <[email protected]>
7399 Date: Fri Jan 22 12:26:47 2016 -0500
7400
7401 Add support for Linux s390x
7402
7403 This resolves gperftools/gperftools#761.
7404
7405 M configure.ac
7406 M m4/pc_from_ucontext.m4
7407 M src/base/basictypes.h
7408 M src/base/linux_syscall_support.h
7409 M src/base/linuxthreads.h
7410 M src/malloc_hook_mmap_linux.h
7411
7412 commit bab7753aad44b3395a063966f32c23f632fee174
7413 Author: Bryan Chan <[email protected]>
7414 Date: Fri Jan 29 02:38:42 2016 -0500
7415
7416 Fix typo in heap-checker-death_unittest.sh
7417
7418 M src/tests/heap-checker-death_unittest.sh
7419
7420 commit 17182e1d3c324e8fe9cc02ce2822caa746d7d17a
7421 Author: Simon Que <[email protected]>
7422 Date: Fri Jan 29 18:17:16 2016 -0800
7423
7424 Fix include of malloc_hook_c.h in malloc_hook.h
7425
7426 malloc_hook.h includes malloc_hook_c.h as
7427 <gperftools/malloc_hook_c.h>.
7428 This requires the compiler to have designated src/gperftools as a
7429 standard include directory (-I), which may not always be the case.
7430
7431 Instead, include it as "malloc_hook_c.h", which will search in
7432 the same
7433 directory first. This will always work, regardless of whether it was
7434 designated a standard include directory.
7435
7436 M src/gperftools/malloc_hook.h
7437
7438 commit c69721b2b2ceae426c36de191dd0a6fa443c5c7a
7439 Author: Andrew Morrow <[email protected]>
7440 Date: Sun Nov 22 15:19:12 2015 -0500
7441
7442 Add support for obtaining cache size of the current thread and
7443 softer idling
7444
7445 M src/gperftools/malloc_extension.h
7446 M src/gperftools/malloc_extension_c.h
7447 M src/malloc_extension.cc
7448 M src/tcmalloc.cc
7449 M src/tests/markidle_unittest.cc
7450 M src/thread_cache.cc
7451 M src/thread_cache.h
7452
7453 commit 5ce42e535d6b3bf6c5e41ceb9cf876a70a7e3318
7454 Author: Brian Silverman <[email protected]>
7455 Date: Tue Jan 26 16:53:48 2016 -0500
7456
7457 Don't always arm the profiling timer.
7458
7459 It causes a noticeable performance hit and can sometimes confuse GDB.
7460
7461 Tested with CPUPROFILE_PER_THREAD_TIMERS=1.
7462
7463 Based on an old version by [email protected].
7464
7465 M src/profile-handler.cc
7466 M src/profile-handler.h
7467 M src/tests/profile-handler_unittest.cc
7468
7469 commit 7f801ea091e03dae7b13160f76fecce15727c9c0
7470 Author: Duncan Sands <[email protected]>
7471 Date: Fri Jan 15 18:12:51 2016 +0100
7472
7473 Make sure the alias is not removed by link-time optimization when
7474 it can prove
7475 that it isn't used by the program, as it might still be needed to
7476 override the
7477 corresponding symbol in shared libraries (or inline assembler for
7478 that matter).
7479 For example, suppose the program uses malloc and free but not calloc
7480 and is
7481 statically linked against tcmalloc (built with -flto) and LTO
7482 is done. Then
7483 before this patch the calloc alias would be deleted by LTO due to
7484 not being
7485 used, but the malloc/free aliases would be kept because they are
7486 used by the
7487 program. Suppose the program is dynamically linked with a shared
7488 library that
7489 allocates memory using calloc and later frees it by calling free.
7490 Then calloc
7491 will use the libc memory allocator, because the calloc alias was
7492 deleted, but
7493 free will call into tcmalloc, resulting in a crash.
7494
7495 M src/libc_override_gcc_and_weak.h
7496
7497 commit 6b3e6ef5e0ab65030d116dd55dac8d64f9c72d33
7498 Author: Aliaksey Kandratsenka <[email protected]>
7499 Date: Sun Jan 24 19:45:16 2016 -0800
7500
7501 don't retain compatibility with old docdir behavior
7502
7503 Since it is not really needed. And since we don't care about too
7504 ancient
7505 autoconfs.
7506
7507 M configure.ac
7508
7509 commit ccffcbd9e988a39654fe6ec5edd6d817f266ded9
7510 Author: Chris Mayo <[email protected]>
7511 Date: Sun Dec 27 18:55:05 2015 +0000
7512
7513 support use of configure --docdir argument
7514
7515 Value of docdir was being overridden in Makefile.
7516
7517 Retain compatibility with old Autoconf versions that do not provide
7518 docdir.
7519
7520 M Makefile.am
7521 M configure.ac
7522
7523 commit 050f2d28be8a63836bd74558b82be7983770a654
7524 Author: Aliaksey Kandratsenka <[email protected]>
7525 Date: Sat Dec 12 18:27:40 2015 -0800
7526
7527 use alias attribute only for elf platforms
7528
7529 It was reported that clang on OSX doesn't support alias
7530 attribute. Most
7531 likely because of executable format limitations.
7532
7533 New code limits use of alias to gcc-compatible compilers on elf
7534 platforms (various gnu and *bsd systems). Elf format is known
7535 to support
7536 aliases.
7537
7538 M src/tcmalloc.cc
7539
7540 commit 07b0b21ddd168d977c04b1395677de71bb3e402e
7541 Author: cyshi <[email protected]>
7542 Date: Wed Dec 2 14:47:15 2015 +0800
7543
7544 fix compilation error in spinlock
7545
7546 M src/base/spinlock.cc
7547
7548 commit e14450366abc51c7a25256615179daf0bae1de34
7549 Author: gshirishfree <[email protected]>
7550 Date: Mon Nov 23 11:34:13 2015 -0800
7551
7552 Added better description for GetStats API
7553
7554 M src/gperftools/malloc_extension.h
7555
7556 commit 64892ae730b704a7f3f23bd8eaeaf206901df201
7557 Author: Aliaksey Kandratsenka <[email protected]>
7558 Date: Sat Nov 21 19:17:36 2015 -0800
7559
7560 lower default transfer batch size down to 512
7561
7562 Some workloads get much slower with too large batch size.
7563
7564 This closes bug #678.
7565
7566 binary_trees benchmark benefits from larger batch size. And I
7567 found that
7568 512 is not much slower than huge value that we had.
7569
7570 M src/common.cc
7571
7572 commit 6fdfc5a7f40ebcff3fdaada1a2994ff54be2f9c7
7573 Author: Aliaksey Kandratsenka <[email protected]>
7574 Date: Sat Oct 24 23:16:45 2015 -0700
7575
7576 implemented enabling sized-delete support at runtime
7577
7578 Under gcc 4.5 or greater we're using ifunc function attribute
7579 to resolve
7580 sized delete operator to either plain delete implementation
7581 (default) or
7582 to sized delete (if enabled via environment variable
7583 TCMALLOC_ENABLE_SIZED_DELETE).
7584
7585 M configure.ac
7586 M src/libc_override_gcc_and_weak.h
7587 M src/tests/tcmalloc_unittest.sh
7588
7589 commit c2a79d063c949584170b3e7dd2939a4548c16079
7590 Author: Aliaksey Kandratsenka <[email protected]>
7591 Date: Fri Oct 23 21:44:04 2015 -0700
7592
7593 use x86 pause in spin loop
7594
7595 This saves power and improves performance, particulary on SMT.
7596
7597 M src/base/spinlock.cc
7598
7599 commit 0fb6dd8aa3c2fe3245c31c138ec4dfb05efd7f6d
7600 Author: Aliaksey Kandratsenka <[email protected]>
7601 Date: Fri Oct 23 22:09:25 2015 -0700
7602
7603 added binary_trees benchmark
7604
7605 M .gitignore
7606 M Makefile.am
7607 A benchmark/binary_trees.cc
7608
7609 commit a8852489e54e915d22abbdad1688d1f25ccb7925
7610 Author: Aliaksey Kandratsenka <[email protected]>
7611 Date: Sun Oct 18 00:40:15 2015 -0700
7612
7613 drop unsupported allocation sampling code in tcmalloc_minimal
7614
7615 M src/tcmalloc.cc
7616 M src/thread_cache.h
7617
7618 commit a9db0ae516c225543976280047b22b1e0ca08b93
7619 Author: Aliaksey Kandratsenka <[email protected]>
7620 Date: Sun Oct 4 21:12:28 2015 -0700
7621
7622 implemented (disabled by default) sized delete support
7623
7624 gcc 5 and clang++-3.7 support sized deallocation from C++14. We are
7625 taking advantage of that by defining sized versions of operator
7626 delete.
7627
7628 This is off by default so that if some existing programs that
7629 define own
7630 global operator delete without sized variant are not broken by
7631 tcmalloc's sized delete operator.
7632
7633 There is also risk of breaking exiting code that deletes objects using
7634 wrong class (i.e. base class) without having virtual destructors.
7635
7636 M configure.ac
7637 M src/gperftools/tcmalloc.h.in
7638 M src/libc_override_gcc_and_weak.h
7639 M src/libc_override_redefine.h
7640 M src/tcmalloc.cc
7641 M src/windows/gperftools/tcmalloc.h
7642 M src/windows/gperftools/tcmalloc.h.in
7643
7644 commit 88686972b9caf8fd132c4e30d89da78c3d324c27
7645 Author: Aliaksey Kandratsenka <[email protected]>
7646 Date: Sun Oct 4 11:15:37 2015 -0700
7647
7648 pass -fsized-deallocation to gcc 5
7649
7650 Otherwise it gives warning for declaration of sized delete operator.
7651
7652 M Makefile.am
7653 M configure.ac
7654
7655 commit 0a18fab3af2f1b59b851836d319a1eff0db5d44e
7656 Author: Aliaksey Kandratsenka <[email protected]>
7657 Date: Sun Oct 4 21:07:54 2015 -0700
7658
7659 implemented sized free support via tc_free_sized
7660
7661 M src/debugallocation.cc
7662 M src/gperftools/tcmalloc.h.in
7663 M src/tcmalloc.cc
7664 M src/windows/gperftools/tcmalloc.h
7665 M src/windows/gperftools/tcmalloc.h.in
7666 M src/windows/patch_functions.cc
7667
7668 commit 464688ab6ddb82db22ea9b934ddea192478e4535
7669 Author: Aliaksey Kandratsenka <[email protected]>
7670 Date: Sun Oct 4 21:27:56 2015 -0700
7671
7672 speedup free code path by dropping "fast path allowed check"
7673
7674 M src/tcmalloc.cc
7675
7676 commit 10f7e207166bd6652ed8803b9885917bfbbd90d0
7677 Author: Aliaksey Kandratsenka <[email protected]>
7678 Date: Sat Oct 3 18:47:14 2015 -0700
7679
7680 added SizeMap::MaybeSizeClass
7681
7682 Because it allows us to first check for smaller sizes, which is most
7683 likely.
7684
7685 M src/common.h
7686
7687 commit 436e1dea43dc097525e43a43e6d79810ba6921b5
7688 Author: Aliaksey Kandratsenka <[email protected]>
7689 Date: Sat Oct 3 15:40:21 2015 -0700
7690
7691 slightly faster GetCacheIfPresent
7692
7693 M src/thread_cache.h
7694
7695 commit 04df911915ecfcbc5e6b13fdab1b86f825ed1938
7696 Author: Aliaksey Kandratsenka <[email protected]>
7697 Date: Sat Oct 3 15:38:33 2015 -0700
7698
7699 tell compiler that non-empty hooks are unlikely
7700
7701 M src/malloc_hook-inl.h
7702
7703 commit 8cc75acd1f4024ca978858b1cf85b45204ed32b2
7704 Author: Aliaksey Kandratsenka <[email protected]>
7705 Date: Sun Oct 4 11:04:12 2015 -0700
7706
7707 correctly test for -Wno-unused-result support
7708
7709 gcc is only giving warning for unknown -Wno-XXX flags so test never
7710 fails on gcc even if -Wno-XXX is not supported. By using
7711 -Wunused-result we're able to test if gcc actually supports it.
7712
7713 This fixes issue #703.
7714
7715 M configure.ac
7716
7717 commit 7753d8239b1e427d83729b6d5c0401e1d80308fd
7718 Author: Aliaksey Kandratsenka <[email protected]>
7719 Date: Sun Oct 4 18:26:59 2015 -0700
7720
7721 fixed clang warning about shifting negative values
7722
7723 M src/base/basictypes.h
7724
7725 commit ae09ebb3833f3f207b7623245e360ddf122b823e
7726 Author: Jens Rosenboom <[email protected]>
7727 Date: Thu Nov 19 17:45:44 2015 +0100
7728
7729 Fix tmpdir usage in heap-profiler_unittest.sh
7730
7731 Using a single fixed directory would break when tests were being
7732 run in
7733 parallel with "make -jN".
7734
7735 Also, the cleanup at the end of the test didn't work because it
7736 referred
7737 to the wrong variable.
7738
7739 M src/tests/heap-profiler_unittest.sh
7740
7741 commit df34e71b575daa6e4fa7588ae0deb9d529566f08
7742 Author: Aliaksey Kandratsenka <[email protected]>
7743 Date: Sat Nov 21 16:03:09 2015 -0800
7744
7745 use $0 when referring to pprof
7746
7747 This fixed debian bug #805536. Debian ships pprof under google-pprof
7748 name so it is handy when google-pprof --help refers to itself
7749 correctly.
7750
7751 M src/pprof
7752
7753 commit 7773ea64ee254700ef0ffc5673539f0e111a4f86
7754 Author: Adhemerval Zanella <[email protected]>
7755 Date: Fri Nov 6 16:29:12 2015 -0200
7756
7757 Alignment fix to static variables for system allocators
7758
7759 This patch the placement new for some system allocator to force the
7760 static buffer to pointer value.
7761
7762 M src/memfs_malloc.cc
7763 M src/system-alloc.cc
7764
7765 commit c46eb1f3d2f7a2bdc54a52ff7cf5e7392f5aa668
7766 Author: Boris Sazonov <[email protected]>
7767 Date: Thu Oct 15 17:39:08 2015 +0300
7768
7769 Fixed printf misuse in pprof - printed string was passed as
7770 format. Better use print instead
7771
7772 M src/pprof
7773
7774 commit 9bbed8b1a8f1229cf0f1b4185683f309267ee1b9
7775 Author: Boris Sazonov <[email protected]>
7776 Date: Wed Sep 23 19:46:13 2015 +0300
7777
7778 Fixed assembler argument passing inside _syscall6 on MIPS - it was
7779 causing 'Expression too complex' compilation errors in spinlock
7780
7781 M src/base/linux_syscall_support.h
7782
7783 commit 962aa53c55968f62f548f6f6f4849a384b505ca0
7784 Author: Aliaksey Kandratsenka <[email protected]>
7785 Date: Sat Oct 10 20:31:50 2015 -0700
7786
7787 added more fastpath microbenchmarks
7788
7789 This also makes them output nicer results. I.e. every benchmark is
7790 run 3
7791 times and iteration duration is printed for every run.
7792
7793 While this is still very synthetic and unrepresentave of malloc
7794 performance
7795 as a whole, it is exercising more situations in tcmalloc fastpath. So
7796 it a
7797 step forward.
7798
7799 M Makefile.am
7800 M benchmark/malloc_bench.cc
7801 A benchmark/run_benchmark.c
7802 A benchmark/run_benchmark.h
7803
7804 commit 347a830689e4dba2adc1368d00fe4723ba726b4a
7805 Author: Aliaksey Kandratsenka <[email protected]>
7806 Date: Sat Oct 10 15:58:11 2015 -0700
7807
7808 Ensure that PPROF_PATH is set for debugallocation_test
7809
7810 Which fixes issue #728.
7811
7812 M src/tests/debugallocation_test.sh
7813
7814 commit a9059b7c30308a4f4ab25f00e55b67c649614ed3
7815 Author: Aliaksey Kandratsenka <[email protected]>
7816 Date: Sat Oct 10 15:35:54 2015 -0700
7817
7818 prevent clang from inlining Mallocer in heap checker unittest
7819
7820 Looks like existing "trick" to avoid inlining doesn't really prevent
7821 sufficiently smart compiler from inlining Mallocer function. Which
7822 breaks tests, since test relies Mallocer having it's own separate
7823 stack
7824 frame.
7825
7826 Making mallocer_addr variable volatile is seemingly enough to
7827 stop that.
7828
7829 M src/tests/heap-checker_unittest.cc
7830
7831 commit 6627f9217d8897b297c6da038cfbcff6a3086cfa
7832 Author: Aliaksey Kandratsenka <[email protected]>
7833 Date: Sat Sep 26 19:52:37 2015 -0700
7834
7835 drop cycleclock
7836
7837 M Makefile.am
7838 D src/base/cycleclock.h
7839 M src/base/sysinfo.cc
7840 M vsprojects/addressmap_unittest/addressmap_unittest.vcproj
7841 M
7842 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
7843 M vsprojects/packed-cache_test/packed-cache_test.vcproj
7844
7845 commit f985abc29607b8a5662ddac7bd1ed99c71ceeb11
7846 Author: Aliaksey Kandratsenka <[email protected]>
7847 Date: Sat Sep 26 19:50:03 2015 -0700
7848
7849 amputate unportable and unused stuff from sysinfo
7850
7851 We still check number of cpus in the system (in spinlock code),
7852 but old
7853 code was built under assumption of "no calls malloc" which is
7854 not needed
7855 in tcmalloc. Which caused it to be far more complicated than
7856 necessary (parsing procfs files, ifdefs for different OSes and
7857 arch-es).
7858
7859 Also we don't need clock cycle frequency measurement.
7860
7861 So I've removed all complexity of ald code and NumCPUs function and
7862 replaced it with GetSystemCPUsCount which is straightforward and
7863 portable call to sysconf.
7864
7865 Renaming of cpus count function was made so that any further code that
7866 we might port from Google that depends on old semantics of NumCPUs
7867 will
7868 be detected at compile time. And has to be inspected for whether it
7869 really needs that semantics.
7870
7871 M src/base/spinlock.cc
7872 M src/base/sysinfo.cc
7873 M src/base/sysinfo.h
7874
7875 commit 16408eb4d71ecbb1dd3b3c7ff22c94bd254d7b58
7876 Author: Aliaksey Kandratsenka <[email protected]>
7877 Date: Tue Sep 29 11:59:25 2015 -0700
7878
7879 amputated wait_cycles accounting in spinlocks
7880
7881 This is not needed and pulls in CycleClock dependency that lowers
7882 code portability.
7883
7884 M src/base/spinlock.cc
7885 M src/base/spinlock.h
7886
7887 commit fedceef40cd217ef406bc2522e9a8f879d60b0c0
7888 Author: Aliaksey Kandratsenka <[email protected]>
7889 Date: Tue Sep 29 11:55:35 2015 -0700
7890
7891 drop cycleclock reference in ThreadCache
7892
7893 M src/thread_cache.cc
7894
7895 commit d7fdc3fc9ddc0937eb7961b1d8c864cc8182f0d9
7896 Author: Aliaksey Kandratsenka <[email protected]>
7897 Date: Tue Sep 29 11:43:09 2015 -0700
7898
7899 dropped unused and unsupported synchronization profiling facility
7900
7901 Spinlock usage of cycle counter is due do tracking of time it's spent
7902 waiting for lock. But this tracking is only useful we actually have
7903 synchronization profiling working, which dont have. Thus I'm dropping
7904 calls to this facility with eye towards further removal of cycle clock
7905 usage.
7906
7907 M Makefile.am
7908 M src/base/spinlock.cc
7909 D src/base/synchronization_profiling.h
7910
7911 commit 3a054d37c1f5323462bd77f55be02c5b0d764611
7912 Author: Aliaksey Kandratsenka <[email protected]>
7913 Date: Mon Sep 28 15:40:15 2015 -0700
7914
7915 dropped unused SpinLockWait function
7916
7917 M src/base/spinlock.cc
7918 M src/base/spinlock_internal.cc
7919 M src/base/spinlock_internal.h
7920
7921 commit 5b62d38329f04ee30fb475a36b552c1ebc4e1e79
7922 Author: Aliaksey Kandratsenka <[email protected]>
7923 Date: Mon Sep 28 10:27:49 2015 -0700
7924
7925 avoid checking for dup. entries on empty backtrace
7926
7927 This might fix issue #721. But it is right thing to do
7928 regardless. Since
7929 if depth is 0 we'll be reading random "garbage" on the stack.
7930
7931 M src/profiler.cc
7932
7933 commit 7b9ded722e4cef9a44b8d8bfa80d3e1e108cf590
7934 Author: Aliaksey Kandratsenka <[email protected]>
7935 Date: Sat Sep 12 16:26:54 2015 -0700
7936
7937 fixed compiler warning in memory_region_map.cc
7938
7939 M src/memory_region_map.cc
7940
7941 commit 4194e485cbb5d8c59f65aba49da63c08ecc573da
7942 Author: Aliaksey Kandratsenka <[email protected]>
7943 Date: Sat Sep 12 16:20:53 2015 -0700
7944
7945 Don't link libtcmalloc_minimal.so to libpthread.so
7946
7947 So that LD_PRELOAD-ing doesn't force loading libpthread.so which may
7948 slow down some single-threaded apps.
7949
7950 tcmalloc already has maybe_threads facility that can detect if
7951 libpthread.so is loaded (via weak symbols) and provide 'simulations'
7952 of
7953 some pthread functions that tcmalloc needs.
7954
7955 M Makefile.am
7956 M src/maybe_threads.cc
7957 M src/maybe_threads.h
7958 M src/static_vars.cc
7959
7960 commit 121038308d8c5b34707614c44de265816a322563
7961 Author: Fredrik Mellbin <[email protected]>
7962 Date: Sun Sep 27 01:18:31 2015 +0200
7963
7964 Check if _MSC_VER is defined to avoid warnings
7965
7966 M src/windows/port.h
7967
7968 commit 73673229955cf35c5b3046ee3100d94e82d33bc2
7969 Author: Fredrik Mellbin <[email protected]>
7970 Date: Sun Sep 27 01:11:14 2015 +0200
7971
7972 Make default config.h work with VS2015
7973
7974 M src/windows/config.h
7975
7976 commit ae0a444db06b2327441e6160eafa33d8b7b95629
7977 Author: Dair Grant <[email protected]>
7978 Date: Thu Oct 1 13:46:22 2015 +0100
7979
7980 Ensure ThreadCache objects are CACHELINE_ALIGNED.
7981
7982 M src/base/basictypes.h
7983 M src/common.cc
7984 M src/system-alloc.cc
7985
7986 commit ea0b1d315497771d9d077c754fb9d6a92d1c01dc
7987 Author: Aliaksey Kandratsenka <[email protected]>
7988 Date: Sat Sep 26 11:13:13 2015 -0700
7989
7990 unbreak TestErrno again
7991
7992 Somehow in previous commit I managed to break assignment of memalign
7993 result to variable being tested. Luckily gcc gave me warning.
7994
7995 M src/tests/tcmalloc_unittest.cc
7996
7997 commit e53aef24add50bdee5ab2943d96e5c5b2b3bf596
7998 Author: Aliaksey Kandratsenka <[email protected]>
7999 Date: Sat Sep 26 09:58:45 2015 -0700
8000
8001 don't try to test memalign on windows
8002
8003 This unbreaks tcmalloc_unittest.cc on windows.
8004
8005 M src/tests/tcmalloc_unittest.cc
8006
8007 commit 7707582448d5bdf311d88726c5f77fcbb0bedd09
8008 Merge: 5078abdb 9eb63bdd
8009 Author: Aliaksey Kandratsenka (aka Aliaksei Kandratsenka)
8010 <[email protected]>
8011 Date: Sat Sep 26 09:43:50 2015 -0700
8012
8013 Merge pull request #717 from myrsloik/master
8014
8015 Use correct mangled new and delete symbols on windows x64
8016
8017 commit 9eb63bddfb7a8a7b086bb3daceeeaa2e3b1d8f67
8018 Author: Fredrik Mellbin <[email protected]>
8019 Date: Thu Sep 24 19:57:26 2015 +0200
8020
8021 Use correct mangled new and delete symbols on windows x64
8022
8023 M src/windows/patch_functions.cc
8024
8025 commit 5078abdb331e63d7a216994f186eb736861f8df7
8026 Author: fdeweerdt <[email protected]>
8027 Date: Tue Aug 18 09:27:39 2015 -0700
8028
8029 Don't discard curl options if timeout is not defined.
8030
8031 Editing the options passed to curl via 'my @URL_FETCHER = ("curl",
8032 "-s");' (in particular to add a -k to ignore self signed certs) fails
8033 for some invocations of curl. In FetchDynamicProfile, 'my @fetcher =
8034 AddFetchTimeout($fetch_timeout, @URL_FETCHER);' ends up being just
8035 'curl' if timeout is not defined.
8036
8037 This happens because AddFetchTimeout doesn't retrieve all the
8038 arguments
8039 from the caller.
8040
8041 [[email protected]: updated commit message]
8042 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
8043
8044 M src/pprof
8045
8046 commit 54505f1d50c2d1f4676f5e87090b64a117fd980e
8047 Author: Aliaksey Kandratsenka <[email protected]>
8048 Date: Sun Aug 2 19:28:03 2015 -0700
8049
8050 help clang with inlining important fast-path functions
8051
8052 Clang's recent focus on code size doesn't help us in malloc fast-path
8053 because somehow clang completely ignores inline directives.
8054
8055 In order to help clang generate code that was actually intended by
8056 original authors, we're adding always_inline attribute to key
8057 fast-path functions.
8058
8059 Clang also guessed likely branch "wrong" in couple places. Which is
8060 now addressed by UNLIKELY declarations there.
8061
8062 M src/tcmalloc.cc
8063 M src/thread_cache.h
8064
8065 commit 73c0c8c61b84e268bafd961bf304b2e4d296142f
8066 Author: Aliaksey Kandratsenka <[email protected]>
8067 Date: Sun Aug 2 19:09:02 2015 -0700
8068
8069 moved do_mallor_or_cpp_alloc in better place
8070
8071 M src/tcmalloc.cc
8072
8073 commit 41aca070e85258d9d47b0ac47f5eddece8bf45ba
8074 Author: Aliaksey Kandratsenka <[email protected]>
8075 Date: Sun Aug 2 10:53:32 2015 -0700
8076
8077 always set errno to ENOMEM on OOM condition and in single place
8078
8079 While standards do not require us to set errno to ENOMEM in certain
8080 places (like posix_memalign), existing code may sometimes set it
8081 (i.e. because mmap or sbrk couldn't get memory from kernel)
8082 anyways. And from my reading of glibc, it's malloc is doing more or
8083 less same by just always setting ENOMEM on OOM condition.
8084
8085 This commit also eliminates some functions (XXX_no_errno) that are not
8086 needed anymore.
8087
8088 M src/debugallocation.cc
8089 M src/page_heap.cc
8090 M src/tcmalloc.cc
8091
8092 commit c4493874cd3b662d2778f3b79a3096ae61569b67
8093 Author: Aliaksey Kandratsenka <[email protected]>
8094 Date: Sat Aug 1 22:20:15 2015 -0700
8095
8096 deal with OOM handling in one place and prior to returning result
8097
8098 This commit removes 4 (four!) duplicates of C++ OOM handling. And
8099 introduces one helper for that.
8100
8101 Other change is that malloc doesn't have to check tc_new_mode anymore
8102 until it _actually_ deals with OOM condition. Which shaves off couple
8103 instructions from fast-path.
8104
8105 M src/debugallocation.cc
8106 M src/tcmalloc.cc
8107
8108 commit 09448a8fe977eaa083340c2504caac8820832179
8109 Author: Aliaksey Kandratsenka <[email protected]>
8110 Date: Sun Aug 2 18:11:04 2015 -0700
8111
8112 added tcmalloc_unittest path with TCMALLOC_HEAP_LIMIT_MB=512
8113
8114 I.e. so that we can exercise "voluntary" OOM conditions better.
8115
8116 M src/tests/tcmalloc_unittest.sh
8117
8118 commit 73fb7c7eb37e488bab8298be4110e30d83af2184
8119 Author: Aliaksey Kandratsenka <[email protected]>
8120 Date: Sun Aug 2 01:47:55 2015 -0700
8121
8122 added test on errno = ENOMEM on out of memory
8123
8124 M src/tests/tcmalloc_unittest.cc
8125
8126 commit eb725ff26371dfb5ae2523802c6abe75833cacef
8127 Author: Aliaksey Kandratsenka <[email protected]>
8128 Date: Sat Aug 1 20:54:39 2015 -0700
8129
8130 unbreak heap-profiler-unittest on gcc 5
8131
8132 gcc 5 has got nice new optimization (-fipa-icf) which merges identical
8133 functions into one. And that causes heap-profiler_unittest to fail
8134 since it expects to see both Allocate and Allocate2 in heap
8135 profiles. And smart GCC detects that they are same function and makes
8136 one function out of two and thus breaks this test.
8137
8138 New code simply adds (disabled) logging calls to make those functions
8139 non-identical.
8140
8141 M src/tests/heap-profiler_unittest.cc
8142
8143 commit 53833298f3822b2b8b78c9dc85160d65d78a6857
8144 Author: Aliaksey Kandratsenka <[email protected]>
8145 Date: Sat Aug 1 19:46:34 2015 -0700
8146
8147 unbreak heap_checker_unittest on gcc 5
8148
8149 GCC 5 ended up too smart and optimized out assignment of allocated
8150 block to global variable. Which caused test to fail since it triggered
8151 unexpected "leak".
8152
8153 M src/tests/heap-checker_unittest.cc
8154
8155 commit 024bae96ce8e1591993fc0da191ce0a92d609481
8156 Author: Aliaksey Kandratsenka <[email protected]>
8157 Date: Sat Aug 1 17:41:42 2015 -0700
8158
8159 dropped support for PREANSINEW define which nobody needs anymore
8160
8161 M src/debugallocation.cc
8162 M src/tcmalloc.cc
8163
8164 commit 64e0133901a20f83c41adb36748fd19d21228515
8165 Author: Aliaksey Kandratsenka <[email protected]>
8166 Date: Sun May 3 12:55:47 2015 -0700
8167
8168 added trivial malloc fast-path benchmark
8169
8170 While this is not good representation of real-world production malloc
8171 behavior, it is representative of length (instruction-wise and well as
8172 cycle-wise) of fast-path. So this is better than nothing.
8173
8174 M .gitignore
8175 M Makefile.am
8176 A benchmark/malloc_bench.cc
8177
8178 commit e1d1311cfb6312cd44e086c879f3e95cbfa0eb9d
8179 Author: Tom Conerly <[email protected]>
8180 Date: Mon Jul 27 11:35:29 2015 -0700
8181
8182 issue-699: Free list on error paths
8183
8184 M src/heap-profile-table.cc
8185
8186 commit b5b79860fd2b8e1a9b0573e93f942695f2992b59
8187 Author: Aliaksey Kandratsenka <[email protected]>
8188 Date: Sat Aug 1 11:24:56 2015 -0700
8189
8190 issue-702: correctly declare arg-less functions in profiler.h
8191
8192 This is patch by user mitchblank.
8193
8194 From his words:
8195
8196 The problem is pretty simple. Ancient C code allowed declarations
8197 without argument prototypes, i.e.
8198
8199 int foo();
8200
8201 For compatibility this is still accepted. If you want to declare a
8202 function with zero prototypes the correct way to do it is:
8203
8204 int foo(void);
8205
8206 C++ also accepts this syntax, but it's not needed there.
8207
8208 Normally compilers still accept the old-style entries, but with
8209 sufficient warning flags gcc will complain about them. It is good for
8210 header files to have the explicit "void" argument so all compilers are
8211 kept happy.
8212
8213 I'm attaching a simple patch to add the "void" parameter to that file.
8214 I haven't checked if other headers have the same problem (I'm just
8215 using the profiler at the moment)
8216
8217 <end of quote>
8218
8219 In fact "int foo()" means "foo accepts any args" and we really want
8220 "foo has no args". For which int foo (void) is right declaration.
8221
8222 M src/gperftools/profiler.h
8223
8224 commit 7df7f14c949d89d9c3f5c7c339bbdda81fb8abc7
8225 Author: Aliaksey Kandratsenka <[email protected]>
8226 Date: Sat Jun 13 21:35:06 2015 -0700
8227
8228 issue-693: enable futex usage on arm
8229
8230 This patch was contributed by user spotrh.
8231
8232 M src/base/linux_syscall_support.h
8233 M src/base/spinlock_linux-inl.h
8234
8235 commit cb998e56d763cfe901cf30a692d4cfd4f85259ae
8236 Author: Aliaksey Kandratsenka <[email protected]>
8237 Date: Sat Jun 13 21:28:28 2015 -0700
8238
8239 issue-693: convert sys_futex to it's 6-arg form
8240
8241 Because sys_futex actually takes 6 args in more recent kernels (even
8242 though last two args are unused for FUTEX_{WAKE,WAIT}.
8243
8244 This is patch contributed by user spotrh.
8245
8246 M src/base/linux_syscall_support.h
8247
8248 commit 36066b8df4bc516ade5209a1f60bd84d6448b531
8249 Author: Aliaksey Kandratsenka <[email protected]>
8250 Date: Sat Jun 27 21:15:30 2015 -0700
8251
8252 issue-695: implementated TCMALLOC_TRACE_FILE variable
8253
8254 This is contributed by Paolo Bonzini.
8255
8256 This commit adds TCMALLOC_TRACE_FILE environment variable, which if
8257 defined overrides location of malloc trace file.
8258
8259 M src/debugallocation.cc
8260
8261 commit c4069d2d37b67296d675c2d1de42a46dc6d43efc
8262 Author: Brian Silverman <[email protected]>
8263 Date: Tue Jun 16 16:56:48 2015 -0700
8264
8265 Add empty virtual destructor to class with virtual methods.
8266
8267 Clang 3.5 has a warning about deleting objects with virtual methods
8268 through non-virtual destructors which was triggered. I'm not sure
8269 whether this actually creates any undefined or otherwise incorrect
8270 behavior, but it seems like a good thing to fix regardless.
8271
8272 Example compiler warning:
8273 third_party/gperftools/src/tests/profile-handler_unittest.cc:282:5:
8274 error:
8275 delete called on '(anonymous namespace)::BusyThread' that has
8276 virtual
8277 functions but non-virtual destructor [-Wdelete-non-virtual-dtor]
8278 delete busy_worker_;
8279 ^
8280
8281 M src/tests/profile-handler_unittest.cc
8282
8283 commit 019362fefcdca39a46d1a831ed46266c24b9a273
8284 Author: Patrick LoPresti <[email protected]>
8285 Date: Mon May 25 11:56:51 2015 -0700
8286
8287 Add support for CPUPROFILE_TIMER_SIGNAL environment variable.
8288
8289 Which both enables per-thread timers and allows the signal number for
8290 the timer to be selected.
8291
8292 [[email protected]: reformatted commit message for subject line length]
8293 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
8294
8295 M src/profile-handler.cc
8296 M src/tests/profile-handler_unittest.cc
8297
8298 commit 81d8d2a9e7f941a2051781fd0fe62c683c32f1ef
8299 Author: Milton Chiang <[email protected]>
8300 Date: Wed May 13 21:53:05 2015 +0800
8301
8302 Add "ARMv8-A" to the supporting list of ARM architecture.
8303
8304 M src/base/arm_instruction_set_select.h
8305
8306 commit 64d1a86cb8da245c982d470a7dfdd635197e6e5e
8307 Author: Aliaksey Kandratsenka <[email protected]>
8308 Date: Sat May 9 15:38:12 2015 -0700
8309
8310 include time.h for struct timespec on Visual Studio 2015
8311
8312 This patch was submitted by user wmamrak.
8313
8314 M src/windows/port.h
8315
8316 commit 7013b219970a329d1db58fbd7fa7c907bec8dbba
8317 Author: Aliaksey Kandratsenka <[email protected]>
8318 Date: Sat May 9 12:48:11 2015 -0700
8319
8320 hook mi_force_{un,}lock on OSX instead of pthread_atfork
8321
8322 This is patch by Anton Samokhvalov.
8323
8324 Apparently it helps with locking around forking on OSX.
8325
8326 M src/libc_override_osx.h
8327 M src/static_vars.cc
8328
8329 commit f25f8e0bf2d361f852929848d79d7ba54586c352
8330 Author: Angus Gratton <[email protected]>
8331 Date: Mon May 4 08:18:43 2015 +1000
8332
8333 Clarify that only tcmalloc_minimal is supported on Windows.
8334
8335 M INSTALL
8336 M README_windows.txt
8337
8338 commit 772a686c45455893708178d3b59b1d3b571015aa
8339 Author: Aliaksey Kandratsenka <[email protected]>
8340 Date: Sun May 3 13:15:16 2015 -0700
8341
8342 issue-683: fix compile error in clang with -m32 and 64-bit off_t
8343
8344 M src/malloc_hook_mmap_linux.h
8345
8346 commit 0a3bafd645764250732f3c0627534142568f6f1f
8347 Author: Aliaksey Kandratsenka <[email protected]>
8348 Date: Sat Apr 11 10:35:53 2015 -0700
8349
8350 fix typo in PrintAvailability code
8351
8352 This is patch contributed by user ssubotin.
8353
8354 M src/windows/nm-pdb.c
8355
8356 commit 6ce10a2a05f13803061538d5c77e89695de59be4
8357 Author: Matt Cross <[email protected]>
8358 Date: Thu Mar 26 16:13:56 2015 -0400
8359
8360 Add support for printing collapsed stacks for generating flame graphs.
8361
8362 M src/pprof
8363
8364 commit 2c1a165fa56a6f8dff8fa2662ceda971ad676ead
8365 Author: Matt Cross <[email protected]>
8366 Date: Thu Mar 26 12:10:23 2015 -0400
8367
8368 Add support for reading debug symbols automatically on systems
8369 where shared libraries with debug symbols are installed at
8370 "/usr/lib/debug/<originalpath>.debug", such as RHEL and CentOS.
8371
8372 M src/pprof
8373
8374 commit 2e654956287043a30aeaec20e5c19650358af618
8375 Author: Jonathan Lambrechts <[email protected]>
8376 Date: Fri Feb 13 18:52:21 2015 +0100
8377
8378 callgrind : handle inlined functions
8379
8380 M src/pprof
8381
8382 commit 90d7408d381cf2ad68ce4974cbccd51ed5222ca3
8383 Author: Jonathan Lambrechts <[email protected]>
8384 Date: Fri Feb 13 18:51:33 2015 +0100
8385
8386 pprof : callgrind : fix unknown files
8387
8388 M src/pprof
8389
8390 commit aa963a24ae7a74b095a631ea6a86cd071c453911
8391 Author: Aliaksey Kandratsenka <[email protected]>
8392 Date: Mon Feb 9 08:35:03 2015 -0800
8393
8394 issue-672: fixed date of news entry of gperftools 2.4 release
8395
8396 It is 2015 and not 2014. Spotted and reported by Armin Rigo.
8397
8398 M NEWS
8399
8400 commit c66aeabdbacbfd3aff7a6633f34526ca32642f67
8401 Author: Aliaksey Kandratsenka <[email protected]>
8402 Date: Sat Jan 10 14:35:54 2015 -0800
8403
8404 fixed default value of HEAP_PROFILER_TIME_INTERVAL in .html doc
8405
8406 M doc/heapprofile.html
8407
8408 commit 689e4a5bb4b2a8afecb85e83b8e4f294f80b6124
8409 Author: Aliaksey Kandratsenka <[email protected]>
8410 Date: Sat Jan 10 12:26:51 2015 -0800
8411
8412 bumped version to 2.4
8413
8414 M NEWS
8415 M configure.ac
8416 M src/windows/config.h
8417 M src/windows/gperftools/tcmalloc.h
8418
8419 commit 3f5f1bba0c2cb4862c38b0c9050f1cede8a5c344
8420 Author: Aliaksey Kandratsenka <[email protected]>
8421 Date: Sun Dec 28 18:18:27 2014 -0800
8422
8423 bumped version to 2.4rc
8424
8425 M configure.ac
8426 M src/windows/config.h
8427 M src/windows/gperftools/tcmalloc.h
8428
8429 commit c4dfdebc796c2e802db686a7eea483f3d31edbcf
8430 Author: Aliaksey Kandratsenka <[email protected]>
8431 Date: Sun Dec 28 17:53:02 2014 -0800
8432
8433 updated NEWS for gperftools 2.4rc
8434
8435 M NEWS
8436
8437 commit 0096be5f6f0bb2a01438ee78b7bb6158ffa5a1fb
8438 Author: Aliaksey Kandratsenka <[email protected]>
8439 Date: Sat Dec 20 17:17:41 2014 -0800
8440
8441 pprof: allow disabling auto-removal of "constant 2nd frame"
8442
8443 "constand 2nd frame" feature is supposed to detect and workaround
8444 incorrect cpu profile stack captures where parts of or whole cpu
8445 profiling signal handler frames are not skipped.
8446
8447 I've seen programs where this feature incorrectly removes non-signal
8448 frames.
8449
8450 Plus it actually hides bugs in stacktrace capturing which we want be
8451 able to spot.
8452
8453 There is now --no-auto-signal-frm option for disabling it.
8454
8455 M src/pprof
8456
8457 commit 4859d8020579cd8db8b1f8bc6af382104b38f96f
8458 Author: Aliaksey Kandratsenka <[email protected]>
8459 Date: Sat Dec 20 19:56:13 2014 -0800
8460
8461 cpuprofiler: drop correct number of signal handler frames
8462
8463 We actually have 3 and not 2 of them.
8464
8465 M src/profiler.cc
8466
8467 commit 812ab1ee7e4365e6a9568834c7e8f4aef10018fb
8468 Author: Aliaksey Kandratsenka <[email protected]>
8469 Date: Sat Dec 20 19:53:12 2014 -0800
8470
8471 pprof: eliminate duplicate top frames if dropping signal frames
8472
8473 In cpu profiles that had parts of signal handler we could have
8474 situation like that:
8475
8476 * PC
8477 * signal handler frame
8478 * PC
8479
8480 Specifically when capturing stacktraces via libunwind.
8481
8482 For such stacktraces pprof used to draw self-cycle in functions
8483 confusing everybody. Given that me might have a number of such
8484 profiles in the wild it makes sense to treat that duplicate PC issue.
8485
8486 M src/pprof
8487
8488 commit e6e78315e4761ad121a5eeb4fdffe3571d81ac17
8489 Author: Aliaksey Kandratsenka <[email protected]>
8490 Date: Sat Dec 20 19:04:43 2014 -0800
8491
8492 cpuprofiler: better explain deduplication of top stacktrace entry
8493
8494 M src/profiler.cc
8495
8496 commit 24b8ec28464712bc124af5655ebf877fb3f79032
8497 Author: Aliaksey Kandratsenka <[email protected]>
8498 Date: Sat Dec 20 17:10:38 2014 -0800
8499
8500 cpuprofiler: disable capturing stacktrace from signal's ucontext
8501
8502 This was reported to cause problems due to libunwind occasionally
8503 returning top level pc that is 1 smaller than real pc which causes
8504 problems.
8505
8506 M src/stacktrace_libunwind-inl.h
8507
8508 commit 83588de7204124c8d60703f169c2f3af8aa0ce5f
8509 Author: Aliaksey Kandratsenka <[email protected]>
8510 Date: Sat Dec 20 17:04:10 2014 -0800
8511
8512 pprof: added support for dumping stacks in --text mode
8513
8514 Which is very useful for diagnosing stack capturing and processing
8515 bugs.
8516
8517 M src/pprof
8518
8519 commit 2f29c9b06220ce6ad75d5ab2b98e9f63ade79eea
8520 Author: Aliaksey Kandratsenka <[email protected]>
8521 Date: Sat Dec 20 17:03:36 2014 -0800
8522
8523 pprof: made --show-addresses work
8524
8525 M src/pprof
8526
8527 commit b8b027d09a53dd6292d2f7c45aec40198628a808
8528 Author: Raphael Moreira Zinsly <[email protected]>
8529 Date: Tue Dec 23 10:55:22 2014 -0200
8530
8531 Make PPC64 use 64K of internal page size for tcmalloc by default
8532
8533 This patch set the default tcmalloc internal page size to 64K when
8534 built on PPC.
8535
8536 M configure.ac
8537
8538 commit 3f55d874be8812aef9f0f567048188584962b4c1
8539 Author: Raphael Moreira Zinsly <[email protected]>
8540 Date: Tue Dec 23 10:29:49 2014 -0200
8541
8542 New configure flags to set the alignment and page size of tcmalloc
8543
8544 Added two new configure flags, --with-tcmalloc-pagesize and
8545 --with-tcmalloc-alignment, in order to set the tcmalloc internal page
8546 size and tcmalloc allocation alignment without the need of a compiler
8547 directive and to make the choice of the page size independent of the
8548 allocation alignment.
8549
8550 M INSTALL
8551 M configure.ac
8552 M src/common.h
8553
8554 commit 1035d5c18f64d114ac790b92a96f3b3a1a301eb9
8555 Author: Aliaksey Kandratsenka <[email protected]>
8556 Date: Sun Dec 21 19:52:34 2014 -0800
8557
8558 start building malloc_extension_c_test even with static linking
8559
8560 Comment in Makefile.am stating that it doesn't work with static
8561 linking is not accurate anymore.
8562
8563 M Makefile.am
8564
8565 commit d570a6391cf4c5a5570e22ada5cf0b324c7b8dfd
8566 Author: Aliaksey Kandratsenka <[email protected]>
8567 Date: Sun Dec 21 19:33:25 2014 -0800
8568
8569 unbreak malloc_extension_c_test on clang
8570
8571 Looks like even force_malloc trick was not enough to force clang to
8572 actually call malloc. I'm now calling tc_malloc directly to prevent
8573 that smartness.
8574
8575 M src/tests/malloc_extension_c_test.c
8576
8577 commit 4ace8dbbe2e04da1029a539c72b90dee1724c33f
8578 Author: Aliaksey Kandratsenka <[email protected]>
8579 Date: Sun Dec 21 18:49:47 2014 -0800
8580
8581 added subdir-objects automake options
8582
8583 This is suggested by automake itself regarding future-compat.
8584
8585 M .gitignore
8586 M Makefile.am
8587
8588 commit f72e37c3f99d942dd648e392abc65f294aa94fa8
8589 Author: Aliaksey Kandratsenka <[email protected]>
8590 Date: Sun Dec 21 18:27:03 2014 -0800
8591
8592 fixed C++ comment warning in malloc_extension_c.h from C compiler
8593
8594 M src/gperftools/malloc_extension_c.h
8595
8596 commit f94ff0cc0943c7b7bb9390f9d83a57c9d9e11fd0
8597 Author: Aliaksey Kandratsenka <[email protected]>
8598 Date: Sun Nov 9 14:41:36 2014 -0800
8599
8600 made AtomicOps_x86CPUFeatureStruct hidden
8601
8602 So that access to has_sse2 is faster under -fPIC.
8603
8604 M src/base/atomicops-internals-x86.h
8605
8606 commit 987a724c23df4b29eb457e96bb5c1bbab15dc90e
8607 Author: Aliaksey Kandratsenka <[email protected]>
8608 Date: Sun Nov 9 14:20:20 2014 -0800
8609
8610 dropped atopmicops workaround for irrelevant Opteron locking bug
8611
8612 It's not cheap at all when done in this way (i.e. without runtime
8613 patching) and apparently useless.
8614
8615 It looks like Linux kernel never got this workaround at all. See
8616 bugzilla ticket: https://bugzilla.kernel.org/show_bug.cgi?id=11305
8617
8618 And I see no traces of this workaround in glibc either.
8619
8620 On the other hand, opensolaris folks apparently still have it (or
8621 something similar, based on comments on linux bugzilla) in their code:
8622 https://github.com/illumos/illumos-gate/blob/32842aabdc7c6f8f0c6140a256cf42cf5404fefb/usr/src/uts/i86pc/os/mp_startup.c#L1136
8623
8624 And affected CPUs (if any) are from year 2008 (that's 6 years now).
8625
8626 Plus even if somebody still uses those cpus (which is unlikely), they
8627 won't have working kernel and glibc anyways.
8628
8629 M src/base/atomicops-internals-x86.cc
8630 M src/base/atomicops-internals-x86.h
8631
8632 commit 7da5bd014d77ddaf694054b1e3ae0a3ef92ab384
8633 Author: Aliaksey Kandratsenka <[email protected]>
8634 Date: Sat Dec 20 21:18:07 2014 -0800
8635
8636 enabled aggressive decommit by default
8637
8638 TCMALLOC_AGGRESSIVE_DECOMMIT=f is one way to disable it and
8639 SetNumericProperty is another.
8640
8641 M src/static_vars.cc
8642 M src/tests/tcmalloc_unittest.sh
8643
8644 commit 51b0ad55b3267caff6cd2d25815bfb913179b526
8645 Author: Aliaksey Kandratsenka <[email protected]>
8646 Date: Sun Nov 9 17:36:49 2014 -0800
8647
8648 added basic unit test for singular malloc hooks
8649
8650 M src/tests/malloc_extension_c_test.c
8651
8652 commit bce72dda078d2cb3e9745077f9903e642a966131
8653 Author: Aliaksey Kandratsenka <[email protected]>
8654 Date: Sun Nov 9 17:12:05 2014 -0800
8655
8656 inform compiler that tcmalloc allocation sampling is unlikely
8657
8658 Now compiler generates slightly better code which produces jump-less
8659 code for common case of not sampling allocations.
8660
8661 M src/tcmalloc.cc
8662
8663 commit 4f051fddcd7af53e1607b6a4866ffa461a0033ef
8664 Author: Aliaksey Kandratsenka <[email protected]>
8665 Date: Sat Mar 15 13:18:28 2014 -0700
8666
8667 eliminated CheckIfKernelSupportsTLS
8668
8669 We don't care about pre-2.6.0 kernels anymore. So we can assume that
8670 if compile time check worked, then at runtime it'll work.
8671
8672 M src/tcmalloc.cc
8673 M src/thread_cache.cc
8674 M src/thread_cache.h
8675 M src/windows/port.cc
8676
8677 commit 81291ac3992ec7500faf2026c72feb80c9634dc3
8678 Author: Aliaksey Kandratsenka <[email protected]>
8679 Date: Sun Nov 9 16:48:45 2014 -0800
8680
8681 set elf visibility to hidden for malloc hooks
8682
8683 To speed up access to them under -fPIC.
8684
8685 M src/malloc_hook-inl.h
8686
8687 commit 105c004d0c84137f32cc71b8d3f7899fcc8c2e72
8688 Author: Aliaksey Kandratsenka <[email protected]>
8689 Date: Sun Nov 9 14:40:30 2014 -0800
8690
8691 introduced ATTRIBUTE_VISIBILITY_HIDDEN
8692
8693 So that we can disable elf symbol interposition for certain
8694 perf-sensitive symbols.
8695
8696 M src/base/basictypes.h
8697
8698 commit 6a6c49e1f53df3d00a0661a86090534146ce686c
8699 Author: Aliaksey Kandratsenka <[email protected]>
8700 Date: Sun Nov 9 16:23:06 2014 -0800
8701
8702 replaced separate singular malloc hooks with faster HookList
8703
8704 Specifically, we can now check in one place if hooks are set at all,
8705 instead of two places. Which makes fast path shorter.
8706
8707 M src/malloc_hook-inl.h
8708 M src/malloc_hook.cc
8709
8710 commit ba0441785bae214566008e69adcd915800d9dbb3
8711 Author: Aliaksey Kandratsenka <[email protected]>
8712 Date: Sun Nov 9 16:13:54 2014 -0800
8713
8714 removed extra barriers in malloc hooks mutation methods
8715
8716 Because those are already done under spinlock and read-only and
8717 lockless Traverse is already tolerant to slight inconsistencies.
8718
8719 M src/malloc_hook.cc
8720
8721 commit 890f34c77ef79dfe1e00ce36a3f91aee7fe759b7
8722 Author: Aliaksey Kandratsenka <[email protected]>
8723 Date: Sun Nov 9 16:08:59 2014 -0800
8724
8725 introduced support for deprecated singular hooks into HookList
8726
8727 So that we can later drop separate singular hooks.
8728
8729 M src/malloc_hook-inl.h
8730 M src/malloc_hook.cc
8731
8732 commit 81ed7dff11de915b12c4111d403e52c81c786f82
8733 Author: Aliaksey Kandratsenka <[email protected]>
8734 Date: Sun Dec 7 13:33:40 2014 -0800
8735
8736 returned date of 2.3rc in NEWS back
8737
8738 M NEWS
8739
8740 commit 463a619408219fff8955d47a71de1aab31ebd129
8741 Author: Aliaksey Kandratsenka <[email protected]>
8742 Date: Sun Dec 7 12:53:35 2014 -0800
8743
8744 bumped version to 2.3
8745
8746 M configure.ac
8747 M src/windows/config.h
8748 M src/windows/gperftools/tcmalloc.h
8749
8750 commit 76e8138e12567e2ee1f638e8fcffadc2ca1de83d
8751 Author: Aliaksey Kandratsenka <[email protected]>
8752 Date: Sun Dec 7 12:46:49 2014 -0800
8753
8754 updated NEWS for gperftools 2.3
8755
8756 M NEWS
8757
8758 commit 8eb4ed785ae883acb6425fd980e9d3e6bdcab89d
8759 Author: Raphael Moreira Zinsly <[email protected]>
8760 Date: Thu Nov 27 14:11:09 2014 -0200
8761
8762 Added option to disable libunwind linking
8763
8764 This patch adds a configure option to enable or disable libunwind
8765 linking.
8766 The patch also disables libunwind on ppc by default.
8767
8768 M configure.ac
8769
8770 commit 3b94031d21fac39ce5be820f4f8bc37626ed4c08
8771 Author: Aliaksey Kandratsenka <[email protected]>
8772 Date: Thu Nov 27 11:48:08 2014 -0800
8773
8774 compile libunwind unwinder only of __thread is supported
8775
8776 This fixed build on certain OSX that I have access to.
8777
8778 M src/stacktrace.cc
8779
8780 commit 3ace468202a8647a16134d0bb42fa497fa3ec0d4
8781 Author: Aliaksey Kandratsenka <[email protected]>
8782 Date: Thu Nov 27 10:43:11 2014 -0800
8783
8784 issue-658: correctly close socketpair fds when socketpair fails
8785
8786 This applies patch by glider.
8787
8788 M src/symbolize.cc
8789
8790 commit e7d5e512b068fb76f5eeed7985bebd9c0d14e226
8791 Author: Aliaksey Kandratsenka <[email protected]>
8792 Date: Sun Nov 2 20:02:12 2014 -0800
8793
8794 bumped version to 2.3rc
8795
8796 M configure.ac
8797 M src/windows/config.h
8798 M src/windows/gperftools/tcmalloc.h
8799
8800 commit 1d44d378513a8782c058833118b664f38214810b
8801 Author: Aliaksey Kandratsenka <[email protected]>
8802 Date: Sun Nov 2 19:38:29 2014 -0800
8803
8804 updated NEWS for gperftools 2.3rc
8805
8806 M NEWS
8807
8808 commit 1108d83cf4a1692fce3b736e16d3e98c33329177
8809 Author: Aliaksey Kandratsenka <[email protected]>
8810 Date: Sun Sep 7 13:09:14 2014 -0700
8811
8812 implemented cpu-profiling mode that profiles threads separately
8813
8814 Default mode of operation of cpu profiler uses itimer and
8815 SIGPROF. This timer is by definition per-process and no spec defines
8816 which thread is going to receive SIGPROF. And it provides correct
8817 profiles only if we assume that probability of picking threads will be
8818 proportional to cpu time spent by threads.
8819
8820 It is easy to see, that recent Linux (at least on common SMP hardware)
8821 doesn't satisfy that assumption. Quite big skews of SIGPROF ticks
8822 between threads is visible. I.e. I could see as big as 70%/20%
8823 division instead of 50%/50% for pair of cpu-hog threads. (And I do see
8824 it become 50/50 with new mode)
8825
8826 Fortunately POSIX provides mechanism to track per-thread cpu time via
8827 posix timers facility. And even more fortunately, Linux also provides
8828 mechanism to deliver timer ticks to specific threads.
8829
8830 Interestingly, it looks like FreeBSD also has very similar facility
8831 and seems to suffer from same skew. But due to difference in a way
8832 how threads are identified, I haven't bothered to try to support this
8833 mode on FreeBSD.
8834
8835 This commit implements new profiling mode where every thread creates
8836 posix timer which tracks thread's cpu time. Threads also also set up
8837 signal delivery to itself on overflows of that timer.
8838
8839 This new mode requires every thread to be registered in cpu
8840 profiler. Existing ProfilerRegisterThread function is used for that.
8841
8842 Because registering threads requires application support (or suitable
8843 LD_PRELOAD-able wrapper for thread creation API), new mode is off by
8844 default. And it has to be manually activated by setting environment
8845 variable CPUPROFILE_PER_THREAD_TIMERS.
8846
8847 New mode also requires librt symbols to be available. Which we do not
8848 link to due to librt's dependency on libpthread. Which we avoid due
8849 to perf impact of bringing in libpthread to otherwise single-threaded
8850 programs. So it has to be either already loaded by profiling program
8851 or LD_PRELOAD-ed.
8852
8853 M Makefile.am
8854 M configure.ac
8855 M src/maybe_threads.cc
8856 M src/maybe_threads.h
8857 M src/profile-handler.cc
8858 M src/tests/profile-handler_unittest.cc
8859
8860 commit 714bd93e42535e759716324a90fbb395506499d2
8861 Author: Aliaksey Kandratsenka <[email protected]>
8862 Date: Sun Aug 10 20:28:11 2014 -0700
8863
8864 drop workaround for too old redhat 7
8865
8866 Note that this is _not_ RHEL7 but original redhat 7 from early 2000s.
8867
8868 M configure.ac
8869
8870 commit 8de46e66fcd2577758ab297b553bb0f468d8a97a
8871 Author: Aliaksey Kandratsenka <[email protected]>
8872 Date: Sun Aug 10 19:38:07 2014 -0700
8873
8874 don't add leaf function twice to profile under libunwind
8875
8876 M src/profiler.cc
8877
8878 commit 2e5ee0488996437aeef2028ad95d969b56abcad1
8879 Author: Aliaksey Kandratsenka <[email protected]>
8880 Date: Sun Aug 10 12:19:45 2014 -0700
8881
8882 pprof: indicate if using remote profile
8883
8884 Missing profile file is common source of confusion. So a bit more
8885 clarify is useful.
8886
8887 M src/pprof
8888
8889 commit 6efe96b41c9531f68f806faa0464445f884178ce
8890 Author: Aliaksey Kandratsenka <[email protected]>
8891 Date: Sat Oct 11 15:09:18 2014 -0700
8892
8893 issue-493: correctly detect __ARM_ARCH_6ZK__ for MemoryBarrier
8894
8895 Which should fix issue reported by user pedronavf
8896
8897 M src/base/atomicops-internals-arm-v6plus.h
8898
8899 commit 8e97626378d5c4151a480dea1964e25969c9311d
8900 Author: Aliaksey Kandratsenka <[email protected]>
8901 Date: Sun Nov 2 11:28:30 2014 -0800
8902
8903 issue-655: use safe getenv for aggressive decommit mode flag
8904
8905 Because otherwise we risk deadlock due to too early use of getenv on
8906 windows.
8907
8908 M src/base/commandlineflags.h
8909 M src/static_vars.cc
8910
8911 commit 8c3dc52fcfe02412a529769a22cbc75388a5d368
8912 Author: Aliaksey Kandratsenka <[email protected]>
8913 Date: Sat Oct 18 16:35:57 2014 -0700
8914
8915 issue-654: [pprof] handle split text segments
8916
8917 This applies patch by user simonb.
8918
8919 Quoting:
8920
8921 Relocation packing splits a single executable load segment into two.
8922 Before:
8923
8924 LOAD 0x000000 0x00000000 0x00000000 0x2034d28 0x2034d28
8925 R E 0x1000
8926 LOAD 0x2035888 0x02036888 0x02036888 0x182d38 0x1a67d0
8927 RW 0x1000
8928
8929 After:
8930 LOAD 0x000000 0x00000000 0x00000000 0x14648 0x14648 R
8931 E 0x1000
8932 LOAD 0x014648 0x0020c648 0x0020c648 0x1e286e0 0x1e286e0
8933 R E 0x1000
8934 ...
8935 LOAD 0x1e3d888 0x02036888 0x02036888 0x182d38 0x1a67d0
8936 RW 0x1000
8937
8938 The .text section is in the second LOAD, and this is not at
8939 offset/address zero. The result is that this library shows up in
8940 /proc/self/maps as multiple executable entries, for example (note:
8941 this trace is not from the library dissected above, but rather from an
8942 earlier version of it):
8943
8944 73b0c000-73b21000 r-xp 00000000 b3:19 786460
8945 /data/.../libchrome.2160.0.so
8946 73b21000-73d12000 ---p 00000000 00:00 0
8947 73d12000-75a90000 r-xp 00014000 b3:19 786460
8948 /data/.../libchrome.2160.0.so
8949 75a90000-75c0d000 rw-p 01d91000 b3:19 786460
8950 /data/.../libchrome.2160.0.so
8951
8952 When parsing this, pprof needs to merge the two r-xp entries above
8953 into a single entry, otherwise the addresses it prints are incorrect.
8954
8955 The following fix against 2.2.1 was sufficient to make pprof --text
8956 print the correct output. Untested with other pprof options.
8957
8958 M src/pprof
8959
8960 commit 44c61ce6c4c713b194330641f400bbf64fd2abec
8961 Author: Ricardo M. Correia <[email protected]>
8962 Date: Wed Oct 8 04:39:14 2014 +0200
8963
8964 Fix parsing /proc/pid/maps dump in CPU profile data file
8965
8966 When trying to use pprof on my machine, the symbols of my program were
8967 not being recognized.
8968
8969 It turned out that pprof, when calculating the offset of the text list
8970 of mapped objects (the last section of the CPU profile data file), was
8971 assuming that the slot size was always 4 bytes, even on 64-bit
8972 machines.
8973
8974 This led to ParseLibraries() reading a lot of garbage data at the
8975 beginning of the map, and consequently the regex was failing to
8976 match on
8977 the first line of the real (non-garbage) map.
8978
8979 M src/pprof
8980
8981 commit 2a28ef24ddf8013bff59914b10902f1fb07bf9b2
8982 Author: Aliaksey Kandratsenka <[email protected]>
8983 Date: Sat Sep 6 16:49:24 2014 -0700
8984
8985 Added remaining memory allocated info to 'Exiting' dump message
8986
8987 This applies patch by user yurivict.
8988
8989 M src/heap-profiler.cc
8990
8991 commit bbf346a856d4a7c5c2ab0e65d7cccf3dc1f23f13
8992 Author: Adam McNeeney <[email protected]>
8993 Date: Fri Aug 22 10:01:24 2014 +0100
8994
8995 Cope with new addr2line outputs for DWARF4
8996
8997 Copes with ? for line number (converts to 0).
8998 Copes with (discriminator <num>) suffixes to file/linenum (removes).
8999
9000 Change-Id: I96207165e4852c71d3512157864f12d101cdf44a
9001
9002 M src/pprof
9003
9004 commit b08d760958dbacc8da822f7f3db76d58c95a114d
9005 Author: Aliaksey Kandratsenka <[email protected]>
9006 Date: Sat Aug 23 14:45:08 2014 -0700
9007
9008 issue-641: Added --show_addresses option
9009
9010 This applies patch by user yurivict.
9011
9012 M src/pprof
9013
9014 commit 3c326d9f200a527bba45f1d222aa6aff20d42bc3
9015 Author: Aliaksey Kandratsenka <[email protected]>
9016 Date: Tue Aug 19 08:14:08 2014 -0700
9017
9018 issue-644: fix possible out-of-bounds access in GetenvBeforeMain
9019
9020 As suggested by user Ivan L.
9021
9022 M src/base/sysinfo.cc
9023
9024 commit f1ae3c446f34f6e8ac901caba281051ec269e4df
9025 Author: jiakai <[email protected]>
9026 Date: Mon Jul 28 11:28:03 2014 -0700
9027
9028 Add an option to allow disabling stripping template argument in pprof
9029
9030 M src/pprof
9031
9032 commit a12890df2519d254d1c497b8e0a65bb8fc9e1ab2
9033 Author: Aliaksey Kandratsenka <[email protected]>
9034 Date: Sat Jul 26 14:12:42 2014 -0700
9035
9036 issue-635: allow whitespace in libraries paths
9037
9038 This applies change suggested by user [email protected]
9039
9040 M src/pprof
9041
9042 commit d5e36788d8bc626403dc020a86213cfc740ee73d
9043 Author: Aliaksey Kandratsenka <[email protected]>
9044 Date: Sat Jul 26 13:59:22 2014 -0700
9045
9046 issue-636: fix prof/web command on Windows/MinGW
9047
9048 This applies patch sent by user chaishushan.
9049
9050 M src/pprof
9051
9052 commit 4b788656bb9c480640d917d27d8a94a5eae436f5
9053 Author: Michael Pasieka <[email protected]>
9054 Date: Sun Jul 13 18:09:35 2014 -0700
9055
9056 added option to display stack traces in output for heap checker
9057
9058 Quoting from email:
9059
9060 I had the same question as William posted to stack overflow back on
9061 Dec 9,2013: How to display symbols in stack trace of google-perftools
9062 heap profiler (*). I dug into the source and realized the
9063 functionality was not there but could be added. I am hoping that
9064 someone else will find this useful/helpful.
9065
9066 The patch I created will not attach so I am adding below.
9067
9068 Enjoy!
9069
9070 -- Michael
9071
9072 *
9073 http://stackoverflow.com/questions/20476918/how-to-display-symbols-in-stack-trace-of-google-perftools-heap-profiler
9074
9075 M src/pprof
9076
9077 commit 3abb5cb819bafe7004363f041c194afd827cb053
9078 Author: WenSheng He <[email protected]>
9079 Date: Mon Jun 30 14:39:13 2014 +0800
9080
9081 issue-630: The env var should be "CPUPROFILE"
9082
9083 To enable cpu profile, the env var should be "CPUPROFILE", not
9084 "PROFILE"
9085 actually.
9086
9087 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
9088
9089 M src/gperftools/profiler.h
9090
9091 commit fd81ec257883c6d5486e4568c955dff86dbed5c8
9092 Author: Aliaksey Kandratsenka <[email protected]>
9093 Date: Sat Jun 28 13:05:12 2014 -0700
9094
9095 issue-631: fixed miscompilation of debugallocation without mmap
9096
9097 This applies patch sent by user iamxujian.
9098
9099 Clearly, when I updated debugallocation to fix issue-464 I've broken
9100 no-mmap path by forgetting closing brace.
9101
9102 M src/debugallocation.cc
9103
9104 commit 2e90b6fd72fec33aedf547d1977bdee6b77645b9
9105 Author: Aliaksey Kandratsenka <[email protected]>
9106 Date: Sat Jun 21 15:45:44 2014 -0700
9107
9108 bumped version to 2.2.1
9109
9110 M configure.ac
9111 M src/windows/config.h
9112 M src/windows/gperftools/tcmalloc.h
9113
9114 commit 577b940cc0a0ef207115d071a81127f93f02c083
9115 Author: Aliaksey Kandratsenka <[email protected]>
9116 Date: Sat Jun 21 15:39:46 2014 -0700
9117
9118 updated NEWS for 2.2.1
9119
9120 M NEWS
9121
9122 commit 2fe4b329ad4d3f242b0bd73295375e70be79187b
9123 Author: Aliaksey Kandratsenka <[email protected]>
9124 Date: Sun May 18 10:59:06 2014 -0700
9125
9126 applied chromium patch fixing some build issue on android
9127
9128 This applies patch from: https://codereview.chromium.org/284843002/ by
9129 [email protected]
9130
9131 M src/libc_override_gcc_and_weak.h
9132
9133 commit c009398e3239be8ae6185ad83685caf7458c49bb
9134 Author: Aliaksey Kandratsenka <[email protected]>
9135 Date: Sun Jun 15 12:53:14 2014 -0700
9136
9137 issue-628:package missing stacktrace_powerpc-{linux,darwin}-inl.h
9138
9139 This headers were missing in .tar.gz because they were not mentioned
9140 anywhere in Makefile.am.
9141
9142 M Makefile.am
9143
9144 commit 81d99f21ede78ab8d5fec15d0055416ac1b581f3
9145 Author: Adhemerval Zanella <[email protected]>
9146 Date: Tue Jun 3 07:50:56 2014 -0500
9147
9148 issue-626: Fix SetupAggressiveDecommit initialization
9149
9150 This patch fixes the SetupAggressiveDecommit initialization to
9151 run after
9152 pageheap_ creation. Current code it not enforcing it, since
9153 InitStaticVars is being called outside the static_vars module.
9154
9155 M src/static_vars.cc
9156
9157 commit 846b775dfadb77901202ae7ddbac30ad1de7df01
9158 Author: Aliaksey Kandratsenka <[email protected]>
9159 Date: Sat May 3 17:50:11 2014 -0700
9160
9161 bumped version to 2.2
9162
9163 M configure.ac
9164 M src/windows/config.h
9165 M src/windows/gperftools/tcmalloc.h
9166
9167 commit cdf8e1e932016bd5f1737e1f10bce07226228d15
9168 Author: Aliaksey Kandratsenka <[email protected]>
9169 Date: Sat May 3 17:44:38 2014 -0700
9170
9171 updated NEWS for 2.2
9172
9173 M NEWS
9174
9175 commit 0807476f56b9dc024375467be0f0b07fae3ee7fb
9176 Author: Aliaksey Kandratsenka <[email protected]>
9177 Date: Sat May 3 17:38:14 2014 -0700
9178
9179 issue-620: windows dll patching: fixed delete of old stub code
9180
9181 After code for issue 359 was applied PreamblePatcher started using
9182 it's own code to manage memory of stub code fragments. It's not using
9183 new[] anymore. And it automatically frees stub code memory on
9184 Unpatch.
9185
9186 Clearly, author of that code forgot to remote that no more needed
9187 delete call. With that delete call we end up trying to free memory
9188 that was never allocated with any of known allocators and crash.
9189
9190 M src/windows/patch_functions.cc
9191
9192 commit facd7e83b341e069875b7c1cd7774cf671d932b5
9193 Author: Aliaksey Kandratsenka <[email protected]>
9194 Date: Sat Apr 19 11:08:51 2014 -0700
9195
9196 bumped version to 2.1.90
9197
9198 M configure.ac
9199 M src/windows/config.h
9200 M src/windows/gperftools/tcmalloc.h
9201
9202 commit e8e082af25952e9bd10c37c3f028c91fcbbd1cbd
9203 Author: Aliaksey Kandratsenka <[email protected]>
9204 Date: Sat Apr 19 13:16:02 2014 -0700
9205
9206 updated NEWS for 2.2rc
9207
9208 M NEWS
9209
9210 commit 802fdb739e7aabcb15443030842a2137a5559338
9211 Author: Aliaksey Kandratsenka <[email protected]>
9212 Date: Sat Apr 12 18:07:11 2014 -0700
9213
9214 issue-610: use TCMallocGetenvSafe from inside malloc
9215
9216 Instead of plain getenv. So that windows getenv implementation that
9217 may call malloc does not deadlock.
9218
9219 M src/common.cc
9220 M src/thread_cache.cc
9221
9222 commit 6b83516adefcf0806825f6dba2eb2232615d744b
9223 Author: Aliaksey Kandratsenka <[email protected]>
9224 Date: Sat Apr 12 18:06:32 2014 -0700
9225
9226 issue-610: made dynamic_annotations.c use TCMallocGetenvSafe
9227
9228 M src/base/dynamic_annotations.c
9229
9230 commit aeef3b44201ba663c19ca1c97792b724346e84a9
9231 Author: Aliaksey Kandratsenka <[email protected]>
9232 Date: Sat Apr 12 18:05:59 2014 -0700
9233
9234 issue-610: introduced TCMallocGetenvSafe
9235
9236 This is version of GetenvBeforeMain that's available to C code.
9237
9238 M Makefile.am
9239 M src/base/sysinfo.cc
9240 A src/getenv_safe.h
9241
9242 commit 125e5ed58b72550e6b74e0139c2c3dd5e262a6ec
9243 Author: Aliaksey Kandratsenka <[email protected]>
9244 Date: Sat Apr 12 12:38:19 2014 -0700
9245
9246 don't enable backtrace() for stacktrace capturing by default
9247
9248 Because we don't yet have a treatment for deadlocks that are caused by
9249 (recursive) use of malloc from within that facility.
9250
9251 M configure.ac
9252
9253 commit 75b65f96b0bb44d70e0a461a03a3f8c928390283
9254 Author: Raphael Moreira Zinsly <[email protected]>
9255 Date: Tue Apr 8 17:45:13 2014 -0300
9256
9257 PowerPC: stacktrace function refactor and fixes
9258
9259 This patch fixes the stacktrace creating when the function is
9260 interrupted by a signal. For Linux, the vDSO signal trampoline
9261 symbol is
9262 compared against LR from stack backchain and handled different in that
9263 case (since the signal trampoline layout a different stack frame).
9264
9265 Because of this extensive change the PowerPC stacktrace code has now
9266 been refactored to split in Linux and Darwin specific codes.
9267
9268 M src/stacktrace.cc
9269 A src/stacktrace_powerpc-darwin-inl.h
9270 A src/stacktrace_powerpc-linux-inl.h
9271
9272 commit 8deea9ff2a3e6eef8af64ea0727d6cb840c13769
9273 Author: Raphael Moreira Zinsly <[email protected]>
9274 Date: Tue Apr 8 17:44:08 2014 -0300
9275
9276 VDSOsupport cleanup
9277
9278 This patch cleans up unused VDSO getcpu racking from VDSOsupport
9279 class,
9280 since the code is not used anywhere in gperftools and symbol name
9281 is not
9282 architecture independent.
9283
9284 M src/base/vdso_support.cc
9285 M src/base/vdso_support.h
9286
9287 commit 9d5e1a0aa5a6ad1c7af18b65016fa5c9b7fccb47
9288 Author: Raphael Moreira Zinsly <[email protected]>
9289 Date: Tue Apr 8 17:43:11 2014 -0300
9290
9291 Fixed issues with heap checker on PPC64 LE.
9292
9293 Fixed the wrapper for the syscall sys_clone and the test for heap
9294 checker on PPC64 LE. Both use the ODP structure, which is only
9295 used on BE architectures.
9296
9297 M src/base/linux_syscall_support.h
9298 M src/tests/heap-checker_unittest.cc
9299
9300 commit 49237462c8ae7920332c1034c623e57b50a6109c
9301 Author: Raphael Moreira Zinsly <[email protected]>
9302 Date: Tue Apr 8 17:42:17 2014 -0300
9303
9304 Fixed the way that pprof packed profile data in BE.
9305
9306 pprof was writing profile data in a way that only works for
9307 little-endian
9308 files, this patch verifies if the system is big-endian and writes
9309 packed
9310 data correctly.
9311
9312 M src/pprof
9313
9314 commit a1ae66ef110bd87ff97903e86fd84c745db24646
9315 Author: Raphael Moreira Zinsly <[email protected]>
9316 Date: Tue Apr 8 17:37:29 2014 -0300
9317
9318 Fixed the use of addr2line to discover the separator symbol.
9319
9320 In systems where addr2line has a version greater than 2.22 pprof fails
9321 in discover the separator symbol (_fini). This patch identifies if
9322 addr2line can find the symbol, otherwise pprof uses objdump to recover
9323 a address that addr2line's newer versions can recognize as the
9324 separator
9325 function.
9326
9327 M src/pprof
9328
9329 commit 8b2e5ee831760a94bc407dc255a522eda242b04f
9330 Author: Aliaksey Kandratsenka <[email protected]>
9331 Date: Tue Apr 1 22:29:11 2014 -0700
9332
9333 issue-614: use tc_memalign in ReallocAfterMemalloc test
9334
9335 Because some OSes lack plain memalign. And we really need to test our
9336 implementation which is always available via tc_malloc.
9337
9338 M configure.ac
9339 M src/tests/debugallocation_test.cc
9340
9341 commit 0399af1019240e2d9127a588ddc8e31ff4656df0
9342 Author: Aliaksey Kandratsenka <[email protected]>
9343 Date: Tue Apr 1 21:30:27 2014 -0700
9344
9345 added tc_malloc_skip_new_handler
9346
9347 This is port of corresponding chromium change at:
9348 https://codereview.chromium.org/55333002/
9349
9350 Basic idea is that sometimes apps that use tc_set_new_mode in order to
9351 have C++ out-of-memory handler catch OOMs in malloc, need to invoke
9352 usual malloc that returns 0 on OOM.
9353
9354 That new API is exactly for that. It'll always return NULL on OOM even
9355 if tc_new_mode is set to true.
9356
9357 M src/debugallocation.cc
9358 M src/gperftools/tcmalloc.h.in
9359 M src/tcmalloc.cc
9360 M src/tests/tcmalloc_unittest.cc
9361 M src/windows/gperftools/tcmalloc.h
9362 M src/windows/gperftools/tcmalloc.h.in
9363
9364 commit d77317247e1c8de1ea4e0419318b26f636e64431
9365 Author: Aliaksey Kandratsenka <[email protected]>
9366 Date: Sat Mar 1 12:13:02 2014 -0800
9367
9368 issue deprecation warning on use of google/ headers
9369
9370 M src/google/heap-checker.h
9371 M src/google/heap-profiler.h
9372 M src/google/malloc_extension.h
9373 M src/google/malloc_extension_c.h
9374 M src/google/malloc_hook.h
9375 M src/google/malloc_hook_c.h
9376 M src/google/profiler.h
9377 M src/google/stacktrace.h
9378 M src/google/tcmalloc.h
9379
9380 commit e7297f0c14b0c09ec52707ed7477c7aee43a1882
9381 Author: Aliaksey Kandratsenka <[email protected]>
9382 Date: Sat Mar 15 12:44:11 2014 -0700
9383
9384 speed up MallocExtension::instance()
9385
9386 It was reported that pthread_once is expensive, especially on ppc.
9387
9388 In new implementation in hot path instead of doing potentially
9389 expensive atomic read with barrier, we do just plain read.
9390
9391 It's slightly less robust than older implementation, but it should be
9392 faster.
9393
9394 New code is making assumption that programs do not spawn threads
9395 before main() is called. And therefore all variables & modules are
9396 initialized before threads are created. Which looks like pretty safe
9397 assumption. With that assumption, doing plain read is safe, because
9398 current_instance is initialized as part of module init and therefore
9399 before threads are spawned.
9400
9401 This patch is based on feedback of Adhemerval Zanella.
9402
9403 M src/malloc_extension.cc
9404
9405 commit df227794361f9725eca7420d95e65ab997c79716
9406 Author: Adhemerval Zanella <[email protected]>
9407 Date: Sun Feb 2 19:32:37 2014 -0200
9408
9409 Fix getpc_test for PPC64v2 LE
9410
9411 This patch fix the PPC64 guard to get the function address for
9412 PPC64v2.
9413 It removes the use of an indirection (to get the ODP text address),
9414 since the PPCv2 does not have function descriptors.
9415
9416 M src/tests/getpc_test.cc
9417
9418 commit e3deb42d5dcdeeb8a340d03f207f2488ef3fb2eb
9419 Author: Aliaksey Kandratsenka <[email protected]>
9420 Date: Sat Mar 29 13:28:01 2014 -0700
9421
9422 issue-613: remove friend declaration from HeapLeakChecker
9423
9424 This applies patch by [email protected]:
9425
9426 heap-checker.h contains the following friend declaration of main:
9427 friend int main(int, char**).
9428
9429 C99 allows another declaration of main, i.e. int main(int, char**,
9430 char**), and if code uses it and includes the heap-checker header,
9431 this might result in a conflict, e.g.
9432
9433 error: declaration of C function 'int main(int, char**, char**)'
9434 conflicts with
9435 int main(int argc, char* argv[], char** envp)
9436
9437 Actually the comment above the friend declaration of main() mentions
9438 that this is required to get the unittest working and for other
9439 internal usage, but I'm not completely sure if this is true as long as
9440 I'm able to build and run the unittest removing the declaration.
9441
9442 M src/gperftools/heap-checker.h
9443
9444 commit 1a28754656bd6dfc1297d62a4c5811b0a5f03180
9445 Author: Aliaksey Kandratsenka <[email protected]>
9446 Date: Sat Mar 29 13:23:00 2014 -0700
9447
9448 issue-612: added missing include for std::min
9449
9450 Otherwise Visual Studio 2013 rightfully complains
9451
9452 M src/windows/system-alloc.cc
9453
9454 commit fe566314267fc3f67c5068d0afe004b1733ff19a
9455 Author: Aliaksey Kandratsenka <[email protected]>
9456 Date: Sat Mar 1 12:38:08 2014 -0800
9457
9458 unbreak building with libunwind
9459
9460 Caused by premature merging of previous patch.
9461
9462 When we're searching for backtrace in libexecinfo and don't find it,
9463 we should not reset UNWIND_LIBS to empty value.
9464
9465 Correct fix is to first search for backtrace in libunwind and then to
9466 search for it in libexecinfo.
9467
9468 M configure.ac
9469
9470 commit 91179961490a601f2c611889b3075e995fa5437e
9471 Author: Thomas Klausner <[email protected]>
9472 Date: Tue Feb 25 21:41:28 2014 +0100
9473
9474 Look for backtrace() in libexecinfo as well.
9475
9476 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
9477
9478 M configure.ac
9479
9480 commit fd3379a213b2a99a72a7544b6208056daf15d0d2
9481 Author: Thomas Klausner <[email protected]>
9482 Date: Tue Feb 25 21:41:07 2014 +0100
9483
9484 Fix unportable test(1) construct.
9485
9486 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
9487
9488 M configure.ac
9489
9490 commit a7223c2a14431ffd783117c55f84a2c84423d313
9491 Author: Thomas Klausner <[email protected]>
9492 Date: Tue Feb 25 21:40:11 2014 +0100
9493
9494 Test for memalign in configure.ac. Disable a test that uses memalign
9495 if it is not found.
9496
9497 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
9498
9499 M configure.ac
9500 M src/tests/debugallocation_test.cc
9501
9502 commit bd9665ebbe1aca5e1ba993de4214e64ee8bbe43a
9503 Author: Aliaksey Kandratsenka <[email protected]>
9504 Date: Sat Feb 22 13:46:42 2014 -0800
9505
9506 issue-489: added tcmalloc test pass with chromium-style decommit
9507
9508 M src/tests/tcmalloc_unittest.sh
9509
9510 commit 6a000d6dd5968ac29f8fc43f7dfc736338e11781
9511 Author: Aliaksey Kandratsenka <[email protected]>
9512 Date: Sat Feb 22 13:46:11 2014 -0800
9513
9514 issue-489: added unit test for chromium-style decommitting
9515
9516 M src/tests/tcmalloc_unittest.cc
9517
9518 commit eb2d69014cb3e163f2ab3ed676fbedf5c3a97caa
9519 Author: Aliaksey Kandratsenka <[email protected]>
9520 Date: Sat Feb 22 13:10:08 2014 -0800
9521
9522 issue-489: made tests pass on enabled chromium-style decommitting
9523
9524 M src/tests/tcmalloc_unittest.cc
9525
9526 commit a92fc76f72318f7a46e91d9ef6dd24f2bcf44802
9527 Author: Aliaksey Kandratsenka <[email protected]>
9528 Date: Sat Feb 22 13:09:05 2014 -0800
9529
9530 issue-489: enable chromium-style decommitting on env variable
9531
9532 TCMALLOC_AGGRESSIVE_DECOMMIT=t now enables aggressive decommitting by
9533 default.
9534
9535 M src/static_vars.cc
9536
9537 commit c7ce50cd04ea08bd20d4ea4b2924e6a4451d2565
9538 Author: Aliaksey Kandratsenka <[email protected]>
9539 Date: Sat Nov 9 17:35:24 2013 -0800
9540
9541 issue-489: implemented API to set chromium-style de-committing
9542
9543 Chrome has code to decommit (release back to OS) every span that's
9544 released. I don't want to make it default, but indeed some
9545 applications may want to enable this mode.
9546
9547 The code itself is taken from 2-way-merging of code from Chromium
9548 fork.
9549
9550 M src/page_heap.cc
9551 M src/page_heap.h
9552 M src/tcmalloc.cc
9553
9554 commit 1d707cd4a3dfe6f238a530f945291acfd5995042
9555 Author: Aliaksey Kandratsenka <[email protected]>
9556 Date: Sat Nov 9 17:19:34 2013 -0800
9557
9558 issue-489: fixed warning
9559
9560 Computing certain values just for ASSERT raises just warning from
9561 compiler because if NDEBUG is set those are dead code.
9562
9563 M src/page_heap.cc
9564
9565 commit 91bffcbad60d84beebe8b69a1db6c85c10fc04bf
9566 Author: Aliaksey Kandratsenka <[email protected]>
9567 Date: Sat Nov 9 16:22:35 2013 -0800
9568
9569 issue-489: ported chromium windows decommitting code
9570
9571 I tried to do it cleanly with merges but chromium code has so many
9572 relevant commits (with frequent reverts) that makes it near
9573 impossible.
9574
9575 Simpler 2-way emerge-files worked in the end. I've removed chromium's
9576 aggressive 'always decommit' behavior which I want to make optional
9577 later.
9578
9579 Majority of this work is the following commits (but there are more,
9580 particularly against port.cc):
9581
9582 commit 9c92338c5f8770c440799d24387c3733fd6d826b
9583 Author: [email protected]
9584 <[email protected]@0039d316-1c4b-4281-b951-d872f2087c98>
9585 Date: Tue Oct 6 18:33:31 2009 +0000
9586
9587 Tracks the amount of committed vs uncommitted memory in tcmalloc's
9588 page heap's freelists
9589
9590 Keeps track of the number of reserved but not committed pages in
9591 the freelist and uses that to calculate a waste metric, which is
9592 the ratio of committed pages vs pages used by the application.
9593 This is exposed in the GetStats() call (which is used for
9594 about:tcmalloc) and through GetNumericalProperty() in Malloc
9595
9596 BUG=none
9597 TEST=open about:tcmalloc and monitor 'WASTE' columns while using
9598 the browser
9599
9600 Review URL: http://codereview.chromium.org/251065
9601
9602 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@28133
9603 0039d316-1c4b-4281-b951-d872f2087c98
9604
9605 commit aef4f1be3eec2059a7c6e2c106050a5f3d6ccf12
9606 Author: [email protected]
9607 <[email protected]@0039d316-1c4b-4281-b951-d872f2087c98>
9608 Date: Mon Oct 5 17:58:51 2009 +0000
9609
9610 Revert further back to MBelshe's baseline forking TCMalloc
9611
9612 This changes to decommitting in all paths through the
9613 page_heap delete method (which adds spans to the free lists).
9614
9615 r=mbelshe,jamesr
9616 Review URL: http://codereview.chromium.org/255067
9617
9618 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@28006
9619 0039d316-1c4b-4281-b951-d872f2087c98
9620
9621 commit e94afbb913b95f512cb8745a2729c73f82b15ae7
9622 Author: [email protected]
9623 <[email protected]@0039d316-1c4b-4281-b951-d872f2087c98>
9624 Date: Thu Oct 1 00:25:41 2009 +0000
9625
9626 Rollback Scavenge implemetation and rely on existing functionality
9627 to free
9628
9629 This is a landing of a patch provided by antonm. See:
9630 http://codereview.chromium.org/235022
9631
9632 Also included change to browser_about_handler.cc to fix build,
9633 and I set
9634 TCMALLOC_RELEASE_RATE to 1.0 on line 40 of page_heap.cc (I
9635 think this
9636 was an inadvertent rollback element).
9637
9638 r=antonm
9639 Review URL: http://codereview.chromium.org/257009
9640
9641 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@27692
9642 0039d316-1c4b-4281-b951-d872f2087c98
9643
9644 commit c585892d2c42a47c95d06a684a6685156c545403
9645 Author: [email protected]
9646 <[email protected]@0039d316-1c4b-4281-b951-d872f2087c98>
9647 Date: Wed Sep 2 17:33:23 2009 +0000
9648
9649 Landing for Anton Muhin's tcmalloc patch:
9650 http://codereview.chromium.org/180021/show
9651
9652 Restore decommitting in IncrementalScavenge and draft Scavenge
9653 method to
9654 be invoked periodically
9655 to reduce amount of committed pages.
9656
9657 BUG=none
9658 TEST=none
9659
9660 Review URL: http://codereview.chromium.org/187008
9661
9662 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@25188
9663 0039d316-1c4b-4281-b951-d872f2087c98
9664
9665 commit 14239acc00731e94736ac62e80fc6b17c31ea131
9666 Author: [email protected]
9667 <[email protected]@0039d316-1c4b-4281-b951-d872f2087c98>
9668 Date: Wed Aug 12 02:17:14 2009 +0000
9669
9670 Major changes to the Chrome allocator.
9671
9672 Changes include:
9673 * Fix tcmalloc to release memory. Implements the
9674 TCMalloc_SystemCommit()
9675 mechanism so that tcmalloc can implement SystemRelease()
9676 and later
9677 reuse that memory.
9678 * Enable dynamic switching of allocators based on an environment
9679 variable.
9680 Users can now switch between tcmalloc, jemalloc, the default
9681 windows
9682 heap, and the windows low-fragmentation heap.
9683 * Implements set_new_mode() across all allocators so that we
9684 can be sure
9685 that out-of-memory conditions are handled safely.
9686
9687 BUG=18345
9688 TEST=none; plan to get all unit tests running through these
9689 allocators.
9690
9691 Review URL: http://codereview.chromium.org/165275
9692
9693 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@23140
9694 0039d316-1c4b-4281-b951-d872f2087c98
9695
9696 M src/page_heap.cc
9697 M src/page_heap.h
9698 M src/system-alloc.cc
9699 M src/system-alloc.h
9700 M src/windows/system-alloc.cc
9701
9702 commit 7be2edfe7f09e7e8c123e958742815784a718880
9703 Author: Aliaksey Kandratsenka <[email protected]>
9704 Date: Sat Feb 22 12:14:11 2014 -0800
9705
9706 issue-525: only warn if --enable-frame-pointers are not given
9707
9708 This fixes issue when frame pointers warning is given even if frame
9709 pointers are enabled
9710
9711 M configure.ac
9712
9713 commit 7e24b6ca2aa2c15a24504c02fad09f47e1ece8b5
9714 Author: Aliaksey Kandratsenka <[email protected]>
9715 Date: Sun Feb 16 17:04:26 2014 -0800
9716
9717 added debugallocation check for offset_ corruption
9718
9719 It was previously possible (although unlikely) for damaged offset_
9720 field
9721 to lead FromRawPointer implementation into different MallocBlock.
9722
9723 As is usual with any damage, it's best to catch errors at earliest
9724 possible time.
9725
9726 M src/debugallocation.cc
9727
9728 commit 6dcd73f1eb5ce2cc44ab918c53cd42c472f44c52
9729 Author: Aliaksey Kandratsenka <[email protected]>
9730 Date: Sun Feb 16 16:59:43 2014 -0800
9731
9732 avoid crash in DebugMallocImplementation::GetOwnership
9733
9734 It was possible that if GetOwnership is passed pointer to memory not
9735 owned by tcmalloc, it would crash. Or incorrectly return
9736 owned. I.e. due to indirection in FromRawPointer.
9737
9738 New implementation prevents that, but introduces different bug
9739 instead. New implementation incorrectly returns "not owned" for
9740 memalign chunks with big alignment. But in can be argued that passing
9741 pointer returned from different memalign implementation did not work
9742 previously too.
9743
9744 M src/debugallocation.cc
9745
9746 commit 33280ffb71fc0e4eb75e455d53824c344d011e35
9747 Author: Aliaksey Kandratsenka <[email protected]>
9748 Date: Sun Feb 16 16:25:14 2014 -0800
9749
9750 removed unused "using" in malloc_extension_test.cc
9751
9752 M src/tests/malloc_extension_test.cc
9753
9754 commit 066e524d6e33e9e3364bb6819f98a02b347c14ef
9755 Author: Aliaksey Kandratsenka <[email protected]>
9756 Date: Sun Feb 16 14:58:35 2014 -0800
9757
9758 eliminated useless BASE_XXX defines in debugallocation.cc
9759
9760 And closed TODO entry for that.
9761
9762 M src/debugallocation.cc
9763
9764 commit a2375a1f360c0451ec2a2b852ea26a71fb731727
9765 Author: Aliaksey Kandratsenka <[email protected]>
9766 Date: Sun Feb 16 14:49:57 2014 -0800
9767
9768 issue-464: correctly handle realloc after memalign in debugalloc
9769
9770 debug memalign is creating special header block to allow us to find
9771 real allocated block. And previous implementation of data copying
9772 wasn't taking that into account and was copying that "alignment
9773 header" into newly allocated block.
9774
9775 M src/debugallocation.cc
9776 M src/tests/debugallocation_test.cc
9777
9778 commit d31f522f0e1b0e87ee3d5941e331ba745a0f5c23
9779 Author: Riku Voipio <[email protected]>
9780 Date: Mon Feb 3 16:31:32 2014 +0200
9781
9782 Add aarch64 defines
9783
9784 With atomic operations and system call support in place, enable
9785 with __aarch64__ defines Aarch64 support in other files around
9786 the google-perftools header files. After these, google-perftools
9787 testsuite (make check) results:
9788
9789 8 of 46 tests failed.
9790
9791 FAIL: sampling_test.sh
9792 FAIL: heap-profiler_unittest.sh
9793 FAIL: heap-checker_unittest.sh
9794 FAIL: heap-checker-death_unittest.sh
9795 FAIL: sampling_debug_test.sh
9796 FAIL: heap-profiler_debug_unittest.sh
9797 FAIL: heap-checker_debug_unittest.sh
9798 FAIL: profiler_unittest.sh
9799
9800 While it indicates that there is still work to do, This is still
9801 better than the result I get on ARMv7:
9802
9803 12 of 46 tests failed.
9804
9805 M src/base/basictypes.h
9806 M src/base/cycleclock.h
9807 M src/base/linuxthreads.h
9808 M src/malloc_hook_mmap_linux.h
9809
9810 commit 15b5e7a35c83ce5d38fa523f2c291a2ac30bb8ed
9811 Author: Riku Voipio <[email protected]>
9812 Date: Tue Feb 4 11:09:47 2014 +0200
9813
9814 linux_syscall_support.h: add aarch64 support
9815
9816 Aarch64 support for linux_syscall_support.h. Since Aarch64 is a brand
9817 new architecture, none of the legacy system calls are neccesarily
9818 available. Thus some changes were neccesary affect other architectures
9819 as well:
9820
9821 1) use getdents64 where available and else getdents (for ppc64)
9822
9823 2) other legacy system calls, pipe, waitpid and open replaced
9824 by pipe2,
9825 wait4 and openat where available.
9826
9827 3) use fstatat if stat is not available.
9828
9829 The aarch64 system call interface follows the Aarch64 calling
9830 convention
9831 (regs x0-x5 for arguments and x8 system call number - return in x0).
9832 Clone implementation is adapted from glibc.
9833
9834 v2: step back in getdents removal due to ppc64
9835
9836 M src/base/linux_syscall_support.h
9837 M src/base/linuxthreads.cc
9838
9839 commit b5e584dfdc22328f713488311707f502aa77ff5f
9840 Author: Aliaksey Kandratsenka <[email protected]>
9841 Date: Sun Feb 16 19:41:37 2014 -0800
9842
9843 issue-525: warn user on lack of libunwind but don't fail
9844
9845 Because we now have access to other backtrace capturing method(s) at
9846 runtime.
9847
9848 M configure.ac
9849
9850 commit 90ba15d1f2f6704af96f62ce1e8c5f214697bab1
9851 Author: Aliaksey Kandratsenka <[email protected]>
9852 Date: Sat Feb 8 15:30:36 2014 -0800
9853
9854 issue-604: implement runtime-selectable stacktrace capturing
9855
9856 We're now building all supported stacktrace capturing methods. And
9857 there's now a way to select at runtime which method is used.
9858
9859 M Makefile.am
9860 M configure.ac
9861 M src/stacktrace.cc
9862 M src/stacktrace_arm-inl.h
9863 D src/stacktrace_config.h
9864 M src/stacktrace_generic-inl.h
9865 C055 src/stacktrace.cc src/stacktrace_impl_setup-inl.h
9866 M src/stacktrace_instrument-inl.h
9867 M src/stacktrace_libunwind-inl.h
9868 M src/stacktrace_powerpc-inl.h
9869 M src/stacktrace_win32-inl.h
9870 M src/stacktrace_x86-inl.h
9871
9872 commit 33f6781d64af88ea23698a084188d8c2ab94ecb1
9873 Author: Aliaksey Kandratsenka <[email protected]>
9874 Date: Sun Feb 16 18:47:04 2014 -0800
9875
9876 issue-605: avoid compilation errors if pthread_key_t is pointer
9877
9878 Which seems to be the case on later cygwin
9879
9880 M src/maybe_threads.cc
9881
9882 commit 100f310088aa30d347cd5d07bbe9d9f80cf2c2d0
9883 Author: Aliaksey Kandratsenka <[email protected]>
9884 Date: Sun Feb 16 18:27:14 2014 -0800
9885
9886 unbreak make dist
9887
9888 M Makefile.am
9889
9890 commit a0ed9ace534372fbeee9bfefd6e26b922f0a1835
9891 Author: Wang YanQing <[email protected]>
9892 Date: Tue Feb 11 23:20:31 2014 +0800
9893
9894 debugallocation: fix bus error on mipsel-linux platform when enable
9895 use_malloc_page_fence
9896
9897 Fix below "BUS ERROR" issue:
9898
9899 a0 hold start address of memory block allocated by DebugAllocate
9900 in debugallocation.cc
9901
9902 gdb) info registers
9903 zero at v0 v1 a0 a1
9904 a2 a3
9905 R0 00000000 10008700 772f62a0 00084d40 766dcfef 7fb5f420 00000000
9906 004b4dd8
9907 t0 t1 t2 t3 t4 t5
9908 t6 t7
9909 R8 7713c1a0 7712dbc0 ffffffff 777bc000 f0000000 00000001 00000000
9910 00403d10
9911 s0 s1 s2 s3 s4 s5
9912 s6 s7
9913 R16 7fb5ff1c 00401b9c 77050020 7fb5fb18 00000000 004cb008 004ca748
9914 ffffffff
9915 t8 t9 k0 k1 gp sp
9916 s8 ra
9917 R24 0000002f 771adcd4 00000000 00000000 771f4140 7fb5f408 7fb5f430
9918 771add6c
9919 sr lo hi bad cause pc
9920 00008713 0000e9fe 00000334 766dcff7 00800010 771adcfc
9921 fsr fir
9922 00000004 00000000
9923
9924 (gdb) disassemble
9925 Dump of assembler code for function _ZNSs4_Rep10_M_disposeERKSaIcE:
9926 0x771adcd4 <+0>: lui gp,0x4
9927 0x771adcd8 <+4>: addiu gp,gp,25708
9928 0x771adcdc <+8>: addu gp,gp,t9
9929 0x771adce0 <+12>: lw v0,-28696(gp)
9930 0x771adce4 <+16>: beq a0,v0,0x771add38
9931 <_ZNSs4_Rep10_M_disposeERKSaIcE+100>
9932 0x771adce8 <+20>: nop
9933 0x771adcec <+24>: lw v0,-30356(gp)
9934 0x771adcf0 <+28>: beqzl v0,0x771add1c
9935 <_ZNSs4_Rep10_M_disposeERKSaIcE+72>
9936 0x771adcf4 <+32>: lw v0,8(a0)
9937 0x771adcf8 <+36>: sync
9938 => 0x771adcfc <+40>: ll v0,8(a0)
9939 0x771add00 <+44>: addiu at,v0,-1
9940 0x771add04 <+48>: sc at,8(a0)
9941 0x771add08 <+52>: beqz at,0x771adcfc
9942 <_ZNSs4_Rep10_M_disposeERKSaIcE+40>
9943 0x771add0c <+56>: nop
9944 0x771add10 <+60>: sync
9945 0x771add14 <+64>: b 0x771add24
9946 <_ZNSs4_Rep10_M_disposeERKSaIcE+80>
9947 0x771add18 <+68>: nop
9948 0x771add1c <+72>: addiu v1,v0,-1
9949 0x771add20 <+76>: sw v1,8(a0)
9950 0x771add24 <+80>: bgtz v0,0x771add38
9951 <_ZNSs4_Rep10_M_disposeERKSaIcE+100>
9952 0x771add28 <+84>: nop
9953 0x771add2c <+88>: lw t9,-27072(gp)
9954 0x771add30 <+92>: jr t9
9955 0x771add34 <+96>: nop
9956 0x771add38 <+100>: jr ra
9957 0x771add3c <+104>: nop
9958 End of assembler dump.
9959
9960 ll instruction manual:
9961 Load Linked:
9962 Loads the destination register with the contents of the word
9963 that is at the memory location. This instruction implicity performs
9964 a SYNC operation; all loads and stores to shared memory fetched prior
9965 to the ll must access memory before the ll, and loads and stores to
9966 shared memory fetched subsequent to the ll must access memory
9967 after ll.
9968 Load Linked and Store Conditional can be use to automatically update
9969 memory locations. *This instruction is not valid in the mips1
9970 architectures.
9971 The machine signals an address exception when the effective address
9972 is not
9973 divisible by four.
9974
9975 Signed-off-by: Wang YanQing <[email protected]>
9976 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
9977 [[email protected]: removed addition of unused #include]
9978
9979 M src/debugallocation.cc
9980
9981 commit 38bfc7a1c2f9ba718bcaa9b82fdcd9c429c8f85f
9982 Author: Aliaksey Kandratsenka <[email protected]>
9983 Date: Sat Feb 8 14:10:11 2014 -0800
9984
9985 removed irrelevant comment
9986
9987 M src/base/atomicops.h
9988
9989 commit d03c467a3446088b229f0106d9f47fab6d7c52a4
9990 Author: Aliaksey Kandratsenka <[email protected]>
9991 Date: Sat Feb 8 13:35:16 2014 -0800
9992
9993 allow asking for gcc atomics on all platforms
9994
9995 I.e. by doing ./configure CPPFLAGS=-DTCMALLOC_PREFER_GCC_ATOMICS
9996
9997 M src/base/atomicops.h
9998
9999 commit 6de1f38b687e3974d67adad45217f35a2c376049
10000 Author: Aliaksey Kandratsenka <[email protected]>
10001 Date: Sat Feb 8 13:43:04 2014 -0800
10002
10003 chmod -x configure.ac
10004
10005 Because configure.ac is not really executable. And because it
10006 interferes with tab completion of configure.
10007
10008 M configure.ac
10009
10010 commit e8fe990fa06e337a07059c55adc6ed2a0888de95
10011 Author: Riku Voipio <[email protected]>
10012 Date: Wed Jan 29 10:54:29 2014 +0200
10013
10014 implement atomics with gcc intrinsics
10015
10016 Gcc after 4.7 provides atomic builtins[1]. Use these instead of adding
10017 yet-another-assembly port for Aarch64 (64-bit ARM). This patch enables
10018 succesfully building and running atomicops unittest on Aarch64.
10019
10020 This patch enables using gcc builtins only when no assembly
10021 implementation is provided. But as a quick check, atomicops_unittest
10022 and rest of testsuite passes with atomicops-internals-gcc also
10023 ARMv7 and X86_64 if the ifdef in atomicops is adjusted to prefer
10024 the generic implementation.
10025
10026 [1] http://gcc.gnu.org/onlinedocs/gcc/_005f_005fatomic-Builtins.html
10027
10028 A src/base/atomicops-internals-gcc.h
10029 M src/base/atomicops.h
10030
10031 commit fa4b1c401da1ac381d4d72172825231b3d5518d9
10032 Author: Aliaksey Kandratsenka <[email protected]>
10033 Date: Sun Jan 19 22:37:44 2014 -0800
10034
10035 issue-599: fixing FreeBSD issue with sbrk
10036
10037 Applied patch by yurivict.
10038
10039 It was wrong assembly specifically for FreeBSD in sbrk overriding
10040 code.
10041
10042 M src/malloc_hook_mmap_freebsd.h
10043
10044 commit 71a239e559c9ea2300ad0511f8de7077db5369c3
10045 Author: Aliaksey Kandratsenka <[email protected]>
10046 Date: Sun Jan 19 12:30:53 2014 -0800
10047
10048 check debug_malloc_implementation_space via COMPILE_ASSERT
10049
10050 Because we can and because compile-time is always better.
10051
10052 M src/debugallocation.cc
10053
10054 commit 54568e32fc2321e0adef15fb1eab4e3a7f8ce5b0
10055 Author: Aliaksey Kandratsenka <[email protected]>
10056 Date: Sat Sep 21 18:14:00 2013 -0700
10057
10058 issue-565: don't pollute global namespace with thread lister API
10059
10060 Instead those functions that are original taken from google's "base"
10061 code now have prefix TCMalloc_. So that they don't conflict with other
10062 google's libraries having same functions.
10063
10064 M src/base/linuxthreads.cc
10065 M src/base/thread_lister.c
10066 M src/base/thread_lister.h
10067 M src/gperftools/heap-checker.h
10068 M src/heap-checker.cc
10069 M src/memory_region_map.h
10070
10071 commit 64bc1baa1f4723d73ba40cd730b72896bd45a810
10072 Author: Aliaksey Kandratsenka <[email protected]>
10073 Date: Sat May 18 17:11:58 2013 -0700
10074
10075 issue-{66,547}: use signal's ucontext when unwinding backtrace
10076
10077 In issue-66 (and readme) it is pointed out that sometimes there are
10078 some issues grabbing backtrace across signal handler boundary.
10079
10080 This code attempts to fix it by grabbing backtrace from signal's
10081 ucontext which clearly does not include signal handler boundary.
10082
10083 We're using "feature" of libunwind that for some important platforms
10084 libunwind's context is same as libc's ucontext_t which is given to us
10085 as part of calling signal handler.
10086
10087 M src/stacktrace_libunwind-inl.h
10088
10089 commit 185bf3fcc36f8cb3839abdfe652f615bfb5306d1
10090 Author: Aliaksey Kandratsenka <[email protected]>
10091 Date: Sat Jan 11 12:46:02 2014 -0800
10092
10093 issue-581: avoid destructing DebugMallocImplementation
10094
10095 Because otherwise destructor might be invoked well before other places
10096 that might touch malloc extension instance.
10097
10098 We're using placement new to initialize it and pass pointer to
10099 MallocExtension::Register. Which ensures that destructor for it is
10100 never run.
10101
10102 Based on idea suggested by Andrew C. Morrow.
10103
10104 M src/debugallocation.cc
10105
10106 commit 48a0d131c1aa088c6075e9c4676ee430f81d8600
10107 Author: Aliaksey Kandratsenka <[email protected]>
10108 Date: Sat Jan 18 13:27:41 2014 -0800
10109
10110 issue-548: pass -fno-builtin to compiler for unittests
10111
10112 Because clang doesn't understand -fno-builtin-malloc and friends. And
10113 otherwise new/delete pairs get optimized away causing our tests that
10114 expect hooks to be called to fail.
10115
10116 M Makefile.am
10117
10118 commit e98371540d63dde53ce4b7c772d78c1da6b59ea8
10119 Author: Aliaksey Kandratsenka <[email protected]>
10120 Date: Sat Jan 11 16:28:15 2014 -0800
10121
10122 eliminated gcc warning on __thread configure snippet
10123
10124 gcc complained about lack of matching ' in code that force-fails
10125 __thread detection on mingw
10126
10127 M configure.ac
10128
10129 commit 60b12171bc73117c0108b847bb310af095cd2778
10130 Author: xiaoyur347 <[email protected]>
10131 Date: Sat Jan 11 11:39:53 2014 +0800
10132
10133 fix GCC version detect for platforms other than X86/X64
10134
10135 [[email protected]: commented why we're disabling __thread not just for x86]
10136
10137 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
10138
10139 M configure.ac
10140
10141 commit 764d304222f0c3057ab99babd06246016cbfa505
10142 Author: Aliaksey Kandratsenka <[email protected]>
10143 Date: Sun Jan 5 12:49:23 2014 -0800
10144
10145 don't re-define strtoq for VS2013
10146
10147 Which is part of previous change that wasn't correctly applied.
10148
10149 M src/windows/port.h
10150
10151 commit 1fc768864d506466b45f3f38474912bcb69bb772
10152 Author: Aliaksey Kandratsenka <[email protected]>
10153 Date: Sat Jan 4 18:28:36 2014 -0800
10154
10155 fix compilation under VS 2013
10156
10157 This is essentially a copy of corresponding chromium change from:
10158 https://codereview.chromium.org/27017003
10159
10160 M src/windows/port.cc
10161 M src/windows/port.h
10162
10163 commit 4c274b9e20132230e62117ff583ebadd83081d90
10164 Author: Aliaksey Kandratsenka <[email protected]>
10165 Date: Sat Jan 4 18:28:36 2014 -0800
10166
10167 issue-592: handle recent mingw with C++11 threads
10168
10169 Somehow it's c++ headers (like string) define pthread symbols without
10170 even us asking for. That breaks old assumption that pthread symbols
10171 are not available on windows.
10172
10173 In order to fix that we detect this condition in configure.ac and
10174 avoid defining windows versions of pthread symbols.
10175
10176 M Makefile.am
10177 M configure.ac
10178 M src/windows/mingw.h
10179 M src/windows/port.h
10180
10181 commit 1458ee2239e0791567e69112931dc17eb0456cf8
10182 Author: Aliaksey Kandratsenka <[email protected]>
10183 Date: Sat Jan 4 13:54:24 2014 -0800
10184
10185 issue-596: removed unused AtomicIncrement operation
10186
10187 There's no need for us to attempt to maintain Google's atomic ops code
10188 in era of C++11.
10189
10190 M src/base/atomicops-internals-arm-generic.h
10191 M src/base/atomicops-internals-arm-v6plus.h
10192 M src/base/atomicops-internals-linuxppc.h
10193 M src/base/atomicops-internals-macosx.h
10194 M src/base/atomicops-internals-mips.h
10195 M src/base/atomicops-internals-windows.h
10196 M src/base/atomicops-internals-x86.h
10197 M src/base/atomicops.h
10198 M src/tests/atomicops_unittest.cc
10199
10200 commit 6630b24e27c6a62727fe73aaae21dcc7364b8fee
10201 Author: Aliaksey Kandratsenka <[email protected]>
10202 Date: Sat Jan 4 13:07:35 2014 -0800
10203
10204 Removed unused AtomicPtr::CompareAndSwap
10205
10206 M src/malloc_hook-inl.h
10207 M src/malloc_hook.cc
10208
10209 commit a15115271cc475509b17bf7fecbe1ac4966baf2e
10210 Author: xiaoyur347 <[email protected]>
10211 Date: Fri Dec 20 09:41:08 2013 +0800
10212
10213 add "-finstrument-functions" support for MIPS uclibc.
10214 should configure with CXXFLAGS="-finstrument-functions"
10215
10216 M src/stacktrace_config.h
10217 A src/stacktrace_instrument-inl.h
10218
10219 commit 7c4888515ed93347d4793fc066cd6048e519a197
10220 Author: xiaoyur347 <[email protected]>
10221 Date: Fri Dec 20 09:02:49 2013 +0800
10222
10223 add uclibc support
10224 * some variables defined with "char *" should be modified to
10225 "const char*"
10226 * For uclibc, glibc's "void malloc_stats(void)" should be "void
10227 malloc_stats(FILE *)", is commented now.
10228 * For uclibc, __sbrk is with attribute "hidden", so we use mmap
10229 allocator for uclibc.
10230
10231 M Makefile.am
10232 M src/heap-checker.cc
10233 M src/heap-profiler.cc
10234 M src/libc_override_gcc_and_weak.h
10235 M src/malloc_hook_mmap_linux.h
10236 M src/memory_region_map.cc
10237 M src/symbolize.cc
10238 M src/system-alloc.cc
10239
10240 commit 7bd193bca97d93b43ff6c824bc9f39227329312f
10241 Author: Aliaksey Kandratsenka <[email protected]>
10242 Date: Sat Dec 14 12:03:02 2013 -0800
10243
10244 issue-586: detect main executable even if PIE is active
10245
10246 Previous logic of detecting main program addresses is to assume that
10247 main executable is at least addressess. With PIE (active by default on
10248 Ubuntus) it doesn't work.
10249
10250 In order to deal with that, we're attempting to find main executable
10251 mapping in /proc/[pid]/maps. And old logic is preserved too just in
10252 case.
10253
10254 M src/pprof
10255
10256 commit f8a2163b5131050765ea877e2573f4930e41f630
10257 Author: Aliaksey Kandratsenka <[email protected]>
10258 Date: Fri Dec 6 12:23:48 2013 -0800
10259
10260 Added AM_MAINTAINER_MODE to disable Makefile rebuild rules
10261
10262 Some people might want to check-in unpacked result on make dist into
10263 git. But because git doesn't preserve timestamps it would cause those
10264 automatic "auto-retool" rules to trigger. Sometimes even causing build
10265 breakage if system's autotools version don't match autotools version
10266 used for make dist.
10267
10268 Easiest way around this problem is to simply disable those unnecessary
10269 "maintainer" rebuild rules. Especially given that source is always
10270 freely available via git and therefore there should be no reason to
10271 regenerate any of autotools products in 'make dist'-produced sources.
10272
10273 M configure.ac
10274
10275 commit 925bbaea76b91bd307634908cfd6902f99804544
10276 Author: Aliaksey Kandratsenka <[email protected]>
10277 Date: Sat Nov 16 14:01:38 2013 -0800
10278
10279 actually check result of CheckAddressBits
10280
10281 Previously call to CheckAddressBits was made but nothing was done to
10282 it's result.
10283
10284 I've also make sure that actual size is used in checks and in bumping
10285 up of TCMalloc_SystemTaken.
10286
10287 M src/system-alloc.cc
10288
10289 commit f216317a879e972ceafe77e61b2d66fd5f29922e
10290 Author: Aliaksey Kandratsenka <[email protected]>
10291 Date: Sat Nov 16 15:05:45 2013 -0800
10292
10293 use AC_PROG_LIBTOOL to summon libtool
10294
10295 So that older autotools of rhel 5 can be used
10296
10297 M configure.ac
10298
10299 commit d4f4c5a3104e30b14b1090241fb4d0fac6a0e357
10300 Author: Aliaksey Kandratsenka <[email protected]>
10301 Date: Sat Nov 16 14:00:19 2013 -0800
10302
10303 assert that ClassSize(0) is 0 instead >=0
10304
10305 Because it's return value being size_t cannot be negative
10306 anyways. This fixes clang warning
10307
10308 M src/common.cc
10309
10310 commit 946203d60e50488a0b9e0fe003c13662873fa17d
10311 Author: Aliaksey Kandratsenka <[email protected]>
10312 Date: Sat Nov 16 13:35:59 2013 -0800
10313
10314 assert key size in way that is clearer to gcc
10315
10316 Both new and old asserts are checking same condition, however new
10317 assert helps gcc see that out of bounds access is not possible in
10318 root_ array.
10319
10320 M src/pagemap.h
10321
10322 commit bf2d7bd3f8c1fb6a0843e55e652e37f4ce0fae3d
10323 Author: Aliaksey Kandratsenka <[email protected]>
10324 Date: Sat Nov 16 13:31:34 2013 -0800
10325
10326 fixed gcc warning
10327
10328 We've recently changed old_signal_handler to by integer, so comparing
10329 it with NULL is not good idea.
10330
10331 M src/heap-profiler.cc
10332
10333 commit dd5f979c5e6e7e3127835a659a5af89ac21597d5
10334 Author: Aliaksey Kandratsenka <[email protected]>
10335 Date: Sat Nov 16 13:31:08 2013 -0800
10336
10337 fixed -Wreorder warning in HeapProfileTable constructor
10338
10339 M src/heap-profile-table.cc
10340
10341 commit e4ea98f147a7602979f70be022de068b04e68060
10342 Author: Aliaksey Kandratsenka <[email protected]>
10343 Date: Sat Nov 9 14:19:16 2013 -0800
10344
10345 issue-585: fixed use of TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES
10346
10347 In order to apply that, we're now doing explicit EnvToInt64 call as
10348 part of initializing thread cache module.
10349
10350 M src/thread_cache.cc
10351
10352 commit e0102230ec7e8304155798bf7f03d6abcd5991ee
10353 Author: Aliaksey Kandratsenka <[email protected]>
10354 Date: Sat Nov 16 12:03:35 2013 -0800
10355
10356 issue-588: Fix profiler_unittest.cc fork()
10357
10358 As suggested by Hannes Weisbach.
10359
10360 Call heap-profiler_unittest with the arguments 1 -2 (one iteration, 2
10361 fork()ed children).
10362
10363 Instead of running the test, the program crashes with a std::bad_alloc
10364 exception. This is caused by unconditionally passing the
10365 number-of-threads-argument (0 or positive for threads, negative for
10366 fork()s) in RunManyThreads(), thus allocating an array of pthread_t of
10367 size -2. Depending on the sign of the thread number argument either
10368 RunManyThreads or fork() should be called.
10369
10370 M src/tests/profiler_unittest.cc
10371
10372 commit 2bf83af65664a2badbaebdb722ad498e8b38548c
10373 Author: Aliaksey Kandratsenka <[email protected]>
10374 Date: Sat Nov 16 11:35:32 2013 -0800
10375
10376 issue-587: fix typos in unit test scripts
10377
10378 As proposed by Hannes Weisbach.
10379
10380 The argument will be garbled because of a misplaced brace, for example
10381 (heap-checker_unittest.sh):
10382
10383 HEAP_CHECKER="${1:-$BINDIR}/heap-checker_unittest"
10384 which should be:
10385 HEAP_CHECKER="${1:-$BINDIR/heap-checker_unittest}"
10386
10387 This unit test is used to check the binaries heap-checker_unittest and
10388 heap-checker_debug_unittest. With the typo, the executable
10389 heap-checker_debug_unittest is never actually run.
10390
10391 M src/tests/heap-checker-death_unittest.sh
10392 M src/tests/heap-checker_unittest.sh
10393 M src/tests/heap-profiler_unittest.sh
10394 M src/tests/tcmalloc_unittest.sh
10395
10396 commit b3b19269783cf1ed163bdb447cef9ca11b10851c
10397 Author: Aliaksey Kandratsenka <[email protected]>
10398 Date: Sat Nov 9 12:28:55 2013 -0800
10399
10400 issue-584: added license note to files without explicit license
10401
10402 As suggested at corresponding chromium issue discussion it's seemingly
10403 sufficient to simply refer to project-wide LICENSE file.
10404
10405 M src/tests/page_heap_test.cc
10406 M src/tests/profile-handler_unittest.cc
10407 M src/tests/raw_printer_test.cc
10408 M src/tests/stack_trace_table_test.cc
10409 M src/windows/config.h
10410
10411 commit 7be35fb0d844c73f5c5301f62074294df5acbc5c
10412 Author: Joonsoo Kim <[email protected]>
10413 Date: Thu Oct 10 14:26:47 2013 +0900
10414
10415 central_freelist: change fetch ordering
10416
10417 When we fetch objects from the span for thread cache, we make
10418 reverse-ordered list against original list on the span and suppy
10419 this list
10420 to thread cache. This algorithm has trouble with newly created span.
10421 Newly created span has ascending ordered objects list. Since thread
10422 cache
10423 will get reverse-ordered list against it, user gets objects as
10424 descending order.
10425
10426 Following example shows what occurs in this algorithm.
10427
10428 new span: object list: 1 -> 2 -> 3 -> 4 -> 5 -> ...
10429 fetch N items: N -> N-1 -> N-2 -> ... -> 2 -> 1 -> NULL
10430 thread cache: N -> N-1 -> N-2 -> ... -> 2 -> 1 -> NULL
10431
10432 user's 1st malloc: N
10433 user's 2nd malloc: N-1
10434 ...
10435 user's Nth malloc: 1
10436
10437 In general, access memory with ascending order is better than
10438 descending
10439 order in terms of the performance. So this patch fix this situation.
10440
10441 I run below program to measure performance effect.
10442
10443 #define MALLOC_SIZE (512)
10444 #define CACHE_SIZE (64)
10445 #define TOUCH_SIZE (512 / CACHE_SIZE)
10446
10447 array = malloc(sizeof(void *) * count);
10448
10449 for (i = 0; i < 1; i++) {
10450 for (j = 0; j < count; j++) {
10451 x = malloc(MALLOC_SIZE);
10452 array[j] = x;
10453 }
10454 }
10455
10456 repeat = 10;
10457 for (i = 0; i < repeat; i++) {
10458 for (j = 0; j < count; j++) {
10459 x = array[j];
10460 for (k = 0; k < TOUCH_SIZE; k++) {
10461 *(x + (k * CACHE_SIZE)) = '1';
10462 }
10463 }
10464 }
10465
10466 LD_PRELOAD=libtcmalloc_minimal.so perf stat -r 10 ./a.out 1000000
10467
10468 **** Before ****
10469 Performance counter stats for './a.out 1000000' (10 runs):
10470
10471 2.715161299 seconds time elapsed
10472 ( +- 0.07% )
10473
10474 **** After ****
10475 Performance counter stats for './a.out 1000000' (10 runs):
10476
10477 2.259366428 seconds time elapsed
10478 ( +- 0.08% )
10479
10480 M src/central_freelist.cc
10481
10482 commit 7315b45c28564afdc2699beff934a3b45457edc6
10483 Author: Joonsoo Kim <[email protected]>
10484 Date: Thu Oct 10 11:46:12 2013 +0900
10485
10486 central_freelist: fetch objects as much as possible during each trial
10487
10488 It is better to reduce function call if possible. If we try to fetch
10489 objects from one span as much as possible during each function call,
10490 number of function call would be reduced and this would help
10491 performance.
10492
10493 M src/central_freelist.cc
10494 M src/central_freelist.h
10495
10496 commit cc002ea19363e1ebbd7f3e809d116ab81a6862cd
10497 Author: Joonsoo Kim <[email protected]>
10498 Date: Thu Oct 10 10:06:33 2013 +0900
10499
10500 skip unnecessary check during double-check SizeClass intergrity
10501
10502 On initialization step, tcmalloc double-checks SizeClass integrity
10503 with
10504 all possible size values, 0 to kMaxSize. This causes tremendous
10505 overhead
10506 for short-lived applications.
10507
10508 For example, consider following command.
10509 'find -exec grep something {} \;'
10510
10511 Actual work of each grep is really small, but double-check requires
10512 more work. To reduce this overhead, it is best to remove double-check
10513 entirely. But we cannot be sure the integrity without double-checking,
10514 so alternative is needed.
10515
10516 This patch doesn't remove double-check, instead, try to skip
10517 unnecessary
10518 check based on ClassIndex() implementation. This reduce much
10519 overhead and
10520 the code has same coverage as previous double-check. Following is
10521 the result of this patch.
10522
10523 time LD_PRELOAD=libtcmalloc_minimal.so find ./ -exec grep "SOMETHING"
10524 {} \;
10525
10526 * Before
10527 real 0m3.675s
10528 user 0m1.000s
10529 sys 0m0.640s
10530
10531 * This patch
10532 real 0m2.833s
10533 user 0m0.056s
10534 sys 0m0.220s
10535
10536 * Remove double-check entirely
10537 real 0m2.675s
10538 user 0m0.072s
10539 sys 0m0.184s
10540
10541 M src/common.cc
10542
10543 commit 3e9a33e8c708ccf3ec91e3a3b14e924f5f79e4a6
10544 Author: Aliaksey Kandratsenka <[email protected]>
10545 Date: Sat Oct 26 16:54:03 2013 -0700
10546
10547 issue-583: include pthread.h into static_var.cc
10548
10549 Because we're doing pthread_atfork.
10550
10551 Fix suggested by user named drussel.
10552
10553 M src/static_vars.cc
10554
10555 commit db0d5730ee059d72b895fbead5237f9cb5bbf98a
10556 Author: Aliaksey Kandratsenka <[email protected]>
10557 Date: Sat Jun 22 13:48:11 2013 -0700
10558
10559 issue-579: ensure order between memory region and libunwind locks
10560
10561 I.e. to prevent possible deadlock when this locks are taked by
10562 different threads in different order.
10563
10564 This particular problem was also reported as part of issue 66.
10565
10566 M src/memory_region_map.cc
10567
10568 commit 42ddc8d42c82ba6f5137c26b4e7f752b1a022831
10569 Author: Aliaksey Kandratsenka <[email protected]>
10570 Date: Sat Sep 21 19:03:52 2013 -0700
10571
10572 added emacs -*- mode lines for google coding style
10573
10574 M src/addressmap-inl.h
10575 M src/base/atomicops-internals-arm-generic.h
10576 M src/base/atomicops-internals-arm-v6plus.h
10577 M src/base/atomicops-internals-linuxppc.h
10578 M src/base/atomicops-internals-macosx.h
10579 M src/base/atomicops-internals-mips.h
10580 M src/base/atomicops-internals-windows.h
10581 M src/base/atomicops-internals-x86.cc
10582 M src/base/atomicops-internals-x86.h
10583 M src/base/atomicops.h
10584 M src/base/basictypes.h
10585 M src/base/commandlineflags.h
10586 M src/base/cycleclock.h
10587 M src/base/elf_mem_image.cc
10588 M src/base/elf_mem_image.h
10589 M src/base/elfcore.h
10590 M src/base/googleinit.h
10591 M src/base/linux_syscall_support.h
10592 M src/base/linuxthreads.cc
10593 M src/base/logging.cc
10594 M src/base/logging.h
10595 M src/base/low_level_alloc.cc
10596 M src/base/low_level_alloc.h
10597 M src/base/simple_mutex.h
10598 M src/base/spinlock.cc
10599 M src/base/spinlock.h
10600 M src/base/spinlock_internal.cc
10601 M src/base/spinlock_internal.h
10602 M src/base/spinlock_linux-inl.h
10603 M src/base/spinlock_posix-inl.h
10604 M src/base/spinlock_win32-inl.h
10605 M src/base/stl_allocator.h
10606 M src/base/synchronization_profiling.h
10607 M src/base/sysinfo.cc
10608 M src/base/sysinfo.h
10609 M src/base/thread_lister.h
10610 M src/central_freelist.cc
10611 M src/central_freelist.h
10612 M src/common.cc
10613 M src/common.h
10614 M src/config_for_unittests.h
10615 M src/debugallocation.cc
10616 M src/getpc.h
10617 M src/gperftools/heap-checker.h
10618 M src/gperftools/heap-profiler.h
10619 M src/gperftools/malloc_extension.h
10620 M src/gperftools/malloc_hook.h
10621 M src/gperftools/profiler.h
10622 M src/gperftools/stacktrace.h
10623 M src/gperftools/tcmalloc.h.in
10624 M src/heap-checker-bcad.cc
10625 M src/heap-checker.cc
10626 M src/heap-profile-stats.h
10627 M src/heap-profile-table.cc
10628 M src/heap-profile-table.h
10629 M src/heap-profiler.cc
10630 M src/internal_logging.cc
10631 M src/internal_logging.h
10632 M src/libc_override.h
10633 M src/libc_override_gcc_and_weak.h
10634 M src/libc_override_glibc.h
10635 M src/libc_override_osx.h
10636 M src/libc_override_redefine.h
10637 M src/linked_list.h
10638 M src/malloc_extension.cc
10639 M src/malloc_hook-inl.h
10640 M src/malloc_hook.cc
10641 M src/malloc_hook_mmap_freebsd.h
10642 M src/malloc_hook_mmap_linux.h
10643 M src/maybe_threads.cc
10644 M src/maybe_threads.h
10645 M src/memfs_malloc.cc
10646 M src/memory_region_map.cc
10647 M src/memory_region_map.h
10648 M src/packed-cache-inl.h
10649 M src/page_heap.cc
10650 M src/page_heap.h
10651 M src/page_heap_allocator.h
10652 M src/pagemap.h
10653 M src/profile-handler.cc
10654 M src/profile-handler.h
10655 M src/profiledata.cc
10656 M src/profiledata.h
10657 M src/profiler.cc
10658 M src/raw_printer.cc
10659 M src/raw_printer.h
10660 M src/sampler.cc
10661 M src/sampler.h
10662 M src/span.cc
10663 M src/span.h
10664 M src/stack_trace_table.cc
10665 M src/stack_trace_table.h
10666 M src/stacktrace.cc
10667 M src/stacktrace_arm-inl.h
10668 M src/stacktrace_config.h
10669 M src/stacktrace_generic-inl.h
10670 M src/stacktrace_libunwind-inl.h
10671 M src/stacktrace_powerpc-inl.h
10672 M src/stacktrace_win32-inl.h
10673 M src/stacktrace_x86-inl.h
10674 M src/static_vars.cc
10675 M src/static_vars.h
10676 M src/symbolize.cc
10677 M src/symbolize.h
10678 M src/system-alloc.cc
10679 M src/system-alloc.h
10680 M src/tcmalloc.cc
10681 M src/tcmalloc.h
10682 M src/tcmalloc_guard.h
10683 M src/tests/addressmap_unittest.cc
10684 M src/tests/atomicops_unittest.cc
10685 M src/tests/current_allocated_bytes_test.cc
10686 M src/tests/debugallocation_test.cc
10687 M src/tests/frag_unittest.cc
10688 M src/tests/getpc_test.cc
10689 M src/tests/heap-checker_unittest.cc
10690 M src/tests/heap-profiler_unittest.cc
10691 M src/tests/low_level_alloc_unittest.cc
10692 M src/tests/malloc_extension_c_test.c
10693 M src/tests/malloc_extension_test.cc
10694 M src/tests/malloc_hook_test.cc
10695 M src/tests/markidle_unittest.cc
10696 M src/tests/memalign_unittest.cc
10697 M src/tests/packed-cache_test.cc
10698 M src/tests/page_heap_test.cc
10699 M src/tests/pagemap_unittest.cc
10700 M src/tests/profile-handler_unittest.cc
10701 M src/tests/profiledata_unittest.cc
10702 M src/tests/profiler_unittest.cc
10703 M src/tests/raw_printer_test.cc
10704 M src/tests/realloc_unittest.cc
10705 M src/tests/sampler_test.cc
10706 M src/tests/sampling_test.cc
10707 M src/tests/simple_compat_test.cc
10708 M src/tests/stack_trace_table_test.cc
10709 M src/tests/system-alloc_unittest.cc
10710 M src/tests/tcmalloc_large_unittest.cc
10711 M src/tests/tcmalloc_unittest.cc
10712 M src/tests/testutil.cc
10713 M src/tests/testutil.h
10714 M src/tests/thread_dealloc_unittest.cc
10715 M src/thread_cache.cc
10716 M src/thread_cache.h
10717 M src/windows/auto_testing_hook.h
10718 M src/windows/get_mangled_names.cc
10719 M src/windows/gperftools/tcmalloc.h
10720 M src/windows/gperftools/tcmalloc.h.in
10721 M src/windows/mingw.h
10722 M src/windows/mini_disassembler.cc
10723 M src/windows/mini_disassembler.h
10724 M src/windows/mini_disassembler_types.h
10725 M src/windows/override_functions.cc
10726 M src/windows/port.cc
10727 M src/windows/port.h
10728 M src/windows/preamble_patcher.cc
10729 M src/windows/preamble_patcher.h
10730 M src/windows/preamble_patcher_test.cc
10731 M src/windows/preamble_patcher_with_stub.cc
10732
10733 commit 799a22624c85e8749f8bed0bfa63282b03e75bdd
10734 Author: Aliaksey Kandratsenka <[email protected]>
10735 Date: Sat Sep 28 19:32:20 2013 -0700
10736
10737 issue-575: do not use cycle count register on arm6
10738
10739 Apparently not all arm6 implementations implement it in this
10740 particular way.
10741
10742 This applies patch by Ben Avison.
10743
10744 M src/base/cycleclock.h
10745
10746 commit 2a2d6596f8c3d62b7ec444761f4edf0c85d10d92
10747 Author: Petr Hosek <[email protected]>
10748 Date: Fri Aug 23 23:58:47 2013 -0700
10749
10750 Adds system-alloc_unittest Visual Studio project
10751
10752 M gperftools.sln
10753 A vsprojects/system-alloc_unittest/system-alloc_unittest.vcproj
10754
10755 commit 83aed118e009b92ea88645ef1f7b842a921612c5
10756 Author: Petr Hosek <[email protected]>
10757 Date: Fri Aug 23 13:53:35 2013 -0700
10758
10759 issue-567: Allows for overriding system allocator on Windows
10760
10761 [[email protected]: minor changes to make mingw build work]
10762 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
10763
10764 M Makefile.am
10765 M src/windows/port.cc
10766 A src/windows/system-alloc.cc
10767 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
10768 M vsprojects/tmu-static/tmu-static.vcproj
10769
10770 commit 4ad16873a0a2d8861a0bfe8234d45e31cc70ee90
10771 Author: Petr Hosek <[email protected]>
10772 Date: Sat Aug 24 14:24:47 2013 -0700
10773
10774 Exports SysAllocator class to avoid .dll build errors
10775
10776 M src/gperftools/malloc_extension.h
10777
10778 commit 326990b5c30d249c3cf4688a88fc415b05494aca
10779 Author: Aliaksey Kandratsenka <[email protected]>
10780 Date: Sat Sep 14 14:38:53 2013 -0700
10781
10782 issue-557: added support for dumping heap profile via signal
10783
10784 This applies patch from Jean Lee.
10785
10786 I've reformatted it to match surronding code style and changed
10787 validation logic a bit. I.e. we're not checking signal for range
10788 anymore given we're not sure what different platforms support, but
10789 we're checking return value of signal() for SIG_ERR instead.
10790
10791 M src/heap-profiler.cc
10792
10793 commit cb65e49b83c84bc205203c12793f2dd00c4a7721
10794 Author: Aliaksey Kandratsenka <[email protected]>
10795 Date: Sat Sep 14 16:45:42 2013 -0700
10796
10797 issue-536: do not PrintStats if running under valgrind
10798
10799 When we detect running under valgrind we do not initialize our own
10800 malloc. So trying to print malloc stats when asked via MALLOCSTATS
10801 cannot work.
10802
10803 This does fix proposed by Philippe Waroquiers. In which we detect
10804 running under valgrind prior to checking MALLOCSTATS environment
10805 variable and refuse printing stats if we detect valgrind.
10806
10807 M src/tcmalloc.cc
10808
10809 commit 6979583592df555a369a2c975f5117a1f61911af
10810 Author: Aliaksey Kandratsenka <[email protected]>
10811 Date: Mon Sep 9 07:59:25 2013 -0700
10812
10813 issue-564: added atomic ops support for mips{,64}
10814
10815 This merges patch contributed by Jovan Zelincevic.
10816
10817 And with that patch tcmalloc build with --enable-minimal (just malloc
10818 replacement) appears to work (passes unit tests).
10819
10820 M Makefile.am
10821 A src/base/atomicops-internals-mips.h
10822 M src/base/atomicops.h
10823 M src/base/basictypes.h
10824 M src/base/linux_syscall_support.h
10825 M src/stacktrace_config.h
10826
10827 commit 28dd85e2825af71138621a4417e6ab004631924d
10828 Author: Aliaksey Kandratsenka <[email protected]>
10829 Date: Fri Aug 30 16:57:14 2013 +0300
10830
10831 implement pc from ucontext access for mips
10832
10833 M m4/pc_from_ucontext.m4
10834
10835 commit 819a2b051f1dba9526f2338098fff6dd1700bdb6
10836 Author: Aliaksey Kandratsenka <[email protected]>
10837 Date: Thu Aug 29 19:00:31 2013 +0300
10838
10839 issue-413: disable __thread usage on OSX
10840
10841 Because it was found that __thread variables access is compiled into
10842 calls to tlv_get_addr which was found to call malloc. Because we
10843 actually use thread-local storage from inside malloc it leads to stack
10844 overflow. So we'll continue using pthreads API for that which is known
10845 to work on OSX.
10846
10847 M configure.ac
10848
10849 commit 43809080931127037ce6e748f37a28ce7489387d
10850 Author: Aliaksey Kandratsenka <[email protected]>
10851 Date: Thu Aug 29 18:51:48 2013 +0300
10852
10853 lowered autoconf requirement
10854
10855 Autoconf 2.59 works. And most notably it will not affect our releases
10856 which are all prepared with newer autoconf.
10857
10858 M configure.ac
10859
10860 commit 7a178d472737d988583bade3f0345201651cd43f
10861 Author: Joon-Sung Um <[email protected]>
10862 Date: Tue Aug 27 22:16:18 2013 +0900
10863
10864 Update document for tcmalloc
10865
10866 Update tcmalloc.html for new parameters.
10867
10868 * kMaxSize = 256k
10869 * kNumClasses = 88
10870 * kPageShift = 13
10871
10872 Signed-off-by: Aliaksey Kandratsenka <[email protected]>
10873
10874 M doc/tcmalloc.html
10875
10876 commit 313e08b5a1a951f710b0323b3ca4e1e8d869520e
10877 Author: Aliaksey Kandratsenka <[email protected]>
10878 Date: Sun Aug 4 20:44:06 2013 +0300
10879
10880 issue-560: Revert "issue-481: ...
10881
10882 ...Replaced test mechanism for distinct address spaces with a more
10883 reliable mechanism"
10884
10885 This reverts commit 5dd53ab6cbf9d98f2d60546835e84785a104da46 (svn
10886 revision 167)
10887
10888 With this commit rhel 6.2 fails heap-checker-death_unittest and
10889 without it passes.
10890
10891 Ticket refers to 2 things and both are invalid:
10892
10893 * that ptrace PEEKDATA ignores data argument. I've checked kernel
10894 source and found it to be wrong
10895
10896 * something about distinct address spaces
10897
10898 And in addition to all that original ticket admits that it doesn't fix
10899 anything.
10900
10901 It looks like, compared to original code that "fix" is not succesfully
10902 wait-ing on parent's ptrace request. I.e. by adding some additional
10903 diagnostics I'm seeing this sys_waitpid returning ECHILD.
10904
10905 M src/base/linuxthreads.cc
10906
10907 commit 6d00cbce92c5576b4bcf07a2b5634b4dcfa14f8a
10908 Author: Aliaksey Kandratsenka <[email protected]>
10909 Date: Sat Aug 17 15:52:53 2013 +0300
10910
10911 issue-561: don't cast function pointer to void *
10912
10913 Which gcc-3.4 (as shipped in rhel 4) doesn't like.
10914
10915 Cast to void * was originally added to avoid issue on OSX which
10916 doesn't have sighandler_t.
10917
10918 In that place we only need to know if it's null or not. So casting to
10919 intptr_t looks like simplest possible way to achieve that.
10920
10921 M src/profiler.cc
10922
10923 commit 7d8d522d737f6c55169264b35ebe0be1f9b23efd
10924 Author: Aliaksey Kandratsenka <[email protected]>
10925 Date: Sat Aug 17 15:52:22 2013 +0300
10926
10927 add heap-profile-stats.h to dist .tar.gz
10928
10929 M Makefile.am
10930
10931 commit d76cfa6d6ca5944d6300765fcb9160e889423750
10932 Author: Aliaksey Kandratsenka <[email protected]>
10933 Date: Sun Aug 4 18:35:55 2013 +0300
10934
10935 issue-502: Count m(un)map for each stacktrace in MemoryRegionMap
10936
10937 ..instead of HeapProfileTable
10938
10939 This upstreams chromium commit reviewed at:
10940 https://codereview.chromium.org/12388070
10941
10942 Original and upstreaming author is: Dai MIKURUBE
10943
10944 This patch fixes a bug that gperftools(TCMalloc)'s mmap profiler
10945 (HEAP_PROFILE_MMAP) doesn't hook some memory pages used by the
10946 profiler itself.
10947
10948 This problem has been lived in gperftools for a long time.
10949 It is discussed in gperftools' issue 502.
10950 https://code.google.com/p/gperftools/issues/detail?id=502
10951
10952 Some bugs in the mmap profiler were fixed by
10953 https://code.google.com/p/gperftools/issues/detail?id=383,
10954 but the patch in the issue 383 didn't fix the bug mentioned in
10955 the issue 502.
10956
10957 This change reverts the previous patch and http://crrev.com/132771
10958 at first. Then, it modifies MemoryRegionMap to count m(un)map
10959 calls for each stacktrace in itself instead of merging the counts
10960 for each stacktrace in HeapProfileTable.
10961
10962 This change also cleans up heap-profiler, heap-profile-table and
10963 deep-heap-profile.
10964
10965 Chromium-BUG=https://code.google.com/p/chromium/issues/detail?id=181517
10966 Chromium-Committed:
10967 https://src.chromium.org/viewvc/chrome?view=rev&revision=188176
10968
10969 M src/heap-checker.cc
10970 A src/heap-profile-stats.h
10971 M src/heap-profile-table.cc
10972 M src/heap-profile-table.h
10973 M src/heap-profiler.cc
10974 M src/memory_region_map.cc
10975 M src/memory_region_map.h
10976
10977 commit 89b163a0883d40a612331ed8565b8ab1219fdbd0
10978 Author: Aliaksey Kandratsenka <[email protected]>
10979 Date: Sun Aug 4 18:27:07 2013 +0300
10980
10981 added .gitignore
10982
10983 A .gitignore
10984
10985 commit 18fbc316eebea9db2d7ec41e161c0a3fbb09fa42
10986 Author: [email protected]
10987 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
10988 Date: Wed Jul 31 04:02:19 2013 +0000
10989
10990 Fix a C++11 error.
10991
10992 This merges patch by Nico Weber.
10993
10994 New clang versions complain in C++11 mode that:
10995
10996 error: cannot initialize a variable of type 'void *' with an rvalue
10997 of type 'uintptr_t' (aka 'unsigned long')
10998
10999 This same change was done for the google-internal version of
11000 tcmalloc too.
11001
11002 Reviewed-at: https://codereview.appspot.com/12132043
11003
11004
11005
11006 git-svn-id: http://gperftools.googlecode.com/svn/trunk@238
11007 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11008
11009 M src/stack_trace_table.cc
11010
11011 commit 674fcd94a8a0a3595f64e13762ba3a6529e09926
11012 Author: [email protected]
11013 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11014 Date: Tue Jul 30 09:14:20 2013 +0000
11015
11016 bump version to 2.1
11017
11018
11019
11020 git-svn-id: http://gperftools.googlecode.com/svn/trunk@236
11021 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11022
11023 M NEWS
11024 M configure.ac
11025 M packages/deb/changelog
11026 M src/windows/config.h
11027 M src/windows/gperftools/tcmalloc.h
11028
11029 commit 805a6601939edd9bca60a8911e56b217e477c75e
11030 Author: [email protected]
11031 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11032 Date: Tue Jul 30 08:30:45 2013 +0000
11033
11034 issue-559: don't setup fork handler in InitStaticVars
11035
11036 Because on OSX it calls malloc which leads to deadlock.
11037
11038 Given that we don't really need that fork handler _that_ early, it's
11039 fine to change it to normal static initializer
11040
11041
11042
11043 git-svn-id: http://gperftools.googlecode.com/svn/trunk@235
11044 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11045
11046 M src/static_vars.cc
11047
11048 commit c583a5babb54aedf6ab6e7746fd976ef2469839d
11049 Author: [email protected]
11050 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11051 Date: Tue Jul 30 08:30:19 2013 +0000
11052
11053 issue-559: don't depend on sighandler_t
11054
11055 Which is not available on OSX.
11056
11057 I've also fixed style around this place.
11058
11059
11060
11061 git-svn-id: http://gperftools.googlecode.com/svn/trunk@234
11062 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11063
11064 M src/profiler.cc
11065
11066 commit 1af20041bf5821167e69bf905596b2d45d1d94da
11067 Author: [email protected]
11068 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11069 Date: Tue Jul 30 08:29:46 2013 +0000
11070
11071 issue-559: AtomicWord xxx_AtomicExchange are all inline
11072
11073 Otherwise OSX correctly complains about duplicate definitions
11074
11075
11076
11077 git-svn-id: http://gperftools.googlecode.com/svn/trunk@233
11078 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11079
11080 M src/base/atomicops.h
11081
11082 commit 7a9c4e075fe2572307cddd11bc2f5c7d6b3d29ee
11083 Author: [email protected]
11084 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11085 Date: Fri Jul 26 19:36:08 2013 +0000
11086
11087 issue-556: update windows PACKAGE_{VERSION,STRING}
11088
11089
11090
11091 git-svn-id: http://gperftools.googlecode.com/svn/trunk@232
11092 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11093
11094 M src/windows/config.h
11095
11096 commit 11d46f0aac5ef941b74114412bda4747a349c60d
11097 Author: [email protected]
11098 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11099 Date: Fri Jul 26 19:35:38 2013 +0000
11100
11101 issue-556: dll-export getpagesize
11102
11103 Because page_heap_test needs this.
11104
11105
11106
11107 git-svn-id: http://gperftools.googlecode.com/svn/trunk@231
11108 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11109
11110 M src/windows/port.cc
11111 M src/windows/port.h
11112
11113 commit 6d4ad4428bb8f587d890fa5b7a76ba65dcf48dcb
11114 Author: [email protected]
11115 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11116 Date: Fri Jul 26 19:35:20 2013 +0000
11117
11118 issue-556: drop malloc limit at the end of page_heap_test
11119
11120 Which otherwise causes somewhat weird stack overflow on release
11121 windows builds.
11122
11123
11124
11125 git-svn-id: http://gperftools.googlecode.com/svn/trunk@230
11126 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11127
11128 M src/tests/page_heap_test.cc
11129
11130 commit be54c3f3ed192eabab9b075ec924d2479a956b9d
11131 Author: [email protected]
11132 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11133 Date: Fri Jul 26 19:34:54 2013 +0000
11134
11135 issue-556: dll-export TCMalloc_System{Alloc,Release}
11136
11137 because page_heap_test is using this stuff
11138
11139
11140
11141 git-svn-id: http://gperftools.googlecode.com/svn/trunk@229
11142 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11143
11144 M src/system-alloc.h
11145 M src/windows/port.cc
11146
11147 commit e3716145cbfa1247fa37dd627b2136a061944255
11148 Author: [email protected]
11149 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11150 Date: Fri Jul 26 18:19:05 2013 +0000
11151
11152 issue-552: Fix page_heap_test for system with different page size
11153
11154 This is patch by Adhemerval Zanella.
11155
11156 PowerPC uses 64K page size instead of 4k for x86 and x86_64. It
11157 makes the
11158 page_heap_test fails because the following test:
11159
11160 static bool HaveSystemRelease =
11161 TCMalloc_SystemRelease(TCMalloc_SystemAlloc(kPageSize, NULL,
11162 0), kPageSize);
11163
11164 will always fail if kPageSize is less than getpagesize() (the default
11165 configuration).
11166
11167 The following patch fixes it by trying to allocate/deallocate
11168 an entire
11169 page instead.
11170
11171
11172
11173 git-svn-id: http://gperftools.googlecode.com/svn/trunk@228
11174 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11175
11176 M src/tests/page_heap_test.cc
11177
11178 commit f45133e75c09ca7d5e86bda2db16e30c6fa348c0
11179 Author: [email protected]
11180 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11181 Date: Fri Jul 26 18:04:03 2013 +0000
11182
11183 issue-553: Fix syscall wrapper for PowerPC
11184
11185 This is patch by Adhemerval Zanella.
11186
11187 * src/stacktrace_powerpc-inl.h: It is just a cleanup for the
11188 stacktrace
11189 functions for PowerPC. The idea is to simplify the code.
11190 * src/tests/heap-checker_unittest.cc: Handles the PPC64 function
11191 descriptor
11192 correctly in malloc tracers. Different from other architecture,
11193 for PPC64
11194 the address returned in function pointers are the ODP entry, not the
11195 symbol address in .text segment. This leads the comparison bogus,
11196 since
11197 it will compare a ODP entry with a .text address.
11198 * src/heap-checker.cc: Add support for PPC in ptrace.
11199 * src/base/elfcore.h: Likewise.
11200 * src/base/linuxthreads.cc: Fix the thread creation using the clone
11201 wrapper.
11202 * src/base/linux_syscall_support.h: Various fixes for PPC32 and PPC64:
11203 fixes the kernel_stat[64] struct layout, and sys_clone and
11204 sys_socket
11205 implementation.
11206
11207
11208
11209 git-svn-id: http://gperftools.googlecode.com/svn/trunk@227
11210 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11211
11212 M src/base/elfcore.h
11213 M src/base/linux_syscall_support.h
11214 M src/base/linuxthreads.cc
11215 M src/heap-checker.cc
11216 M src/stacktrace_powerpc-inl.h
11217 M src/tests/heap-checker_unittest.cc
11218
11219 commit ee2bf097133b115e3da249e43507e02645e46e59
11220 Author: [email protected]
11221 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11222 Date: Sat Jul 20 21:37:26 2013 +0000
11223
11224 updated NEWS and packages/ for 2.1rc
11225
11226
11227
11228 git-svn-id: http://gperftools.googlecode.com/svn/trunk@226
11229 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11230
11231 M NEWS
11232 M packages/deb/changelog
11233 M packages/deb/control
11234 M packages/deb/copyright
11235 M packages/rpm/rpm.spec
11236
11237 commit 60ab178aeb827378d78cae8f1c6f893b14deaaf0
11238 Author: [email protected]
11239 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11240 Date: Sat Jul 20 21:37:01 2013 +0000
11241
11242 bump version number for 2.1rc
11243
11244
11245
11246 git-svn-id: http://gperftools.googlecode.com/svn/trunk@225
11247 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11248
11249 M configure.ac
11250 M src/windows/gperftools/tcmalloc.h
11251
11252 commit 7c2aa2f7c2cf901c17b9b9215224f725e798d1f2
11253 Author: [email protected]
11254 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11255 Date: Sat Jul 20 21:35:56 2013 +0000
11256
11257 issue-546: fixed return type of Release_AtomicExchange on arm6+
11258
11259
11260
11261 git-svn-id: http://gperftools.googlecode.com/svn/trunk@224
11262 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11263
11264 M src/base/atomicops-internals-arm-v6plus.h
11265
11266 commit d8e12e94ea09eaf8b040c6887f75cdf832b5619b
11267 Author: [email protected]
11268 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11269 Date: Sat Jul 20 21:35:14 2013 +0000
11270
11271 issue-549: handle most recent mingw that has sleep and nanosleep
11272
11273 I.e. we have to check their presence in configure and in case of their
11274 presence we have to avoid re-defining then in window's port.h
11275
11276
11277
11278 git-svn-id: http://gperftools.googlecode.com/svn/trunk@223
11279 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11280
11281 M configure.ac
11282 M src/windows/port.h
11283
11284 commit ac354636de8a4f11e4fde679b52e9f58fda0e079
11285 Author: [email protected]
11286 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11287 Date: Sun Jul 14 04:19:02 2013 +0000
11288
11289 issue-550: remove config.h.in
11290
11291 Which is autoconf product and thus is not needed in source repository
11292
11293
11294
11295 git-svn-id: http://gperftools.googlecode.com/svn/trunk@222
11296 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11297
11298 D src/config.h.in
11299
11300 commit e54971d58641853a9515d9f0313645729eab113a
11301 Author: [email protected]
11302 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11303 Date: Sat Jul 6 21:54:34 2013 +0000
11304
11305 issue-534: fixed a number of gcc warnings
11306
11307 This applies patch from Adhemerval Zanella.
11308
11309
11310
11311 git-svn-id: http://gperftools.googlecode.com/svn/trunk@221
11312 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11313
11314 M src/base/basictypes.h
11315 M src/heap-profiler.cc
11316 M src/profiler.cc
11317 M src/tests/heap-checker_unittest.cc
11318 M src/tests/page_heap_test.cc
11319
11320 commit 7dd038d7c58c9de889e3fcc552161533ea1baab1
11321 Author: [email protected]
11322 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11323 Date: Sat Jul 6 20:48:18 2013 +0000
11324
11325 issue-545: corrected README for CPUPROFILE_FREQUENCY variable
11326
11327 Applied patch by Mikhail Veltishchev
11328
11329
11330
11331 git-svn-id: http://gperftools.googlecode.com/svn/trunk@220
11332 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11333
11334 M README
11335
11336 commit a833a146b7de93774f0a9428edc5eda410d9a7dc
11337 Author: [email protected]
11338 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11339 Date: Sat Jul 6 20:46:50 2013 +0000
11340
11341 issue-528: fixed spelling
11342
11343 This simply applies patch by Lajos Veres
11344
11345
11346
11347 git-svn-id: http://gperftools.googlecode.com/svn/trunk@219
11348 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11349
11350 M doc/heapprofile.html
11351 M doc/pprof_remote_servers.html
11352 M src/base/spinlock.cc
11353 M src/base/sysinfo.cc
11354 M src/base/sysinfo.h
11355 M src/gperftools/profiler.h
11356 M src/pprof
11357 M src/stacktrace_x86-inl.h
11358 M src/windows/patch_functions.cc
11359 M src/windows/preamble_patcher.cc
11360
11361 commit 205abf1e7cb860224cbf391c7f69a6192ccc5076
11362 Author: [email protected]
11363 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11364 Date: Sun May 12 19:15:13 2013 +0000
11365
11366 issue-518: add support for x32 ABI to linux_syscall_support.h
11367
11368 As pointed out in the ticket this is taken from chromium review system
11369 here: https://codereview.chromium.org/13648012
11370
11371
11372
11373 git-svn-id: http://gperftools.googlecode.com/svn/trunk@218
11374 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11375
11376 M src/base/linux_syscall_support.h
11377
11378 commit 376a3107aa06063d72132318ac5ad266466e4325
11379 Author: [email protected]
11380 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11381 Date: Tue May 7 22:28:05 2013 +0000
11382
11383 issue-368: allocate enough bytes for large metadata allocations
11384
11385 During issue-368 review it was correctly pointed out then in place
11386 where I compare metadata allocation size to threshold I should pass
11387 that size down to TCMalloc_SystemAlloc instead of threshold.
11388
11389
11390
11391 git-svn-id: http://gperftools.googlecode.com/svn/trunk@217
11392 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11393
11394 M src/common.cc
11395
11396 commit 58d39a1b2190b9d0b719ee33f7bc5383a9b27462
11397 Author: [email protected]
11398 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11399 Date: Tue May 7 22:23:37 2013 +0000
11400
11401 issue-368: added missing large_heap_fragmentation_unittest.cc
11402
11403
11404
11405 git-svn-id: http://gperftools.googlecode.com/svn/trunk@216
11406 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11407
11408 A src/tests/large_heap_fragmentation_unittest.cc
11409
11410 commit c45bb7d6033952ea33bb181c1f96999317b535a1
11411 Author: [email protected]
11412 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11413 Date: Tue May 7 19:22:45 2013 +0000
11414
11415 issue-511: fixed negative offset handling for conditional jumps
11416
11417 While doing and testing issue-511 I've found one subtle bug which is
11418 incorrect handling of short offsets. They are defined to be signed but
11419 previous code used unsigned char for them which caused negative
11420 offsets to look like larger positive offsets. Fix is trivial.
11421
11422
11423
11424 git-svn-id: http://gperftools.googlecode.com/svn/trunk@215
11425 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11426
11427 M src/windows/preamble_patcher.cc
11428
11429 commit 3567b1701aa55a6421aefa5f3de1ca5507cbdf79
11430 Author: [email protected]
11431 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11432 Date: Tue May 7 19:15:35 2013 +0000
11433
11434 issue-511: recognise rex.w jmpq *<literal>(%rip) as iat jump
11435
11436 Apparently Windows Server 2012 (and presumably windows 8) now has this
11437 form of iat jump. Which is quite useless (rex.w is according to my
11438 understanding is not needed at all) but because of rex.w our code to
11439 recognize jumps like that didn't work.
11440
11441 Fix is just skip this prefix.
11442
11443
11444
11445 git-svn-id: http://gperftools.googlecode.com/svn/trunk@214
11446 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11447
11448 M src/windows/preamble_patcher.cc
11449
11450 commit 7fcb5ac0696e7ef7f7e7e51c18745af25de7da96
11451 Author: [email protected]
11452 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11453 Date: Tue May 7 19:13:29 2013 +0000
11454
11455 issue-511: implemented rel8 jump patching
11456
11457 I've found that Visual Studio 2012 release 32-bit C runtime library
11458 patching fails because array new has rel8 jmp which previous code
11459 could not handle.
11460
11461 Implementation is largely copied from conditional jumps handling code.
11462
11463
11464
11465 git-svn-id: http://gperftools.googlecode.com/svn/trunk@213
11466 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11467
11468 M src/windows/preamble_patcher.cc
11469 M src/windows/preamble_patcher.h
11470 M src/windows/preamble_patcher_with_stub.cc
11471
11472 commit 8cb4086a0aee539869bd087a85881788545b23d6
11473 Author: [email protected]
11474 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11475 Date: Mon May 6 19:50:59 2013 +0000
11476
11477 issue-368: test that we don't fragment large spans too badly
11478
11479 This adds unit test that does essentially same things as code to
11480 reproduce bug in
11481 https://code.google.com/p/gperftools/issues/detail?id=368
11482
11483
11484
11485
11486 git-svn-id: http://gperftools.googlecode.com/svn/trunk@212
11487 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11488
11489 M Makefile.am
11490
11491 commit f25ac4421f9efb635b88105e9c0830293ce19f9d
11492 Author: [email protected]
11493 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11494 Date: Mon May 6 19:50:23 2013 +0000
11495
11496 issue-{368,443}: allocate metadata in big batches
11497
11498 It uses same approach as PageHeapAllocator. Namely allocates big chunk
11499 which is then used to satisfy smaller allocations.
11500
11501 In issue-443 gradual heap grows causes old code that allocates
11502 metadata in smaller pieces and thus more frequently to fragment the
11503 heap. It's also causing most of 368 heap fragmentation too.
11504
11505 New code allocates 8 megs of address space at once for metadata
11506 allocations. Most OSes will allocate actual memory only when
11507 corresponding pages are touched. Thus this change should not cause
11508 increased memory usage.
11509
11510 I've also made sure metadata is always properly aligned in case we
11511 ever allocate something that breaks natural alignment. E.g. strings.
11512
11513
11514
11515 git-svn-id: http://gperftools.googlecode.com/svn/trunk@211
11516 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11517
11518 M src/common.cc
11519
11520 commit 4fd762cead660d4661359ad507083ac4f4967ee4
11521 Author: [email protected]
11522 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11523 Date: Mon May 6 19:49:48 2013 +0000
11524
11525 issue-368: unmap free spans and retry before growing heap
11526
11527 Because unmapped spans are not coalesced with normal spans it's
11528 possible that we indeed have a large enough free span, but we fail to
11529 see that because we always consider unmapped and normal spans
11530 separately. That behavior is more likely for larger spans.
11531
11532 In order to protect programs that grow heap frequently and by small
11533 amounts from much more frequent minor page faults, there's limit of
11534 running that force pages unmap path once per 128 megs of heap growth.
11535
11536
11537
11538 git-svn-id: http://gperftools.googlecode.com/svn/trunk@210
11539 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11540
11541 M src/page_heap.cc
11542
11543 commit 99fe9944de32046fd954399e60415fba7d03eeb0
11544 Author: [email protected]
11545 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11546 Date: Mon Apr 1 05:32:22 2013 +0000
11547
11548 issue-510: remove duplicate arm #elif
11549
11550
11551
11552 git-svn-id: http://gperftools.googlecode.com/svn/trunk@209
11553 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11554
11555 M src/base/basictypes.h
11556
11557 commit 6354e2c8cdaaaeffdfe7d5b347b125394e2d55fa
11558 Author: [email protected]
11559 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11560 Date: Mon Mar 11 19:24:49 2013 +0000
11561
11562 issue-506: fixed bogus unit test failure
11563
11564 Looks like my version of GCC is aware that free(malloc(X)) is a
11565 no-op. So it optimizes that away completely ignoring simple fact that
11566 we're observing malloc hooks invocations. By adding check that malloc
11567 succeeded we force gcc to actually preserve that malloc call.
11568
11569
11570
11571 git-svn-id: http://gperftools.googlecode.com/svn/trunk@208
11572 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11573
11574 M src/tests/malloc_extension_c_test.c
11575
11576 commit 7896dcb9db2375fcd1d8d19052985a12ad113ba9
11577 Author: [email protected]
11578 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11579 Date: Mon Mar 11 19:16:47 2013 +0000
11580
11581 issue-504: disable tls on mingw
11582
11583 Which is known to fail.
11584
11585
11586
11587 git-svn-id: http://gperftools.googlecode.com/svn/trunk@207
11588 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11589
11590 M configure.ac
11591
11592 commit 8bf4522aa53a7b612a865bc10716192f0f05314b
11593 Author: [email protected]
11594 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11595 Date: Mon Mar 11 19:15:46 2013 +0000
11596
11597 issue-504: do not define HAVE_MMAP on mingw
11598
11599 Because, clearly, windows doesn't have one
11600
11601
11602
11603 git-svn-id: http://gperftools.googlecode.com/svn/trunk@206
11604 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11605
11606 M configure.ac
11607
11608 commit 9c24d255cdd8dab7482d968acc69d493680233ce
11609 Author: [email protected]
11610 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11611 Date: Mon Mar 11 19:15:23 2013 +0000
11612
11613 issue-504: add AM_LDFLAGS to all *_la targets
11614
11615 Because automake will not automatically add AM_LDFLAGS if there's
11616 per-target LDFLAGS. See their good info manual.
11617
11618 This fixes .dll compilation of tcmalloc
11619
11620
11621
11622 git-svn-id: http://gperftools.googlecode.com/svn/trunk@205
11623 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11624
11625 M Makefile.am
11626
11627 commit a2275fc6b48b98d5265388bb8586faa7c8cc581f
11628 Author: [email protected]
11629 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11630 Date: Mon Mar 11 19:14:18 2013 +0000
11631
11632 issue-504: don't try to typedef pid_t on mingw
11633
11634 Because recent mingws (more then few years ago seemingly) do that
11635 already.
11636
11637
11638
11639 git-svn-id: http://gperftools.googlecode.com/svn/trunk@204
11640 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11641
11642 M src/windows/mingw.h
11643 M src/windows/port.h
11644
11645 commit f00977533de6fd27a2f796b2e8e9adb7fcd965e8
11646 Author: [email protected]
11647 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11648 Date: Mon Mar 11 19:13:53 2013 +0000
11649
11650 issue-504: use gcc inline assembly atomic ops on mingw
11651
11652 Because those are well tested and can be trusted
11653
11654
11655
11656 git-svn-id: http://gperftools.googlecode.com/svn/trunk@203
11657 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11658
11659 M Makefile.am
11660 M src/base/atomicops.h
11661
11662 commit 34cdd6821b637b124722687578aa119e8714266c
11663 Author: [email protected]
11664 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11665 Date: Mon Mar 11 19:13:13 2013 +0000
11666
11667 issue-504: use lowercase windows includes for mingw x-compilation
11668
11669 I.e. so that I can build tcmalloc.dll using comfortable environment of
11670 my GNU/Linux box and without having to touch actual windows box or VM.
11671
11672
11673
11674 git-svn-id: http://gperftools.googlecode.com/svn/trunk@202
11675 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11676
11677 M Makefile.am
11678 M src/base/sysinfo.h
11679 M src/windows/patch_functions.cc
11680
11681 commit 0971d80d0ef0c881dcf46453e576a7e289d86975
11682 Author: [email protected]
11683 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11684 Date: Mon Mar 11 18:35:16 2013 +0000
11685
11686 issue-503: updated svn:ignore to ignore generated files
11687
11688
11689
11690 git-svn-id: http://gperftools.googlecode.com/svn/trunk@201
11691 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11692
11693 commit 05b100d9d4ffeb1342eb5ecf50bd0d2339e626c5
11694 Author: [email protected]
11695 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11696 Date: Mon Mar 11 18:32:05 2013 +0000
11697
11698 issue-503: removed checked in configure and other auto* products
11699
11700 So that it's easier to contribute and review contributions.
11701
11702 People wishing to build gperftools from svn checkout should run
11703 autogen.sh first which will create those files and then ./configure &&
11704 make as usual
11705
11706 INSTALL file has updated instructions too.
11707
11708
11709
11710 git-svn-id: http://gperftools.googlecode.com/svn/trunk@200
11711 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11712
11713 M INSTALL
11714 D Makefile.in
11715 D aclocal.m4
11716 M autogen.sh
11717 D compile
11718 D config.guess
11719 D config.sub
11720 D configure
11721 M configure.ac
11722 D depcomp
11723 D install-sh
11724 D ltmain.sh
11725 D m4/libtool.m4
11726 D m4/ltoptions.m4
11727 D m4/ltsugar.m4
11728 D m4/ltversion.m4
11729 D m4/lt~obsolete.m4
11730 D missing
11731 D mkinstalldirs
11732
11733 commit dcf55b976f19a186ac04d60d95e03ed2795285c4
11734 Author: [email protected]
11735 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11736 Date: Mon Mar 11 18:29:00 2013 +0000
11737
11738 issue-503: recover original INSTALL file
11739
11740 That was accidently overwritten by autotools files update in r196
11741
11742
11743
11744 git-svn-id: http://gperftools.googlecode.com/svn/trunk@199
11745 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11746
11747 M INSTALL
11748
11749 commit 5037f687d1de80722f634e84763572caf2725f98
11750 Author: [email protected]
11751 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11752 Date: Mon Mar 11 01:35:37 2013 +0000
11753
11754 issue-425: Fixed tcmalloc unittest crashes for windows builds
11755
11756 Missing use of volatile was causing vs2010 to perform unwanted
11757 optimization resulting in the crash. See issue for more details.
11758
11759
11760 git-svn-id: http://gperftools.googlecode.com/svn/trunk@198
11761 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11762
11763 M src/tests/tcmalloc_unittest.cc
11764
11765 commit beb78cc05babf0a49d21aed0ec789f19fc0f2d28
11766 Author: [email protected]
11767 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11768 Date: Mon Mar 11 00:23:03 2013 +0000
11769
11770 issue-493: Fix for building against ARM targets
11771
11772 gperftools was failing to build for arm targets for the following
11773 reasons:
11774 1. Some ARMv7 instructions used when the target is ARMv6 so those
11775 fail to assemble
11776 2. The cache line length is undefined for ARM architectures
11777
11778
11779 git-svn-id: http://gperftools.googlecode.com/svn/trunk@197
11780 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11781
11782 M src/base/atomicops-internals-arm-v6plus.h
11783 M src/base/basictypes.h
11784
11785 commit bd3b3a7e9a67fac846cf574f5bfd241157bdfe3c
11786 Author: [email protected]
11787 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11788 Date: Sun Mar 10 20:17:21 2013 +0000
11789
11790 issue-496: Fixes an issue where tcmalloc might deadlock in a thread
11791 calling fork
11792
11793 tcmalloc contained a bug where some internal locks were left in a
11794 undefined state
11795 between fork, leaving the child process in a deadlock state. This
11796 patch fixes the
11797 issue by introducing stricter locking between the parent nd child
11798 while forking.
11799
11800
11801 git-svn-id: http://gperftools.googlecode.com/svn/trunk@196
11802 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11803
11804 M INSTALL
11805 M Makefile.in
11806 M aclocal.m4
11807 M config.guess
11808 M config.sub
11809 M configure
11810 M configure.ac
11811 M depcomp
11812 M install-sh
11813 M missing
11814 M mkinstalldirs
11815 M src/central_freelist.h
11816 M src/config.h.in
11817 M src/static_vars.cc
11818
11819 commit 560ca8650c8a9d2971420970f0ed5e17848150eb
11820 Author: [email protected]
11821 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11822 Date: Sun Mar 10 20:02:46 2013 +0000
11823
11824 issue-491: Significant performance improvement for spin lock
11825 contention
11826
11827 This patch fixes issues where spinlocks under contention were
11828 failing to
11829 wakeup waiters, sometimes resulting in blow ups from 13ns to as high
11830 as 256ms.
11831 Under heavy contention, applications were observed sleeping for
11832 minutes at a
11833 time giving the appearance of a hang.
11834
11835
11836 git-svn-id: http://gperftools.googlecode.com/svn/trunk@195
11837 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11838
11839 M src/base/spinlock.h
11840
11841 commit b591d53af951eac60683237204464ebfec2c3afa
11842 Author: [email protected]
11843 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11844 Date: Sun Mar 10 19:54:39 2013 +0000
11845
11846 issue-490: Added support for acquire/release atomic exchange
11847
11848 The atomic ops API was missing support for all of the various
11849 flavors of Acquire_AtomicExchange and Release_AtomicExchange
11850 which have now been added.
11851
11852
11853 git-svn-id: http://gperftools.googlecode.com/svn/trunk@194
11854 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11855
11856 M src/base/atomicops-internals-arm-generic.h
11857 M src/base/atomicops-internals-arm-v6plus.h
11858 M src/base/atomicops-internals-linuxppc.h
11859 M src/base/atomicops-internals-macosx.h
11860 M src/base/atomicops-internals-windows.h
11861 M src/base/atomicops-internals-x86.h
11862 M src/base/atomicops.h
11863 M src/tests/atomicops_unittest.cc
11864
11865 commit 687207b5fcfadc821a3d5b94f2ea112ca933e476
11866 Author: [email protected]
11867 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11868 Date: Sun Mar 10 19:44:43 2013 +0000
11869
11870 issue-443: Increase the number of objects transferred between thread
11871 cache and central free list
11872
11873 This fix is a result of a performance degradation observed in
11874 multi-threaded programs where large
11875 amounts of memory (30GB) are consumed, released by a pool of threads
11876 in a cyclic manner. This was
11877 mainly due to the amount of time we were spending in the slow path
11878 consolidating memory between
11879 the thread cache and central free list. The default has been bumped
11880 up to 32768 and is now also
11881 controllable through the TCMALLOC_TRANSFER_NUM_OBJ environment
11882 setting.
11883
11884
11885 git-svn-id: http://gperftools.googlecode.com/svn/trunk@193
11886 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11887
11888 M Makefile.am
11889 M Makefile.in
11890 M src/common.cc
11891 M src/common.h
11892 A src/tests/tcmalloc_unittest.sh
11893
11894 commit 19eff86f850e54683c8868316e362af8919df18a
11895 Author: [email protected]
11896 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11897 Date: Sat Mar 9 01:16:17 2013 +0000
11898
11899 actually made svn:ignore work
11900
11901
11902
11903 git-svn-id: http://gperftools.googlecode.com/svn/trunk@192
11904 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11905
11906 commit 177a896a8273ae73affc48a126fcf9a3ba91f502
11907 Author: [email protected]
11908 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11909 Date: Sat Mar 9 01:06:58 2013 +0000
11910
11911 added proper svn:ignore
11912
11913
11914
11915 git-svn-id: http://gperftools.googlecode.com/svn/trunk@191
11916 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11917
11918 commit b96203b591f63d24cd2979241b18534d22358bca
11919 Author: [email protected]
11920 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11921 Date: Sat Dec 22 20:14:08 2012 +0000
11922
11923 issue-461: Fix to malloc_extension.h so that it builds with
11924 -std=gnu++98
11925
11926
11927 git-svn-id: http://gperftools.googlecode.com/svn/trunk@190
11928 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11929
11930 M src/gperftools/malloc_extension.h
11931
11932 commit 84b983c8d43f43a3c7f71d45d51fc4adcc688cd9
11933 Author: [email protected]
11934 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11935 Date: Sat Dec 22 20:06:47 2012 +0000
11936
11937 issue-465: Adding automagic support for __builtin_expect
11938
11939 Previously __builtin_ expect was based on a macro check against
11940 gcc version.
11941 Now we perform the check via AM which is a cleaner approach. There
11942 are also
11943 a number of code changes here to utilize LIKELY/UNLIKELY macros
11944 based on
11945 __builtin_expect to improve performance.
11946
11947
11948 git-svn-id: http://gperftools.googlecode.com/svn/trunk@189
11949 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11950
11951 M configure
11952 M configure.ac
11953 M src/base/basictypes.h
11954 M src/common.h
11955 M src/config.h.in
11956 M src/tcmalloc.cc
11957
11958 commit a5dacccd6ae4cbfedb5263bfe0f325f03c7f0db8
11959 Author: [email protected]
11960 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11961 Date: Sat Dec 22 19:06:35 2012 +0000
11962
11963 issue-466: Clarified stats output and comments for ExtractStats()
11964 and GetThreadStats()
11965
11966
11967 git-svn-id: http://gperftools.googlecode.com/svn/trunk@188
11968 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11969
11970 M src/tcmalloc.cc
11971 M src/thread_cache.h
11972
11973 commit 09d97533b09e473c0cdd269e8cf4e9a9737e49fa
11974 Author: [email protected]
11975 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11976 Date: Sat Dec 22 19:02:52 2012 +0000
11977
11978 issue-467: Fixed issue with allocation size being narrowed to 32-bit
11979
11980
11981
11982 git-svn-id: http://gperftools.googlecode.com/svn/trunk@187
11983 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
11984
11985 M src/tcmalloc.cc
11986
11987 commit 990889e6232ff3787f1d42d4091a0478ffb93988
11988 Author: [email protected]
11989 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
11990 Date: Sat Dec 22 18:53:18 2012 +0000
11991
11992 issue-470: Allows the default system allocator to be overidden
11993
11994 This is useful when you run into cases where you need an allocator
11995 before tcmalloc has been fully initialized.
11996
11997
11998 git-svn-id: http://gperftools.googlecode.com/svn/trunk@186
11999 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12000
12001 M src/system-alloc.cc
12002
12003 commit 3bf5f55d78769d2f8525626dd662457d2f014151
12004 Author: [email protected]
12005 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12006 Date: Sat Dec 22 18:38:48 2012 +0000
12007
12008 issue-482: Adds support for CACHELINE_ALIGNED value for arm based
12009 devices
12010
12011 Hardcoded the cacheline size to 32 because this value is supported
12012 by several
12013 arm devices, however a better logic needs to be implemented at some
12014 point to
12015 handle more arm devices.
12016
12017
12018 git-svn-id: http://gperftools.googlecode.com/svn/trunk@185
12019 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12020
12021 M src/base/basictypes.h
12022
12023 commit 6856d1d1b2a807efd84dcafa10b865b6de22bc28
12024 Author: [email protected]
12025 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12026 Date: Sat Dec 22 18:34:43 2012 +0000
12027
12028 issue-488: Performance improvement for initialization checks
12029
12030 These came in from the Google-internal version of tcmalloc. They
12031 saw some
12032 decent speed improvements, both on microbenchmarks and big
12033 programs. The
12034 idea is to improve the speed of the "Is everything initialized?" type
12035 of
12036 code, that's at the start of all allocations and deallocations.
12037
12038
12039 git-svn-id: http://gperftools.googlecode.com/svn/trunk@184
12040 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12041
12042 M src/static_vars.h
12043 M src/tcmalloc.cc
12044 M src/thread_cache.cc
12045 M src/thread_cache.h
12046
12047 commit ad5aa05838121d52ad1fde5463a796c3320fe067
12048 Author: [email protected]
12049 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12050 Date: Sat Dec 22 18:25:58 2012 +0000
12051
12052 issue-483: Speed up accesses to ClassIndex()
12053
12054 Making its return type unsigned can save a conversion from
12055 signed to unsigned, and getting rid of the ternary operators
12056 seems to help a little bit as well. Various gcc versions weren't
12057 generating conditional moves for them as one would expect.
12058
12059
12060 git-svn-id: http://gperftools.googlecode.com/svn/trunk@183
12061 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12062
12063 M src/base/basictypes.h
12064 M src/common.h
12065
12066 commit 8de78fd85b69bc569ac8fc9e75144e02f5cae851
12067 Author: [email protected]
12068 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12069 Date: Mon Nov 5 04:45:01 2012 +0000
12070
12071 issue-452: Adds a control mechanism to the cpu profiler to be switched
12072 on and off using a user defined signal.
12073
12074 CPUPROFILESIGNAL - Takes a signal number between the value of 1 and
12075 64 inclusive which represents a signal
12076 number as defined by signal.h. The signal must
12077 not be in use by the program. Sending this
12078 signal to the program turns profiling on and off
12079 like a switch. By default the switch is
12080 off when the program starts. Successive profiles
12081 are suffixed with a monotonically increasing
12082 number.
12083
12084
12085 git-svn-id: http://gperftools.googlecode.com/svn/trunk@182
12086 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12087
12088 M doc/cpuprofile.html
12089 M src/profiler.cc
12090
12091 commit 46f06ca0db41d3f598750caefdaf3c6063969e64
12092 Author: [email protected]
12093 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12094 Date: Sun Nov 4 23:08:17 2012 +0000
12095
12096 issue-451: Fixed incorrect assembly for 64-bit barrier load and
12097 store on windows platforms.
12098
12099
12100 git-svn-id: http://gperftools.googlecode.com/svn/trunk@181
12101 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12102
12103 M src/base/atomicops-internals-windows.h
12104
12105 commit 5fe91d5623c2351ba4675db71822fc6be5e2cbce
12106 Author: [email protected]
12107 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12108 Date: Sun Nov 4 22:53:01 2012 +0000
12109
12110 issue-450: Move includes for struct mallinfo from tcmalloc.cc to
12111 tcmalloc.h to fix compiler warnings from clang.
12112
12113
12114 git-svn-id: http://gperftools.googlecode.com/svn/trunk@180
12115 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12116
12117 M src/gperftools/tcmalloc.h.in
12118 M src/tcmalloc.cc
12119
12120 commit 86a55316baf2a7c0718670fd7342b648e0ef4a87
12121 Author: [email protected]
12122 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12123 Date: Sun Nov 4 19:12:42 2012 +0000
12124
12125 issue-449: Disables mmap check when cross compiling since it is
12126 meaningless anyways
12127
12128 git-svn-id: http://gperftools.googlecode.com/svn/trunk@179
12129 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12130
12131 M configure
12132 M configure.ac
12133
12134 commit 6f6c2bf68576e43491884364fc05c9ed752adb4f
12135 Author: [email protected]
12136 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12137 Date: Sun Nov 4 19:02:15 2012 +0000
12138
12139 issue-448: New environment setting that allows you to set an explicit
12140 heap limit
12141
12142 TCMALLOC_HEAP_LIMIT_MB - The maximum amount of heap memory that
12143 tcmalloc can use.
12144 TCMALLOC_DISABLE_MEMORY_RELEASE - emulate platform with no
12145 MADV_DONTNEED support (generally for testing purposes).
12146
12147
12148 git-svn-id: http://gperftools.googlecode.com/svn/trunk@178
12149 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12150
12151 M src/page_heap.cc
12152 M src/page_heap.h
12153 M src/system-alloc.cc
12154 M src/system-alloc.h
12155 M src/tests/page_heap_test.cc
12156 M src/tests/tcmalloc_unittest.cc
12157 M src/windows/port.cc
12158
12159 commit 644b1c6e355a0f5dd948ca482a575f49a4bd2032
12160 Author: [email protected]
12161 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12162 Date: Sun Nov 4 18:30:05 2012 +0000
12163
12164 issue-445: Fix for wrong cmpx instruction in OSAtomicCompareAndSwap64
12165 for ppc64
12166
12167
12168 git-svn-id: http://gperftools.googlecode.com/svn/trunk@177
12169 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12170
12171 M src/base/atomicops-internals-linuxppc.h
12172
12173 commit e32bb2d9a76dc5ee42ee41a2ca2c6e5caece0150
12174 Author: [email protected]
12175 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12176 Date: Sun Nov 4 18:24:46 2012 +0000
12177
12178 issue-444: Fix for invalid conversion build error in signal handler
12179 code
12180
12181
12182 git-svn-id: http://gperftools.googlecode.com/svn/trunk@176
12183 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12184
12185 M src/base/linux_syscall_support.h
12186
12187 commit abeaf46028c8dfab7e7867ee7a3a49ebe21cf129
12188 Author: [email protected]
12189 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12190 Date: Sun Nov 4 18:15:11 2012 +0000
12191
12192 issue-430: Introduces 8-byte alignment support for tcmalloc
12193
12194
12195 git-svn-id: http://gperftools.googlecode.com/svn/trunk@175
12196 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12197
12198 M src/common.cc
12199 M src/common.h
12200 M src/tests/tcmalloc_unittest.cc
12201
12202 commit baaf0188295582ca68df03f70baa13d96a88e2eb
12203 Author: [email protected]
12204 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12205 Date: Sun Nov 4 17:41:47 2012 +0000
12206
12207 issue-473: Fix for make install error regarding
12208 src/windows/google/tcmalloc.h
12209
12210
12211 git-svn-id: http://gperftools.googlecode.com/svn/trunk@174
12212 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12213
12214 M Makefile.am
12215 M Makefile.in
12216
12217 commit 1363bc69550f43af693f6455118432988042a8dd
12218 Author: [email protected]
12219 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12220 Date: Sun Nov 4 17:21:00 2012 +0000
12221
12222 issue-475: Re-applying changes made originally for r150 and r151
12223
12224 - Used aclocal, autoupdate, autoconf, and automake to
12225 correctly apply changes made to Makefile.am. Detailed
12226 instructions on this procedure can be found here
12227 http://inti.sourceforge.net/tutorial/libinti/autotoolsproject.html.
12228 - Fixed a number of error/warning messages due to use of newer
12229 aclocal, autoconf, and automake utilities.
12230 - Directly and indirectly related to issue-385 and issue-480.
12231
12232
12233
12234 git-svn-id: http://gperftools.googlecode.com/svn/trunk@173
12235 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12236
12237 M Makefile.am
12238 M Makefile.in
12239 M aclocal.m4
12240 M configure
12241 M configure.ac
12242 M m4/libtool.m4
12243
12244 commit 1243295cdcc20f2f47df9bdc9b8bb50b69c085b0
12245 Author: [email protected]
12246 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12247 Date: Sun Nov 4 15:59:08 2012 +0000
12248
12249 issue-475: Amend of r151 and r150
12250
12251 In revisions 151 and 150 an attempt was made to enable frame pointers
12252 by default for i386. However, in the process of doing so a number of
12253 files were inadvertently touched as a result of running autogen.sh. As
12254 a result, I have needed to roll back these revisions so that I can
12255 reattempt the change.
12256
12257
12258 git-svn-id: http://gperftools.googlecode.com/svn/trunk@172
12259 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12260
12261 M INSTALL
12262 M Makefile.am
12263 M Makefile.in
12264 M aclocal.m4
12265 M config.guess
12266 M config.sub
12267 M configure
12268 M configure.ac
12269 M depcomp
12270 M install-sh
12271 M ltmain.sh
12272 M m4/libtool.m4
12273 M m4/ltoptions.m4
12274 M m4/ltversion.m4
12275 M m4/lt~obsolete.m4
12276 M missing
12277 M mkinstalldirs
12278
12279 commit 5c17459c5e7b4133a2dec5ffe6a9237a114f5378
12280 Author: [email protected]
12281 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12282 Date: Sat Nov 3 16:14:25 2012 +0000
12283
12284 issue-476: Amendment to r152:issue-437 to fix a build error due to
12285 PR_SET_TRACER not being defined in older glibc releases.
12286
12287 git-svn-id: http://gperftools.googlecode.com/svn/trunk@171
12288 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12289
12290 M src/base/linuxthreads.cc
12291
12292 commit 6287bbbbad8730712cfd1ee28ecc0648cbaa9f94
12293 Author: [email protected]
12294 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12295 Date: Sat Nov 3 15:56:27 2012 +0000
12296
12297 issue-477: Fix clang compilation errors regarding format specifiers
12298
12299 git-svn-id: http://gperftools.googlecode.com/svn/trunk@170
12300 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12301
12302 M src/base/sysinfo.cc
12303 M src/debugallocation.cc
12304 M src/getpc.h
12305 M src/heap-checker.cc
12306 M src/heap-profile-table.cc
12307 M src/heap-profiler.cc
12308 M src/malloc_extension.cc
12309 M src/memory_region_map.cc
12310 M src/symbolize.cc
12311 M src/tcmalloc.cc
12312 M src/tests/markidle_unittest.cc
12313 M src/tests/tcmalloc_unittest.cc
12314
12315 commit 87699c97cc07b30fc2765415853ae685f7bc45f2
12316 Author: [email protected]
12317 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12318 Date: Sat Nov 3 15:40:01 2012 +0000
12319
12320 issue-478: Added the heap profiler tunnable setting
12321 HEAP_PROFILER_TIME_INTERVAL used to dump heap profiles on a specified
12322 time interval.
12323
12324 git-svn-id: http://gperftools.googlecode.com/svn/trunk@169
12325 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12326
12327 M doc/heapprofile.html
12328 M src/heap-profiler.cc
12329
12330 commit 4eb6946d689bbaa74f8be78265f6d6ff9ae04ab2
12331 Author: [email protected]
12332 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12333 Date: Sat Nov 3 15:27:58 2012 +0000
12334
12335 issue-479: Replaced sscanf usage while iterating proc maps entries
12336 to fix a deadlock introduced by glibc-2.15.
12337
12338 git-svn-id: http://gperftools.googlecode.com/svn/trunk@168
12339 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12340
12341 M src/base/sysinfo.cc
12342
12343 commit 5dd53ab6cbf9d98f2d60546835e84785a104da46
12344 Author: [email protected]
12345 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12346 Date: Sat Nov 3 14:52:42 2012 +0000
12347
12348 issue-481: Replaced test mechanism for distinct address spaces with
12349 a more reliable mechanism
12350
12351 Rather than using sys_ptrace+PTRACE_PEEK_DATA to determine whether
12352 address spaces are distinct, we now use sys_waitpid+__WCLONE. See
12353 issue-481 for a more detailed rationale.
12354
12355
12356 git-svn-id: http://gperftools.googlecode.com/svn/trunk@167
12357 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12358
12359 M src/base/linuxthreads.cc
12360
12361 commit 6c436203522156699ef11c6792ed481e02cef366
12362 Author: [email protected]
12363 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12364 Date: Sat Nov 3 14:36:33 2012 +0000
12365
12366 amend-r165: forgot to regenrate configure and Makefile.in
12367
12368 git-svn-id: http://gperftools.googlecode.com/svn/trunk@166
12369 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12370
12371 M Makefile.in
12372 M configure
12373
12374 commit c566266b10d2fc4e0c67fec25e2a45ebad803196
12375 Author: [email protected]
12376 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12377 Date: Sat Nov 3 14:13:21 2012 +0000
12378
12379 issue-480: duplicate of issue-385 fixed in r150 however some of the
12380 comments in the code
12381 regarding frame pointers needed to be clarified.
12382
12383
12384 git-svn-id: http://gperftools.googlecode.com/svn/trunk@165
12385 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12386
12387 M Makefile.am
12388 M configure.ac
12389
12390 commit 4f325fbcc9f0db17fbf108188a32b2c67bcd38e0
12391 Author: [email protected]
12392 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12393 Date: Sun Oct 28 19:45:03 2012 +0000
12394
12395 issue-424: Updated heap profiler documentation that had
12396 mispelling for the heap profiler mmap only environment setting
12397 (HEAP_PROFILE_ONLY_MMAP)
12398
12399 git-svn-id: http://gperftools.googlecode.com/svn/trunk@164
12400 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12401
12402 M doc/heapprofile.html
12403
12404 commit a828c0a8ff59300fffce466cc617d40fa39dedea
12405 Author: [email protected]
12406 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12407 Date: Sun Oct 28 19:30:23 2012 +0000
12408
12409 issue-417: This commit removed the fatal error message printed along
12410 with the pprof usage information when pprof is run with no arguments.
12411
12412 git-svn-id: http://gperftools.googlecode.com/svn/trunk@163
12413 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12414
12415 M src/pprof
12416
12417 commit e5b095abdc2368111f53cedf376cd1505a2c7583
12418 Author: [email protected]
12419 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12420 Date: Sun Oct 28 18:28:21 2012 +0000
12421
12422 issue-411: This commit adds additional logging to the cpu profiler
12423 to warn when the profiler is run and no CPUPROFILE environment
12424 setting can be found. It also adds a new environment variable
12425 PERFTOOLS_UNITTEST to allow certain modules to take action when
12426 running under the umbrella of a unit test.
12427
12428 git-svn-id: http://gperftools.googlecode.com/svn/trunk@162
12429 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12430
12431 M src/profiler.cc
12432 M src/tests/profiler_unittest.sh
12433
12434 commit 57c48e9b5ffdb0f177c07ea8420c4072e816e1a7
12435 Author: [email protected]
12436 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12437 Date: Sun Oct 28 14:51:41 2012 +0000
12438
12439 issue-405: backed out original commit made in revision 147 due to
12440 the fact that it didn't work as intended
12441
12442 git-svn-id: http://gperftools.googlecode.com/svn/trunk@161
12443 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12444
12445 M src/heap-checker.cc
12446
12447 commit 3e296c28c3c2b67df624d372b75650ae65c87b04
12448 Author: [email protected]
12449 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12450 Date: Tue Sep 18 02:47:24 2012 +0000
12451
12452 issue-442: Fixes an issue encountered on PPC with undefined
12453 conversions from 'base::subtle::Atomic64*' to 'const volatile
12454 Atomic32*'
12455
12456 git-svn-id: http://gperftools.googlecode.com/svn/trunk@160
12457 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12458
12459 M src/base/spinlock_internal.cc
12460
12461 commit 1f59bb34d3ce03df51c8356dd44fcb0b15de4769
12462 Author: [email protected]
12463 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12464 Date: Tue Sep 18 02:33:00 2012 +0000
12465
12466 Adding support for CPU frequencies for Linux on PowerPC
12467
12468 git-svn-id: http://gperftools.googlecode.com/svn/trunk@159
12469 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12470
12471 M src/base/sysinfo.cc
12472
12473 commit a2fec7c694ce992a3b7d3fe4205bb8710b017bab
12474 Author: [email protected]
12475 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12476 Date: Tue Sep 18 01:43:02 2012 +0000
12477
12478 issue-446: Fixes compatability issues in malloc_hook_test for PPC64
12479
12480 git-svn-id: http://gperftools.googlecode.com/svn/trunk@158
12481 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12482
12483 M src/tests/malloc_hook_test.cc
12484
12485 commit b05b6fc7655ba7fee8397da7013db007f0c47e30
12486 Author: [email protected]
12487 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12488 Date: Tue Sep 18 01:27:34 2012 +0000
12489
12490 issue-458: Optimizes malloc hooks for weakly consistent memory
12491 architectures
12492
12493 git-svn-id: http://gperftools.googlecode.com/svn/trunk@157
12494 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12495
12496 M src/malloc_hook-inl.h
12497
12498 commit 00edf667f62e17a268f03c6aaef19946add75dcf
12499 Author: [email protected]
12500 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12501 Date: Tue Sep 18 01:18:15 2012 +0000
12502
12503 issue-456: Adds support for CACHELINE_ALIGNED values for PPC
12504
12505 git-svn-id: http://gperftools.googlecode.com/svn/trunk@156
12506 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12507
12508 M src/base/basictypes.h
12509
12510 commit 731d0932a081d521a36462df7dcbc45fa4221011
12511 Author: [email protected]
12512 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12513 Date: Tue Sep 18 00:59:16 2012 +0000
12514
12515 issue-455: Adds another size for large pages in tcmalloc through
12516 the new define TCMALLOC_LARGE_PAGES64K
12517
12518 git-svn-id: http://gperftools.googlecode.com/svn/trunk@155
12519 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12520
12521 M src/common.h
12522
12523 commit 771b6bcf689bd089b802ddcdb27a40b4fb4a8e3b
12524 Author: [email protected]
12525 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12526 Date: Tue Sep 18 00:52:13 2012 +0000
12527
12528 issue-457: Fixed an issue where the 'getpc' testcase was using the
12529 wrong PPC64 preprocessor directive
12530
12531 git-svn-id: http://gperftools.googlecode.com/svn/trunk@154
12532 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12533
12534 M src/tests/getpc_test.cc
12535
12536 commit 711232a1ef4997305a681287191de74f6597dabc
12537 Author: [email protected]
12538 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12539 Date: Tue Sep 18 00:42:23 2012 +0000
12540
12541 issue-453 Added support to get the timebase register value using
12542 just one instruction and also adjusts the PPC32 code to the recent
12543 GLIBC one that implements the same functionality
12544
12545 git-svn-id: http://gperftools.googlecode.com/svn/trunk@153
12546 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12547
12548 M src/base/cycleclock.h
12549
12550 commit fa0209f261c5e065d523bb1858f84fd91eb2f39a
12551 Author: [email protected]
12552 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12553 Date: Tue Sep 18 00:00:20 2012 +0000
12554
12555 issue-437 Fixed issues related to new glibc shipped with Ubuntu 10.10
12556
12557 1. ptrace permissions were modifed to be a bit more strict which
12558 required
12559 us to programatically set the permissions while syncing up to
12560 the profiling
12561 thread.
12562
12563 2. Order of destructors registered with atexit changed which was
12564 casuing us to
12565 miss generating the backtrace when heap checker was finished. Seems
12566 that we
12567 initially fixed this for FreeBSD and now linux has changed their
12568 behaviour
12569 to be the same. We are now a bit stricter on the rules here
12570 accross all
12571 platforms.
12572
12573
12574 git-svn-id: http://gperftools.googlecode.com/svn/trunk@152
12575 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12576
12577 M src/base/linuxthreads.cc
12578 M src/heap-checker.cc
12579
12580 commit cd723b43ff783a05321f0c0ba79a82494185b23c
12581 Author: [email protected]
12582 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12583 Date: Wed May 16 12:24:13 2012 +0000
12584
12585 issue-385: updates to automake to pick up the explicit setting of
12586 the -fno-omit-frame-pointer for i386
12587
12588 git-svn-id: http://gperftools.googlecode.com/svn/trunk@151
12589 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12590
12591 M INSTALL
12592 M Makefile.am
12593 M Makefile.in
12594 M aclocal.m4
12595 M config.guess
12596 M config.sub
12597 M configure
12598 M configure.ac
12599 M depcomp
12600 M install-sh
12601 M ltmain.sh
12602 M m4/libtool.m4
12603 M m4/ltoptions.m4
12604 M m4/ltversion.m4
12605 M m4/lt~obsolete.m4
12606 M missing
12607 M mkinstalldirs
12608
12609 commit 0c0382234d5812474b070b6eccdfd99c98490548
12610 Author: [email protected]
12611 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12612 Date: Wed May 16 01:28:08 2012 +0000
12613
12614 issue-385: as of gcc 4.6 we now need to enable frame pointers for
12615 i386 platforms since the default is to now omit frame pointers
12616
12617 git-svn-id: http://gperftools.googlecode.com/svn/trunk@150
12618 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12619
12620 M Makefile.am
12621
12622 commit bafd0f8b7cec4dd0b74a4a8e5456b96744640679
12623 Author: [email protected]
12624 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12625 Date: Sat Apr 21 17:27:54 2012 +0000
12626
12627 Changing internal behaviour of tcmalloc to return an
12628 allocated size of 0 when the target pointer is null (refer to
12629 TCMallocImplementation::GetAllocatedSize). This change was reviewed
12630 as https://codereview.appspot.com/5833050/.
12631
12632 git-svn-id: http://gperftools.googlecode.com/svn/trunk@149
12633 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12634
12635 M src/tcmalloc.cc
12636
12637 commit 9be2f8d802a19ab6e4b2abed3f7cd84780628d16
12638 Author: [email protected]
12639 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12640 Date: Sat Apr 21 16:15:20 2012 +0000
12641
12642 Adding in further support for 64-bit NoBarrier_Load on windows
12643
12644 git-svn-id: http://gperftools.googlecode.com/svn/trunk@148
12645 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12646
12647 M src/base/atomicops-internals-windows.h
12648
12649 commit 68df7fa6057a29b3130d41a27592f74d58c2df57
12650 Author: [email protected]
12651 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12652 Date: Sat Mar 3 19:45:49 2012 +0000
12653
12654 issue-405: Added a tunable to the heap checker that allows the heap
12655 checker error exit code to be overridden. For larger projects this
12656 allows certain heap checked programs that have intentional/known
12657 leaks to be managed.
12658
12659 git-svn-id: http://gperftools.googlecode.com/svn/trunk@147
12660 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12661
12662 M src/heap-checker.cc
12663
12664 commit 0a9d3f8af373c71e5fabdc8c01239c47df4743ed
12665 Author: [email protected]
12666 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12667 Date: Fri Mar 2 06:04:08 2012 +0000
12668
12669 issue-410: This set of changes exposes the central cache free list
12670 size, transfer cache free list size, and the meta data overhead size
12671 through the getNumericProperty() function.
12672
12673 git-svn-id: http://gperftools.googlecode.com/svn/trunk@146
12674 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12675
12676 M src/gperftools/malloc_extension.h
12677 M src/tcmalloc.cc
12678
12679 commit 1233f0f9101c3f4e85e04711c7223dc7f73ec4fb
12680 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12681 Date: Tue Feb 21 03:02:43 2012 +0000
12682
12683 This file is in the tarball, but I forgot to put it in svn as well.
12684 (It's tricky because it's auto-generated from tcmalloc.h.in at
12685 tarball-making time.)
12686
12687
12688 git-svn-id: http://gperftools.googlecode.com/svn/trunk@145
12689 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12690
12691 A src/windows/gperftools/tcmalloc.h
12692
12693 commit ce2891234584069f90e6e0a446e28bf7ca3e031d
12694 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12695 Date: Sat Feb 4 00:10:11 2012 +0000
12696
12697 Url-typo in the latest NEWS message.
12698
12699
12700 git-svn-id: http://gperftools.googlecode.com/svn/trunk@144
12701 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12702
12703 M NEWS
12704
12705 commit 4e9432c5090d5b8354b0b57fca8b96b2efef37e3
12706 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12707 Date: Sat Feb 4 00:07:36 2012 +0000
12708
12709 Fri Feb 03 15:40:45 2012 Google
12710 Inc. <[email protected]>
12711
12712 * gperftools: version 2.0
12713 * Renamed the project from google-perftools to gperftools
12714 (csilvers)
12715 * Renamed the .deb/.rpm packagse from google-perftools to
12716 gperftools too
12717 * Renamed include directory from google/ to gperftools/
12718 (csilvers)
12719 * Changed the 'official' perftools email in setup.py/etc
12720 * Renamed google-perftools.sln to gperftools.sln
12721 * PORTING: Removed bash-isms & grep -q in
12722 heap-checker-death_unittest.sh
12723 * Changed copyright text to reflect Google's relinquished
12724 ownership
12725
12726
12727 git-svn-id: http://gperftools.googlecode.com/svn/trunk@142
12728 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12729
12730 M AUTHORS
12731 M ChangeLog
12732 M INSTALL
12733 M Makefile.am
12734 M Makefile.in
12735 M NEWS
12736 M README
12737 M README_windows.txt
12738 M aclocal.m4
12739 M config.guess
12740 M config.sub
12741 M configure
12742 M configure.ac
12743 M depcomp
12744 M doc/cpuprofile.html
12745 M doc/designstyle.css
12746 M doc/heap_checker.html
12747 M doc/heapprofile.html
12748 M doc/index.html
12749 M doc/pprof.1
12750 M doc/pprof.see_also
12751 M doc/pprof_remote_servers.html
12752 R100 google-perftools.sln gperftools.sln
12753 M install-sh
12754 M missing
12755 M mkinstalldirs
12756 M packages/deb/changelog
12757 M packages/deb/control
12758 M packages/deb/copyright
12759 R071 packages/deb/libgoogle-perftools-dev.dirs
12760 packages/deb/libgperftools-dev.dirs
12761 R079 packages/deb/libgoogle-perftools-dev.install
12762 packages/deb/libgperftools-dev.install
12763 R100 packages/deb/libgoogle-perftools0.dirs
12764 packages/deb/libgperftools0.dirs
12765 R100 packages/deb/libgoogle-perftools0.install
12766 packages/deb/libgperftools0.install
12767 R100 packages/deb/libgoogle-perftools0.manpages
12768 packages/deb/libgperftools0.manpages
12769 M packages/rpm/rpm.spec
12770 M src/base/low_level_alloc.cc
12771 M src/config.h.in
12772 M src/debugallocation.cc
12773 M src/google/heap-checker.h
12774 M src/google/heap-profiler.h
12775 M src/google/malloc_extension.h
12776 M src/google/malloc_extension_c.h
12777 M src/google/malloc_hook.h
12778 M src/google/malloc_hook_c.h
12779 M src/google/profiler.h
12780 M src/google/stacktrace.h
12781 A src/google/tcmalloc.h
12782 C099 src/google/heap-checker.h src/gperftools/heap-checker.h
12783 C100 src/google/heap-profiler.h src/gperftools/heap-profiler.h
12784 C100 src/google/malloc_extension.h src/gperftools/malloc_extension.h
12785 C100 src/google/malloc_extension_c.h
12786 src/gperftools/malloc_extension_c.h
12787 C099 src/google/malloc_hook.h src/gperftools/malloc_hook.h
12788 C100 src/google/malloc_hook_c.h src/gperftools/malloc_hook_c.h
12789 C100 src/google/profiler.h src/gperftools/profiler.h
12790 C100 src/google/stacktrace.h src/gperftools/stacktrace.h
12791 R098 src/google/tcmalloc.h.in src/gperftools/tcmalloc.h.in
12792 M src/heap-checker-bcad.cc
12793 M src/heap-checker.cc
12794 M src/heap-profile-table.cc
12795 M src/heap-profiler.cc
12796 M src/internal_logging.cc
12797 M src/libc_override.h
12798 M src/libc_override_gcc_and_weak.h
12799 M src/libc_override_glibc.h
12800 M src/libc_override_osx.h
12801 M src/malloc_extension.cc
12802 M src/malloc_hook-inl.h
12803 M src/malloc_hook.cc
12804 M src/memfs_malloc.cc
12805 M src/memory_region_map.cc
12806 M src/page_heap.cc
12807 M src/page_heap.h
12808 M src/pprof
12809 M src/profiler.cc
12810 M src/stacktrace.cc
12811 M src/stacktrace_arm-inl.h
12812 M src/stacktrace_generic-inl.h
12813 M src/stacktrace_libunwind-inl.h
12814 M src/stacktrace_powerpc-inl.h
12815 M src/stacktrace_win32-inl.h
12816 M src/stacktrace_x86-inl.h
12817 M src/system-alloc.cc
12818 M src/tcmalloc.cc
12819 M src/tests/current_allocated_bytes_test.cc
12820 M src/tests/debugallocation_test.cc
12821 M src/tests/frag_unittest.cc
12822 M src/tests/heap-checker-death_unittest.sh
12823 M src/tests/heap-checker_unittest.cc
12824 M src/tests/heap-profiler_unittest.cc
12825 M src/tests/low_level_alloc_unittest.cc
12826 M src/tests/malloc_extension_c_test.c
12827 M src/tests/malloc_extension_test.cc
12828 M src/tests/malloc_hook_test.cc
12829 M src/tests/markidle_unittest.cc
12830 M src/tests/profiler_unittest.cc
12831 M src/tests/sampling_test.cc
12832 C059 src/tests/current_allocated_bytes_test.cc
12833 src/tests/simple_compat_test.cc
12834 M src/tests/stacktrace_unittest.cc
12835 M src/tests/system-alloc_unittest.cc
12836 M src/tests/tcmalloc_unittest.cc
12837 M src/tests/thread_dealloc_unittest.cc
12838 A src/windows/auto_testing_hook.h
12839 M src/windows/config.h
12840 A src/windows/google/tcmalloc.h
12841 R098 src/windows/google/tcmalloc.h.in
12842 src/windows/gperftools/tcmalloc.h.in
12843 M src/windows/patch_functions.cc
12844 M
12845 vsprojects/current_allocated_bytes_test/current_allocated_bytes_test.vcproj
12846 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
12847 M
12848 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
12849 M vsprojects/malloc_extension_test/malloc_extension_test.vcproj
12850 M vsprojects/malloc_hook_test/malloc_hook_test.vcproj
12851 M vsprojects/tmu-static/tmu-static.vcproj
12852
12853 commit e10b720797359e1855a5e4769424988b4e76c3e7
12854 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12855 Date: Sat Feb 4 00:06:48 2012 +0000
12856
12857 Last of the remnant empty directories that I don't know why they're
12858 there.
12859
12860
12861 git-svn-id: http://gperftools.googlecode.com/svn/trunk@141
12862 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12863
12864 commit e55949bba69b7a6eede3468765e86c5ed000be52
12865 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12866 Date: Sat Feb 4 00:06:06 2012 +0000
12867
12868 This directory got renamed tmu-static a long time ago. Get rid of
12869 this ancient remnant.
12870
12871
12872 git-svn-id: http://gperftools.googlecode.com/svn/trunk@140
12873 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12874
12875 commit 0f588815f979721346a6cb05a1cbd83d670b975a
12876 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12877 Date: Sat Feb 4 00:05:39 2012 +0000
12878
12879 Apparently I'm not good at deleting files from svn after they're no
12880 longer used anymore. Get rid of a few more that fall into this
12881 category.
12882
12883
12884 git-svn-id: http://gperftools.googlecode.com/svn/trunk@139
12885 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12886
12887 D src/stacktrace_nacl-inl.h
12888 D src/stacktrace_x86_64-inl.h
12889
12890 commit 2b2016f6844a88b9448d445f90ad1c849a56ea23
12891 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12892 Date: Sat Feb 4 00:03:35 2012 +0000
12893
12894 I have no idea why I created this directory, but it's never been
12895 used. Delete it.
12896
12897
12898 git-svn-id: http://gperftools.googlecode.com/svn/trunk@138
12899 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12900
12901 commit ace2da06732ee549b6d64c1f59b62181f5110230
12902 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12903 Date: Sat Feb 4 00:02:15 2012 +0000
12904
12905 This file has been obsolete for almost a year now; I just forgot to
12906 delete it from svn.
12907
12908
12909 git-svn-id: http://gperftools.googlecode.com/svn/trunk@137
12910 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12911
12912 D src/base/atomicops-internals-arm-gcc.h
12913
12914 commit 0afb078b34700859cef8d708463fc5c955e26c31
12915 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12916 Date: Tue Jan 31 19:11:26 2012 +0000
12917
12918 Tue Jan 31 10:43:50 2012 Google Inc. <[email protected]>
12919
12920 * google-perftools: version 1.10 release
12921 * PORTING: Support for patching assembly on win
12922 x86_64! (scott.fr...)
12923 * PORTING: Work around atexit-execution-order bug on freebsd
12924 (csilvers)
12925 * PORTING: Patch _calloc_crt for windows (roger orr)
12926 * PORTING: Add C++11 compatibility method for stl allocator
12927 (jdennett)
12928 * PORTING: use MADV_FREE, not MADV_DONTNEED, on freebsd
12929 (csilvers)
12930 * PORTING: Don't use SYS_open when not supported on solaris
12931 (csilvers)
12932 * PORTING: Do not assume uname() returns 0 on success
12933 (csilvers)
12934 * LSS: Improved ARM support in linux-syscall-support
12935 (dougkwan)
12936 * LSS: Get rid of unused syscalls in linux-syscall-support
12937 (csilvers)
12938 * LSS: Fix broken mmap wrapping for ppc (markus)
12939 * LSS: Emit .cfi_adjust_cfa_offset when appropriate
12940 (ppluzhnikov)
12941 * LSS: Be more accurate in register use in __asm__ (markus)
12942 * LSS: Fix __asm__ calls to compile under clang (chandlerc)
12943 * LSS: Fix ARM inline assembly bug around r7 and swi (lcwu)
12944 * No longer log when an allocator fails (csilvers)
12945 * void* -> const void* for MallocExtension methods (llib)
12946 * Improve HEAP_PROFILE_MMAP and fix bugs with it (dmikurube)
12947 * Replace int-based abs with more correct fabs in a test
12948 (pmurin)
12949
12950
12951 git-svn-id: http://gperftools.googlecode.com/svn/trunk@135
12952 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
12953
12954 M ChangeLog
12955 M Makefile.am
12956 M Makefile.in
12957 M NEWS
12958 M README_windows.txt
12959 M configure
12960 M configure.ac
12961 M google-perftools.sln
12962 M packages/deb/changelog
12963 M packages/deb/copyright
12964 M src/pprof
12965 M src/system-alloc.cc
12966 M src/tcmalloc.cc
12967 M src/windows/config.h
12968 M src/windows/mini_disassembler.h
12969 M src/windows/patch_functions.cc
12970 M src/windows/preamble_patcher.h
12971 M src/windows/preamble_patcher_test.cc
12972 M src/windows/shortproc.asm
12973 A vsprojects/preamble_patcher_test/preamble_patcher_test.vcproj
12974
12975 commit e0eaf5981eede6311e311ac5054489b85015c5d7
12976 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
12977 Date: Sat Jan 28 00:13:43 2012 +0000
12978
12979 * PORTING: Support for patching assembly on win
12980 x86_64! (scott.fr...)
12981 * PORTING: Work around atexit-execution-order bug on freebsd
12982 (csilvers)
12983 * PORTING: Patch _calloc_crt for windows (roger orr)
12984 * PORTING: Add C++11 compatibility method for stl allocator
12985 (jdennett)
12986 * PORTING: use MADV_FREE, not MADV_DONTNEED, on freebsd
12987 (csilvers)
12988 * PORTING: Don't use SYS_open when not supported on solaris
12989 (csilvers)
12990 * PORTING: Do not assume uname() returns 0 on success
12991 (csilvers)
12992 * LSS: Improved ARM support in linux-syscall-support
12993 (dougkwan)
12994 * LSS: Get rid of unused syscalls in linux-syscall-support
12995 (csilvers)
12996 * LSS: Fix broken mmap wrapping for ppc (markus)
12997 * LSS: Emit .cfi_adjust_cfa_offset when appropriate
12998 (ppluzhnikov)
12999 * LSS: Be more accurate in register use in __asm__ (markus)
13000 * LSS: Fix __asm__ calls to compile under clang (chandlerc)
13001 * LSS: Fix ARM inline assembly bug around r7 and swi (lcwu)
13002 * No longer log when an allocator fails (csilvers)
13003 * void* -> const void* for MallocExtension methods (llib)
13004 * Improve HEAP_PROFILE_MMAP and fix bugs with it (dmikurube)
13005 * Replace int-based abs with more correct fabs in a test
13006 (pmurin)
13007
13008
13009 git-svn-id: http://gperftools.googlecode.com/svn/trunk@134
13010 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13011
13012 M NEWS
13013 M configure
13014 M configure.ac
13015 M doc/cpuprofile.html
13016 M src/base/atomicops-internals-arm-v6plus.h
13017 M src/base/linux_syscall_support.h
13018 M src/base/stl_allocator.h
13019 M src/base/sysinfo.cc
13020 M src/debugallocation.cc
13021 M src/google/malloc_extension.h
13022 M src/google/malloc_extension_c.h
13023 M src/heap-checker.cc
13024 M src/heap-profile-table.cc
13025 M src/heap-profile-table.h
13026 M src/heap-profiler.cc
13027 M src/malloc_extension.cc
13028 M src/memory_region_map.cc
13029 M src/memory_region_map.h
13030 M src/system-alloc.cc
13031 M src/tcmalloc.cc
13032 M src/tests/sampler_test.cc
13033 M src/thread_cache.cc
13034 M src/windows/ia32_opcode_map.cc
13035 M src/windows/mini_disassembler.cc
13036 M src/windows/mini_disassembler.h
13037 M src/windows/mini_disassembler_types.h
13038 M src/windows/patch_functions.cc
13039 M src/windows/preamble_patcher.cc
13040 M src/windows/preamble_patcher.h
13041 A src/windows/preamble_patcher_test.cc
13042 M src/windows/preamble_patcher_with_stub.cc
13043 A src/windows/shortproc.asm
13044
13045 commit 4fa02db374683d3c69c1b36158534c011513ed31
13046 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13047 Date: Sat Dec 24 01:00:32 2011 +0000
13048
13049 Ready to release perftools 1.9.1 :-/
13050
13051 Added the ARM stacktrace file to the tarball (for 'make dist')
13052
13053
13054 git-svn-id: http://gperftools.googlecode.com/svn/trunk@132
13055 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13056
13057 M Makefile.am
13058 M Makefile.in
13059 M NEWS
13060 M configure
13061 M configure.ac
13062
13063 commit 2a7b3b88371c2f78ab31be202f384fd0ed466890
13064 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13065 Date: Fri Dec 23 00:45:49 2011 +0000
13066
13067 Thu Dec 22 16:22:45 2011 Google Inc. <[email protected]>
13068
13069 * google-perftools: version 1.9 release
13070 * Lightweight check for double-frees (blount)
13071 * BUGFIX: Fix pprof to exit properly if run with no args
13072 (dagitses)
13073 * Suggest ASan as a way to diagnose buggy code (ppluzhnikov)
13074 * Get rid of unused CACHELINE_SIZE (csilvers)
13075 * Replace atexit() calls with global dtors; helps freebsd
13076 (csilvers)
13077 * Disable heap-checker under AddressSanitizer (kcc)
13078 * Fix bug in powerpc stacktracing (ppluzhnikov)
13079 * PERF: Use exponential backoff waiting for spinlocks (m3b)
13080 * Fix 64-bit nm on 32-bit binaries in pprof (csilvers)
13081 * Add ProfileHandlerDisallowForever (rsc)
13082 * BUGFIX: Shell escape when forking in pprof (csilvers)
13083 * No longer combine overloaded functions in pprof (csilvers)
13084 * Fix address-normalizing bug in pprof (csilvers)
13085 * More consistently call abort() instead of exit() on failure
13086 (csilvers)
13087 * Allow NoGlobalLeaks to be safely called more than once
13088 (csilvers)
13089 * PORTING/BUGFIX: Fix ARM cycleclock to use volatile asm
13090 (dougkwan)
13091 * PORTING: 64-bit atomic ops for ARMv7 (dougkwan)
13092 * PORTING: Implement stacktrace for ARM (dougkwan)
13093 * PORTING: Fix malloc_hook_mmap_linux for ARM (dougkwan)
13094 * PORTING: Update linux_syscall_support.h for ARM/etc
13095 (evannier, sanek)
13096 * PORTING: Fix freebsd to work on x86_64 ([email protected])
13097 * PORTING: Added additional SYS_mmap fixes for FreeBSD
13098 (chappedm)
13099 * PORTING: Allow us to compile on OS X 10.6 and run on 10.5
13100 (raltherr)
13101 * PORTING: Check for mingw compilers that *do* define timespec
13102 * PORTING: Add "support" for MIPS cycletimer
13103 * PORTING: Fix fallback cycle-timer to work with Now
13104 (dougkwan)
13105 * PERF: Move stack trace collecting out of the mutex (taylorc)
13106 * PERF: Get the deallocation stack trace outside the mutex
13107 (sean)
13108 * Make PageHeap dynamically allocated for leak checks (maxim)
13109 * BUGFIX: Fix probing of nm -f behavior in pprof (dpeng)
13110 * BUGFIX: Fix a race with the CentralFreeList lock before main
13111 (sanjay)
13112 * Support /pprof/censusprofile url arguments (rajatjain)
13113 * Change IgnoreObject to return its argument (nlewycky)
13114 * Update malloc-hook files to support more CPUs
13115 * BUGFIX: write our own strstr to avoid libc problems
13116 (csilvers)
13117 * Use simple callgrind compression facility in pprof
13118 * Print an error message when we can't run pprof to symbolize
13119 (csilvers)
13120 * Die in configure when g++ is't installed (csilvers)
13121 * DOC: Beef up the documentation a bit about using libunwind
13122 (csilvers)
13123
13124
13125 git-svn-id: http://gperftools.googlecode.com/svn/trunk@130
13126 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13127
13128 M ChangeLog
13129 M INSTALL
13130 M NEWS
13131 M configure
13132 M configure.ac
13133 M packages/deb/changelog
13134 M src/pprof
13135 M src/windows/config.h
13136
13137 commit 75f5a791bc1c0af0ddf43453c18f44723884f37e
13138 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13139 Date: Fri Dec 23 00:45:14 2011 +0000
13140
13141 Add the magical incantations to get perftools to compile
13142 cleanly under windows. The only non-trivial one is the #undef
13143 of small, which I needed to compile under the latest cygwin --
13144 I ran g++ with -dD under cygwin and saw the '#define small
13145 char' with my own eyes. I wouldn't have thought it...
13146
13147
13148 git-svn-id: http://gperftools.googlecode.com/svn/trunk@129
13149 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13150
13151 M src/internal_logging.h
13152 M src/tcmalloc.cc
13153 M src/windows/port.h
13154
13155 commit 76131b04cf41a7f0fed99503f985847fc261b930
13156 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13157 Date: Fri Dec 23 00:44:53 2011 +0000
13158
13159 In my testing for a new release, I was noticing hangs on
13160 cygwin, that went away when I turned off tls support. Since I
13161 didn't use to have such problems, I am being conservative and
13162 only turning off TLS for the particular version of cygwin I'm
13163 testing on (as returned by uname()).
13164
13165
13166 git-svn-id: http://gperftools.googlecode.com/svn/trunk@128
13167 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13168
13169 M src/thread_cache.cc
13170
13171 commit 3d30fbcc85021b267f7e3faf4c2280d70681920c
13172 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13173 Date: Fri Dec 23 00:44:30 2011 +0000
13174
13175 We use mmx instructions now in perftools, so specify -xmms for
13176 gcc on i386, where it's not on by default (it is for
13177 gcc/x86_64, in my tests). This could potentially cause an
13178 error for embedded systems, which can have i386 but no mms,
13179 but the code wouldn't run properly on them anyway without
13180 tweaks.
13181
13182
13183 git-svn-id: http://gperftools.googlecode.com/svn/trunk@127
13184 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13185
13186 M Makefile.am
13187 M Makefile.in
13188
13189 commit f622d261e9a4331724d222cc2a41b517607c07d7
13190 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13191 Date: Tue Dec 20 18:43:38 2011 +0000
13192
13193 * Fix ARM cycleclock to use volatile asm (dougkwan)
13194 * BUGFIX: Define BASE_HAS_ATOMIC64 for ARMv7 (dougkwan)
13195 * Fix fallback cycle-timer to work with Now (dougkwan)
13196 * Fix pprof to exit properly if run with no args (dagitses)
13197 * Suggest ASan as a way to diagnose buggy code (ppluzhnikov)
13198 * Get rid of unused CACHELINE_SIZE (csilvers)
13199
13200
13201
13202 git-svn-id: http://gperftools.googlecode.com/svn/trunk@126
13203 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13204
13205 M src/base/atomicops-internals-arm-v6plus.h
13206 M src/base/basictypes.h
13207 M src/base/cycleclock.h
13208 M src/debugallocation.cc
13209 M src/malloc_hook_mmap_freebsd.h
13210 M src/pprof
13211 M src/windows/port.h
13212
13213 commit 7da02003014644f73893df8530ee788dc0e626de
13214 Author: [email protected]
13215 <[email protected]@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13216 Date: Wed Nov 30 02:54:35 2011 +0000
13217
13218 Added additional SYS_mmap fixes for FreeBSD due to syscall differences
13219 between os versions (pulled from differences between FreeBSD6 and
13220 FreeBSD8 mmap implementations)
13221
13222 git-svn-id: http://gperftools.googlecode.com/svn/trunk@125
13223 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13224
13225 M src/malloc_hook_mmap_freebsd.h
13226
13227 commit 8c8329390cd05a1bf1bf9613712521a9ee9fe260
13228 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13229 Date: Tue Nov 22 01:06:33 2011 +0000
13230
13231 * Eliminate some TSAN warnings (wilsonh)
13232 * Better fix for freebsd on x86_64 ([email protected])
13233 * Lightweight check for double-frees (blount)
13234 * 64-bit atomic ops for ARMv7 (dougkwan)
13235 * Remove dynamic annotations in some cases (dvyukov)
13236 * Update linux_syscall_support.h for ARM, etc (evannier)
13237
13238
13239 git-svn-id: http://gperftools.googlecode.com/svn/trunk@124
13240 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13241
13242 M src/base/atomicops-internals-arm-v6plus.h
13243 M src/base/dynamic_annotations.c
13244 M src/malloc_hook_mmap_freebsd.h
13245 M src/profile-handler.cc
13246 M src/thread_cache.h
13247
13248 commit ca23f57679e2bd4e0390d6891aa765b7e9daca22
13249 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13250 Date: Thu Nov 17 00:11:42 2011 +0000
13251
13252 Fix bad #include paths.
13253
13254
13255 git-svn-id: http://gperftools.googlecode.com/svn/trunk@123
13256 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13257
13258 M src/stacktrace.cc
13259 M src/stacktrace_config.h
13260
13261 commit e580d7888154fa7f95b3cef9e18f1ce69182212b
13262 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13263 Date: Wed Nov 16 21:20:45 2011 +0000
13264
13265 Some linux_syscall_support.h improvements:
13266
13267 * Fix socketcall functions (mec)
13268 * Add ARM related defs and syscall asembly code (sanek)
13269 * Use __arm__ instead of particular ARM arch macro (dougkwan)
13270
13271
13272 git-svn-id: http://gperftools.googlecode.com/svn/trunk@122
13273 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13274
13275 M src/base/linux_syscall_support.h
13276
13277 commit 6c3eaabd7306173b6b71b63797ebf050675046cf
13278 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13279 Date: Wed Nov 16 05:21:54 2011 +0000
13280
13281 * Check for mingw compilers that *do* define timespec
13282 * Replace atexit() calls with global dtors; helps freebsd
13283 (csilvers)
13284 * Fix malloc_hook_mmap_linux for ARM (dougkwan)
13285 * Disalbe heap-checker under AddressSanitizer (kcc)
13286 * Fix bug in powerpc stacktracing (ppluzhnikov)
13287 * Use exponential backoff waiting for spinlocks (m3b)
13288 * Fix 64-bit nm on 32-bit binaries in pprof (csilvers)
13289 * Implement stacktrace for ARM (dougkwan)
13290 * Add ProfileHandlerDisallowForever (rsc)
13291 * Shell escape when forking in pprof (csilvers)
13292 * Fix freebsd to work on x86_64 ([email protected])
13293 * No longer combine overloaded functions in pprof (csilvers)
13294 * Fix address-normalizing bug in pprof (csilvers)
13295 * More consistently call abort() instead of exit() on failure
13296 (csilvers)
13297 * Allow NoGlobalLeaks to be safely called more than once
13298 (csilvers)
13299 * Beef up the documentation a bit about using libunwind
13300 (csilvers)
13301
13302
13303 git-svn-id: http://gperftools.googlecode.com/svn/trunk@121
13304 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13305
13306 M INSTALL
13307 M Makefile.am
13308 M Makefile.in
13309 M src/base/googleinit.h
13310 M src/base/linux_syscall_support.h
13311 M src/base/logging.h
13312 M src/base/spinlock_internal.cc
13313 M src/base/spinlock_linux-inl.h
13314 M src/base/spinlock_posix-inl.h
13315 M src/base/spinlock_win32-inl.h
13316 M src/debugallocation.cc
13317 M src/google/heap-checker.h
13318 M src/heap-checker.cc
13319 M src/malloc_hook_mmap_freebsd.h
13320 M src/malloc_hook_mmap_linux.h
13321 M src/pprof
13322 M src/profile-handler.cc
13323 M src/profile-handler.h
13324 M src/stacktrace.cc
13325 C059 src/stacktrace_powerpc-inl.h src/stacktrace_arm-inl.h
13326 M src/stacktrace_config.h
13327 M src/stacktrace_powerpc-inl.h
13328 M src/stacktrace_x86-inl.h
13329 M src/tests/heap-checker_unittest.cc
13330 M src/tests/profile-handler_unittest.cc
13331
13332 commit a6076edd177d59e67207753b799ce047a3663cb0
13333 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13334 Date: Tue Oct 18 20:57:45 2011 +0000
13335
13336 * Get the deallocation stack trace outside the lock (sean)
13337 * Make PageHeap dynamically allocated for leak checks (maxim)
13338 * BUGFIX: Fix probing of nm -f behavior in pprof (dpeng)
13339 * PORTING: Add "support" for MIPS cycletimer
13340 * BUGFIX: Fix a race with the CentralFreeList lock (sanjay)
13341 * Allow us to compile on OS X 10.6 and run on 10.5 (raltherr)
13342 * Support /pprof/censusprofile url arguments (rajatjain)
13343 * Die in configure when g++ is't installed (csilvers)
13344 * Change IgnoreObject to return its argument (nlewycky)
13345 * Update malloc-hook files to support more CPUs
13346 * Move stack trace collecting out of the mutex (taylorc)
13347 * BUGFIX: write our own strstr to avoid libc problems
13348 (csilvers)
13349 * use simple callgrind compression facility in pprof
13350 * print an error message when we can't run pprof to symbolize
13351 (csilvers)
13352
13353
13354 git-svn-id: http://gperftools.googlecode.com/svn/trunk@120
13355 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13356
13357 M INSTALL
13358 M Makefile.am
13359 M Makefile.in
13360 M configure
13361 M configure.ac
13362 M packages/rpm.sh
13363 M src/base/basictypes.h
13364 M src/base/cycleclock.h
13365 M src/base/dynamic_annotations.h
13366 M src/base/spinlock_internal.cc
13367 M src/base/spinlock_linux-inl.h
13368 M src/central_freelist.cc
13369 M src/central_freelist.h
13370 M src/common.cc
13371 M src/common.h
13372 M src/debugallocation.cc
13373 M src/google/heap-checker.h
13374 M src/heap-checker.cc
13375 M src/heap-profile-table.cc
13376 M src/heap-profile-table.h
13377 M src/heap-profiler.cc
13378 M src/internal_logging.cc
13379 M src/internal_logging.h
13380 M src/libc_override_osx.h
13381 M src/malloc_hook.cc
13382 M src/malloc_hook_mmap_linux.h
13383 M src/memfs_malloc.cc
13384 M src/page_heap.cc
13385 M src/page_heap.h
13386 M src/page_heap_allocator.h
13387 M src/pprof
13388 M src/profile-handler.cc
13389 M src/profiler.cc
13390 M src/span.cc
13391 M src/span.h
13392 M src/stack_trace_table.cc
13393 M src/stacktrace.cc
13394 M src/stacktrace_config.h
13395 M src/static_vars.cc
13396 M src/static_vars.h
13397 M src/symbolize.cc
13398 M src/system-alloc.cc
13399 M src/tcmalloc.cc
13400 M src/tests/tcmalloc_unittest.cc
13401 M src/thread_cache.cc
13402 M src/thread_cache.h
13403 M src/windows/config.h
13404
13405 commit c2eedce2a718913ed6264ac8e96571c233761e3b
13406 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13407 Date: Fri Aug 26 21:08:59 2011 +0000
13408
13409 Fri Aug 26 13:29:25 2011 Google Inc. <[email protected]>
13410
13411 * google-perftools: version 1.8.3 release
13412 * Added back the 'pthreads unsafe early' #define, needed
13413 for FreeBSD
13414
13415
13416 git-svn-id: http://gperftools.googlecode.com/svn/trunk@117
13417 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13418
13419 M ChangeLog
13420 M NEWS
13421 M configure
13422 M configure.ac
13423 M src/config.h.in
13424 M src/pprof
13425 M src/windows/config.h
13426
13427 commit f633b40ba5d3785c7e31fd9b575c5cf5e7c174f3
13428 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13429 Date: Thu Aug 11 22:06:22 2011 +0000
13430
13431 Thu Aug 11 15:01:47 2011 Google Inc. <[email protected]>
13432
13433 * google-perftools: version 1.8.2 release
13434 * Fixed calculation of patchlevel, 'make check' should all
13435 pass again
13436
13437
13438 git-svn-id: http://gperftools.googlecode.com/svn/trunk@115
13439 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13440
13441 M ChangeLog
13442 M NEWS
13443 M configure
13444 M configure.ac
13445 M src/pprof
13446 M src/windows/config.h
13447
13448 commit ffcdf7bf5a7e97428f7521b3c9e6d854e3a424a9
13449 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13450 Date: Thu Aug 11 22:05:09 2011 +0000
13451
13452 Fix #includes for ARM.
13453
13454
13455 git-svn-id: http://gperftools.googlecode.com/svn/trunk@114
13456 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13457
13458 M src/base/atomicops-internals-arm-generic.h
13459
13460 commit d2faf4646dc79486babfbd63cf5f658b917dd6ab
13461 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13462 Date: Wed Jul 27 04:18:01 2011 +0000
13463
13464 Tue Jul 26 20:57:51 2011 Google Inc. <[email protected]>
13465
13466 * google-perftools: version 1.8 release
13467 * Added an #include to fix compile breakage on latest gcc's
13468 * Removed an extra , in the configure.ac script
13469
13470
13471 git-svn-id: http://gperftools.googlecode.com/svn/trunk@112
13472 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13473
13474 M ChangeLog
13475 M NEWS
13476 M configure
13477 M configure.ac
13478 M src/base/elf_mem_image.cc
13479 M src/config.h.in
13480 M src/pprof
13481 M src/tests/tcmalloc_unittest.cc
13482 M src/windows/config.h
13483
13484 commit 100c38c1a225446c1bbeeaac117902d0fbebfefe
13485 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13486 Date: Sat Jul 16 01:07:10 2011 +0000
13487
13488 Fri Jul 15 16:10:51 2011 Google Inc. <[email protected]>
13489
13490 * google-perftools: version 1.8 release
13491 * PORTING: (Disabled) support for patching mmap on freebsd
13492 (chapp...)
13493 * PORTING: Support volatile __malloc_hook for glibc 2.14
13494 (csilvers)
13495 * PORTING: Use _asm rdtsc and __rdtsc to get cycleclock in
13496 windows (koda)
13497 * PORTING: Fix fd vs. HANDLE compiler error on cygwin
13498 (csilvers)
13499 * PORTING: Do not test memalign or double-linking on OS X
13500 (csilvers)
13501 * PORTING: Actually enable TLS on windows (jontra)
13502 * PORTING: Some work to compile under Native Client (krasin)
13503 * PORTING: deal with pthread_once w/o -pthread on freebsd
13504 (csilvers)
13505 * Rearrange libc-overriding to make it easier to port
13506 (csilvers)
13507 * Display source locations in pprof disassembly (sanjay)
13508 * BUGFIX: Actually initialize allocator name (mec)
13509 * BUGFIX: Keep track of 'overhead' bytes in malloc reporting
13510 (csilvers)
13511 * Allow ignoring one object twice in the leak checker (glider)
13512 * BUGFIX: top10 in pprof should print 10 lines, not 11 (rsc)
13513 * Refactor vdso source files (tipp)
13514 * Some documentation cleanups
13515 * Document MAX_TOTAL_THREAD_CACHE_SIZE <= 1Gb (nsethi)
13516 * Add MallocExtension::GetOwnership(ptr) (csilvers)
13517 * BUGFIX: We were leaving out a needed $(top_srcdir) in
13518 the Makefile
13519 * PORTING: Support getting argv0 on OS X
13520 * Add 'weblist' command to pprof: like 'list' but html
13521 (sanjay)
13522 * Improve source listing in pprof (sanjay)
13523 * Cap cache sizes to reduce fragmentation (ruemmler)
13524 * Improve performance by capping or increasing sizes
13525 (ruemmler)
13526 * Add M{,un}mapReplacmenet hooks into MallocHook (ribrdb)
13527 * Refactored system allocator logic (gangren)
13528 * Include cleanups (csilvers)
13529 * Add TCMALLOC_SMALL_BUT_SLOW support (ruemmler)
13530 * Clarify that tcmalloc stats are MiB (robinson)
13531 * Remove support for non-tcmalloc debugallocation (blount)
13532 * Add a new test: malloc_hook_test (csilvers)
13533 * Change the configure script to be more crosstool-friendly
13534 (mcgrathr)
13535 * PORTING: leading-underscore changes to support win64
13536 (csilvers)
13537 * Improve debugallocation tc_malloc_size (csilvers)
13538 * Extend atomicops.h and cyceclock to use ARM V6+ optimized
13539 code (sanek)
13540 * Change malloc-hook to use a list-like structure (llib)
13541 * Add flag to use MAP_PRIVATE in memfs_malloc (gangren)
13542 * Windows support for pprof: nul and /usr/bin/file (csilvers)
13543 * TESTING: add test on strdup to tcmalloc_test (csilvers)
13544 * Augment heap-checker to deal with no-inode maps (csilvers)
13545 * Count .dll/.dylib as shared libs in heap-checker (csilvers)
13546 * Disable sys_futex for arm; it's not always reliable (sanek)
13547 * PORTING: change lots of windows/port.h macros to functions
13548 * BUGFIX: Generate correct version# in tcmalloc.h on windows
13549 (csilvers)
13550 * PORTING: Some casting to make solaris happier about types
13551 (csilvers)
13552 * TESTING: Disable debugallocation_test in 'minimal' mode
13553 (csilvers)
13554 * Rewrite debugallocation to be more modular (csilvers)
13555 * Don't try to run the heap-checker under valgrind
13556 (ppluzhnikov)
13557 * BUGFIX: Make focused stat %'s relative, not absolute
13558 (sanjay)
13559 * BUGFIX: Don't use '//' comments in a C file (csilvers)
13560 * Quiet new-gcc compiler warnings via -Wno-unused-result,
13561 etc (csilvers)
13562
13563
13564 git-svn-id: http://gperftools.googlecode.com/svn/trunk@110
13565 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13566
13567 M ChangeLog
13568 M Makefile.am
13569 M Makefile.in
13570 M NEWS
13571 M README
13572 M configure
13573 M configure.ac
13574 M doc/heapprofile.html
13575 M doc/tcmalloc.html
13576 M google-perftools.sln
13577 M packages/deb/changelog
13578 M src/base/atomicops-internals-windows.h
13579 M src/base/atomicops.h
13580 M src/base/cycleclock.h
13581 C067 src/base/vdso_support.cc src/base/elf_mem_image.cc
13582 A src/base/elf_mem_image.h
13583 M src/base/low_level_alloc.cc
13584 M src/base/spinlock_internal.cc
13585 M src/base/vdso_support.cc
13586 M src/base/vdso_support.h
13587 M src/central_freelist.cc
13588 M src/central_freelist.h
13589 M src/common.cc
13590 M src/common.h
13591 M src/config.h.in
13592 M src/debugallocation.cc
13593 M src/google/heap-checker.h
13594 M src/google/malloc_extension.h
13595 M src/google/malloc_extension_c.h
13596 M src/google/malloc_hook.h
13597 M src/heap-checker.cc
13598 A src/libc_override.h
13599 A src/libc_override_gcc_and_weak.h
13600 A src/libc_override_glibc.h
13601 A src/libc_override_osx.h
13602 A src/libc_override_redefine.h
13603 M src/malloc_extension.cc
13604 M src/malloc_hook-inl.h
13605 M src/malloc_hook.cc
13606 A src/malloc_hook_mmap_freebsd.h
13607 A src/malloc_hook_mmap_linux.h
13608 M src/maybe_threads.cc
13609 M src/memfs_malloc.cc
13610 M src/memory_region_map.cc
13611 M src/pprof
13612 M src/profile-handler.cc
13613 M src/profiler.cc
13614 M src/stack_trace_table.h
13615 M src/stacktrace.cc
13616 M src/stacktrace_config.h
13617 C053 src/stack_trace_table.h src/stacktrace_nacl-inl.h
13618 M src/symbolize.cc
13619 M src/system-alloc.cc
13620 M src/system-alloc.h
13621 M src/tcmalloc.cc
13622 M src/tests/debugallocation_test.cc
13623 M src/tests/malloc_extension_c_test.c
13624 M src/tests/malloc_extension_test.cc
13625 M src/tests/malloc_hook_test.cc
13626 M src/tests/profiler_unittest.cc
13627 M src/tests/system-alloc_unittest.cc
13628 M src/tests/tcmalloc_unittest.cc
13629 M src/thread_cache.cc
13630 M src/windows/config.h
13631 M src/windows/google/tcmalloc.h.in
13632 M src/windows/mingw.h
13633 M src/windows/port.cc
13634 M src/windows/port.h
13635 M
13636 vsprojects/current_allocated_bytes_test/current_allocated_bytes_test.vcproj
13637 M vsprojects/malloc_hook_test/malloc_hook_test.vcproj
13638
13639 commit 8c7d2289d24f7a49f1f6f60d4a6eaee06fa04c60
13640 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13641 Date: Thu May 19 21:37:12 2011 +0000
13642
13643 * Fix typos in comment in profiler.h (nrhodes)
13644 * #include fixes (jyrki)
13645 * Add missing stddef.h for ptrdiff_t (mec)
13646 * Add M{,un}mapReplacement hooks into MallocHook (ribrdb)
13647 * Force big alloc in frag test (ruemmler)
13648 * PERF: Increase the size class cache to 64K entries
13649 (ruemmler)
13650 * PERF: Increase the transfer cache by 16x (ruemmler)
13651 * Use windows intrinsic to get the tsc (csilvers)
13652 * Rename atomicops-internals-x86-msvc.h->windows.h (csilvers)
13653 * Remove flaky DEATH test in malloc_hook_test (ppluzhnikov)
13654 * Expose internal ReadStackTraces()/etc (lantran)
13655 * Refactored system allocator logic (gangren)
13656 * Include-what-you-use: cleanup tcmalloc #includes (csilvers)
13657 * Don't set kAddressBits to 48 on 32-bit systems (csilvers)
13658 * Add declaration for __rdtsc() for windows (koda)
13659 * Don't revert to system alloc for expected errors (gangren)
13660 * Add TCMALLOC_SMALL_BUT_SLOW support (ruemmler)
13661 * Clarify that tcmalloc stats are MiB (robinson)
13662 * Avoid setting cpuinfo_cycles_per_second to 0 (koda)
13663 * Fix frag_unittest memory calculations (ruemmler)
13664 * Remove support for non-tcmalloc debugallocation (blount)
13665 * Add malloc_hook_test (llib)
13666 * Change the objcopy -W test to be cross-friendly (mcgrathr)
13667 * Export __tcmalloc in addition to _tcmalloc, for 86_64
13668 (csilvers)
13669
13670
13671 git-svn-id: http://gperftools.googlecode.com/svn/trunk@109
13672 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13673
13674 M INSTALL
13675 M Makefile.am
13676 M Makefile.in
13677 M README
13678 M README_windows.txt
13679 M configure
13680 M configure.ac
13681 M google-perftools.sln
13682 R096 src/base/atomicops-internals-x86-msvc.h
13683 src/base/atomicops-internals-windows.h
13684 M src/base/atomicops.h
13685 M src/base/cycleclock.h
13686 M src/base/sysinfo.cc
13687 M src/base/sysinfo.h
13688 M src/base/vdso_support.h
13689 M src/central_freelist.cc
13690 M src/central_freelist.h
13691 M src/common.cc
13692 M src/common.h
13693 M src/config.h.in
13694 M src/debugallocation.cc
13695 M src/google/malloc_extension.h
13696 M src/google/malloc_hook.h
13697 M src/google/malloc_hook_c.h
13698 M src/google/profiler.h
13699 M src/google/tcmalloc.h.in
13700 M src/heap-profiler.cc
13701 M src/internal_logging.cc
13702 M src/internal_logging.h
13703 M src/malloc_extension.cc
13704 M src/malloc_hook-inl.h
13705 M src/malloc_hook.cc
13706 M src/memfs_malloc.cc
13707 M src/memory_region_map.cc
13708 M src/memory_region_map.h
13709 M src/packed-cache-inl.h
13710 M src/page_heap.cc
13711 M src/page_heap.h
13712 M src/page_heap_allocator.h
13713 M src/pagemap.h
13714 M src/sampler.cc
13715 M src/sampler.h
13716 M src/span.cc
13717 M src/stack_trace_table.cc
13718 M src/stack_trace_table.h
13719 M src/static_vars.cc
13720 M src/system-alloc.cc
13721 M src/system-alloc.h
13722 M src/tcmalloc.cc
13723 M src/tests/frag_unittest.cc
13724 M src/tests/heap-checker_unittest.cc
13725 A src/tests/malloc_hook_test.cc
13726 M src/tests/memalign_unittest.cc
13727 M src/tests/page_heap_test.cc
13728 M src/tests/realloc_unittest.cc
13729 M src/tests/system-alloc_unittest.cc
13730 M src/tests/tcmalloc_large_unittest.cc
13731 M src/thread_cache.cc
13732 M src/thread_cache.h
13733 M src/windows/patch_functions.cc
13734 M src/windows/port.cc
13735 A vsprojects/malloc_hook_test/malloc_hook_test.vcproj
13736
13737 commit 1d30e525ae6ac38ae381bb3118f7f47998af2942
13738 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13739 Date: Mon Mar 21 21:41:55 2011 +0000
13740
13741 * Improve debugallocation tc_malloc_size (csilvers)
13742 * Extend atomicops.h to use ARM V6+ optimized code (sanek)
13743 * Fix failure in Ranges test (ppluzhnikov)
13744 * Change malloc-hook to use a list-like structure (llib)
13745 * Update tcmalloc_regtest to use new malloc hooks (llib)
13746 * PARTIAL: Keep track of 'overhead' bytes in the page cache
13747 (csilvers)
13748
13749
13750 git-svn-id: http://gperftools.googlecode.com/svn/trunk@108
13751 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13752
13753 M Makefile.am
13754 M Makefile.in
13755 M google-perftools.sln
13756 A src/base/atomicops-internals-arm-generic.h
13757 A src/base/atomicops-internals-arm-v6plus.h
13758 M src/base/atomicops.h
13759 M src/central_freelist.cc
13760 M src/central_freelist.h
13761 M src/google/malloc_hook.h
13762 M src/google/malloc_hook_c.h
13763 M src/heap-checker-bcad.cc
13764 M src/heap-checker.cc
13765 M src/heap-profiler.cc
13766 M src/malloc_hook-inl.h
13767 M src/malloc_hook.cc
13768 M src/memory_region_map.cc
13769 M src/memory_region_map.h
13770 M src/tcmalloc.cc
13771 A src/tests/current_allocated_bytes_test.cc
13772 M src/tests/low_level_alloc_unittest.cc
13773 M src/tests/malloc_extension_c_test.c
13774 M src/tests/tcmalloc_unittest.cc
13775 A
13776 vsprojects/current_allocated_bytes_test/current_allocated_bytes_test.vcproj
13777
13778 commit 6fe07cd2c0527e18276cc79a57e2212a4b048746
13779 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13780 Date: Fri Mar 4 23:52:33 2011 +0000
13781
13782 * add a flag to use MAP_PRIVATE in memfs_malloc (gangren)
13783 * pthread_self() is now safe to use early (ppluzhnikov)
13784 * windows support for pprof: nul and /usr/bin/file (csilvers)
13785 * fix tc_malloc_size for debugallocation (csilvers)
13786 * add test on strdup to tcmalloc_test (csilvers)
13787 * augment heap-checker to deal with no-inode maps (csilvers)
13788 * Get rid of -Wno-unused-result: not all gcc's support it
13789 (csilvers)
13790 * /bin/true -> ':', which is faster and more portable
13791 (csilvers)
13792
13793
13794 git-svn-id: http://gperftools.googlecode.com/svn/trunk@107
13795 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13796
13797 M Makefile.am
13798 M Makefile.in
13799 M doc/tcmalloc.html
13800 M src/debugallocation.cc
13801 M src/heap-checker.cc
13802 M src/memfs_malloc.cc
13803 M src/pprof
13804 M src/tcmalloc.cc
13805 M src/tests/debugallocation_test.sh
13806 M src/tests/memalign_unittest.cc
13807 M src/tests/tcmalloc_unittest.cc
13808 M src/thread_cache.cc
13809
13810 commit 75584139e40c9d6c952d9c5339c52e5b58302fc8
13811 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13812 Date: Wed Mar 2 08:10:05 2011 +0000
13813
13814 * Enhance cycleclock on ARM v6 and above (sanek)
13815 * Reduce object copying by using a reference (nherring)
13816 * Modified lock annotations a bit (lcwu)
13817 * Make debugallocation test a bit more forgiving (csilvers)
13818 * Count .dll/.dylib as shared libs in heapchecker (csilvers)
13819 * Disable sys_futex for arm (sanek)
13820 * Don't use macros as much in windows/port.h (andrey.s...)
13821 * Update #includes in case malloc.h is in weird places
13822 (csilvers)
13823 * Turn off some not-so-useful warnings in gcc 4 (csilvers)
13824 * Do some casting to make solaris happier about types
13825 (csilvers)
13826 * Disable debugallocation_test in 'minimal' mode (csilvers)
13827 * Rewrite debugallocation to be more modular (csilvers)
13828 * We can't run the heap-checker under valgrind (ppluzhnikov)
13829
13830
13831 git-svn-id: http://gperftools.googlecode.com/svn/trunk@106
13832 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13833
13834 M Makefile.am
13835 M Makefile.in
13836 M configure
13837 M configure.ac
13838 C055 src/raw_printer.cc src/base/arm_instruction_set_select.h
13839 M src/base/cycleclock.h
13840 M src/base/logging.h
13841 M src/base/low_level_alloc.cc
13842 M src/base/spinlock_linux-inl.h
13843 M src/base/stl_allocator.h
13844 M src/base/sysinfo.cc
13845 M src/config.h.in
13846 M src/debugallocation.cc
13847 M src/google/tcmalloc.h.in
13848 M src/heap-checker.cc
13849 M src/internal_logging.cc
13850 M src/raw_printer.cc
13851 M src/tcmalloc.cc
13852 M src/tests/debugallocation_test.cc
13853 M src/tests/heap-checker_unittest.cc
13854 M src/tests/memalign_unittest.cc
13855 M src/tests/sampler_test.cc
13856 M src/windows/config.h
13857 R094 src/windows/google/tcmalloc.h src/windows/google/tcmalloc.h.in
13858 M src/windows/port.cc
13859 M src/windows/port.h
13860
13861 commit c1abbfae802af5bf949c78e0bfdfd58d5c669a86
13862 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13863 Date: Sat Feb 19 00:11:09 2011 +0000
13864
13865 * avoid unused-var warnings with annotalysis (lcwu)
13866 * make focused stats %'s relative, not absolute (sanjay)
13867
13868
13869 git-svn-id: http://gperftools.googlecode.com/svn/trunk@105
13870 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13871
13872 M src/base/dynamic_annotations.h
13873 M src/pprof
13874
13875 commit dd3d9d969e9b889e96c2af497e436856fac1a2a7
13876 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13877 Date: Tue Feb 8 01:03:37 2011 +0000
13878
13879 * Fix tcmalloc_unittest on MSVC 10 in release mode (csilvers)
13880 * Fix malloc_hook_c.h to compile with -ansi under gcc
13881 (csilvers)
13882
13883
13884 git-svn-id: http://gperftools.googlecode.com/svn/trunk@104
13885 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13886
13887 M NEWS
13888 M src/google/malloc_hook_c.h
13889 M src/tests/tcmalloc_unittest.cc
13890
13891 commit 7375b4f3cb3ab4471d0016017be5e18ba5451c5f
13892 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13893 Date: Sat Feb 5 00:19:37 2011 +0000
13894
13895 Fri Feb 04 15:54:31 2011 Google Inc. <[email protected]>
13896
13897 * google-perftools: version 1.7 release
13898 * Reduce page map key size under x86_64 by 4.4MB (rus)
13899 * Remove a flaky malloc-extension test (fdabek)
13900 * Improve the performance of PageHeap::New (ond..., csilvers)
13901 * Improve sampling_test with no-inline additions/etc (fdabek)
13902 * 16-byte align debug allocs (jyasskin)
13903 * Change FillProcSelfMaps to detect out-of-buffer-space
13904 (csilvers)
13905 * Document the need for sampling to use GetHeapSample
13906 (csilvers)
13907 * Try to read TSC frequency from tsc_freq_khs (adurbin)
13908 * Do better at figuring out if tests are running under gdb
13909 (ppluzhnikov)
13910 * Improve spinlock contention performance (ruemmler)
13911 * Better internal-function list for pprof's /contention
13912 (ruemmler)
13913 * Speed up GoogleOnce (m3b)
13914 * Limit number of incoming/outgoing edges in pprof (sanjay)
13915 * Add pprof --evince to go along with --gv (csilvers)
13916 * Document the various ways to get heap-profiling information
13917 (csilvers)
13918 * Separate out synchronization profiling routines (ruemmler)
13919 * Improve malloc-stats output to be more understandable
13920 (csilvers)
13921 * Add support for census profiler in pporf (nabeelmian)
13922 * Document how pprof's /symbol must support GET requests
13923 (csilvers)
13924 * Improve acx_pthread.m4 (ssuomi, liujisi)
13925 * Speed up pprof's ExtractSymbols (csilvers)
13926 * Ignore some known-leaky (java) libraries in the heap checker
13927 (davidyu)
13928 * Make kHideMask use all 64 bits in tests (ppluzhnikov)
13929 * Clean up pprof input-file handling (csilvers)
13930 * BUGFIX: Don't crash if __environ is NULL (csilvers)
13931 * BUGFIX: Fix totally broken debugallocation tests (csilvers)
13932 * BUGFIX: Fix up fake_VDSO handling for unittest (ppluzhnikov)
13933 * BUGFIX: Suppress all large allocs when report threshold is 0
13934 (lexie)
13935 * BUGFIX: mmap2 on i386 takes an off_t, not off64_t (csilvers)
13936 * PORTING: Add missing PERFTOOLS_DLL_DECL (csilvers)
13937 * PORTING: Add stddef.h to make newer gcc's happy (csilvers)
13938 * PORTING: Document some tricks for working under OS X
13939 (csilvers)
13940 * PORTING: Don't try to check valgrind for windows (csilvers)
13941 * PORTING: Make array-size a var to compile under clang
13942 (chandlerc)
13943 * PORTING: No longer hook _aligned_malloc and _aligned_free
13944 (csilvers)
13945 * PORTING: Quiet some gcc warnings (csilvers)
13946 * PORTING: Replace %PRIxPTR with %p to be more portable
13947 (csilvers)
13948 * PORTING: Support systems that capitalize /proc weirdly
13949 (sanek)
13950 * PORTING: Treat arm3 the same as arm5t in cycletimer
13951 (csilvers)
13952 * PORTING: Update windows logging to not allocate memory
13953 (csilvers)
13954 * PORTING: avoid double-patching newer windows DLLs
13955 (roger.orr)
13956 * PORTING: get dynamic_annotations.c to work on windows
13957 (csilvers)
13958 * Add pkg-config .pc files for the 5 libraries we produce
13959 (csilvers)
13960 * Added proper libtool versioning, so this lib will be 0.1.0
13961 (csilvers)
13962 * Moved from autoconf 2.64 to 2.65
13963
13964
13965 git-svn-id: http://gperftools.googlecode.com/svn/trunk@102
13966 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
13967
13968 M ChangeLog
13969 M Makefile.am
13970 M Makefile.in
13971 M NEWS
13972 M README
13973 M configure
13974 M configure.ac
13975 M m4/acx_pthread.m4
13976 M packages/deb/changelog
13977 M src/base/cycleclock.h
13978 M src/base/dynamic_annotations.c
13979 M src/base/logging.h
13980 M src/base/stl_allocator.h
13981 M src/base/vdso_support.cc
13982 M src/debugallocation.cc
13983 M src/heap-checker.cc
13984 M src/page_heap.cc
13985 M src/page_heap.h
13986 M src/pprof
13987 M src/symbolize.h
13988 M src/system-alloc.cc
13989 M src/tests/debugallocation_test.cc
13990 M src/tests/debugallocation_test.sh
13991 M src/tests/malloc_extension_test.cc
13992 M src/windows/port.h
13993
13994 commit 3d77cbf7d569a7c7f0ce39a83f6c98da1718f1c4
13995 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
13996 Date: Wed Jan 19 21:37:15 2011 +0000
13997
13998 * Make kHideMask use all 64 bits (ppluzhnikov)
13999 * Add new IsDebuggerAttached method (ppluzhnikov)
14000 * Document some tricks for maybe getting perftools to work
14001 on OS X
14002 * Redo file-top pprof commands (csilvers)
14003 * Clean up pprof input-file handling (csilvers)
14004 * 16-byte align debug allocs (jyasskin)
14005 * Ignore JVM memory leakage in the heap checker (davidyu,
14006 kkurimoto)
14007 * Better internal-function list for contentionz (ruemmler)
14008 * mmap2 on i386 takes an off_t, not an off64_t (csilvers)
14009 * Fix up fake-VDSO handling for unittest (ppluzhnikov)
14010 * Don't try to check valgrind for windows (csilvers)
14011
14012
14013 git-svn-id: http://gperftools.googlecode.com/svn/trunk@101
14014 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14015
14016 M README
14017 M src/base/dynamic_annotations.c
14018 M src/base/linux_syscall_support.h
14019 M src/base/vdso_support.cc
14020 M src/base/vdso_support.h
14021 M src/debugallocation.cc
14022 M src/heap-checker.cc
14023 M src/malloc_hook.cc
14024 M src/pprof
14025 M src/tcmalloc.cc
14026 M src/tests/heap-checker_unittest.cc
14027 M src/tests/tcmalloc_unittest.cc
14028 M src/windows/config.h
14029
14030 commit a0a2ff3b493481ff52e4b4deec2fcc494756b64a
14031 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14032 Date: Wed Nov 24 00:33:20 2010 +0000
14033
14034 * PORTING: Add PERFTOOLS_DLL_DECL to malloc_hook_c.h
14035 (csilvers)
14036 * Add new .h files to Makefile (csilvers)
14037
14038
14039 git-svn-id: http://gperftools.googlecode.com/svn/trunk@100
14040 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14041
14042 M Makefile.am
14043 M Makefile.in
14044 M src/google/malloc_hook_c.h
14045
14046 commit 3014cf142e5a2409c88ab4559f3274434ed9a29b
14047 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14048 Date: Thu Nov 18 01:07:25 2010 +0000
14049
14050 * Suppress all large allocs when report threshold==0
14051 * Clarified meaning of various malloc stats
14052 * Change from ATTRIBUTED_DEPRECATED to comments
14053 * Make array-size a var to compile under clang
14054 * Reduce page map key size under x86_64 by 4.4MB
14055 * Added full qualification to MemoryBarrier
14056 * Support systems that capitalize /proc weirdly
14057 * Avoid gcc warning: exporting type in unnamed ns
14058 * Add some dynamic annotations for gcc attributes
14059 * Add support for census profiler in pprof
14060 * Speed up pprof's ExtractSymbols
14061 * Speed up GoogleOnce
14062 * Add pkg-config (.pc) files
14063 * Detect when __environ exists but is NULL
14064 * Improve spinlock contention performance
14065 * Add GetFreeListSizes
14066 * Improve sampling_test, eg by adding no-inline
14067 * Relax malloc_extension test-check for big pages
14068 * Add proper library version number information
14069 * Update from autoconf 2.64 to 2.65
14070 * Better document how to write a server that works with pprof
14071 * Change FillProcSelfMaps to better handle out-of-space
14072 * No longer hook _aligned_malloc/free in windows
14073 * Handle function-forwarding in DLLs when patching (in
14074 windows)
14075 * Update .vcproj files that had wrong .cc files in them (!)
14076 * get rid of unnecessary 'size < 0'
14077 * fix comments a bit in sysinfo.cc
14078 * another go at improving malloc-stats output
14079 * fix comment typo in profiler.cc
14080 * Add a few more thread annotations
14081 * Try to read TSC frequency from 'tsc_freq_khz'
14082 * Fix annotalysis/TSAN incompatibility
14083 * Add pprof --evince to go along with --gv
14084 * Document need for sampling to use GetHeapSample
14085 * Fix flakiness in malloc_extension_test
14086 * Separate out synchronization profiling routines
14087
14088
14089 git-svn-id: http://gperftools.googlecode.com/svn/trunk@99
14090 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14091
14092 M Makefile.am
14093 M Makefile.in
14094 M NEWS
14095 M README
14096 M aclocal.m4
14097 M configure
14098 M configure.ac
14099 M doc/pprof_remote_servers.html
14100 M ltmain.sh
14101 A m4/libtool.m4
14102 A m4/ltoptions.m4
14103 A m4/ltsugar.m4
14104 A m4/ltversion.m4
14105 A m4/lt~obsolete.m4
14106 M packages/deb/libgoogle-perftools-dev.dirs
14107 M packages/deb/libgoogle-perftools-dev.install
14108 M packages/rpm/rpm.spec
14109 A src/base/atomicops-internals-arm-gcc.h
14110 M src/base/basictypes.h
14111 M src/base/dynamic_annotations.h
14112 M src/base/logging.h
14113 M src/base/low_level_alloc.cc
14114 M src/base/spinlock.cc
14115 M src/base/spinlock.h
14116 A src/base/spinlock_internal.cc
14117 C057 src/base/spinlock_win32-inl.h src/base/spinlock_internal.h
14118 M src/base/spinlock_linux-inl.h
14119 M src/base/spinlock_posix-inl.h
14120 M src/base/spinlock_win32-inl.h
14121 C067 src/base/spinlock_win32-inl.h
14122 src/base/synchronization_profiling.h
14123 M src/base/sysinfo.cc
14124 M src/base/sysinfo.h
14125 M src/base/thread_annotations.h
14126 M src/base/vdso_support.cc
14127 M src/common.h
14128 M src/config.h.in
14129 M src/debugallocation.cc
14130 M src/google/heap-checker.h
14131 M src/google/malloc_extension.h
14132 M src/heap-profile-table.cc
14133 M src/malloc_extension.cc
14134 M src/memory_region_map.cc
14135 M src/page_heap.cc
14136 M src/page_heap.h
14137 M src/pprof
14138 M src/profiler.cc
14139 M src/system-alloc.cc
14140 M src/tcmalloc.cc
14141 M src/tests/debugallocation_test.cc
14142 M src/tests/malloc_extension_test.cc
14143 M src/tests/sampling_test.cc
14144 M src/tests/sampling_test.sh
14145 M src/tests/system-alloc_unittest.cc
14146 M src/tests/tcmalloc_unittest.cc
14147 M src/windows/config.h
14148 M src/windows/patch_functions.cc
14149 M src/windows/port.cc
14150 M src/windows/port.h
14151 M vsprojects/addressmap_unittest/addressmap_unittest.vcproj
14152 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
14153 M
14154 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
14155 M vsprojects/packed-cache_test/packed-cache_test.vcproj
14156 M vsprojects/tmu-static/tmu-static.vcproj
14157
14158 commit 682ff7da1205398376ee725b4ce3219c107b3f8a
14159 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14160 Date: Thu Aug 5 20:36:47 2010 +0000
14161
14162 Thu Aug 5 12:48:03 PDT 2010
14163
14164 * google-perftools: version 1.6 release
14165 * Add tc_malloc_usable_size for compatibility with glibc
14166 (csilvers)
14167 * Override malloc_usable_size with tc_malloc_usable_size
14168 (csilvers)
14169 * Default to no automatic heap sampling in tcmalloc (csilvers)
14170 * Add -DTCMALLOC_LARGE_PAGES, a possibly faster tcmalloc (rus)
14171 * Make some functions extern "C" to avoid false ODR warnings
14172 (jyasskin)
14173 * pprof: Add SVG-based output (rsc)
14174 * pprof: Extend pprof --tools to allow per-tool configs
14175 (csilvers)
14176 * pprof: Improve support of 64-bit and big-endian profiles
14177 (csilvers)
14178 * pprof: Add interactive callgrind suport (weidenri...)
14179 * pprof: Improve address->function mapping a bit (dpeng)
14180 * Better detection of when we're running under valgrind
14181 (csilvers)
14182 * Better CPU-speed detection under valgrind (saito)
14183 * Use, and recommend, -fno-builtin-malloc when compiling
14184 (csilvers)
14185 * Avoid false-sharing of memory between caches (bmaurer)
14186 * BUGFIX: Fix heap sampling to use correct alloc size (bmauer)
14187 * BUGFIX: Avoid gcc 4.0.x bug by making hook-clearing atomic
14188 (csilvers)
14189 * BUGFIX: Avoid gcc 4.5.x optimization bug (csilvers)
14190 * BUGFIX: Work around deps-determining bug in libtool 1.5.26
14191 (csilvers)
14192 * BUGFIX: Fixed test to use HAVE_PTHREAD, not HAVE_PTHREADS
14193 (csilvers)
14194 * BUGFIX: Fix tls callback behavior on windows when using wpo
14195 (wtc)
14196 * BUGFIX: properly align allocation sizes on Windows (antonm)
14197 * BUGFIX: Fix prototypes for tcmalloc/debugalloc wrt throw()
14198 (csilvers)
14199 * DOC: Updated heap-checker doc to match reality better
14200 (fischman)
14201 * DOC: Document ProfilerFlush, ProfilerStartWithOptions
14202 (csilvers)
14203 * DOC: Update docs for heap-profiler functions (csilvers)
14204 * DOC: Clean up documentation around tcmalloc.slack_bytes
14205 (fikes)
14206 * DOC: Renamed README.windows to README_windows.txt (csilvers)
14207 * DOC: Update the NEWS file to be non-empty (csilvers)
14208 * PORTING: Fix windows addr2line and nm with proper rc code
14209 (csilvers)
14210 * PORTING: Add CycleClock and atomicops support for arm 5
14211 (sanek)
14212 * PORTING: Improve PC finding on cygwin and redhat 7
14213 (csilvers)
14214 * PORTING: speed up function-patching under windows (csilvers)
14215
14216
14217 git-svn-id: http://gperftools.googlecode.com/svn/trunk@97
14218 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14219
14220 M ChangeLog
14221 M INSTALL
14222 M NEWS
14223 M README
14224 M configure
14225 M configure.ac
14226 M doc/heap_checker.html
14227 M m4/pc_from_ucontext.m4
14228 M packages/deb/changelog
14229 M src/base/atomicops.h
14230 M src/base/cycleclock.h
14231 M src/base/dynamic_annotations.c
14232 M src/base/dynamic_annotations.h
14233 M src/base/sysinfo.cc
14234 M src/base/thread_annotations.h
14235 M src/base/vdso_support.h
14236 M src/common.cc
14237 M src/common.h
14238 M src/google/malloc_extension.h
14239 M src/google/tcmalloc.h.in
14240 M src/heap-checker.cc
14241 M src/heap-profiler.cc
14242 M src/malloc_hook-inl.h
14243 M src/malloc_hook.cc
14244 M src/pprof
14245 M src/stacktrace_x86-inl.h
14246 M src/tcmalloc.cc
14247 M src/tests/sampler_test.cc
14248 M src/tests/tcmalloc_unittest.cc
14249 M src/windows/google/tcmalloc.h
14250 M src/windows/port.cc
14251 M vsprojects/addressmap_unittest/addressmap_unittest.vcproj
14252 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
14253 M
14254 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
14255 M vsprojects/tmu-static/tmu-static.vcproj
14256
14257 commit 488eee994d571da216ef105d4144282c801f0eee
14258 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14259 Date: Wed Jun 23 16:42:31 2010 +0000
14260
14261 Oops, I thought I had done this last time, but maybe not. Submit the
14262 name change.
14263
14264
14265 git-svn-id: http://gperftools.googlecode.com/svn/trunk@96
14266 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14267
14268 R092 README.windows README_windows.txt
14269
14270 commit cb7393cbe2d737009001fd9d30dad568bac7a3d8
14271 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14272 Date: Mon Jun 21 15:59:56 2010 +0000
14273
14274 * Default to not sampling in tcmalloc (csilvers)
14275 * Add -DTCMALLOC_LARGE_PAGES: better perf for some workloads
14276 (rus)
14277 * Extend pprof --tools to allow per-tool configs (csilvers)
14278 * Have STL_Allocator pass on # bytes to free (richardfang)
14279 * Add a header guard to config.h (csilvers)
14280 * DOC: Clean up documentation around tcmalloc.slack_bytes
14281 (fikes)
14282 * DOC: Document ProfilerFlush, ProfilerStartWithOptions
14283 (csilvers)
14284 * PORTING: Work around a gcc 4.5.0 optimization bug (csilvers)
14285 * PORTING: Use -fno-builtin-malloc and friends when compiling
14286 tcmalloc
14287 * PORTING: Define _WIN32_WINNT high enough for mingw
14288 (csilvers)
14289 * PORTING: Work around libtool bug getting deps wrong in
14290 some cases
14291 * Update README.windows to emphasize $IncludeDir more
14292 (csilvers)
14293 * Rename README.windows to README_windows.txt (csilvers)
14294
14295
14296 git-svn-id: http://gperftools.googlecode.com/svn/trunk@95
14297 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14298
14299 M Makefile.am
14300 M Makefile.in
14301 M README
14302 M configure
14303 M configure.ac
14304 M doc/cpuprofile.html
14305 M doc/tcmalloc.html
14306 M src/base/dynamic_annotations.c
14307 M src/base/dynamic_annotations.h
14308 M src/base/stl_allocator.h
14309 M src/central_freelist.cc
14310 M src/common.h
14311 M src/config.h.in
14312 M src/google/malloc_extension.h
14313 M src/heap-checker.cc
14314 M src/internal_logging.h
14315 M src/linked_list.h
14316 M src/memory_region_map.h
14317 M src/page_heap.cc
14318 M src/page_heap.h
14319 M src/pprof
14320 M src/sampler.cc
14321 M src/span.h
14322 M src/tcmalloc.cc
14323 M src/tests/frag_unittest.cc
14324 M src/tests/page_heap_test.cc
14325 M src/tests/testutil.cc
14326 M src/thread_cache.cc
14327 M src/thread_cache.h
14328 M src/windows/config.h
14329 M src/windows/mingw.h
14330
14331 commit d8c02761689ba909f474b85618f99ac6dfc9a168
14332 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14333 Date: Fri May 7 21:53:24 2010 +0000
14334
14335 * Update docs for heap-profiler fns (csilvers)
14336 * In pprof, accept URLs without ports but with http:// (rsc)
14337 * Refactor sizeclass handling in tcmalloc (bmaurer)
14338 * Always log failed calls to FindAllocDetails (mec)
14339 * Clarify comments for ProfilerStart* (malcolmr)
14340 * Add #include guards to stacktrace_win32-inl.h (glider)
14341 * Add ANNOTATE_ENABLE_RACE_DETECTION(enable) (kcc)
14342 * Make a contentful NEWS file (csilvers)
14343 * Fix addr2line --help (which pprof relies on) for windows
14344 (csilvers)
14345 * Fixes a bug in tcmalloc's TLS callback on windows -static
14346 (wtc)
14347
14348
14349 git-svn-id: http://gperftools.googlecode.com/svn/trunk@94
14350 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14351
14352 M NEWS
14353 M doc/heapprofile.html
14354 M packages/deb/control
14355 M packages/rpm/rpm.spec
14356 M src/base/dynamic_annotations.c
14357 M src/base/dynamic_annotations.h
14358 M src/base/vdso_support.cc
14359 M src/central_freelist.cc
14360 M src/common.h
14361 M src/google/profiler.h
14362 M src/heap-checker.cc
14363 M src/internal_logging.h
14364 M src/page_heap.cc
14365 M src/page_heap.h
14366 M src/pprof
14367 M src/span.h
14368 M src/stacktrace_win32-inl.h
14369 M src/tcmalloc.cc
14370 M src/tests/page_heap_test.cc
14371 M src/windows/addr2line-pdb.c
14372 M src/windows/nm-pdb.c
14373 M src/windows/port.cc
14374
14375 commit b0fe220d503eb23830e622939c2e14f084392d1e
14376 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14377 Date: Thu Apr 22 17:29:02 2010 +0000
14378
14379 * Fix pthread_once extern declarations (willchan)
14380 * Add missing closing brace that resulted in compile failure
14381 (willchan)
14382 * Handle closed stdin/out/err properly when forking (glider)
14383
14384
14385 git-svn-id: http://gperftools.googlecode.com/svn/trunk@93
14386 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14387
14388 M src/maybe_threads.cc
14389 M src/profile-handler.cc
14390 M src/symbolize.cc
14391 M src/tcmalloc.cc
14392
14393 commit 21c65ecb037d7d41dcd574c70cd7b7e9d5045462
14394 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14395 Date: Mon Apr 12 21:18:48 2010 +0000
14396
14397 * Speed up IsSymbolizedProfile by checking for NUL (csilvers)
14398 * Fix heap sampling to use correct alloc size (bmaurer)
14399 * Make pprof ignore tc_new{,array}_nothrow (csilvers)
14400 * PORTING: Have IsHeapProfilerRunning return an int, for C
14401 (csilvers)
14402 * Avoid false-sharing of memory between caches (bmaurer)
14403 * Fix some variable shadowing (rt)
14404 * SVG-based ouptut in pprof; also, wget->curl (rsc)
14405 * Allow arbitrary prefix before obvious handler (rsc)
14406 * Advertise when using libunwind (ppluzhnikov)
14407
14408 Also, the install file seems to have reverted back to the default at
14409 some previous point in time (autotools will do that occasionally).
14410 Change that back to have the perftools-specific text in there.
14411
14412
14413 git-svn-id: http://gperftools.googlecode.com/svn/trunk@92
14414 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14415
14416 M INSTALL
14417 M src/base/basictypes.h
14418 M src/google/heap-profiler.h
14419 M src/heap-checker.cc
14420 M src/heap-profiler.cc
14421 M src/page_heap_allocator.h
14422 M src/pprof
14423 M src/stacktrace_config.h
14424 M src/system-alloc.cc
14425 M src/system-alloc.h
14426 M src/tcmalloc.cc
14427 M src/thread_cache.h
14428
14429 commit 92beff88437b31f4a618640b88487e0f8dfb7017
14430 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14431 Date: Tue Mar 23 20:39:55 2010 +0000
14432
14433 * Add new (std::nothrow) to debugallocation (corrado)
14434 * Add a flag to ingore unaligned-ptr leaks (archanakannan)
14435 * PORTING: Add get-pc capabilities for a new OS (csilvers)
14436 * Don't register malloc extension under valgrind (csilvers)
14437 * Fix throw specs for our global operator new (chandlerc)
14438 * PORTING: link to instructions on windows static overrides
14439 (mbelshe)
14440 * Fix prototype differences in debugalloc (chandlerc,
14441 csilvers, wan)
14442 * Change pprof to handle big-endian input files (csilvers)
14443 * Properly align allocation sizes on Windows (antonm)
14444 * Improve IsRunningOnValgrind, using valgrind.h (csilvers,
14445 kcc)
14446 * Improve the accuracy of system_alloc actual_size (csilvers)
14447 * Add interactive callgrind support to pprof (weidenri...)
14448 * Fix off-by-one problems when symbolizing in pprof (dpeng)
14449 * Be more permissive in allowed library names, in pprof
14450 (csilvers)
14451 * PORTING: Fix pc_from_ucontext to handle cygwin and redhat7
14452 (csilvers)
14453 * Fix stacktrace to avoid inlining (ppluzhnikov)
14454
14455
14456 git-svn-id: http://gperftools.googlecode.com/svn/trunk@91
14457 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14458
14459 M Makefile.am
14460 M Makefile.in
14461 M README.windows
14462 M configure
14463 M configure.ac
14464 M doc/heap_checker.html
14465 M m4/pc_from_ucontext.m4
14466 A src/base/dynamic_annotations.c
14467 D src/base/dynamic_annotations.cc
14468 M src/base/dynamic_annotations.h
14469 M src/base/low_level_alloc.cc
14470 M src/config.h.in
14471 M src/debugallocation.cc
14472 M src/google/stacktrace.h
14473 M src/google/tcmalloc.h.in
14474 M src/heap-checker.cc
14475 M src/heap-profile-table.cc
14476 M src/heap-profile-table.h
14477 M src/malloc_extension.cc
14478 M src/malloc_hook.cc
14479 M src/memory_region_map.cc
14480 M src/pprof
14481 M src/stacktrace.cc
14482 M src/stacktrace_generic-inl.h
14483 M src/stacktrace_libunwind-inl.h
14484 M src/stacktrace_powerpc-inl.h
14485 D src/stacktrace_with_context.cc
14486 M src/stacktrace_x86-inl.h
14487 M src/system-alloc.cc
14488 M src/tcmalloc.cc
14489 M src/tests/debugallocation_test.cc
14490 M src/tests/heap-checker-death_unittest.sh
14491 M src/tests/profiler_unittest.cc
14492 M src/tests/profiler_unittest.sh
14493 M src/tests/tcmalloc_unittest.cc
14494 A src/third_party/valgrind.h
14495 M src/windows/google/tcmalloc.h
14496 M src/windows/port.cc
14497
14498 commit 23dd124970bc11636feaa240394063ba5889ca54
14499 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14500 Date: Thu Feb 11 01:32:42 2010 +0000
14501
14502 * Speed up patching by not patching modules that are already
14503 loaded
14504
14505
14506 git-svn-id: http://gperftools.googlecode.com/svn/trunk@88
14507 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14508
14509 M src/windows/config.h
14510 M src/windows/patch_functions.cc
14511
14512 commit 8f8a010cab7d3070069a31990cb0e269f03dceb0
14513 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14514 Date: Wed Jan 20 23:00:51 2010 +0000
14515
14516 Oops, I submitted the wrong version of patch_functions.cc with the
14517 previous commit. This is really the one that goes with release 1.5.
14518 .
14519 ?\027[D
14520
14521 git-svn-id: http://gperftools.googlecode.com/svn/trunk@86
14522 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14523
14524 M src/windows/patch_functions.cc
14525
14526 commit fef86cfeba8915e34575f09499e18b184d7254a9
14527 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14528 Date: Wed Jan 20 22:47:29 2010 +0000
14529
14530 * google-perftools: version 1.5 release
14531 * Add tc_set_new_mode (willchan)
14532 * Make memalign functions + realloc respect tc_set_new_mode
14533 (willchan)
14534 * Add ReleaseToSystem(num_bytes) (kash)
14535 * Handle zero-length symbols a bit better in pprof (csilvers)
14536 * Prefer __environ to /proc/self/environ in cpu profiler
14537 (csilvers)
14538 * Add HEAP_CHECK_MAX_LEAKS flag to control #leaks to report
14539 (glider)
14540 * Add two new numeric pageheap properties to MallocExtension
14541 (fikes)
14542 * Print alloc size when mmap fails (hakon)
14543 * Add ITIMER_REAL support to cpu profiler (csilvers,
14544 nabeelmian)
14545 * Speed up symbolizer in heap-checker reporting (glider)
14546 * Speed up futexes with FUTEX_PRIVATE_FLAG (m3b)
14547 * Speed up tcmalloc but doing better span coalescing (sanjay)
14548 * Better support for different wget's and addr2maps in pprof
14549 (csilvres)
14550 * Implement a nothrow version of delete and delete[]
14551 (csilvers)
14552 * BUGFIX: fix a race on module_libcs[i] in windows patching
14553 (csilvers)
14554 * BUGFIX: Fix debugallocation to call cpp_alloc for new
14555 (willchan)
14556 * BUGFIX: A simple bugfix for --raw mode (mrabkin)
14557 * BUGFIX: Fix C shims to actually be valid C (csilvers)
14558 * BUGFIX: Fix recursively-unmapped-region accounting
14559 (ppluzhnikov)
14560 * BUGFIX: better distinguish real and fake vdso (ppluzhnikov)
14561 * WINDOWS: replace debugmodule with more reliable psai
14562 (andrey)
14563 * PORTING: Add .bundle as another shared library extension
14564 (csilvers)
14565 * PORTING: Fixed a typo bug in the ocnfigure PRIxx m4 macro
14566 (csilvers)
14567 * PORTING: Augment sysinfo to work on 64-bit OS X (csilvers)
14568 * PORTING: Use sys/ucontext.h to fix compiing on OS X 10.6
14569 (csilvers)
14570 * PORTING: Fix sysinfo libname reporting for solaris x86
14571 (jeffrey)
14572 * PORTING: Use libunwind for i386 when using --omitfp
14573 (ppluzhnikov)
14574
14575 NOTE: This release uses an older version of
14576 src/windows/patch_functions.cc
14577 because I decided the latest optimizations were not well enough tested
14578 for a release. I'll aim to get them into next release.
14579
14580
14581 git-svn-id: http://gperftools.googlecode.com/svn/trunk@84
14582 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14583
14584 M ChangeLog
14585 M INSTALL
14586 M autogen.sh
14587 M configure
14588 M configure.ac
14589 M packages/deb/changelog
14590 M src/pprof
14591 M src/windows/patch_functions.cc
14592
14593 commit 63b8d63beb7d771713774f9a5d57381cbd29bf19
14594 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14595 Date: Thu Jan 14 16:26:05 2010 +0000
14596
14597 * PORTING: Revised patch_functions to avoid deadlock
14598 (csilvers, andrey)
14599 * PORTING: Revised patch_functions to speed up .dll loads
14600 (csilvers)
14601 * PORTING: Build and run sampling_test for windows (csilvers)
14602 * Correctly init tc structs even when libc isn't patched
14603 (csilvers)
14604 * Make low-level allocs async-signal-safe (saito)
14605
14606
14607 git-svn-id: http://gperftools.googlecode.com/svn/trunk@83
14608 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14609
14610 M Makefile.am
14611 M Makefile.in
14612 M README.windows
14613 M src/base/low_level_alloc.cc
14614 M src/base/low_level_alloc.h
14615 M src/google/malloc_hook.h
14616 M src/malloc_hook.cc
14617 M src/pprof
14618 M src/stacktrace_x86-inl.h
14619 M src/tcmalloc.cc
14620 M src/tests/sampling_test.sh
14621 M src/thread_cache.cc
14622 M src/windows/patch_functions.cc
14623 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
14624 M vsprojects/tmu-static/tmu-static.vcproj
14625
14626 commit eeeacd5ec4fa36256091f45e5b3af81cee2a4d86
14627 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14628 Date: Wed Jan 6 00:34:23 2010 +0000
14629
14630 * PORTING: Fix a race condition in windows patching
14631 * PORTING: Use Psapi instead of debugmodule to get windows
14632 module info
14633
14634
14635 git-svn-id: http://gperftools.googlecode.com/svn/trunk@82
14636 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14637
14638 M README.windows
14639 M google-perftools.sln
14640 M src/page_heap.h
14641 M src/windows/patch_functions.cc
14642 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
14643 M vsprojects/tmu-static/tmu-static.vcproj
14644
14645 commit 6e7479331c751bdfe04d272dbb1bbbe877f0e86a
14646 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14647 Date: Tue Dec 15 01:41:30 2009 +0000
14648
14649 * Fix a memory leak with repeated Patch() calls on windows
14650 (csilvers)
14651 * Fix a bug when we re-Patch() a previously unpatched lib
14652 (csilvers)
14653 * Add .bundle as another .so extension in pprof (csilvers)
14654
14655
14656 git-svn-id: http://gperftools.googlecode.com/svn/trunk@81
14657 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14658
14659 M src/base/spinlock.h
14660 M src/heap-checker.cc
14661 M src/heap-profile-table.h
14662 M src/memfs_malloc.cc
14663 M src/memory_region_map.h
14664 M src/packed-cache-inl.h
14665 M src/pprof
14666 M src/profile-handler.cc
14667 M src/profiledata.h
14668 M src/windows/patch_functions.cc
14669 M src/windows/preamble_patcher.cc
14670
14671 commit 1d981b6997f73e6d92b647d042cfe8649d4728c1
14672 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14673 Date: Wed Dec 2 21:42:10 2009 +0000
14674
14675 * Make memalign && posix_memalign respect tc_set_new_mode
14676 (willchan)
14677 * Fix windows patch functions to respect tc_set_new_mode
14678 (willchan)
14679
14680
14681 git-svn-id: http://gperftools.googlecode.com/svn/trunk@80
14682 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14683
14684 M src/tcmalloc.cc
14685 M src/tests/tcmalloc_unittest.cc
14686 M src/windows/patch_functions.cc
14687
14688 commit a94d5f797412bac7b811bf8d69d9298fec54f5cc
14689 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14690 Date: Wed Dec 2 18:15:13 2009 +0000
14691
14692 * Prefer __environ to /proc/self/environ (csilvers)
14693 * Add HEAP_CHECK_MAX_LEAKS envvar (glider)
14694 * BUGFIX: debugallocation now calls cpp_alloc for new
14695 (willchan)
14696 * BUGFIX: tc_set_new_mode() respected for realloc and calloc
14697 (willchan)
14698 * BUGFIX: fix opt-mode maybe-crash on debugallocation_test
14699 (csilvers)
14700 * Print alloc size when mmap fails (hakon)
14701 * Add ITIMER_REAL support (csilvers, nabeelmian)
14702 * BUGFIX: correctly report double-frees (csilvers)
14703 * Export tc_set_new_mode() from the .h file (willchan)
14704 * Restructure Symbolize to make it more efficient (glider)
14705 * PORTING: Augment sysinfo to work on 64-bit OS X (csilvers)
14706 * Add two numeric pageheap properties to MallocExtension
14707 (fikes)
14708 * PORTING: Use libunwind for i386 when using --omitfp
14709 (ppluzhnikov)
14710 * Add ReleaseToSystem(num_bytes) (kash)
14711 * Provide correct library filenames under solaris (jeffrey)
14712 * BUGFIX: simple fix in pprof --raw mode (mrabkin)
14713 * PORTING: Prefer sys/ucontext.h to fix OS 10.6 builds
14714 (csilvers)
14715 * Improve support for inlined functions in pprof (sanjay)
14716 * Update wget code to not use keepalive (mrabkin, csilvers)
14717 * PORTING: correctly handle x86_64 machines that use fp's
14718 (csilvers)
14719
14720
14721 git-svn-id: http://gperftools.googlecode.com/svn/trunk@79
14722 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14723
14724 M configure
14725 M configure.ac
14726 M doc/cpuprofile.html
14727 M doc/heap_checker.html
14728 M m4/compiler_characteristics.m4
14729 M src/base/logging.h
14730 M src/base/sysinfo.cc
14731 M src/config.h.in
14732 M src/debugallocation.cc
14733 M src/google/malloc_extension.h
14734 M src/google/malloc_extension_c.h
14735 M src/google/tcmalloc.h.in
14736 M src/heap-profile-table.cc
14737 M src/malloc_extension.cc
14738 M src/memfs_malloc.cc
14739 M src/pprof
14740 M src/profile-handler.cc
14741 M src/symbolize.cc
14742 M src/symbolize.h
14743 M src/tcmalloc.cc
14744 M src/tests/atomicops_unittest.cc
14745 M src/tests/debugallocation_test.cc
14746 M src/tests/malloc_extension_test.cc
14747 M src/tests/profile-handler_unittest.cc
14748 M src/tests/profiler_unittest.sh
14749 M src/tests/tcmalloc_unittest.cc
14750 M src/windows/config.h
14751 M src/windows/google/tcmalloc.h
14752
14753 commit 5b80f01df1137337131b4c50ce97faaff9973e90
14754 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14755 Date: Tue Nov 10 16:24:57 2009 +0000
14756
14757 * Replace usleep() and poll() with nanosleep() (glider)
14758 * Document problems with _recalloc (csilvers)
14759 * Detect when x86_64 doesn't turn off frame pointers
14760 (csilvers)
14761 * Fix sysinfo.cc/etc to work with 64-bit os x (csilvers)
14762 * BUGFIX: Use __TEXT instead of __DATA to store tcmalloc fns
14763 (csilvers)
14764 * Added two numeric pageheap properties to tcmalloc (fikes)
14765 * Support for mallocranges stats visualization (sanjay)
14766 * Use libunwind for i386, not just x86_64 (ppluzhnikov)
14767 * Add ReleaseToSystem(num_bytes) (kash)
14768 * Provide corect library filenames under solaris (jeffrey)
14769 * BUGFIX: a simple bug in pprof --raw mode (mrabkin)
14770 * Prfer sys/ucontext.h to ucontext.h, to fix OS X 10.6
14771 (csilvers)
14772 * Improve supprot for inlined functions in pprof (sanjay)
14773 * Document inaccuracies in profiling mmap calls (csilvers)
14774 * Update wget code to not use keepalive (mrabkin, csilvers)
14775
14776
14777 git-svn-id: http://gperftools.googlecode.com/svn/trunk@78
14778 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14779
14780 M Makefile.am
14781 M Makefile.in
14782 M README.windows
14783 M configure
14784 M configure.ac
14785 M doc/heapprofile.html
14786 M google-perftools.sln
14787 M m4/pc_from_ucontext.m4
14788 M src/base/basictypes.h
14789 M src/base/dynamic_annotations.h
14790 M src/base/sysinfo.cc
14791 M src/base/sysinfo.h
14792 M src/debugallocation.cc
14793 M src/google/heap-checker.h
14794 M src/google/malloc_extension.h
14795 M src/google/malloc_extension_c.h
14796 M src/heap-checker.cc
14797 M src/malloc_extension.cc
14798 M src/page_heap.cc
14799 M src/page_heap.h
14800 M src/pagemap.h
14801 M src/pprof
14802 M src/stacktrace_config.h
14803 M src/symbolize.cc
14804 M src/symbolize.h
14805 M src/tcmalloc.cc
14806 M src/tests/malloc_extension_c_test.c
14807 A src/tests/page_heap_test.cc
14808 M src/tests/pagemap_unittest.cc
14809 M src/tests/profile-handler_unittest.cc
14810 M src/tests/tcmalloc_unittest.cc
14811 A vsprojects/page_heap_test/page_heap_test.vcproj
14812
14813 commit 25eed16e1b042a80c9a3e83bbf7ed227d04fb45a
14814 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14815 Date: Tue Oct 27 17:30:52 2009 +0000
14816
14817 * Fix Symbolize() to call pprof once, rather than once/symbol
14818 (glider)
14819 * Fix unsetting of hooks before forking, in debug mode (maxim)
14820 * Add some documention for pmuprofile (aruns)
14821 * Speed up futex with FUTEX_PRIVATE_FLAG (m3b)
14822 * Fix os x 10.6: prefer sys/ucontext.h to ucontext.h
14823 (csilvers)
14824 * Fix C shims to be actually valid C: malloc_extension/etc
14825 (csilvers)
14826 * Fix a longtime memset bug (csilvers)
14827 * Implement nothrow versions of delete (csilvers)
14828 * Fix recursively-unmapped-region accounting (ppluzhnikov)
14829 * Better distinguish between real and fake VDSO (ppluzhnikov)
14830 * Modify span coalescing to improve performance (sanjay)
14831 * WINDOWS: Remove unnecessary lock around VirtualAlloc
14832 (mbelshe)
14833 * Remove performance tests for ptmalloc2 (csilvers)
14834
14835
14836 git-svn-id: http://gperftools.googlecode.com/svn/trunk@77
14837 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14838
14839 M Makefile.am
14840 M Makefile.in
14841 M configure
14842 M configure.ac
14843 M doc/pprof_remote_servers.html
14844 M src/base/spinlock_linux-inl.h
14845 M src/base/vdso_support.cc
14846 M src/config.h.in
14847 M src/debugallocation.cc
14848 M src/getpc.h
14849 M src/google/malloc_extension_c.h
14850 M src/google/tcmalloc.h.in
14851 M src/heap-checker.cc
14852 M src/heap-profile-table.cc
14853 M src/malloc_extension.cc
14854 M src/memory_region_map.cc
14855 M src/page_heap.cc
14856 M src/page_heap.h
14857 M src/pprof
14858 M src/profiler.cc
14859 M src/sampler.cc
14860 M src/stacktrace_libunwind-inl.h
14861 M src/stacktrace_x86-inl.h
14862 M src/symbolize.cc
14863 M src/symbolize.h
14864 M src/tcmalloc.cc
14865 M src/tests/heap-checker-death_unittest.sh
14866 M src/tests/heap-checker_unittest.cc
14867 A src/tests/malloc_extension_c_test.c
14868 D src/tests/ptmalloc/COPYRIGHT
14869 D src/tests/ptmalloc/lran2.h
14870 D src/tests/ptmalloc/malloc-machine.h
14871 D src/tests/ptmalloc/t-test.h
14872 D src/tests/ptmalloc/t-test1.c
14873 D src/tests/ptmalloc/t-test2.c
14874 D src/tests/ptmalloc/thread-m.h
14875 D src/tests/ptmalloc/thread-st.h
14876 M src/tests/tcmalloc_unittest.cc
14877 M src/windows/config.h
14878 M src/windows/google/tcmalloc.h
14879 M src/windows/port.cc
14880
14881 commit 19dfa9e3733155e57406fbd082273eb53cb2750e
14882 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
14883 Date: Fri Sep 11 18:42:32 2009 +0000
14884
14885 Thu Sep 10 13:51:15 2009 Google Inc. <[email protected]>
14886
14887 * google-perftools: version 1.4 release
14888 * Add debugallocation library, to catch memory leaks,
14889 stomping, etc
14890 * Add --raw mode to allow for delayed processing of pprof
14891 files
14892 * Use less memory when reading CPU profiles
14893 * New environment variables to control kernel-allocs (sbrk,
14894 memfs, etc)
14895 * Add MarkThreadBusy(): performance improvement
14896 * Remove static thread-cache-size code; all is dynamic now
14897 * Add new HiddenPointer class to heap checker
14898 * BUGFIX: pvalloc(0) allocates now (found by new debugalloc
14899 library)
14900 * BUGFIX: valloc test (not implementation) no longer
14901 overruns memory
14902 * BUGFIX: GetHeapProfile no longer deadlocks
14903 * BUGFIX: Support unmapping memory regions before main
14904 * BUGFIX: Fix some malloc-stats formatting
14905 * BUGFIX: Don't crash as often when freeing libc-allocated
14906 memory
14907 * BUGFIX: Deal better with incorrect PPROF_PATH when
14908 symbolizing
14909 * BUGFIX: weaken new/delete/etc in addition to malloc/free/etc
14910 * BUGFIX: Fix return value of GetAllocatedSize
14911 * PORTING: Fix mmap-#define problem on some 64-bit systems
14912 * PORTING: Call ranlib again (some OS X versions need it)
14913 * PORTING: Fix a leak when building with LLVM
14914 * PORTING: Remove some unneeded bash-ishs from testing scripts
14915 * WINDOWS: Support library unloading as well as loading
14916 * WINDOWS/BUGFIX: Set page to 'xrw' instead of 'rw' when
14917 patching
14918
14919
14920 git-svn-id: http://gperftools.googlecode.com/svn/trunk@76
14921 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
14922
14923 M ChangeLog
14924 M Makefile.am
14925 M Makefile.in
14926 M aclocal.m4
14927 M configure
14928 M configure.ac
14929 M doc/tcmalloc.html
14930 M google-perftools.sln
14931 M m4/install_prefix.m4
14932 A m4/pc_from_ucontext.m4
14933 M packages/deb/changelog
14934 M src/base/atomicops-internals-macosx.h
14935 M src/base/atomicops-internals-x86.cc
14936 M src/base/atomicops-internals-x86.h
14937 M src/base/basictypes.h
14938 M src/base/dynamic_annotations.cc
14939 M src/base/dynamic_annotations.h
14940 M src/base/logging.h
14941 M src/base/low_level_alloc.cc
14942 M src/base/low_level_alloc.h
14943 M src/base/simple_mutex.h
14944 M src/base/spinlock.cc
14945 M src/base/spinlock.h
14946 A src/base/spinlock_linux-inl.h
14947 A src/base/spinlock_posix-inl.h
14948 A src/base/spinlock_win32-inl.h
14949 M src/base/sysinfo.cc
14950 M src/base/sysinfo.h
14951 M src/base/vdso_support.cc
14952 M src/base/vdso_support.h
14953 M src/common.cc
14954 M src/common.h
14955 M src/config.h.in
14956 A src/debugallocation.cc
14957 M src/getpc.h
14958 M src/google/heap-checker.h
14959 M src/google/malloc_extension.h
14960 M src/google/malloc_extension_c.h
14961 M src/google/tcmalloc.h.in
14962 M src/heap-checker.cc
14963 M src/heap-profile-table.cc
14964 M src/heap-profiler.cc
14965 M src/malloc_extension.cc
14966 M src/malloc_hook.cc
14967 M src/memfs_malloc.cc
14968 M src/memory_region_map.cc
14969 M src/packed-cache-inl.h
14970 M src/page_heap.h
14971 M src/pagemap.h
14972 M src/pprof
14973 M src/profiler.cc
14974 M src/sampler.h
14975 M src/stacktrace_win32-inl.h
14976 M src/stacktrace_with_context.cc
14977 M src/stacktrace_x86-inl.h
14978 A src/symbolize.cc
14979 C060 src/stacktrace_with_context.cc src/symbolize.h
14980 M src/system-alloc.cc
14981 M src/tcmalloc.cc
14982 M src/tests/addressmap_unittest.cc
14983 A src/tests/debugallocation_test.cc
14984 A src/tests/debugallocation_test.sh
14985 M src/tests/heap-checker_unittest.cc
14986 M src/tests/heap-profiler_unittest.cc
14987 M src/tests/heap-profiler_unittest.sh
14988 M src/tests/low_level_alloc_unittest.cc
14989 M src/tests/malloc_extension_test.cc
14990 M src/tests/markidle_unittest.cc
14991 M src/tests/memalign_unittest.cc
14992 M src/tests/pagemap_unittest.cc
14993 M src/tests/profile-handler_unittest.cc
14994 M src/tests/profiledata_unittest.cc
14995 M src/tests/profiler_unittest.cc
14996 M src/tests/profiler_unittest.sh
14997 M src/tests/sampler_test.cc
14998 M src/tests/stacktrace_unittest.cc
14999 M src/tests/tcmalloc_unittest.cc
15000 M src/thread_cache.cc
15001 M src/thread_cache.h
15002 M src/windows/config.h
15003 M src/windows/google/tcmalloc.h
15004 M src/windows/patch_functions.cc
15005 M src/windows/port.h
15006 M src/windows/preamble_patcher.cc
15007 M vsprojects/addr2line-pdb/addr2line-pdb.vcproj
15008 M vsprojects/addressmap_unittest/addressmap_unittest.vcproj
15009 M vsprojects/frag_unittest/frag_unittest.vcproj
15010 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
15011 M
15012 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
15013 M vsprojects/malloc_extension_test/malloc_extension_test.vcproj
15014 M vsprojects/markidle_unittest/markidle_unittest.vcproj
15015 M vsprojects/nm-pdb/nm-pdb.vcproj
15016 M vsprojects/packed-cache_test/packed-cache_test.vcproj
15017 M vsprojects/pagemap_unittest/pagemap_unittest.vcproj
15018 M vsprojects/realloc_unittest/realloc_unittest.vcproj
15019 M vsprojects/sampler_test/sampler_test.vcproj
15020 M vsprojects/stack_trace_table_test/stack_trace_table_test.vcproj
15021 M
15022 vsprojects/tcmalloc_minimal_large/tcmalloc_minimal_large_unittest.vcproj
15023 M
15024 vsprojects/tcmalloc_minimal_unittest/tcmalloc_minimal_unittest.vcproj
15025 M vsprojects/thread_dealloc_unittest/thread_dealloc_unittest.vcproj
15026 M vsprojects/tmu-static/tmu-static.vcproj
15027
15028 commit 2197cc670204c583bba3903b765c77620f349609
15029 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15030 Date: Wed Jun 10 02:04:26 2009 +0000
15031
15032 Tue Jun 9 18:19:06 2009 Google Inc. <[email protected]>
15033
15034 * google-perftools: version 1.3 release
15035 * Provide our own name for memory functions: tc_malloc, etc
15036 (csilvers)
15037 * Weaken memory-alloc functions so user can override them
15038 (csilvers)
15039 * Remove meaningless delete(nothrow) and delete[](nothrow)
15040 (csilvers)
15041 * BUILD: replace clever libtcmalloc/profiler.a with a new .a
15042 (csilvers)
15043 * PORTING: improve windows port by using google spinlocks
15044 (csilvers)
15045 * PORTING: Fix RedHat 9 memory allocation in heapchecker
15046 (csilvers)
15047 * PORTING: Rename OS_WINDOWS macro to PLATFORM_WINDOWS
15048 (mbelshe)
15049 * PORTING/BUGFIX: Make sure we don't clobber GetLastError
15050 (mbelshe)
15051 * BUGFIX: get rid of useless data for callgrind (weidenrinde)
15052 * BUGFIX: Modify windows patching to deadlock sometimes
15053 (csilvers)
15054 * BUGFIX: an improved fix for hook handling during fork
15055 (csilvers)
15056 * BUGFIX: revamp profiler_unittest.sh, which was very broken
15057 (csilvers)
15058
15059
15060 git-svn-id: http://gperftools.googlecode.com/svn/trunk@74
15061 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15062
15063 M ChangeLog
15064 M Makefile.am
15065 M Makefile.in
15066 M README
15067 M aclocal.m4
15068 M configure
15069 M configure.ac
15070 D m4/stl_hash.m4
15071 M packages/deb/changelog
15072 M packages/rpm/rpm.spec
15073 M src/base/atomicops-internals-linuxppc.h
15074 M src/base/dynamic_annotations.cc
15075 M src/base/dynamic_annotations.h
15076 M src/base/spinlock.cc
15077 M src/base/sysinfo.cc
15078 M src/base/sysinfo.h
15079 M src/base/vdso_support.cc
15080 M src/base/vdso_support.h
15081 M src/config.h.in
15082 A src/google/tcmalloc.h.in
15083 M src/heap-checker.cc
15084 M src/pprof
15085 M src/profile-handler.cc
15086 M src/tcmalloc.cc
15087 M src/tests/heap-checker-death_unittest.sh
15088 M src/tests/profiler_unittest.sh
15089 M src/tests/stacktrace_unittest.cc
15090 M src/windows/config.h
15091 A src/windows/google/tcmalloc.h
15092 M src/windows/override_functions.cc
15093 M src/windows/patch_functions.cc
15094
15095 commit 104bf697fbd2a0b90b5f01344ee01c8caa1745d0
15096 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15097 Date: Mon May 18 22:50:20 2009 +0000
15098
15099 Use the google spinlock code instead of the built-in windows code.
15100 The main benefit for perftools is that google spinlocks allow for
15101 link-time (static) initialization, which we had to simulate before,
15102 yielding bugs and worse performance.
15103
15104
15105 git-svn-id: http://gperftools.googlecode.com/svn/trunk@73
15106 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15107
15108 M src/base/atomicops-internals-x86-msvc.h
15109 M src/base/spinlock.cc
15110 M src/base/spinlock.h
15111 M src/windows/config.h
15112 M src/windows/patch_functions.cc
15113 M src/windows/port.cc
15114 M src/windows/port.h
15115 M vsprojects/addressmap_unittest/addressmap_unittest.vcproj
15116 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
15117 M
15118 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
15119 M vsprojects/packed-cache_test/packed-cache_test.vcproj
15120 M vsprojects/tmu-static/tmu-static.vcproj
15121
15122 commit ad03b009ef2046cee9dc38afe022b487de37db5c
15123 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15124 Date: Sat Apr 25 01:01:23 2009 +0000
15125
15126 In the case of windows with HAS_EXCEPTIONS turned off we weren't able
15127 to use the std::set_new_handler correctly. Rework the #ifdefs to
15128 allow use of the std_new_handler, but ignore the exceptions.
15129
15130 Patch submitted by mbelshe.
15131
15132
15133 git-svn-id: http://gperftools.googlecode.com/svn/trunk@72
15134 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15135
15136 M src/tcmalloc.cc
15137
15138 commit 1894763f57b00f72a720c4c126815c073d9ed0f3
15139 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15140 Date: Wed Apr 22 22:53:41 2009 +0000
15141
15142 Make sure we don't clobber GetLastError() (for windows).
15143
15144
15145 git-svn-id: http://gperftools.googlecode.com/svn/trunk@71
15146 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15147
15148 M src/windows/port.h
15149
15150 commit a63235c4331445d718d151a2a659ec9687a0b8c8
15151 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15152 Date: Tue Apr 21 17:10:29 2009 +0000
15153
15154 1) Change #include "config.h" to <config.h>. This is what automake
15155 recommends, and makes it easier to override a config file.
15156
15157 2) Rename OS_WINDOWS in sysinfo.cc, to not conflict with a macro
15158 defined in a windows SDK somewhere.
15159
15160
15161 git-svn-id: http://gperftools.googlecode.com/svn/trunk@70
15162 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15163
15164 M src/base/atomicops.h
15165 M src/base/basictypes.h
15166 M src/base/commandlineflags.h
15167 M src/base/elfcore.h
15168 M src/base/logging.cc
15169 M src/base/logging.h
15170 M src/base/low_level_alloc.h
15171 M src/base/simple_mutex.h
15172 M src/base/spinlock.cc
15173 M src/base/spinlock.h
15174 M src/base/stl_allocator.h
15175 M src/base/sysinfo.cc
15176 M src/base/sysinfo.h
15177 M src/base/vdso_support.h
15178 M src/heap-checker.cc
15179 M src/heap-profile-table.cc
15180 M src/heap-profiler.cc
15181 M src/internal_logging.cc
15182 M src/internal_logging.h
15183 M src/malloc_extension.cc
15184 M src/malloc_hook.cc
15185 M src/memfs_malloc.cc
15186 M src/memory_region_map.cc
15187 M src/memory_region_map.h
15188 M src/page_heap.cc
15189 M src/page_heap.h
15190 M src/profiledata.cc
15191 M src/profiledata.h
15192 M src/raw_printer.cc
15193 M src/raw_printer.h
15194 M src/span.cc
15195 M src/span.h
15196 M src/stack_trace_table.cc
15197 M src/stack_trace_table.h
15198 M src/stacktrace.cc
15199 M src/stacktrace_with_context.cc
15200 M src/static_vars.h
15201 M src/system-alloc.cc
15202 M src/system-alloc.h
15203 M src/tcmalloc.cc
15204 M src/tcmalloc.h
15205 M src/thread_cache.cc
15206 M src/thread_cache.h
15207 M src/windows/override_functions.cc
15208 M src/windows/patch_functions.cc
15209 M src/windows/port.cc
15210
15211 commit beb6a9a183c1ca25c99e4401b58266ce73b8c846
15212 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15213 Date: Sat Apr 18 00:02:25 2009 +0000
15214
15215 Fri Apr 17 16:40:48 2009 Google Inc. <[email protected]>
15216
15217 * google-perftools: version 1.2 release
15218 * Allow large_alloc_threshold=0 to turn it off entirely
15219 (csilvers)
15220 * Die more helpfully when out of memory for internal data
15221 (csilvers)
15222 * Refactor profile-data gathering, add a new unittest (cgd,
15223 nabeelmian)
15224 * BUGFIX: fix rounding errors with static thread-size caches
15225 (addi)
15226 * BUGFIX: disable hooks better when forking in leak-checker
15227 (csilvers)
15228 * BUGFIX: fix realloc of crt pointers on windows (csilvers)
15229 * BUGFIX: do a better job of finding binaries in .sh tests
15230 (csilvers)
15231 * WINDOWS: allow overriding malloc/etc instead of patching
15232 (mbelshe)
15233 * PORTING: fix compilation error in a ppc-specific file
15234 (csilvers)
15235 * PORTING: deal with quirks in cygwin's /proc/self/maps
15236 (csilvers)
15237 * PORTING: use 'A' version of functions for ascii input
15238 (mbelshe)
15239 * PORTING: generate .so's on cygwin and mingw (ajenjo)
15240 * PORTING: disable profiler methods on cygwin (jperkins)
15241 * Updated autoconf version to 2.61 and libtool version
15242 to 1.5.26
15243
15244
15245 git-svn-id: http://gperftools.googlecode.com/svn/trunk@68
15246 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15247
15248 M ChangeLog
15249 M INSTALL
15250 M Makefile.am
15251 M Makefile.in
15252 M README.windows
15253 M aclocal.m4
15254 M config.guess
15255 M config.sub
15256 M configure
15257 M configure.ac
15258 M ltmain.sh
15259 M packages/deb/changelog
15260 M packages/rpm.sh
15261 M packages/rpm/rpm.spec
15262 M src/base/atomicops-internals-linuxppc.h
15263 M src/base/logging.cc
15264 M src/base/logging.h
15265 M src/base/simple_mutex.h
15266 M src/base/sysinfo.cc
15267 M src/base/sysinfo.h
15268 M src/config.h.in
15269 M src/google/heap-checker.h
15270 M src/google/profiler.h
15271 M src/heap-checker.cc
15272 M src/heap-profile-table.cc
15273 M src/heap-profile-table.h
15274 M src/page_heap_allocator.h
15275 M src/pprof
15276 A src/profile-handler.cc
15277 A src/profile-handler.h
15278 M src/profiledata.cc
15279 M src/profiledata.h
15280 M src/profiler.cc
15281 M src/sampler.cc
15282 M src/stacktrace_win32-inl.h
15283 M src/stacktrace_x86-inl.h
15284 M src/tcmalloc.cc
15285 M src/tests/frag_unittest.cc
15286 A src/tests/profile-handler_unittest.cc
15287 M src/tests/profiledata_unittest.cc
15288 M src/tests/sampling_test.cc
15289 M src/tests/sampling_test.sh
15290 M src/tests/testutil.cc
15291 M src/thread_cache.cc
15292 M src/windows/addr2line-pdb.c
15293 M src/windows/config.h
15294 M src/windows/mingw.h
15295 M src/windows/nm-pdb.c
15296 A src/windows/override_functions.cc
15297 M src/windows/patch_functions.cc
15298 M src/windows/port.cc
15299 M src/windows/port.h
15300
15301 commit edd03a831f350bc72d76d4fad2b390d43faccb79
15302 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15303 Date: Wed Mar 11 20:50:03 2009 +0000
15304
15305 Wed Mar 11 11:25:34 2009 Google Inc. <[email protected]>
15306
15307 * google-perftools: version 1.1 release
15308 * Dynamically resize thread caches -- nice perf. improvement
15309 (kash)
15310 * Add VDSO support to give better stacktraces in linux
15311 (ppluzhnikov)
15312 * Improve heap-profiling sampling algorithm (ford)
15313 * Rewrite leak-checking code: should be faster and more robust
15314 (sanjay)
15315 * Use ps2 instead of ps for dot: better page cropping for gv
15316 (csilvers)
15317 * Disable malloc-failure warning messages by default
15318 (csilvers)
15319 * Update config/Makefile to disable tests on a per-OS basis
15320 (csilvers)
15321 * PORTING: Get perftools compiling under MSVC 7.1 again
15322 (csilvers)
15323 * PORTING: Get perftools compiling under cygwin again
15324 (csilvers)
15325 * PORTING: automatically set library flags for solaris x86
15326 (csilvers)
15327 * Add TCMALLOC_SKIP_SBRK to mirror TCMALLOC_SKIP_MMAP
15328 (csilvers)
15329 * Add --enable flags to allow selective building (csilvers)
15330 * Put addr2line-pdb and nm-pdb in proper output directory
15331 (csilvers)
15332 * Remove deprecated DisableChecksIn (sanjay)
15333 * DOCUMENTATION: Document most MallocExtension routines
15334 (csilvers)
15335
15336
15337 git-svn-id: http://gperftools.googlecode.com/svn/trunk@66
15338 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15339
15340 M ChangeLog
15341 M INSTALL
15342 M Makefile.am
15343 M Makefile.in
15344 M README
15345 M README.windows
15346 M aclocal.m4
15347 M configure
15348 M configure.ac
15349 M doc/pprof_remote_servers.html
15350 M doc/tcmalloc.html
15351 M google-perftools.sln
15352 A m4/acx_nanosleep.m4
15353 M packages/deb.sh
15354 M packages/deb/changelog
15355 M src/base/basictypes.h
15356 M src/base/cycleclock.h
15357 M src/base/dynamic_annotations.cc
15358 M src/base/dynamic_annotations.h
15359 M src/base/linux_syscall_support.h
15360 M src/base/simple_mutex.h
15361 M src/base/spinlock.h
15362 M src/base/sysinfo.cc
15363 M src/base/sysinfo.h
15364 M src/base/thread_annotations.h
15365 A src/base/vdso_support.cc
15366 A src/base/vdso_support.h
15367 M src/central_freelist.cc
15368 M src/common.cc
15369 M src/common.h
15370 M src/config.h.in
15371 M src/google/heap-checker.h
15372 M src/google/malloc_extension.h
15373 M src/google/malloc_extension_c.h
15374 M src/google/stacktrace.h
15375 M src/heap-checker.cc
15376 M src/heap-profile-table.cc
15377 M src/heap-profile-table.h
15378 M src/heap-profiler.cc
15379 M src/internal_logging.cc
15380 M src/internal_logging.h
15381 M src/malloc_extension.cc
15382 M src/malloc_hook.cc
15383 M src/memory_region_map.cc
15384 M src/page_heap.cc
15385 M src/pprof
15386 M src/profiler.cc
15387 C061 src/static_vars.cc src/raw_printer.cc
15388 A src/raw_printer.h
15389 A src/sampler.cc
15390 A src/sampler.h
15391 M src/span.cc
15392 A src/stack_trace_table.cc
15393 C056 src/static_vars.cc src/stack_trace_table.h
15394 M src/stacktrace.cc
15395 C057 src/stacktrace.cc src/stacktrace_config.h
15396 M src/stacktrace_libunwind-inl.h
15397 M src/stacktrace_win32-inl.h
15398 C059 src/static_vars.cc src/stacktrace_with_context.cc
15399 M src/stacktrace_x86-inl.h
15400 M src/stacktrace_x86_64-inl.h
15401 M src/static_vars.cc
15402 M src/static_vars.h
15403 M src/system-alloc.cc
15404 M src/tcmalloc.cc
15405 M src/tests/heap-checker-death_unittest.sh
15406 M src/tests/heap-checker_unittest.cc
15407 M src/tests/heap-profiler_unittest.sh
15408 C060 src/tests/sampling_test.cc src/tests/malloc_extension_test.cc
15409 A src/tests/pagemap_unittest.cc
15410 A src/tests/raw_printer_test.cc
15411 A src/tests/realloc_unittest.cc
15412 A src/tests/sampler_test.cc
15413 M src/tests/sampling_test.cc
15414 M src/tests/sampling_test.sh
15415 A src/tests/stack_trace_table_test.cc
15416 M src/tests/tcmalloc_unittest.cc
15417 M src/thread_cache.cc
15418 M src/thread_cache.h
15419 M src/windows/addr2line-pdb.c
15420 M src/windows/config.h
15421 M src/windows/patch_functions.cc
15422 M src/windows/port.h
15423 M vsprojects/addr2line-pdb/addr2line-pdb.vcproj
15424 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
15425 M
15426 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
15427 C061 vsprojects/nm-pdb/nm-pdb.vcproj
15428 vsprojects/malloc_extension_test/malloc_extension_test.vcproj
15429 M vsprojects/nm-pdb/nm-pdb.vcproj
15430 C064 vsprojects/nm-pdb/nm-pdb.vcproj
15431 vsprojects/pagemap_unittest/pagemap_unittest.vcproj
15432 C065 vsprojects/nm-pdb/nm-pdb.vcproj
15433 vsprojects/realloc_unittest/realloc_unittest.vcproj
15434 C063 vsprojects/nm-pdb/nm-pdb.vcproj
15435 vsprojects/sampler_test/sampler_test.vcproj
15436 C061 vsprojects/nm-pdb/nm-pdb.vcproj
15437 vsprojects/stack_trace_table_test/stack_trace_table_test.vcproj
15438 M vsprojects/tmu-static/tmu-static.vcproj
15439
15440 commit c75de4d1e91c339fb5142a8a21be8b3ba5224ef7
15441 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15442 Date: Tue Jan 6 19:41:15 2009 +0000
15443
15444 Tue Jan 6 13:58:56 2009 Google Inc. <[email protected]>
15445 * google-perftools: version 1.0 release
15446 * Exactly the same as 1.0rc2
15447
15448
15449 git-svn-id: http://gperftools.googlecode.com/svn/trunk@64
15450 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15451
15452 M ChangeLog
15453 M configure
15454 M configure.ac
15455 M packages/deb/changelog
15456 M src/pprof
15457
15458 commit 49b02736109f9fe3383971e62bda2192ae86a36d
15459 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15460 Date: Mon Dec 15 01:20:31 2008 +0000
15461
15462 Sun Dec 14 17:10:35 2008 Google Inc. <[email protected]>
15463 * google-perftools: version 1.0rc2 release
15464 * Fix compile error on 64-bit systems (casting ptr to int)
15465 (csilvers)
15466
15467 Thu Dec 11 16:01:32 2008 Google Inc. <[email protected]>
15468
15469
15470 git-svn-id: http://gperftools.googlecode.com/svn/trunk@62
15471 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15472
15473 M ChangeLog
15474 M configure
15475 M configure.ac
15476 M src/heap-checker.cc
15477
15478 commit 6fa2a2574ce1c15ac12293e24691d69a41972e54
15479 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15480 Date: Sat Dec 13 01:35:42 2008 +0000
15481
15482 Thu Dec 11 16:01:32 2008 Google Inc. <[email protected]>
15483
15484 * google-perftools: version 1.0rc1 release
15485 * Replace API for selectively disabling heap-checker in code
15486 (sanjay)
15487 * Add a pre-mmap hook (daven, adlr)
15488 * Add MallocExtension interface to set memory-releasing rate
15489 (fikes)
15490 * Augment pprof to allow any string ending in /pprof/profile
15491 (csilvers)
15492 * PORTING: Rewrite -- and fix -- malloc patching for windows
15493 (dvitek)
15494 * PORTING: Add nm-pdb and addr2line-pdb for use by pprof
15495 (dvitek)
15496 * PORTING: Improve cygwin and mingw support (jperkins,
15497 csilvers)
15498 * PORTING: Fix pprof for mac os x, other pprof improvements
15499 (csilvers)
15500 * PORTING: Fix some PPC bugs in our locking code
15501 (anton.blanchard)
15502 * A new unittest, smapling_test, to verify tcmalloc-profiles
15503 (csilvers)
15504 * Turn off TLS for gcc < 4.1.2, due to a TLS + -fPIC bug
15505 (csilvers)
15506 * Prefer __builtin_frame_address to assembly for stacktraces
15507 (nlewycky)
15508 * Separate tcmalloc.cc out into multiple files --
15509 finally! (kash)
15510 * Make our locking code work with -fPIC on 32-bit x86 (aruns)
15511 * Fix an initialization-ordering bug for tcmalloc/profiling
15512 (csilvers)
15513 * Use "initial exec" model of TLS to speed up tcmalloc
15514 (csilvers)
15515 * Enforce 16-byte alignment for tcmalloc, for SSE (sanjay)
15516
15517
15518 git-svn-id: http://gperftools.googlecode.com/svn/trunk@60
15519 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15520
15521 M ChangeLog
15522 M INSTALL
15523 M Makefile.am
15524 M Makefile.in
15525 M README.windows
15526 M configure
15527 M configure.ac
15528 M doc/heap_checker.html
15529 M doc/pprof_remote_servers.html
15530 M google-perftools.sln
15531 M packages/deb.sh
15532 M packages/deb/changelog
15533 M src/base/atomicops-internals-linuxppc.h
15534 M src/base/atomicops-internals-macosx.h
15535 M src/base/atomicops-internals-x86.h
15536 M src/base/basictypes.h
15537 M src/base/linux_syscall_support.h
15538 M src/base/logging.cc
15539 M src/base/logging.h
15540 M src/base/simple_mutex.h
15541 M src/base/sysinfo.cc
15542 M src/base/sysinfo.h
15543 A src/base/thread_annotations.h
15544 M src/base/thread_lister.c
15545 A src/central_freelist.cc
15546 A src/central_freelist.h
15547 A src/common.cc
15548 A src/common.h
15549 M src/getpc.h
15550 M src/google/heap-checker.h
15551 M src/google/heap-profiler.h
15552 M src/google/malloc_extension.h
15553 M src/google/malloc_hook.h
15554 M src/google/malloc_hook_c.h
15555 M src/google/profiler.h
15556 M src/google/stacktrace.h
15557 M src/heap-checker.cc
15558 M src/heap-profile-table.cc
15559 M src/heap-profile-table.h
15560 M src/heap-profiler.cc
15561 A src/linked_list.h
15562 M src/malloc_extension.cc
15563 M src/malloc_hook-inl.h
15564 M src/malloc_hook.cc
15565 M src/memory_region_map.cc
15566 M src/memory_region_map.h
15567 A src/page_heap.cc
15568 A src/page_heap.h
15569 A src/page_heap_allocator.h
15570 M src/pprof
15571 M src/profiler.cc
15572 A src/span.cc
15573 A src/span.h
15574 M src/stacktrace.cc
15575 M src/stacktrace_libunwind-inl.h
15576 A src/stacktrace_win32-inl.h
15577 M src/stacktrace_x86-inl.h
15578 C059 src/base/logging.cc src/static_vars.cc
15579 A src/static_vars.h
15580 M src/tcmalloc.cc
15581 C071 src/base/logging.cc src/tcmalloc_guard.h
15582 M src/tests/atomicops_unittest.cc
15583 M src/tests/frag_unittest.cc
15584 M src/tests/heap-checker_unittest.cc
15585 M src/tests/low_level_alloc_unittest.cc
15586 C052 src/base/logging.cc src/tests/sampling_test.cc
15587 A src/tests/sampling_test.sh
15588 M src/tests/stacktrace_unittest.cc
15589 M src/tests/tcmalloc_unittest.cc
15590 M src/tests/testutil.cc
15591 A src/thread_cache.cc
15592 A src/thread_cache.h
15593 A src/windows/addr2line-pdb.c
15594 A src/windows/get_mangled_names.cc
15595 M src/windows/mingw.h
15596 M src/windows/mini_disassembler.cc
15597 M src/windows/mini_disassembler.h
15598 M src/windows/mini_disassembler_types.h
15599 A src/windows/nm-pdb.c
15600 M src/windows/patch_functions.cc
15601 M src/windows/port.cc
15602 M src/windows/port.h
15603 M src/windows/preamble_patcher.cc
15604 M src/windows/preamble_patcher.h
15605 M src/windows/preamble_patcher_with_stub.cc
15606 D src/windows/vc7and8.def
15607 R059 vsprojects/memalign_unittest/memalign_unittest.vcproj
15608 vsprojects/addr2line-pdb/addr2line-pdb.vcproj
15609 M vsprojects/frag_unittest/frag_unittest.vcproj
15610 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
15611 M
15612 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
15613 M vsprojects/markidle_unittest/markidle_unittest.vcproj
15614 C068 vsprojects/frag_unittest/frag_unittest.vcproj
15615 vsprojects/nm-pdb/nm-pdb.vcproj
15616 M
15617 vsprojects/tcmalloc_minimal_large/tcmalloc_minimal_large_unittest.vcproj
15618 M
15619 vsprojects/tcmalloc_minimal_unittest/tcmalloc_minimal_unittest.vcproj
15620 M vsprojects/thread_dealloc_unittest/thread_dealloc_unittest.vcproj
15621 R074
15622 vsprojects/tcmalloc_minimal_unittest-static/tcmalloc_minimal_unittest-static.vcproj
15623 vsprojects/tmu-static/tmu-static.vcproj
15624
15625 commit 16191f87ff8dc78295c0f617060460664fc444bd
15626 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15627 Date: Tue Sep 23 17:51:05 2008 +0000
15628
15629 Tue Sep 23 08:56:31 2008 Google Inc. <[email protected]>
15630
15631 * google-perftools: version 0.99.2 release
15632 * COMPILE FIX: add #include needed for FreeBSD and OS X
15633 (csilvers)
15634
15635
15636 git-svn-id: http://gperftools.googlecode.com/svn/trunk@58
15637 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15638
15639 M ChangeLog
15640 M configure
15641 M configure.ac
15642 M src/tests/heap-profiler_unittest.cc
15643
15644 commit 8b2dd25dc9d1523ab9a86bd39c4fb798c89db899
15645 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15646 Date: Sat Sep 20 17:15:23 2008 +0000
15647
15648 Sat Sep 20 09:37:18 2008 Google Inc. <[email protected]>
15649
15650 * google-perftools: version 0.99.1 release
15651 * BUG FIX: look for nm, etc in /usr/bin, not /usr/crosstool
15652 (csilvers)
15653
15654
15655 git-svn-id: http://gperftools.googlecode.com/svn/trunk@56
15656 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15657
15658 M ChangeLog
15659 M configure
15660 M configure.ac
15661 M packages/deb/changelog
15662 M src/pprof
15663
15664 commit 106aef86ce5697cf44bdbec90ab6833b9254d273
15665 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15666 Date: Fri Sep 19 20:06:40 2008 +0000
15667
15668 Thu Sep 18 16:00:27 2008 Google Inc. <[email protected]>
15669
15670 * google-perftools: version 0.99 release
15671 * Add IsHeapProfileRunning (csilvers)
15672 * Add C shims for some of the C++ header files (csilvers)
15673 * Fix heap profile file clean-up logic (maxim)
15674 * Rename linuxthreads.c to .cc for better compiler support
15675 (csilvers)
15676 * Add source info to disassembly in pprof (sanjay)
15677 * Use open instead of fopen to avoid memory alloc (csilvers)
15678 * Disable malloc extensions when running under valgrind (kcc)
15679 * BUG FIX: Fix out-of-bound error by reordering a check
15680 (larryz)
15681 * Add Options struct to ProfileData (cgd)
15682 * Correct PC-handling of --base in pprof (csilvers)
15683 * Handle 1 function occurring twice in an image (sanjay)
15684 * Improve stack-data cleaning (maxim)
15685 * Use 'struct Foo' to make header C compatible (csilvers)
15686 * Add 'total' line to pprof --text (csilvers)
15687 * Pre-allocate buffer for heap-profiler to avoid OOM errors
15688 (csilvers)
15689 * Allow a few more env-settings to control tcmalloc (csilvers)
15690 * Document some of the issues involving thread-local storage
15691 (csilvers)
15692 * BUG FIX: Define strtoll and friends for windows (csilvers)
15693
15694
15695 git-svn-id: http://gperftools.googlecode.com/svn/trunk@54
15696 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15697
15698 M ChangeLog
15699 M INSTALL
15700 M Makefile.am
15701 M Makefile.in
15702 M README
15703 M README.windows
15704 M configure
15705 M configure.ac
15706 M doc/heapprofile.html
15707 M doc/tcmalloc.html
15708 M packages/deb/changelog
15709 M src/addressmap-inl.h
15710 M src/base/basictypes.h
15711 M src/base/commandlineflags.h
15712 M src/base/cycleclock.h
15713 M src/base/dynamic_annotations.cc
15714 M src/base/dynamic_annotations.h
15715 R099 src/base/linuxthreads.c src/base/linuxthreads.cc
15716 M src/base/spinlock.h
15717 M src/base/stl_allocator.h
15718 M src/base/sysinfo.cc
15719 M src/google/heap-checker.h
15720 M src/google/heap-profiler.h
15721 M src/google/malloc_extension.h
15722 A src/google/malloc_extension_c.h
15723 M src/google/malloc_hook.h
15724 A src/google/malloc_hook_c.h
15725 M src/google/profiler.h
15726 M src/google/stacktrace.h
15727 M src/heap-checker.cc
15728 M src/heap-profile-table.cc
15729 M src/heap-profile-table.h
15730 M src/heap-profiler.cc
15731 M src/internal_logging.cc
15732 M src/internal_logging.h
15733 M src/malloc_extension.cc
15734 M src/malloc_hook.cc
15735 M src/memory_region_map.cc
15736 M src/pprof
15737 M src/profiledata.cc
15738 M src/profiledata.h
15739 M src/profiler.cc
15740 M src/system-alloc.cc
15741 M src/tcmalloc.cc
15742 M src/tests/heap-checker-death_unittest.sh
15743 M src/tests/heap-checker_unittest.cc
15744 M src/tests/heap-profiler_unittest.cc
15745 M src/tests/markidle_unittest.cc
15746 M src/tests/profiledata_unittest.cc
15747 M src/windows/port.h
15748
15749 commit 100e657c5092bc274424286a728db5116a4bbc54
15750 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15751 Date: Sat Jun 14 02:30:53 2008 +0000
15752
15753 Mon Jun 9 16:47:03 2008 Google Inc. <[email protected]>
15754
15755 * google-perftools: version 0.98 release
15756 * Add ProfilerStartWithOptions() (cgd)
15757 * Change tcmalloc_minimal to not do any stack-tracing at all
15758 (csilvers)
15759 * Prefer mmap to sbrk for 64-buit debug mode (sanjay)
15760 * Fix accounting for some tcmalloc stats (sanjay)
15761 * Use setrlimit() to keep unittests from killing the machine
15762 (odo)
15763 * Fix a bug when sbrk-ing near address 4G (csilvers)
15764 * Make MallocHook thread-safe (jyasskin)
15765 * Fix windows build for MemoryBarrier (jyasskin)
15766 * Fix CPU-profiler docs to mention correct libs (csilvers)
15767 * Fix for GetHeapProfile() when heap-profiling is off (maxim)
15768 * Avoid realloc resizing ping-pongs using hysteresis
15769 (csilvers)
15770 * Add --callgrind output support to pprof (klimek)
15771 * Fix profiler.h and heap-profiler.h to be C-compatible
15772 (csilvers)
15773 * Break malloc_hook.h into two parts to reduce dependencies
15774 (csilvers)
15775 * Better handle systems that don't implement mmap (csilvers)
15776 * PORTING: disable system_alloc_unittest for msvc (csilvers)
15777 * PORTING: Makefile tweaks to build better on cygwin
15778 (csilvers)
15779
15780
15781 git-svn-id: http://gperftools.googlecode.com/svn/trunk@52
15782 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15783
15784 M ChangeLog
15785 M Makefile.am
15786 M Makefile.in
15787 M README.windows
15788 M configure
15789 M configure.ac
15790 M doc/cpuprofile.html
15791 M google-perftools.sln
15792 M m4/acx_pthread.m4
15793 M packages/deb/changelog
15794 M src/addressmap-inl.h
15795 M src/base/atomicops-internals-linuxppc.h
15796 M src/base/atomicops-internals-macosx.h
15797 M src/base/atomicops-internals-x86-msvc.h
15798 M src/base/atomicops-internals-x86.cc
15799 M src/base/atomicops-internals-x86.h
15800 M src/base/atomicops.h
15801 M src/base/commandlineflags.h
15802 M src/base/cycleclock.h
15803 M src/base/dynamic_annotations.cc
15804 M src/base/dynamic_annotations.h
15805 M src/base/linux_syscall_support.h
15806 M src/base/low_level_alloc.cc
15807 M src/base/low_level_alloc.h
15808 M src/base/simple_mutex.h
15809 M src/base/spinlock.h
15810 M src/base/stl_allocator.h
15811 M src/base/sysinfo.h
15812 M src/config.h.in
15813 M src/getpc.h
15814 M src/google/heap-checker.h
15815 M src/google/heap-profiler.h
15816 M src/google/malloc_extension.h
15817 M src/google/malloc_hook.h
15818 M src/google/profiler.h
15819 M src/google/stacktrace.h
15820 M src/heap-checker.cc
15821 M src/heap-profile-table.h
15822 M src/heap-profiler.cc
15823 M src/internal_logging.h
15824 A src/malloc_hook-inl.h
15825 M src/malloc_hook.cc
15826 M src/maybe_threads.cc
15827 M src/maybe_threads.h
15828 M src/memory_region_map.cc
15829 M src/memory_region_map.h
15830 M src/packed-cache-inl.h
15831 M src/pagemap.h
15832 M src/pprof
15833 M src/profiledata.h
15834 M src/profiler.cc
15835 M src/stacktrace_generic-inl.h
15836 M src/system-alloc.cc
15837 M src/system-alloc.h
15838 M src/tcmalloc.cc
15839 M src/tests/maybe_threads_unittest.sh
15840 M src/tests/memalign_unittest.cc
15841 M src/tests/tcmalloc_unittest.cc
15842 M src/tests/testutil.cc
15843 M src/tests/testutil.h
15844 M src/windows/config.h
15845 M src/windows/mingw.h
15846 M src/windows/mini_disassembler.h
15847 M src/windows/mini_disassembler_types.h
15848 M src/windows/patch_functions.cc
15849 M src/windows/port.h
15850 M src/windows/preamble_patcher.h
15851 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
15852 M
15853 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
15854 M vsprojects/memalign_unittest/memalign_unittest.vcproj
15855 D vsprojects/system_alloc_unittest/system_alloc_unittest.vcproj
15856 M
15857 vsprojects/tcmalloc_minimal_unittest-static/tcmalloc_minimal_unittest-static.vcproj
15858
15859 commit 7ec719093b1c9fda979ba0d07eed288e2a7c3c9b
15860 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15861 Date: Tue Apr 22 01:47:16 2008 +0000
15862
15863 Mon Apr 21 15:20:52 2008 Google Inc. <[email protected]>
15864
15865 * google-perftools: version 0.97 release
15866 * Refactor GetHeapProfile to avoid using malloc (maxim)
15867 * Fix heap-checker and heap-profiler hook interactions (maxim)
15868 * Fix a data race in MemoryRegionMap::Lock (jyasskin)
15869 * Improve thread-safety of leak checker (maxim)
15870 * Fix mmap profile to no longer deadlock (maxim)
15871 * Fix rpm to have devel package depend on non-devel (csilvers)
15872 * PORTING: Fix clock-speed detection for Mac OS X (csilvers)
15873
15874
15875 git-svn-id: http://gperftools.googlecode.com/svn/trunk@50
15876 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15877
15878 M ChangeLog
15879 M configure
15880 M configure.ac
15881 M doc/heapprofile.html
15882 M packages/deb/changelog
15883 M packages/rpm/rpm.spec
15884 M src/addressmap-inl.h
15885 M src/base/cycleclock.h
15886 M src/base/elfcore.h
15887 M src/base/low_level_alloc.cc
15888 M src/base/spinlock.h
15889 M src/base/sysinfo.cc
15890 M src/base/sysinfo.h
15891 M src/google/heap-checker.h
15892 M src/google/heap-profiler.h
15893 M src/google/profiler.h
15894 M src/heap-checker.cc
15895 M src/heap-profile-table.cc
15896 M src/heap-profile-table.h
15897 M src/heap-profiler.cc
15898 M src/internal_logging.cc
15899 M src/internal_logging.h
15900 M src/memory_region_map.cc
15901 M src/memory_region_map.h
15902 M src/pprof
15903 M src/stacktrace_libunwind-inl.h
15904 M src/stacktrace_x86-inl.h
15905 M src/tcmalloc.cc
15906 M src/tests/heap-checker-death_unittest.sh
15907 M src/tests/heap-checker_unittest.cc
15908 M src/tests/heap-profiler_unittest.cc
15909 M src/tests/low_level_alloc_unittest.cc
15910 M src/tests/maybe_threads_unittest.sh
15911 M src/windows/port.cc
15912
15913 commit 97fdd4a4f97dd15e8803ed51ac153903c2cdffc2
15914 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15915 Date: Wed Mar 19 23:35:27 2008 +0000
15916
15917 Tue Mar 18 14:30:44 2008 Google Inc. <[email protected]>
15918
15919 * google-perftools: version 0.96 release
15920 * major atomicops rewrite; fixed atomic ops code for linux/ppc
15921 (vchen)
15922 * nix the stacktrace library; now build structure is simpler
15923 (csilvers)
15924 * Speed up heap-checker, and reduce extraneous logging (maxim)
15925 * Improve itimer code for NPTL case (cgd)
15926 * Add source code annotations for use by valgrind, etc (kcc)
15927 * PORTING: Fix high resolution timers for Mac OS X (adlr)
15928
15929
15930 git-svn-id: http://gperftools.googlecode.com/svn/trunk@48
15931 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15932
15933 M ChangeLog
15934 M INSTALL
15935 M Makefile.am
15936 M Makefile.in
15937 M TODO
15938 M configure
15939 M configure.ac
15940 M doc/cpuprofile.html
15941 M doc/heapprofile.html
15942 M packages/deb/changelog
15943 M packages/rpm/rpm.spec
15944 M src/addressmap-inl.h
15945 M src/base/atomicops-internals-linuxppc.h
15946 M src/base/atomicops-internals-macosx.h
15947 M src/base/atomicops-internals-x86-msvc.h
15948 M src/base/atomicops-internals-x86.cc
15949 M src/base/atomicops-internals-x86.h
15950 M src/base/atomicops.h
15951 M src/base/basictypes.h
15952 M src/base/cycleclock.h
15953 A src/base/dynamic_annotations.cc
15954 A src/base/dynamic_annotations.h
15955 M src/base/linux_syscall_support.h
15956 M src/base/low_level_alloc.cc
15957 M src/base/spinlock.cc
15958 M src/base/spinlock.h
15959 M src/base/sysinfo.cc
15960 M src/heap-checker.cc
15961 M src/heap-profile-table.cc
15962 M src/memfs_malloc.cc
15963 M src/pprof
15964 M src/profiler.cc
15965 M src/tests/atomicops_unittest.cc
15966 M src/tests/heap-checker-death_unittest.sh
15967 M src/tests/maybe_threads_unittest.sh
15968 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
15969 M
15970 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
15971 M
15972 vsprojects/tcmalloc_minimal_unittest-static/tcmalloc_minimal_unittest-static.vcproj
15973
15974 commit a644b4f2d61b9610ec6eeb1f09ebce7054aa0762
15975 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15976 Date: Tue Feb 19 22:19:22 2008 +0000
15977
15978 Tue Feb 19 12:01:31 2008 Google Inc. <[email protected]>
15979
15980 * google-perftools: version 0.95.1 release (bugfix release)
15981 * x86_64 compile-fix: nix pread64 and pwrite64 (csilvers)
15982 * more heap-checker debug logging (maxim)
15983 * minor improvement to x86_64 CycleClock (gpike)
15984
15985
15986 git-svn-id: http://gperftools.googlecode.com/svn/trunk@46
15987 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
15988
15989 M ChangeLog
15990 M configure
15991 M configure.ac
15992 M src/base/cycleclock.h
15993 M src/base/linux_syscall_support.h
15994 M src/heap-checker.cc
15995
15996 commit 8a0a3101bc6a7d56ac04b278f28bdf3f95b00a3c
15997 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
15998 Date: Wed Feb 13 00:55:09 2008 +0000
15999
16000 Tue Feb 12 12:28:32 2008 Google Inc. <[email protected]>
16001
16002 * google-perftools: version 0.95 release
16003 * Better -- not perfect -- support for linux-ppc (csilvers)
16004 * Fix race condition in libunwind stacktrace (aruns)
16005 * Speed up x86 spinlock locking (m3b)
16006 * Improve heap-checker performance (maxim)
16007 * Heap checker traverses more ptrs inside heap-alloced objects
16008 (maxim)
16009 * Remove deprecated ProfilerThreadState function (cgd)
16010 * Update libunwind documentation for statically linked
16011 binaries (aruns)
16012
16013
16014 git-svn-id: http://gperftools.googlecode.com/svn/trunk@44
16015 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16016
16017 M ChangeLog
16018 M INSTALL
16019 M Makefile.am
16020 M Makefile.in
16021 M configure
16022 M configure.ac
16023 M doc/heap_checker.html
16024 M doc/tcmalloc.html
16025 M packages/deb/changelog
16026 M packages/deb/docs
16027 M packages/rpm/rpm.spec
16028 M src/addressmap-inl.h
16029 A src/base/atomicops-internals-linuxppc.h
16030 M src/base/atomicops-internals-x86-msvc.h
16031 M src/base/atomicops-internals-x86.h
16032 M src/base/atomicops.h
16033 M src/base/commandlineflags.h
16034 A src/base/cycleclock.h
16035 M src/base/elfcore.h
16036 M src/base/linux_syscall_support.h
16037 M src/base/linuxthreads.c
16038 M src/base/linuxthreads.h
16039 M src/base/logging.h
16040 R090 src/base/mutex.h src/base/simple_mutex.h
16041 M src/base/spinlock.cc
16042 M src/base/spinlock.h
16043 M src/base/sysinfo.cc
16044 M src/base/sysinfo.h
16045 M src/google/heap-checker.h
16046 M src/google/profiler.h
16047 M src/heap-checker.cc
16048 M src/heap-profile-table.cc
16049 M src/heap-profile-table.h
16050 M src/heap-profiler.cc
16051 M src/malloc_extension.cc
16052 M src/malloc_hook.cc
16053 M src/maybe_threads.cc
16054 M src/packed-cache-inl.h
16055 M src/profiledata.h
16056 M src/profiler.cc
16057 M src/stacktrace.cc
16058 M src/stacktrace_libunwind-inl.h
16059 M src/stacktrace_powerpc-inl.h
16060 M src/system-alloc.cc
16061 M src/tcmalloc.cc
16062 M src/tests/addressmap_unittest.cc
16063 M src/tests/heap-checker_unittest.cc
16064 M src/tests/profiler_unittest.cc
16065 M src/tests/stacktrace_unittest.cc
16066 M src/tests/tcmalloc_unittest.cc
16067 M src/windows/preamble_patcher.h
16068
16069 commit b43ba444fcd74fa7c3260f6b2494dcbaa3fdb296
16070 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16071 Date: Wed Dec 5 00:08:28 2007 +0000
16072
16073 Mon Dec 3 23:51:54 2007 Google Inc. <[email protected]>
16074
16075 * google-perftools: version 0.94.1 release (bugfix release)
16076 * Fix missing #includes for x86_64 compile using libunwind
16077 (csilvers)
16078
16079
16080 git-svn-id: http://gperftools.googlecode.com/svn/trunk@42
16081 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16082
16083 M ChangeLog
16084 M configure
16085 M configure.ac
16086 M src/stacktrace_libunwind-inl.h
16087
16088 commit 11b02f7aebd05cf39f6f93bdd48786909f99f34e
16089 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16090 Date: Thu Nov 29 23:39:24 2007 +0000
16091
16092 Thu Nov 29 07:59:43 2007 Google Inc. <[email protected]>
16093
16094 * google-perftools: version 0.94 release
16095 * PORTING: MinGW/Msys support -- runs same code as MSVC does
16096 (csilvers)
16097 * PORTING: Add NumCPUs support for Mac OS X (csilvers)
16098 * Work around a sscanf bug in glibc(?) (waldemar)
16099 * Fix Windows MSVC bug triggered by thread deletion (csilvers)
16100 * Fix bug that triggers in MSVC /O2: missing volatile (gpike)
16101 * March-of-time support: quiet warnings/errors for gcc 4.2,
16102 OS X 10.5
16103 * Modify pprof so it works without nm: useful for windows
16104 (csilvers)
16105 * pprof: Support filtering for CPU profiles (cgd)
16106 * Bugfix: have realloc report to hooks in all situations
16107 (maxim)
16108 * Speed improvement: replace slow memcpy with std::copy
16109 (soren)
16110 * Speed: better iterator efficiency in RecordRegionRemoval
16111 (soren)
16112 * Speed: minor speed improvements via better bitfield
16113 alignment (gpike)
16114 * Documentation: add documentation of binary profile output
16115 (cgd)
16116
16117
16118 git-svn-id: http://gperftools.googlecode.com/svn/trunk@40
16119 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16120
16121 M ChangeLog
16122 M INSTALL
16123 M Makefile.am
16124 M Makefile.in
16125 M configure
16126 M configure.ac
16127 A doc/cpuprofile-fileformat.html
16128 M doc/cpuprofile.html
16129 M packages/deb/changelog
16130 M src/base/linux_syscall_support.h
16131 M src/base/spinlock.cc
16132 M src/base/sysinfo.cc
16133 M src/config.h.in
16134 M src/getpc.h
16135 M src/google/heap-checker.h
16136 M src/google/profiler.h
16137 M src/google/stacktrace.h
16138 M src/heap-checker.cc
16139 M src/heap-profile-table.cc
16140 M src/heap-profiler.cc
16141 M src/malloc_hook.cc
16142 M src/memfs_malloc.cc
16143 M src/memory_region_map.cc
16144 M src/packed-cache-inl.h
16145 M src/pprof
16146 A src/profiledata.cc
16147 A src/profiledata.h
16148 M src/profiler.cc
16149 M src/stacktrace_generic-inl.h
16150 M src/stacktrace_libunwind-inl.h
16151 M src/stacktrace_powerpc-inl.h
16152 M src/stacktrace_x86-inl.h
16153 M src/stacktrace_x86_64-inl.h
16154 M src/system-alloc.cc
16155 M src/system-alloc.h
16156 M src/tcmalloc.cc
16157 M src/tests/addressmap_unittest.cc
16158 M src/tests/heap-checker_unittest.cc
16159 M src/tests/low_level_alloc_unittest.cc
16160 M src/tests/maybe_threads_unittest.sh
16161 A src/tests/profiledata_unittest.cc
16162 M src/tests/system-alloc_unittest.cc
16163 M src/tests/tcmalloc_unittest.cc
16164 M src/tests/testutil.cc
16165 M src/windows/config.h
16166 A src/windows/mingw.h
16167 M src/windows/mini_disassembler.h
16168 M src/windows/patch_functions.cc
16169 M src/windows/port.cc
16170 M src/windows/port.h
16171 M src/windows/preamble_patcher.h
16172 M src/windows/preamble_patcher_with_stub.cc
16173 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
16174
16175 commit 49b74b9508797f8aafe6b86e62e7efc4ec200e48
16176 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16177 Date: Fri Aug 17 20:56:15 2007 +0000
16178
16179 * google-perftools: version 0.93 release
16180 * PORTING: everything compiles on Solaris, OS X, FreeBSD
16181 (see INSTALL)
16182 * PORTING: cpu-profiler works on most platforms (much
16183 better GetPC())
16184 * PORTING: heap-profiler works on most platforms
16185 * PORTING: improved windows support, including release builds
16186 * No longer build or run ptmalloc tests by default
16187 * Add support for using memfs filesystem to allocate memory
16188 in linux
16189 * WINDOWS: give debug library and release library different
16190 names
16191
16192 Tue Jul 17 22:26:27 2007 Google Inc. <[email protected]>
16193
16194
16195 git-svn-id: http://gperftools.googlecode.com/svn/trunk@38
16196 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16197
16198 M ChangeLog
16199 M INSTALL
16200 M Makefile.am
16201 M Makefile.in
16202 M README
16203 M README.windows
16204 M autogen.sh
16205 M configure
16206 M configure.ac
16207 M google-perftools.sln
16208 M m4/program_invocation_name.m4
16209 M packages/deb/changelog
16210 M src/addressmap-inl.h
16211 M src/base/basictypes.h
16212 M src/base/logging.h
16213 M src/base/sysinfo.cc
16214 M src/base/sysinfo.h
16215 M src/base/thread_lister.c
16216 M src/config.h.in
16217 M src/config_for_unittests.h
16218 A src/getpc.h
16219 M src/google/heap-checker.h
16220 M src/google/malloc_hook.h
16221 M src/heap-checker.cc
16222 M src/heap-profile-table.cc
16223 M src/heap-profile-table.h
16224 M src/heap-profiler.cc
16225 M src/malloc_extension.cc
16226 M src/malloc_hook.cc
16227 A src/memfs_malloc.cc
16228 M src/memory_region_map.cc
16229 M src/memory_region_map.h
16230 M src/pprof
16231 M src/profiler.cc
16232 M src/stacktrace_powerpc-inl.h
16233 M src/system-alloc.cc
16234 M src/system-alloc.h
16235 M src/tcmalloc.cc
16236 M src/tests/addressmap_unittest.cc
16237 A src/tests/getpc_test.cc
16238 M src/tests/heap-checker-death_unittest.sh
16239 M src/tests/heap-checker_unittest.cc
16240 M src/tests/heap-profiler_unittest.sh
16241 M src/tests/low_level_alloc_unittest.cc
16242 M src/tests/maybe_threads_unittest.sh
16243 M src/tests/memalign_unittest.cc
16244 M src/tests/system-alloc_unittest.cc
16245 M src/tests/tcmalloc_unittest.cc
16246 M src/windows/TODO
16247 M src/windows/config.h
16248 A src/windows/ia32_modrm_map.cc
16249 A src/windows/ia32_opcode_map.cc
16250 A src/windows/mini_disassembler.cc
16251 A src/windows/mini_disassembler.h
16252 A src/windows/mini_disassembler_types.h
16253 A src/windows/patch_functions.cc
16254 M src/windows/port.cc
16255 M src/windows/port.h
16256 A src/windows/preamble_patcher.cc
16257 A src/windows/preamble_patcher.h
16258 A src/windows/preamble_patcher_with_stub.cc
16259 M src/windows/vc7and8.def
16260 M vsprojects/addressmap_unittest/addressmap_unittest.vcproj
16261 M vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
16262 M
16263 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
16264 C054 vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
16265 vsprojects/tcmalloc_minimal_unittest-static/tcmalloc_minimal_unittest-static.vcproj
16266
16267 commit c437e1fcdd1e6ff3f032928d460cbfc115e2324f
16268 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16269 Date: Wed Jul 18 18:30:50 2007 +0000
16270
16271 Tue Jul 17 22:26:27 2007 Google Inc. <[email protected]>
16272
16273 * google-perftools: version 0.92 release
16274 * PERFORMANCE: use a packed cache to speed up tcmalloc
16275 * PORTING: preliminary windows support! (see README.windows)
16276 * PORTING: better support for solaris, OS X, FreeBSD
16277 (see INSTALL)
16278 * Envvar support for running the heap-checker under gdb
16279 * Add weak declarations to maybe_threads to fix no-pthreads
16280 compile bugs
16281 * Some 64bit fixes, especially with pprof
16282 * Better heap-checker support for some low-level allocations
16283 * Fix bug where heap-profiles would sometimes get truncated
16284 * New documentation about how to handle common heap leak
16285 situations
16286 * Use computed includes for hash_map/set: easier config
16287 * Added all used .m4 templates to the distribution
16288
16289
16290 git-svn-id: http://gperftools.googlecode.com/svn/trunk@36
16291 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16292
16293 M ChangeLog
16294 M INSTALL
16295 M Makefile.am
16296 M Makefile.in
16297 M README
16298 A README.windows
16299 M aclocal.m4
16300 A autogen.sh
16301 M compile
16302 M config.guess
16303 M config.sub
16304 M configure
16305 M configure.ac
16306 M depcomp
16307 M doc/cpuprofile.html
16308 M doc/heap_checker.html
16309 M doc/heapprofile.html
16310 A google-perftools.sln
16311 M install-sh
16312 A m4/ac_have_attribute.m4
16313 A m4/acx_pthread.m4
16314 A m4/compiler_characteristics.m4
16315 A m4/install_prefix.m4
16316 A m4/namespaces.m4
16317 A m4/program_invocation_name.m4
16318 A m4/stl_hash.m4
16319 A m4/stl_namespace.m4
16320 M missing
16321 M mkinstalldirs
16322 M packages/deb/changelog
16323 M packages/rpm/rpm.spec
16324 M src/base/atomicops-internals-x86-msvc.h
16325 M src/base/atomicops-internals-x86.h
16326 M src/base/atomicops.h
16327 M src/base/basictypes.h
16328 M src/base/commandlineflags.h
16329 M src/base/logging.h
16330 M src/base/low_level_alloc.cc
16331 M src/base/low_level_alloc.h
16332 D src/base/mutex.cc
16333 M src/base/mutex.h
16334 M src/base/spinlock.cc
16335 M src/base/spinlock.h
16336 M src/base/stl_allocator.h
16337 M src/base/sysinfo.cc
16338 M src/base/sysinfo.h
16339 M src/config.h.in
16340 C068 src/tests/testutil.h src/config_for_unittests.h
16341 M src/google/heap-checker.h
16342 M src/google/heap-profiler.h
16343 M src/google/malloc_extension.h
16344 M src/google/malloc_hook.h
16345 M src/google/profiler.h
16346 M src/google/stacktrace.h
16347 M src/heap-checker.cc
16348 M src/heap-profile-table.cc
16349 M src/heap-profiler.cc
16350 M src/internal_logging.cc
16351 M src/internal_logging.h
16352 M src/malloc_extension.cc
16353 M src/malloc_hook.cc
16354 M src/maybe_threads.cc
16355 M src/maybe_threads.h
16356 M src/memory_region_map.cc
16357 M src/memory_region_map.h
16358 A src/packed-cache-inl.h
16359 M src/pprof
16360 M src/profiler.cc
16361 M src/solaris/libstdc++.la
16362 M src/stacktrace.cc
16363 A src/stacktrace_powerpc-inl.h
16364 M src/stacktrace_x86-inl.h
16365 M src/system-alloc.cc
16366 M src/system-alloc.h
16367 M src/tcmalloc.cc
16368 C053 src/tests/testutil.cc src/tcmalloc.h
16369 M src/tests/frag_unittest.cc
16370 M src/tests/heap-checker_unittest.cc
16371 M src/tests/heap-profiler_unittest.cc
16372 M src/tests/heap-profiler_unittest.sh
16373 M src/tests/low_level_alloc_unittest.cc
16374 M src/tests/markidle_unittest.cc
16375 A src/tests/maybe_threads_unittest.sh
16376 M src/tests/memalign_unittest.cc
16377 C065 src/tests/testutil.h src/tests/packed-cache_test.cc
16378 M src/tests/profiler_unittest.cc
16379 M src/tests/stacktrace_unittest.cc
16380 C051 src/tests/thread_dealloc_unittest.cc
16381 src/tests/system-alloc_unittest.cc
16382 M src/tests/tcmalloc_unittest.cc
16383 M src/tests/testutil.cc
16384 M src/tests/testutil.h
16385 M src/tests/thread_dealloc_unittest.cc
16386 A src/windows/TODO
16387 C060 src/config.h.in src/windows/config.h
16388 A src/windows/port.cc
16389 A src/windows/port.h
16390 A src/windows/vc7and8.def
16391 A vsprojects/addressmap_unittest/addressmap_unittest.vcproj
16392 A vsprojects/frag_unittest/frag_unittest.vcproj
16393 A vsprojects/libtcmalloc_minimal/libtcmalloc_minimal.vcproj
16394 A
16395 vsprojects/low_level_alloc_unittest/low_level_alloc_unittest.vcproj
16396 A vsprojects/markidle_unittest/markidle_unittest.vcproj
16397 A vsprojects/memalign_unittest/memalign_unittest.vcproj
16398 A vsprojects/packed-cache_test/packed-cache_test.vcproj
16399 A vsprojects/system_alloc_unittest/system_alloc_unittest.vcproj
16400 A
16401 vsprojects/tcmalloc_minimal_large/tcmalloc_minimal_large_unittest.vcproj
16402 A
16403 vsprojects/tcmalloc_minimal_unittest/tcmalloc_minimal_unittest.vcproj
16404 A vsprojects/thread_dealloc_unittest/thread_dealloc_unittest.vcproj
16405
16406 commit 6878379d5bab87c787cdd3487b5620a9c8adf376
16407 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16408 Date: Thu Apr 19 00:53:22 2007 +0000
16409
16410 Wed Apr 18 16:43:55 2007 Google Inc. <[email protected]>
16411
16412 * google-perftools: version 0.91 release
16413 * Brown-paper-bag bugfix: compilation error on some x86-64
16414 machines
16415
16416
16417 git-svn-id: http://gperftools.googlecode.com/svn/trunk@30
16418 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16419
16420 M ChangeLog
16421 M README
16422 M config.guess
16423 M config.sub
16424 M configure
16425 M configure.ac
16426 M src/base/sysinfo.cc
16427 M src/base/sysinfo.h
16428 M src/stacktrace_libunwind-inl.h
16429 M src/tests/ptmalloc/malloc-machine.h
16430
16431 commit 74ad5d57ec08abace386befc6c3c695d85f44d1a
16432 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16433 Date: Mon Apr 16 20:49:32 2007 +0000
16434
16435 Fri Apr 13 14:50:51 2007 Google Inc. <[email protected]>
16436
16437 * google-perftools: version 0.90 release
16438 * (As the version-number jump hints, this is a major new
16439 release:
16440 almost every piece of functionality was rewritten.
16441 I can't do
16442 justice to all the changes, but will concentrate on
16443 highlights.)
16444 *** USER-VISIBLE CHANGES:
16445 * Ability to "release" unused memory added to tcmalloc
16446 * Exposed more tweaking knobs via environment variables
16447 (see docs)
16448 * pprof tries harder to map addresses to functions
16449 * tcmalloc_minimal compiles and runs on FreeBSD 6.0 and
16450 Solaris 10
16451 *** INTERNAL CHANGES:
16452 * Much better 64-bit support
16453 * Better multiple-processor support (e.g. multicore contention
16454 tweaks)
16455 * Support for recent kernel ABI changes (e.g. new arg
16456 to mremap)
16457 * Addition of spinlocks to tcmalloc to reduce contention cost
16458 * Speed up tcmalloc by using __thread on systems that
16459 support TLS
16460 * Total redesign of heap-checker to improve liveness checking
16461 * More portable stack-frame analysis -- no more hard-coded
16462 constants!
16463 * Disentangled heap-profiler code and heap-checker code
16464 * Several new unittests to test, e.g., thread-contention costs
16465 * Lots of small (but important!) bug fixes: e.g., fixing
16466 GetPC on amd64
16467 *** KNOWN PROBLEMS:
16468 * CPU-profiling may crash on x86_64 (64-bit) systems.
16469 See the README
16470 * Profiling/heap-checking may deadlock on x86_64 systems.
16471 See README
16472
16473
16474 git-svn-id: http://gperftools.googlecode.com/svn/trunk@28
16475 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16476
16477 M ChangeLog
16478 M INSTALL
16479 M Makefile.am
16480 M Makefile.in
16481 M README
16482 M TODO
16483 M aclocal.m4
16484 M config.guess
16485 M config.sub
16486 M configure
16487 M configure.ac
16488 R050 doc/cpu_profiler.html doc/cpuprofile.html
16489 A doc/designstyle.css
16490 M doc/heap_checker.html
16491 D doc/heap_profiler.html
16492 A doc/heapprofile.html
16493 M doc/index.html
16494 M doc/tcmalloc.html
16495 M ltmain.sh
16496 M packages/deb/changelog
16497 A packages/deb/docs
16498 M packages/rpm/rpm.spec
16499 M src/addressmap-inl.h
16500 A src/base/atomicops-internals-macosx.h
16501 A src/base/atomicops-internals-x86-msvc.h
16502 A src/base/atomicops-internals-x86.cc
16503 A src/base/atomicops-internals-x86.h
16504 A src/base/atomicops.h
16505 M src/base/basictypes.h
16506 M src/base/commandlineflags.h
16507 M src/base/elfcore.h
16508 M src/base/googleinit.h
16509 M src/base/linux_syscall_support.h
16510 M src/base/linuxthreads.c
16511 M src/base/linuxthreads.h
16512 C074 src/base/googleinit.h src/base/logging.cc
16513 M src/base/logging.h
16514 A src/base/low_level_alloc.cc
16515 A src/base/low_level_alloc.h
16516 A src/base/mutex.cc
16517 A src/base/mutex.h
16518 A src/base/spinlock.cc
16519 A src/base/spinlock.h
16520 A src/base/stl_allocator.h
16521 A src/base/sysinfo.cc
16522 A src/base/sysinfo.h
16523 M src/base/thread_lister.c
16524 M src/base/thread_lister.h
16525 M src/config.h.in
16526 M src/google/heap-checker.h
16527 M src/google/heap-profiler.h
16528 M src/google/malloc_extension.h
16529 M src/google/malloc_hook.h
16530 M src/google/profiler.h
16531 M src/google/stacktrace.h
16532 M src/heap-checker-bcad.cc
16533 M src/heap-checker.cc
16534 A src/heap-profile-table.cc
16535 A src/heap-profile-table.h
16536 D src/heap-profiler-inl.h
16537 M src/heap-profiler.cc
16538 M src/internal_logging.cc
16539 M src/internal_logging.h
16540 D src/internal_spinlock.h
16541 M src/malloc_extension.cc
16542 M src/malloc_hook.cc
16543 M src/maybe_threads.cc
16544 A src/memory_region_map.cc
16545 A src/memory_region_map.h
16546 M src/pprof
16547 M src/profiler.cc
16548 A src/solaris/libstdc++.la
16549 M src/stacktrace.cc
16550 M src/stacktrace_generic-inl.h
16551 M src/stacktrace_libunwind-inl.h
16552 M src/stacktrace_x86-inl.h
16553 M src/stacktrace_x86_64-inl.h
16554 M src/system-alloc.cc
16555 M src/system-alloc.h
16556 M src/tcmalloc.cc
16557 M src/tests/addressmap_unittest.cc
16558 A src/tests/atomicops_unittest.cc
16559 A src/tests/frag_unittest.cc
16560 M src/tests/heap-checker-death_unittest.sh
16561 M src/tests/heap-checker_unittest.cc
16562 M src/tests/heap-checker_unittest.sh
16563 M src/tests/heap-profiler_unittest.cc
16564 M src/tests/heap-profiler_unittest.sh
16565 A src/tests/low_level_alloc_unittest.cc
16566 A src/tests/markidle_unittest.cc
16567 A src/tests/memalign_unittest.cc
16568 M src/tests/profiler_unittest.sh
16569 M src/tests/ptmalloc/malloc-machine.h
16570 M src/tests/stacktrace_unittest.cc
16571 M src/tests/tcmalloc_large_unittest.cc
16572 M src/tests/tcmalloc_unittest.cc
16573 C055 src/stacktrace.cc src/tests/testutil.cc
16574 C074 src/base/googleinit.h src/tests/testutil.h
16575 C055 src/stacktrace.cc src/tests/thread_dealloc_unittest.cc
16576
16577 commit ddbf2f027fb4ca8781fd50820ceb870570f414bc
16578 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16579 Date: Thu Mar 22 05:03:28 2007 +0000
16580
16581 Now that we've uploaded the full source, including the doc/ directory,
16582 we can get rid of docs/
16583
16584 git-svn-id: http://gperftools.googlecode.com/svn/trunk@27
16585 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16586
16587 D docs/html/cpu_profiler.html
16588 D docs/html/heap_checker.html
16589 D docs/html/heap_profiler.html
16590 D docs/html/tcmalloc.html
16591 D docs/images/heap-example1.png
16592 D docs/images/overview.gif
16593 D docs/images/pageheap.gif
16594 D docs/images/pprof-test.gif
16595 D docs/images/pprof-vsnprintf.gif
16596 D docs/images/spanmap.gif
16597 D docs/images/tcmalloc-opspercpusec.png
16598 D docs/images/tcmalloc-opspercpusec_002.png
16599 D docs/images/tcmalloc-opspercpusec_003.png
16600 D docs/images/tcmalloc-opspercpusec_004.png
16601 D docs/images/tcmalloc-opspercpusec_005.png
16602 D docs/images/tcmalloc-opspercpusec_006.png
16603 D docs/images/tcmalloc-opspercpusec_007.png
16604 D docs/images/tcmalloc-opspercpusec_008.png
16605 D docs/images/tcmalloc-opspercpusec_009.png
16606 D docs/images/tcmalloc-opspersec.png
16607 D docs/images/tcmalloc-opspersec_002.png
16608 D docs/images/tcmalloc-opspersec_003.png
16609 D docs/images/tcmalloc-opspersec_004.png
16610 D docs/images/tcmalloc-opspersec_005.png
16611 D docs/images/tcmalloc-opspersec_006.png
16612 D docs/images/tcmalloc-opspersec_007.png
16613 D docs/images/tcmalloc-opspersec_008.png
16614 D docs/images/tcmalloc-opspersec_009.png
16615 D docs/images/threadheap.gif
16616
16617 commit 7ede7d6a9fe772b0bfa05acb2a2a6867405b474a
16618 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16619 Date: Thu Mar 22 05:01:39 2007 +0000
16620
16621 set mime-type for png and gifs
16622
16623 git-svn-id: http://gperftools.googlecode.com/svn/trunk@26
16624 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16625
16626 commit 8e188310f7d8732d81b7b04f193f89964b7af6c5
16627 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16628 Date: Thu Mar 22 04:55:49 2007 +0000
16629
16630 Wed Jun 14 15:11:14 2006 Google Inc. <[email protected]>
16631
16632 * google-perftools: version 0.8 release
16633 * Experimental support for remote profiling added to pprof
16634 (many)
16635 * Fixed race condition in ProfileData::FlushTable (etune)
16636 * Better support for weird /proc maps (maxim, mec)
16637 * Fix heap-checker interaction with gdb (markus)
16638 * Better 64-bit support in pprof (aruns)
16639 * Reduce scavenging cost in tcmalloc by capping NumMoveSize
16640 (sanjay)
16641 * Cast syscall(SYS_mmap); works on more 64-bit systems now
16642 (menage)
16643 * Document the text output of pprof! (csilvers)
16644 * Better compiler support for no-THREADS and for old compilers
16645 (csilvers)
16646 * Make libunwind the default stack unwinder for x86-64 (aruns)
16647 * Somehow the COPYING file got erased. Regenerate it
16648 (csilvers)
16649
16650
16651 git-svn-id: http://gperftools.googlecode.com/svn/trunk@23
16652 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16653
16654 M COPYING
16655 M ChangeLog
16656 M Makefile.am
16657 M aclocal.m4
16658 M configure
16659 M configure.ac
16660 M doc/cpu_profiler.html
16661 A doc/pprof_remote_servers.html
16662 M src/base/linux_syscall_support.h
16663 M src/base/linuxthreads.c
16664 M src/base/thread_lister.c
16665 M src/google/heap-checker.h
16666 M src/heap-checker.cc
16667 M src/malloc_extension.cc
16668 M src/malloc_hook.cc
16669 M src/pprof
16670 M src/profiler.cc
16671 M src/stacktrace.cc
16672 M src/stacktrace_libunwind-inl.h
16673 M src/tcmalloc.cc
16674 M src/tests/heap-checker_unittest.cc
16675 M src/tests/tcmalloc_unittest.cc
16676
16677 commit c3b96b3ac552160abde541bba8ac7b4f8338efa0
16678 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16679 Date: Thu Mar 22 04:48:00 2007 +0000
16680
16681 Thu Apr 13 20:59:09 2006 Google Inc. <[email protected]>
16682
16683 * google-perftools: version 0.7 release
16684 * Major rewrite of thread introspection for new kernels
16685 (markus)
16686 * Major rewrite of heap-checker to use new thread tools
16687 (maxim)
16688 * Add proper support for following data in thread registers
16689 (maxim)
16690 * Syscall support for older kernels, including _syscall6
16691 (markus)
16692 * Support PIC mode (markus, mbland, iant)
16693 * Better support for running in non-threaded contexts
16694 (csilvers)
16695
16696
16697 git-svn-id: http://gperftools.googlecode.com/svn/trunk@21
16698 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16699
16700 M ChangeLog
16701 M configure
16702 M configure.ac
16703 M src/base/elfcore.h
16704 M src/base/linux_syscall_support.h
16705 M src/base/linuxthreads.c
16706 M src/base/thread_lister.h
16707 M src/google/heap-checker.h
16708 M src/heap-checker.cc
16709 M src/heap-profiler.cc
16710 M src/malloc_hook.cc
16711 M src/pprof
16712 M src/profiler.cc
16713 M src/tests/heap-checker-death_unittest.sh
16714 M src/tests/heap-checker_unittest.cc
16715 M src/tests/tcmalloc_unittest.cc
16716
16717 commit 60a3a2ce77ed2713b2eedd20952d9cfc56ff7ccf
16718 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16719 Date: Thu Mar 22 04:46:29 2007 +0000
16720
16721 Fri Jan 27 14:04:27 2006 Google Inc. <[email protected]>
16722
16723 * google-perftools: version 0.6 release
16724 * More sophisticated stacktrace usage, possibly using
16725 libunwind (aruns)
16726 * Update pprof to handle 64-bit profiles (dehnert)
16727 * Fix GetStackTrace to correctly return top stackframe
16728 (sanjay)
16729 * Add ANSI compliance for new and new[], including new_handler
16730 (jkearney)
16731 * More accuracy by reading ELF files directly rather than
16732 objdump (mec)
16733 * Add readline support for pprof (addi)
16734 * Add #includes for PPC (csilvers)
16735 * New PC-detection routine for ibook powerpc (asbestoshead)
16736 * Vastly improved tcmalloc unittest (csilvers)
16737 * Move documentation from /usr/doc to /usr/share/doc
16738
16739
16740 git-svn-id: http://gperftools.googlecode.com/svn/trunk@19
16741 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16742
16743 M ChangeLog
16744 M Makefile.am
16745 M configure
16746 M configure.ac
16747 M packages/rpm/rpm.spec
16748 M src/base/thread_lister.c
16749 M src/config.h.in
16750 M src/heap-checker.cc
16751 M src/heap-profiler.cc
16752 M src/malloc_extension.cc
16753 M src/pprof
16754 M src/profiler.cc
16755 M src/stacktrace.cc
16756 C065 src/tests/tcmalloc_unittest.cc src/stacktrace_generic-inl.h
16757 C064 src/tests/tcmalloc_unittest.cc src/stacktrace_libunwind-inl.h
16758 C068 src/stacktrace.cc src/stacktrace_x86-inl.h
16759 A src/stacktrace_x86_64-inl.h
16760 M src/tcmalloc.cc
16761 M src/tests/stacktrace_unittest.cc
16762 M src/tests/tcmalloc_unittest.cc
16763
16764 commit 298274f8d4f474d2b16a35c8babc58817088c59e
16765 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16766 Date: Thu Mar 22 04:45:23 2007 +0000
16767
16768 Mon Nov 14 17:28:59 2005 Google Inc. <[email protected]>
16769
16770 * google-perftools: version 0.5 release
16771 * Add va_start/va_end calls around vsnprintf() (csilvers)
16772 * Write our own __syscall_return(), since it's not defined
16773 consistently on all 64-bit linux distros (markus)
16774
16775
16776 git-svn-id: http://gperftools.googlecode.com/svn/trunk@17
16777 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16778
16779 M ChangeLog
16780 M configure
16781 M configure.ac
16782 M src/base/linux_syscall_support.h
16783 M src/base/logging.h
16784 M src/config.h.in
16785
16786 commit ee5805f1296f8546c16f90d5427efa347a5f7338
16787 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16788 Date: Thu Mar 22 04:44:18 2007 +0000
16789
16790 Wed Oct 26 15:19:16 2005 Google Inc. <[email protected]>
16791
16792 * Decrease fragmentation in tcmalloc (lefevere)
16793 * Support for ARM in some of the thread-specific code (markus)
16794 * Turn off heap-checker for statically-linked binaries, which
16795 cause error leak reports now (etune)
16796 * Many pprof improvements, including a command-line interface
16797 (jeff)
16798 * CPU profiling now automatically affects all threads in
16799 linux 2.6.
16800 (Kernel bugs break CPU profiling and threads in linux 2.4
16801 a bit.)
16802 ProfilerEnable() and ProfilerDisable() are deprecated.
16803 (sanjay)
16804 * tcmalloc now correctly intercepts memalign (m3b, maxim)
16805 * Syntax fix: added missing va_end()s. Helps non-gcc
16806 compiling (etune)
16807 * Fixed a few coredumper bugs: race condition after
16808 PTRACE_DETACH,
16809 ignore non-aligned stackframe pointers (markus, menage)
16810 * 64-bit cleanup, especially for spinlock code (etune)
16811 and mmap (sanjay)
16812 * Better support for finding threads in linux (markus)
16813 * tcmalloc now tracks those stack traces that allocate memory
16814 (sanjay)
16815 * Work around a weird setspecific problem (sanjay)
16816 * Fix tcmalloc overflow problems when an alloc is close to
16817 2G/4G (sanjay)
16818
16819
16820 git-svn-id: http://gperftools.googlecode.com/svn/trunk@15
16821 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16822
16823 M ChangeLog
16824 M Makefile.am
16825 M aclocal.m4
16826 M configure
16827 M configure.ac
16828 M doc/cpu_profiler.html
16829 D src/base/elfcore.c
16830 M src/base/elfcore.h
16831 A src/base/linux_syscall_support.h
16832 M src/base/linuxthreads.c
16833 M src/base/linuxthreads.h
16834 M src/base/thread_lister.c
16835 M src/base/thread_lister.h
16836 M src/google/heap-checker.h
16837 M src/google/malloc_extension.h
16838 M src/google/profiler.h
16839 M src/heap-checker.cc
16840 M src/heap-profiler.cc
16841 M src/internal_logging.cc
16842 M src/internal_spinlock.h
16843 M src/malloc_extension.cc
16844 M src/malloc_hook.cc
16845 M src/pagemap.h
16846 M src/pprof
16847 M src/profiler.cc
16848 M src/stacktrace.cc
16849 M src/system-alloc.cc
16850 M src/tcmalloc.cc
16851 A src/tests/tcmalloc_large_unittest.cc
16852 M src/tests/tcmalloc_unittest.cc
16853
16854 commit bc455d7b63949fab94ed9518d277866e95f08768
16855 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16856 Date: Thu Mar 22 04:42:30 2007 +0000
16857
16858 Fri Jun 24 18:02:26 2005 Google Inc. <[email protected]>
16859
16860 * Add missing errno include for one of the unittests
16861 (csilvers)
16862 * Reduce tcmalloc startup memory from 5M to 256K (sanjay)
16863 * Add support for mallopt() and mallinfo (sanjay)
16864 * Improve stacktrace's performance on some 64-bit systems
16865 (etune)
16866 * Improve the stacktrace unittest (etune)
16867
16868
16869 git-svn-id: http://gperftools.googlecode.com/svn/trunk@13
16870 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16871
16872 M ChangeLog
16873 M configure
16874 M configure.ac
16875 M src/pagemap.h
16876 M src/stacktrace.cc
16877 M src/tcmalloc.cc
16878 M src/tests/heap-checker_unittest.cc
16879 M src/tests/stacktrace_unittest.cc
16880
16881 commit 91fad389784766782263133c5510976a8f76d89e
16882 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
16883 Date: Thu Mar 22 03:28:56 2007 +0000
16884
16885 Tue May 31 08:14:38 2005 Google Inc. <[email protected]>
16886
16887 * google-perftools: version 0.2 release
16888 * Use mmap2() instead of mmap(), to map more memory (menage)
16889 * Do correct pthread-local checking in heap-checker! (maxim)
16890 * Avoid overflow on 64-bit machines in pprof (sanjay)
16891 * Add a few more GetPC() functions, including for AMD
16892 (csilvers)
16893 * Better method for overriding pthread functions (menage)
16894 * (Hacky) fix to avoid overwriting profile files after fork()
16895 (csilvers)
16896 * Crashing bugfix involving dumping heaps on small-stack
16897 threads (tudor)
16898 * Allow library versions with letters at the end (csilvers)
16899 * Config fixes for systems that don't define PATH_MAX
16900 (csilvers)
16901 * Confix fixes so we no longer need config.h after install
16902 (csilvers)
16903 * Fix to pprof to correctly read very big cpu profiles
16904 (csilvers)
16905 * Fix to pprof to deal with new commandline flags in
16906 modern gv's
16907 * Better error reporting when we can't access /proc/maps
16908 (etune)
16909 * Get rid of the libc-preallocate code (which could crash
16910 on some
16911 systems); no longer needed with local-threads fix (csilvers)
16912
16913
16914 git-svn-id: http://gperftools.googlecode.com/svn/trunk@11
16915 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
16916
16917 M ChangeLog
16918 M Makefile.am
16919 M README
16920 M TODO
16921 M configure
16922 M configure.ac
16923 M doc/cpu_profiler.html
16924 A doc/heap_checker.html
16925 M doc/heap_profiler.html
16926 A doc/index.html
16927 A doc/t-test1.times.txt
16928 A doc/tcmalloc-opspercpusec.vs.threads.1024.bytes.png
16929 A doc/tcmalloc-opspercpusec.vs.threads.128.bytes.png
16930 A doc/tcmalloc-opspercpusec.vs.threads.131072.bytes.png
16931 A doc/tcmalloc-opspercpusec.vs.threads.16384.bytes.png
16932 A doc/tcmalloc-opspercpusec.vs.threads.2048.bytes.png
16933 A doc/tcmalloc-opspercpusec.vs.threads.256.bytes.png
16934 A doc/tcmalloc-opspercpusec.vs.threads.32768.bytes.png
16935 A doc/tcmalloc-opspercpusec.vs.threads.4096.bytes.png
16936 A doc/tcmalloc-opspercpusec.vs.threads.512.bytes.png
16937 A doc/tcmalloc-opspercpusec.vs.threads.64.bytes.png
16938 A doc/tcmalloc-opspercpusec.vs.threads.65536.bytes.png
16939 A doc/tcmalloc-opspercpusec.vs.threads.8192.bytes.png
16940 A doc/tcmalloc-opspersec.vs.size.1.threads.png
16941 A doc/tcmalloc-opspersec.vs.size.12.threads.png
16942 A doc/tcmalloc-opspersec.vs.size.16.threads.png
16943 A doc/tcmalloc-opspersec.vs.size.2.threads.png
16944 A doc/tcmalloc-opspersec.vs.size.20.threads.png
16945 A doc/tcmalloc-opspersec.vs.size.3.threads.png
16946 A doc/tcmalloc-opspersec.vs.size.4.threads.png
16947 A doc/tcmalloc-opspersec.vs.size.5.threads.png
16948 A doc/tcmalloc-opspersec.vs.size.8.threads.png
16949 M doc/tcmalloc.html
16950 M packages/deb/changelog
16951 M packages/deb/control
16952 M packages/deb/copyright
16953 D packages/deb/files
16954 M packages/deb/libgoogle-perftools-dev.install
16955 M packages/deb/libgoogle-perftools0.install
16956 M packages/rpm.sh
16957 M packages/rpm/rpm.spec
16958 M src/addressmap-inl.h
16959 R099 src/google/perftools/basictypes.h src/base/basictypes.h
16960 M src/base/commandlineflags.h
16961 A src/base/elfcore.c
16962 A src/base/elfcore.h
16963 A src/base/linuxthreads.c
16964 A src/base/linuxthreads.h
16965 A src/base/thread_lister.c
16966 A src/base/thread_lister.h
16967 M src/google/heap-checker.h
16968 M src/google/heap-profiler.h
16969 R088 src/google/malloc_interface.h src/google/malloc_extension.h
16970 M src/google/malloc_hook.h
16971 D src/google/perftools/config.h.in
16972 M src/google/stacktrace.h
16973 M src/heap-checker-bcad.cc
16974 M src/heap-checker.cc
16975 M src/heap-profiler-inl.h
16976 M src/heap-profiler.cc
16977 M src/internal_logging.h
16978 M src/internal_spinlock.h
16979 R068 src/malloc_interface.cc src/malloc_extension.cc
16980 M src/malloc_hook.cc
16981 A src/maybe_threads.cc
16982 C070 src/tests/tcmalloc_unittest.cc src/maybe_threads.h
16983 M src/pagemap.h
16984 M src/pprof
16985 M src/profiler.cc
16986 M src/stacktrace.cc
16987 M src/system-alloc.cc
16988 M src/tcmalloc.cc
16989 M src/tests/addressmap_unittest.cc
16990 M src/tests/heap-checker-death_unittest.sh
16991 M src/tests/heap-checker_unittest.cc
16992 C051 src/tests/heap-checker-death_unittest.sh
16993 src/tests/heap-checker_unittest.sh
16994 A src/tests/heap-profiler_unittest.cc
16995 A src/tests/heap-profiler_unittest.sh
16996 M src/tests/profiler_unittest.cc
16997 M src/tests/profiler_unittest.sh
16998 M src/tests/stacktrace_unittest.cc
16999 M src/tests/tcmalloc_unittest.cc
17000
17001 commit 51b4875f8ade3e0930eed2dc2a842ec607a94a2c
17002 Author: csilvers <csilvers@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
17003 Date: Thu Mar 22 03:00:33 2007 +0000
17004
17005 Tue Feb 8 09:57:17 2005 El Goog <[email protected]>
17006
17007 * google-perftools: initial release:
17008 The google-perftools package contains some utilities
17009 to improve
17010 and analyze the performance of C++ programs. This includes
17011 an
17012 optimized thread-caching malloc() and cpu and heap profiling
17013 utilities.
17014
17015
17016 git-svn-id: http://gperftools.googlecode.com/svn/trunk@9
17017 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
17018
17019 A AUTHORS
17020 A COPYING
17021 A ChangeLog
17022 A INSTALL
17023 A Makefile.am
17024 A Makefile.in
17025 A NEWS
17026 A README
17027 A TODO
17028 A aclocal.m4
17029 A compile
17030 A config.guess
17031 A config.sub
17032 A configure
17033 A configure.ac
17034 A depcomp
17035 A doc/cpu_profiler.html
17036 A doc/heap-example1.png
17037 A doc/heap_profiler.html
17038 A doc/overview.dot
17039 A doc/overview.gif
17040 A doc/pageheap.dot
17041 A doc/pageheap.gif
17042 A doc/pprof-test-big.gif
17043 A doc/pprof-test.gif
17044 A doc/pprof-vsnprintf-big.gif
17045 A doc/pprof-vsnprintf.gif
17046 A doc/pprof.1
17047 A doc/pprof.see_also
17048 A doc/spanmap.dot
17049 A doc/spanmap.gif
17050 A doc/tcmalloc.html
17051 A doc/threadheap.dot
17052 A doc/threadheap.gif
17053 A install-sh
17054 A ltmain.sh
17055 A missing
17056 A mkinstalldirs
17057 A packages/deb.sh
17058 A packages/deb/README
17059 A packages/deb/changelog
17060 A packages/deb/compat
17061 A packages/deb/control
17062 A packages/deb/copyright
17063 A packages/deb/files
17064 A packages/deb/libgoogle-perftools-dev.dirs
17065 A packages/deb/libgoogle-perftools-dev.install
17066 A packages/deb/libgoogle-perftools0.dirs
17067 A packages/deb/libgoogle-perftools0.install
17068 A packages/deb/libgoogle-perftools0.manpages
17069 A packages/deb/rules
17070 A packages/rpm.sh
17071 A packages/rpm/rpm.spec
17072 A src/addressmap-inl.h
17073 A src/base/commandlineflags.h
17074 A src/base/googleinit.h
17075 A src/base/logging.h
17076 A src/config.h.in
17077 A src/google/heap-checker.h
17078 A src/google/heap-profiler.h
17079 A src/google/malloc_hook.h
17080 A src/google/malloc_interface.h
17081 A src/google/perftools/basictypes.h
17082 A src/google/perftools/config.h.in
17083 A src/google/profiler.h
17084 A src/google/stacktrace.h
17085 A src/heap-checker-bcad.cc
17086 A src/heap-checker.cc
17087 A src/heap-profiler-inl.h
17088 A src/heap-profiler.cc
17089 A src/internal_logging.cc
17090 A src/internal_logging.h
17091 A src/internal_spinlock.h
17092 A src/malloc_hook.cc
17093 A src/malloc_interface.cc
17094 A src/pagemap.h
17095 A src/pprof
17096 A src/profiler.cc
17097 A src/stacktrace.cc
17098 A src/system-alloc.cc
17099 A src/system-alloc.h
17100 A src/tcmalloc.cc
17101 A src/tests/addressmap_unittest.cc
17102 A src/tests/heap-checker-death_unittest.sh
17103 A src/tests/heap-checker_unittest.cc
17104 A src/tests/profiler_unittest.cc
17105 A src/tests/profiler_unittest.sh
17106 A src/tests/ptmalloc/COPYRIGHT
17107 A src/tests/ptmalloc/lran2.h
17108 A src/tests/ptmalloc/malloc-machine.h
17109 A src/tests/ptmalloc/t-test.h
17110 A src/tests/ptmalloc/t-test1.c
17111 A src/tests/ptmalloc/t-test2.c
17112 A src/tests/ptmalloc/thread-m.h
17113 A src/tests/ptmalloc/thread-st.h
17114 A src/tests/stacktrace_unittest.cc
17115 A src/tests/tcmalloc_unittest.cc
17116
17117 commit e3a8513447d1141f083d4aaced0b240a6e161f47
17118 Author: trowbridge.jon
17119 <trowbridge.jon@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
17120 Date: Thu Dec 28 22:53:59 2006 +0000
17121
17122 Set page mime-type to text/html.
17123
17124
17125 git-svn-id: http://gperftools.googlecode.com/svn/trunk@5
17126 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
17127
17128 commit 66737d1c2519e4a1622f61139bfe2f683ea3696c
17129 Author: trowbridge.jon
17130 <trowbridge.jon@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
17131 Date: Thu Dec 28 22:39:33 2006 +0000
17132
17133 Import of HTML documentation from SourceForge.
17134
17135
17136
17137 git-svn-id: http://gperftools.googlecode.com/svn/trunk@3
17138 6b5cf1ce-ec42-a296-1ba9-69fdba395a50
17139
17140 A docs/html/cpu_profiler.html
17141 A docs/html/heap_checker.html
17142 A docs/html/heap_profiler.html
17143 A docs/html/tcmalloc.html
17144 A docs/images/heap-example1.png
17145 A docs/images/overview.gif
17146 A docs/images/pageheap.gif
17147 A docs/images/pprof-test.gif
17148 A docs/images/pprof-vsnprintf.gif
17149 A docs/images/spanmap.gif
17150 A docs/images/tcmalloc-opspercpusec.png
17151 A docs/images/tcmalloc-opspercpusec_002.png
17152 A docs/images/tcmalloc-opspercpusec_003.png
17153 A docs/images/tcmalloc-opspercpusec_004.png
17154 A docs/images/tcmalloc-opspercpusec_005.png
17155 A docs/images/tcmalloc-opspercpusec_006.png
17156 A docs/images/tcmalloc-opspercpusec_007.png
17157 A docs/images/tcmalloc-opspercpusec_008.png
17158 A docs/images/tcmalloc-opspercpusec_009.png
17159 A docs/images/tcmalloc-opspersec.png
17160 A docs/images/tcmalloc-opspersec_002.png
17161 A docs/images/tcmalloc-opspersec_003.png
17162 A docs/images/tcmalloc-opspersec_004.png
17163 A docs/images/tcmalloc-opspersec_005.png
17164 A docs/images/tcmalloc-opspersec_006.png
17165 A docs/images/tcmalloc-opspersec_007.png
17166 A docs/images/tcmalloc-opspersec_008.png
17167 A docs/images/tcmalloc-opspersec_009.png
17168 A docs/images/threadheap.gif
17169
17170 commit 55d679a05f0518ea73a4bca6e8b71b54fcecf68f
17171 Author: (no author) <(no author)@6b5cf1ce-ec42-a296-1ba9-69fdba395a50>
17172 Date: Thu Jul 27 00:57:14 2006 +0000
17173
17174 Initial directory structure.
17175
17176 git-svn-id: http://gperftools.googlecode.com/svn/trunk@1
17177 6b5cf1ce-ec42-a296-1ba9-69fdba395a50