Debian Codegraph @

upstream/debian-unstable/man1/perldebguts.1

raw
1 .\" -*- mode: troff; coding: utf-8 -*-
2 .\" Automatically generated by Pod::Man 5.0102 (Pod::Simple 3.45)
3 .\"
4 .\" Standard preamble:
5 .\" ========================================================================
6 .de Sp \" Vertical space (when we can't use .PP)
7 .if t .sp .5v
8 .if n .sp
9 ..
10 .de Vb \" Begin verbatim text
11 .ft CW
12 .nf
13 .ne \\$1
14 ..
15 .de Ve \" End verbatim text
16 .ft R
17 .fi
18 ..
19 .\" \*(C` and \*(C' are quotes in nroff, nothing in troff, for use with C<>.
20 .ie n \{\
21 . ds C` ""
22 . ds C' ""
23 'br\}
24 .el\{\
25 . ds C`
26 . ds C'
27 'br\}
28 .\"
29 .\" Escape single quotes in literal strings from groff's Unicode transform.
30 .ie \n(.g .ds Aq \(aq
31 .el .ds Aq '
32 .\"
33 .\" If the F register is >0, we'll generate index entries on stderr for
34 .\" titles (.TH), headers (.SH), subsections (.SS), items (.Ip), and index
35 .\" entries marked with X<> in POD. Of course, you'll have to process the
36 .\" output yourself in some meaningful fashion.
37 .\"
38 .\" Avoid warning from groff about undefined register 'F'.
39 .de IX
40 ..
41 .nr rF 0
42 .if \n(.g .if rF .nr rF 1
43 .if (\n(rF:(\n(.g==0)) \{\
44 . if \nF \{\
45 . de IX
46 . tm Index:\\$1\t\\n%\t"\\$2"
47 ..
48 . if !\nF==2 \{\
49 . nr % 0
50 . nr F 2
51 . \}
52 . \}
53 .\}
54 .rr rF
55 .\" ========================================================================
56 .\"
57 .IX Title "PERLDEBGUTS 1"
58 .TH PERLDEBGUTS 1 2025-11-16 "perl v5.40.1" "Perl Programmers Reference Guide"
59 .\" For nroff, turn off justification. Always turn off hyphenation; it makes
60 .\" way too many mistakes in technical documents.
61 .if n .ad l
62 .nh
63 .SH NAME
64 perldebguts \- Guts of Perl debugging
65 .SH DESCRIPTION
66 .IX Header "DESCRIPTION"
67 This is not perldebug, which tells you how to use
68 the debugger. This manpage describes low-level details concerning
69 the debugger's internals, which range from difficult to impossible
70 to understand for anyone who isn't incredibly intimate with Perl's guts.
71 Caveat lector.
72 .SH "Debugger Internals"
73 .IX Header "Debugger Internals"
74 Perl has special debugging hooks at compile-time and run-time used
75 to create debugging environments. These hooks are not to be confused
76 with the \fIperl \-Dxxx\fR command described in perlrun,
77 which is usable only if a special Perl is built per the instructions in
78 the \fIINSTALL\fR file in the Perl source tree.
79 .PP
80 For example, whenever you call Perl's built-in \f(CW\*(C`caller\*(C'\fR function
81 from the package \f(CW\*(C`DB\*(C'\fR, the arguments that the corresponding stack
82 frame was called with are copied to the \f(CW@DB::args\fR array. These
83 mechanisms are enabled by calling Perl with the \fB\-d\fR switch.
84 Specifically, the following additional features are enabled
85 (cf. "$^P" in perlvar):
86 .IP \(bu 4
87 Perl inserts the contents of \f(CW$ENV{PERL5DB}\fR (or \f(CW\*(C`BEGIN {require
88 \&\*(Aqperl5db.pl\*(Aq}\*(C'\fR if not present) before the first line of your program.
89 .IP \(bu 4
90 Each array \f(CW\*(C`@{"_<$filename"}\*(C'\fR holds the lines of \f(CW$filename\fR for a
91 file compiled by Perl. The same is also true for \f(CW\*(C`eval\*(C'\fRed strings
92 that contain subroutines, or which are currently being executed.
93 The \f(CW$filename\fR for \f(CW\*(C`eval\*(C'\fRed strings looks like \f(CW\*(C`(eval 34)\*(C'\fR.
94 .Sp
95 Values in this array are magical in numeric context: they compare
96 equal to zero only if the line is not breakable.
97 .IP \(bu 4
98 Each hash \f(CW\*(C`%{"_<$filename"}\*(C'\fR contains breakpoints and actions keyed
99 by line number. Individual entries (as opposed to the whole hash)
100 are settable. Perl only cares about Boolean true here, although
101 the values used by \fIperl5db.pl\fR have the form
102 \&\f(CW"$break_condition\e0$action"\fR.
103 .Sp
104 The same holds for evaluated strings that contain subroutines, or
105 which are currently being executed. The \f(CW$filename\fR for \f(CW\*(C`eval\*(C'\fRed strings
106 looks like \f(CW\*(C`(eval 34)\*(C'\fR.
107 .IP \(bu 4
108 Each scalar \f(CW\*(C`${"_<$filename"}\*(C'\fR contains \f(CW$filename\fR. This is
109 also the case for evaluated strings that contain subroutines, or
110 which are currently being executed. The \f(CW$filename\fR for \f(CW\*(C`eval\*(C'\fRed
111 strings looks like \f(CW\*(C`(eval 34)\*(C'\fR.
112 .IP \(bu 4
113 After each \f(CW\*(C`require\*(C'\fRd file is compiled, but before it is executed,
114 \&\f(CWDB::postponed(*{"_<$filename"})\fR is called if the subroutine
115 \&\f(CW\*(C`DB::postponed\*(C'\fR exists. Here, the \f(CW$filename\fR is the expanded name of
116 the \f(CW\*(C`require\*(C'\fRd file, as found in the values of \f(CW%INC\fR.
117 .IP \(bu 4
118 After each subroutine \f(CW\*(C`subname\*(C'\fR is compiled, the existence of
119 \&\f(CW$DB::postponed{subname}\fR is checked. If this key exists,
120 \&\f(CWDB::postponed(subname)\fR is called if the \f(CW\*(C`DB::postponed\*(C'\fR subroutine
121 also exists.
122 .IP \(bu 4
123 A hash \f(CW%DB::sub\fR is maintained, whose keys are subroutine names
124 and whose values have the form \f(CW\*(C`filename:startline\-endline\*(C'\fR.
125 \&\f(CW\*(C`filename\*(C'\fR has the form \f(CW\*(C`(eval 34)\*(C'\fR for subroutines defined inside
126 \&\f(CW\*(C`eval\*(C'\fRs.
127 .IP \(bu 4
128 When the execution of your program reaches a point that can hold a
129 breakpoint, the \f(CWDB::DB()\fR subroutine is called if any of the variables
130 \&\f(CW$DB::trace\fR, \f(CW$DB::single\fR, or \f(CW$DB::signal\fR is true. These variables
131 are not \f(CW\*(C`local\*(C'\fRizable. This feature is disabled when executing
132 inside \f(CWDB::DB()\fR, including functions called from it
133 unless \f(CW\*(C`$^D & (1<<30)\*(C'\fR is true.
134 .IP \(bu 4
135 When execution of the program reaches a subroutine call, a call to
136 \&\f(CW&DB::sub\fR(\fIargs\fR) is made instead, with \f(CW$DB::sub\fR set to identify
137 the called subroutine. (This doesn't happen if the calling subroutine
138 was compiled in the \f(CW\*(C`DB\*(C'\fR package.) \f(CW$DB::sub\fR normally holds the name
139 of the called subroutine, if it has a name by which it can be looked up.
140 Failing that, \f(CW$DB::sub\fR will hold a reference to the called subroutine.
141 Either way, the \f(CW&DB::sub\fR subroutine can use \f(CW$DB::sub\fR as a reference
142 by which to call the called subroutine, which it will normally want to do.
143 .Sp
144 If the call is to an lvalue subroutine, and \f(CW&DB::lsub\fR
145 is defined \f(CW&DB::lsub\fR(\fIargs\fR) is called instead, otherwise falling
146 back to \f(CW&DB::sub\fR(\fIargs\fR).
147 .IX Xref "&DB::lsub"
148 .IP \(bu 4
149 When execution of the program uses \f(CW\*(C`goto\*(C'\fR to enter a non-XS subroutine
150 and the 0x80 bit is set in \f(CW$^P\fR, a call to \f(CW&DB::goto\fR is made, with
151 \&\f(CW$DB::sub\fR set to identify the subroutine being entered. The call to
152 \&\f(CW&DB::goto\fR does not replace the \f(CW\*(C`goto\*(C'\fR; the requested subroutine will
153 still be entered once \f(CW&DB::goto\fR has returned. \f(CW$DB::sub\fR normally
154 holds the name of the subroutine being entered, if it has one. Failing
155 that, \f(CW$DB::sub\fR will hold a reference to the subroutine being entered.
156 Unlike when \f(CW&DB::sub\fR is called, it is not guaranteed that \f(CW$DB::sub\fR
157 can be used as a reference to operate on the subroutine being entered.
158 .PP
159 Note that if \f(CW&DB::sub\fR needs external data for it to work, no
160 subroutine call is possible without it. As an example, the standard
161 debugger's \f(CW&DB::sub\fR depends on the \f(CW$DB::deep\fR variable
162 (it defines how many levels of recursion deep into the debugger you can go
163 before a mandatory break). If \f(CW$DB::deep\fR is not defined, subroutine
164 calls are not possible, even though \f(CW&DB::sub\fR exists.
165 .SS "Writing Your Own Debugger"
166 .IX Subsection "Writing Your Own Debugger"
167 \fIEnvironment Variables\fR
168 .IX Subsection "Environment Variables"
169 .PP
170 The \f(CW\*(C`PERL5DB\*(C'\fR environment variable can be used to define a debugger.
171 For example, the minimal "working" debugger (it actually doesn't do anything)
172 consists of one line:
173 .PP
174 .Vb 1
175 \& sub DB::DB {}
176 .Ve
177 .PP
178 It can easily be defined like this:
179 .PP
180 .Vb 1
181 \& $ PERL5DB="sub DB::DB {}" perl \-d your\-script
182 .Ve
183 .PP
184 Another brief debugger, slightly more useful, can be created
185 with only the line:
186 .PP
187 .Vb 1
188 \& sub DB::DB {print ++$i; scalar <STDIN>}
189 .Ve
190 .PP
191 This debugger prints a number which increments for each statement
192 encountered and waits for you to hit a newline before continuing
193 to the next statement.
194 .PP
195 The following debugger is actually useful:
196 .PP
197 .Vb 5
198 \& {
199 \& package DB;
200 \& sub DB {}
201 \& sub sub {print ++$i, " $sub\en"; &$sub}
202 \& }
203 .Ve
204 .PP
205 It prints the sequence number of each subroutine call and the name of the
206 called subroutine. Note that \f(CW&DB::sub\fR is being compiled into the
207 package \f(CW\*(C`DB\*(C'\fR through the use of the \f(CW\*(C`package\*(C'\fR directive.
208 .PP
209 When it starts, the debugger reads your rc file (\fI./.perldb\fR or
210 \&\fI~/.perldb\fR under Unix), which can set important options.
211 (A subroutine (\f(CW&afterinit\fR) can be defined here as well; it is executed
212 after the debugger completes its own initialization.)
213 .PP
214 After the rc file is read, the debugger reads the PERLDB_OPTS
215 environment variable and uses it to set debugger options. The
216 contents of this variable are treated as if they were the argument
217 of an \f(CW\*(C`o ...\*(C'\fR debugger command (q.v. in "Configurable Options" in perldebug).
218 .PP
219 \fIDebugger Internal Variables\fR
220 .IX Subsection "Debugger Internal Variables"
221 .PP
222 In addition to the file and subroutine-related variables mentioned above,
223 the debugger also maintains various magical internal variables.
224 .IP \(bu 4
225 \&\f(CW@DB::dbline\fR is an alias for \f(CW\*(C`@{"::_<current_file"}\*(C'\fR, which
226 holds the lines of the currently-selected file (compiled by Perl), either
227 explicitly chosen with the debugger's \f(CW\*(C`f\*(C'\fR command, or implicitly by flow
228 of execution.
229 .Sp
230 Values in this array are magical in numeric context: they compare
231 equal to zero only if the line is not breakable.
232 .IP \(bu 4
233 \&\f(CW%DB::dbline\fR is an alias for \f(CW\*(C`%{"::_<current_file"}\*(C'\fR, which
234 contains breakpoints and actions keyed by line number in
235 the currently-selected file, either explicitly chosen with the
236 debugger's \f(CW\*(C`f\*(C'\fR command, or implicitly by flow of execution.
237 .Sp
238 As previously noted, individual entries (as opposed to the whole hash)
239 are settable. Perl only cares about Boolean true here, although
240 the values used by \fIperl5db.pl\fR have the form
241 \&\f(CW"$break_condition\e0$action"\fR.
242 .PP
243 \fIDebugger Customization Functions\fR
244 .IX Subsection "Debugger Customization Functions"
245 .PP
246 Some functions are provided to simplify customization.
247 .IP \(bu 4
248 See "Configurable Options" in perldebug for a description of options parsed by
249 \&\f(CWDB::parse_options(string)\fR.
250 .IP \(bu 4
251 \&\f(CW\*(C`DB::dump_trace(skip[,count])\*(C'\fR skips the specified number of frames
252 and returns a list containing information about the calling frames (all
253 of them, if \f(CW\*(C`count\*(C'\fR is missing). Each entry is reference to a hash
254 with keys \f(CW\*(C`context\*(C'\fR (either \f(CW\*(C`.\*(C'\fR, \f(CW\*(C`$\*(C'\fR, or \f(CW\*(C`@\*(C'\fR), \f(CW\*(C`sub\*(C'\fR (subroutine
255 name, or info about \f(CW\*(C`eval\*(C'\fR), \f(CW\*(C`args\*(C'\fR (\f(CW\*(C`undef\*(C'\fR or a reference to
256 an array), \f(CW\*(C`file\*(C'\fR, and \f(CW\*(C`line\*(C'\fR.
257 .IP \(bu 4
258 \&\f(CW\*(C`DB::print_trace(FH, skip[, count[, short]])\*(C'\fR prints
259 formatted info about caller frames. The last two functions may be
260 convenient as arguments to \f(CW\*(C`<\*(C'\fR, \f(CW\*(C`<<\*(C'\fR commands.
261 .PP
262 Note that any variables and functions that are not documented in
263 this manpages (or in perldebug) are considered for internal
264 use only, and as such are subject to change without notice.
265 .SH "Frame Listing Output Examples"
266 .IX Header "Frame Listing Output Examples"
267 The \f(CW\*(C`frame\*(C'\fR option can be used to control the output of frame
268 information. For example, contrast this expression trace:
269 .PP
270 .Vb 2
271 \& $ perl \-de 42
272 \& Stack dump during die enabled outside of evals.
273 \&
274 \& Loading DB routines from perl5db.pl patch level 0.94
275 \& Emacs support available.
276 \&
277 \& Enter h or \*(Aqh h\*(Aq for help.
278 \&
279 \& main::(\-e:1): 0
280 \& DB<1> sub foo { 14 }
281 \&
282 \& DB<2> sub bar { 3 }
283 \&
284 \& DB<3> t print foo() * bar()
285 \& main::((eval 172):3): print foo() + bar();
286 \& main::foo((eval 168):2):
287 \& main::bar((eval 170):2):
288 \& 42
289 .Ve
290 .PP
291 with this one, once the \f(CW\*(C`o\*(C'\fRption \f(CW\*(C`frame=2\*(C'\fR has been set:
292 .PP
293 .Vb 11
294 \& DB<4> o f=2
295 \& frame = \*(Aq2\*(Aq
296 \& DB<5> t print foo() * bar()
297 \& 3: foo() * bar()
298 \& entering main::foo
299 \& 2: sub foo { 14 };
300 \& exited main::foo
301 \& entering main::bar
302 \& 2: sub bar { 3 };
303 \& exited main::bar
304 \& 42
305 .Ve
306 .PP
307 By way of demonstration, we present below a laborious listing
308 resulting from setting your \f(CW\*(C`PERLDB_OPTS\*(C'\fR environment variable to
309 the value \f(CW\*(C`f=n N\*(C'\fR, and running \fIperl \-d \-V\fR from the command line.
310 Examples using various values of \f(CW\*(C`n\*(C'\fR are shown to give you a feel
311 for the difference between settings. Long though it may be, this
312 is not a complete listing, but only excerpts.
313 .IP 1. 4
314 .Vb 10
315 \& entering main::BEGIN
316 \& entering Config::BEGIN
317 \& Package lib/Exporter.pm.
318 \& Package lib/Carp.pm.
319 \& Package lib/Config.pm.
320 \& entering Config::TIEHASH
321 \& entering Exporter::import
322 \& entering Exporter::export
323 \& entering Config::myconfig
324 \& entering Config::FETCH
325 \& entering Config::FETCH
326 \& entering Config::FETCH
327 \& entering Config::FETCH
328 .Ve
329 .IP 2. 4
330 .Vb 10
331 \& entering main::BEGIN
332 \& entering Config::BEGIN
333 \& Package lib/Exporter.pm.
334 \& Package lib/Carp.pm.
335 \& exited Config::BEGIN
336 \& Package lib/Config.pm.
337 \& entering Config::TIEHASH
338 \& exited Config::TIEHASH
339 \& entering Exporter::import
340 \& entering Exporter::export
341 \& exited Exporter::export
342 \& exited Exporter::import
343 \& exited main::BEGIN
344 \& entering Config::myconfig
345 \& entering Config::FETCH
346 \& exited Config::FETCH
347 \& entering Config::FETCH
348 \& exited Config::FETCH
349 \& entering Config::FETCH
350 .Ve
351 .IP 3. 4
352 .Vb 10
353 \& in $=main::BEGIN() from /dev/null:0
354 \& in $=Config::BEGIN() from lib/Config.pm:2
355 \& Package lib/Exporter.pm.
356 \& Package lib/Carp.pm.
357 \& Package lib/Config.pm.
358 \& in $=Config::TIEHASH(\*(AqConfig\*(Aq) from lib/Config.pm:644
359 \& in $=Exporter::import(\*(AqConfig\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from /dev/null:0
360 \& in $=Exporter::export(\*(AqConfig\*(Aq, \*(Aqmain\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from li
361 \& in @=Config::myconfig() from /dev/null:0
362 \& in $=Config::FETCH(ref(Config), \*(Aqpackage\*(Aq) from lib/Config.pm:574
363 \& in $=Config::FETCH(ref(Config), \*(Aqbaserev\*(Aq) from lib/Config.pm:574
364 \& in $=Config::FETCH(ref(Config), \*(AqPERL_VERSION\*(Aq) from lib/Config.pm:574
365 \& in $=Config::FETCH(ref(Config), \*(AqPERL_SUBVERSION\*(Aq) from lib/Config.pm:574
366 \& in $=Config::FETCH(ref(Config), \*(Aqosname\*(Aq) from lib/Config.pm:574
367 \& in $=Config::FETCH(ref(Config), \*(Aqosvers\*(Aq) from lib/Config.pm:574
368 .Ve
369 .IP 4. 4
370 .Vb 10
371 \& in $=main::BEGIN() from /dev/null:0
372 \& in $=Config::BEGIN() from lib/Config.pm:2
373 \& Package lib/Exporter.pm.
374 \& Package lib/Carp.pm.
375 \& out $=Config::BEGIN() from lib/Config.pm:0
376 \& Package lib/Config.pm.
377 \& in $=Config::TIEHASH(\*(AqConfig\*(Aq) from lib/Config.pm:644
378 \& out $=Config::TIEHASH(\*(AqConfig\*(Aq) from lib/Config.pm:644
379 \& in $=Exporter::import(\*(AqConfig\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from /dev/null:0
380 \& in $=Exporter::export(\*(AqConfig\*(Aq, \*(Aqmain\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from lib/
381 \& out $=Exporter::export(\*(AqConfig\*(Aq, \*(Aqmain\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from lib/
382 \& out $=Exporter::import(\*(AqConfig\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from /dev/null:0
383 \& out $=main::BEGIN() from /dev/null:0
384 \& in @=Config::myconfig() from /dev/null:0
385 \& in $=Config::FETCH(ref(Config), \*(Aqpackage\*(Aq) from lib/Config.pm:574
386 \& out $=Config::FETCH(ref(Config), \*(Aqpackage\*(Aq) from lib/Config.pm:574
387 \& in $=Config::FETCH(ref(Config), \*(Aqbaserev\*(Aq) from lib/Config.pm:574
388 \& out $=Config::FETCH(ref(Config), \*(Aqbaserev\*(Aq) from lib/Config.pm:574
389 \& in $=Config::FETCH(ref(Config), \*(AqPERL_VERSION\*(Aq) from lib/Config.pm:574
390 \& out $=Config::FETCH(ref(Config), \*(AqPERL_VERSION\*(Aq) from lib/Config.pm:574
391 \& in $=Config::FETCH(ref(Config), \*(AqPERL_SUBVERSION\*(Aq) from lib/Config.pm:574
392 .Ve
393 .IP 5. 4
394 .Vb 10
395 \& in $=main::BEGIN() from /dev/null:0
396 \& in $=Config::BEGIN() from lib/Config.pm:2
397 \& Package lib/Exporter.pm.
398 \& Package lib/Carp.pm.
399 \& out $=Config::BEGIN() from lib/Config.pm:0
400 \& Package lib/Config.pm.
401 \& in $=Config::TIEHASH(\*(AqConfig\*(Aq) from lib/Config.pm:644
402 \& out $=Config::TIEHASH(\*(AqConfig\*(Aq) from lib/Config.pm:644
403 \& in $=Exporter::import(\*(AqConfig\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from /dev/null:0
404 \& in $=Exporter::export(\*(AqConfig\*(Aq, \*(Aqmain\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from lib/E
405 \& out $=Exporter::export(\*(AqConfig\*(Aq, \*(Aqmain\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from lib/E
406 \& out $=Exporter::import(\*(AqConfig\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from /dev/null:0
407 \& out $=main::BEGIN() from /dev/null:0
408 \& in @=Config::myconfig() from /dev/null:0
409 \& in $=Config::FETCH(\*(AqConfig=HASH(0x1aa444)\*(Aq, \*(Aqpackage\*(Aq) from lib/Config.pm:574
410 \& out $=Config::FETCH(\*(AqConfig=HASH(0x1aa444)\*(Aq, \*(Aqpackage\*(Aq) from lib/Config.pm:574
411 \& in $=Config::FETCH(\*(AqConfig=HASH(0x1aa444)\*(Aq, \*(Aqbaserev\*(Aq) from lib/Config.pm:574
412 \& out $=Config::FETCH(\*(AqConfig=HASH(0x1aa444)\*(Aq, \*(Aqbaserev\*(Aq) from lib/Config.pm:574
413 .Ve
414 .IP 6. 4
415 .Vb 10
416 \& in $=CODE(0x15eca4)() from /dev/null:0
417 \& in $=CODE(0x182528)() from lib/Config.pm:2
418 \& Package lib/Exporter.pm.
419 \& out $=CODE(0x182528)() from lib/Config.pm:0
420 \& scalar context return from CODE(0x182528): undef
421 \& Package lib/Config.pm.
422 \& in $=Config::TIEHASH(\*(AqConfig\*(Aq) from lib/Config.pm:628
423 \& out $=Config::TIEHASH(\*(AqConfig\*(Aq) from lib/Config.pm:628
424 \& scalar context return from Config::TIEHASH: empty hash
425 \& in $=Exporter::import(\*(AqConfig\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from /dev/null:0
426 \& in $=Exporter::export(\*(AqConfig\*(Aq, \*(Aqmain\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from lib/Exporter.pm:171
427 \& out $=Exporter::export(\*(AqConfig\*(Aq, \*(Aqmain\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from lib/Exporter.pm:171
428 \& scalar context return from Exporter::export: \*(Aq\*(Aq
429 \& out $=Exporter::import(\*(AqConfig\*(Aq, \*(Aqmyconfig\*(Aq, \*(Aqconfig_vars\*(Aq) from /dev/null:0
430 \& scalar context return from Exporter::import: \*(Aq\*(Aq
431 .Ve
432 .PP
433 In all cases shown above, the line indentation shows the call tree.
434 If bit 2 of \f(CW\*(C`frame\*(C'\fR is set, a line is printed on exit from a
435 subroutine as well. If bit 4 is set, the arguments are printed
436 along with the caller info. If bit 8 is set, the arguments are
437 printed even if they are tied or references. If bit 16 is set, the
438 return value is printed, too.
439 .PP
440 When a package is compiled, a line like this
441 .PP
442 .Vb 1
443 \& Package lib/Carp.pm.
444 .Ve
445 .PP
446 is printed with proper indentation.
447 .SH "Debugging Regular Expressions"
448 .IX Header "Debugging Regular Expressions"
449 There are two ways to enable debugging output for regular expressions.
450 .PP
451 If your perl is compiled with \f(CW\*(C`\-DDEBUGGING\*(C'\fR, you may use the
452 \&\fB\-Dr\fR flag on the command line, and \f(CW\*(C`\-Drv\*(C'\fR for more verbose
453 information.
454 .PP
455 Otherwise, one can \f(CW\*(C`use re \*(Aqdebug\*(Aq\*(C'\fR, which has effects at both
456 compile time and run time. Since Perl 5.9.5, this pragma is lexically
457 scoped.
458 .SS "Compile-time Output"
459 .IX Subsection "Compile-time Output"
460 The debugging output at compile time looks like this:
461 .PP
462 .Vb 10
463 \& Compiling REx \*(Aq[bc]d(ef*g)+h[ij]k$\*(Aq
464 \& size 45 Got 364 bytes for offset annotations.
465 \& first at 1
466 \& rarest char g at 0
467 \& rarest char d at 0
468 \& 1: ANYOF[bc](12)
469 \& 12: EXACT <d>(14)
470 \& 14: CURLYX[0] {1,32767}(28)
471 \& 16: OPEN1(18)
472 \& 18: EXACT <e>(20)
473 \& 20: STAR(23)
474 \& 21: EXACT <f>(0)
475 \& 23: EXACT <g>(25)
476 \& 25: CLOSE1(27)
477 \& 27: WHILEM[1/1](0)
478 \& 28: NOTHING(29)
479 \& 29: EXACT <h>(31)
480 \& 31: ANYOF[ij](42)
481 \& 42: EXACT <k>(44)
482 \& 44: EOL(45)
483 \& 45: END(0)
484 \& anchored \*(Aqde\*(Aq at 1 floating \*(Aqgh\*(Aq at 3..2147483647 (checking floating)
485 \& stclass \*(AqANYOF[bc]\*(Aq minlen 7
486 \& Offsets: [45]
487 \& 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
488 \& 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
489 \& 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
490 \& 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
491 \& Omitting $\` $& $\*(Aq support.
492 .Ve
493 .PP
494 The first line shows the pre-compiled form of the regex. The second
495 shows the size of the compiled form (in arbitrary units, usually
496 4\-byte words) and the total number of bytes allocated for the
497 offset/length table, usually 4+\f(CW\*(C`size\*(C'\fR*8. The next line shows the
498 label \fIid\fR of the first node that does a match.
499 .PP
500 The
501 .PP
502 .Vb 2
503 \& anchored \*(Aqde\*(Aq at 1 floating \*(Aqgh\*(Aq at 3..2147483647 (checking floating)
504 \& stclass \*(AqANYOF[bc]\*(Aq minlen 7
505 .Ve
506 .PP
507 line (split into two lines above) contains optimizer
508 information. In the example shown, the optimizer found that the match
509 should contain a substring \f(CW\*(C`de\*(C'\fR at offset 1, plus substring \f(CW\*(C`gh\*(C'\fR
510 at some offset between 3 and infinity. Moreover, when checking for
511 these substrings (to abandon impossible matches quickly), Perl will check
512 for the substring \f(CW\*(C`gh\*(C'\fR before checking for the substring \f(CW\*(C`de\*(C'\fR. The
513 optimizer may also use the knowledge that the match starts (at the
514 \&\f(CW\*(C`first\*(C'\fR \fIid\fR) with a character class, and no string
515 shorter than 7 characters can possibly match.
516 .PP
517 The fields of interest which may appear in this line are
518 .ie n .IP """anchored"" \fISTRING\fR ""at"" \fIPOS\fR" 4
519 .el .IP "\f(CWanchored\fR \fISTRING\fR \f(CWat\fR \fIPOS\fR" 4
520 .IX Item "anchored STRING at POS"
521 .PD 0
522 .ie n .IP """floating"" \fISTRING\fR ""at"" \fIPOS1..POS2\fR" 4
523 .el .IP "\f(CWfloating\fR \fISTRING\fR \f(CWat\fR \fIPOS1..POS2\fR" 4
524 .IX Item "floating STRING at POS1..POS2"
525 .PD
526 See above.
527 .ie n .IP """matching floating/anchored""" 4
528 .el .IP "\f(CWmatching floating/anchored\fR" 4
529 .IX Item "matching floating/anchored"
530 Which substring to check first.
531 .ie n .IP """minlen""" 4
532 .el .IP \f(CWminlen\fR 4
533 .IX Item "minlen"
534 The minimal length of the match.
535 .ie n .IP """stclass"" \fITYPE\fR" 4
536 .el .IP "\f(CWstclass\fR \fITYPE\fR" 4
537 .IX Item "stclass TYPE"
538 Type of first matching node.
539 .ie n .IP """noscan""" 4
540 .el .IP \f(CWnoscan\fR 4
541 .IX Item "noscan"
542 Don't scan for the found substrings.
543 .ie n .IP """isall""" 4
544 .el .IP \f(CWisall\fR 4
545 .IX Item "isall"
546 Means that the optimizer information is all that the regular
547 expression contains, and thus one does not need to enter the regex engine at
548 all.
549 .ie n .IP """GPOS""" 4
550 .el .IP \f(CWGPOS\fR 4
551 .IX Item "GPOS"
552 Set if the pattern contains \f(CW\*(C`\eG\*(C'\fR.
553 .ie n .IP """plus""" 4
554 .el .IP \f(CWplus\fR 4
555 .IX Item "plus"
556 Set if the pattern starts with a repeated char (as in \f(CW\*(C`x+y\*(C'\fR).
557 .ie n .IP """implicit""" 4
558 .el .IP \f(CWimplicit\fR 4
559 .IX Item "implicit"
560 Set if the pattern starts with \f(CW\*(C`.*\*(C'\fR.
561 .ie n .IP """with eval""" 4
562 .el .IP "\f(CWwith eval\fR" 4
563 .IX Item "with eval"
564 Set if the pattern contain eval-groups, such as \f(CW\*(C`(?{ code })\*(C'\fR and
565 \&\f(CW\*(C`(??{ code })\*(C'\fR.
566 .ie n .IP anchored(TYPE) 4
567 .el .IP \f(CWanchored(TYPE)\fR 4
568 .IX Item "anchored(TYPE)"
569 If the pattern may match only at a handful of places, with \f(CW\*(C`TYPE\*(C'\fR
570 being \f(CW\*(C`SBOL\*(C'\fR, \f(CW\*(C`MBOL\*(C'\fR, or \f(CW\*(C`GPOS\*(C'\fR. See the table below.
571 .PP
572 If a substring is known to match at end-of-line only, it may be
573 followed by \f(CW\*(C`$\*(C'\fR, as in \f(CW\*(C`floating \*(Aqk\*(Aq$\*(C'\fR.
574 .PP
575 The optimizer-specific information is used to avoid entering (a slow) regex
576 engine on strings that will not definitely match. If the \f(CW\*(C`isall\*(C'\fR flag
577 is set, a call to the regex engine may be avoided even when the optimizer
578 found an appropriate place for the match.
579 .PP
580 Above the optimizer section is the list of \fInodes\fR of the compiled
581 form of the regex. Each line has format
582 .PP
583 \&\f(CW\*(C` \*(C'\fR\fIid\fR: \fITYPE\fR \fIOPTIONAL-INFO\fR (\fInext-id\fR)
584 .SS "Types of Nodes"
585 .IX Subsection "Types of Nodes"
586 Here are the current possible types, with short descriptions:
587 .PP
588 .Vb 1
589 \& # TYPE arg\-description [regnode\-struct\-suffix] [longjump\-len] DESCRIPTION
590 \&
591 \& # Exit points
592 \&
593 \& END no End of program.
594 \& SUCCEED no Return from a subroutine, basically.
595 \&
596 \& # Line Start Anchors:
597 \& SBOL no Match "" at beginning of line: /^/, /\eA/
598 \& MBOL no Same, assuming multiline: /^/m
599 \&
600 \& # Line End Anchors:
601 \& SEOL no Match "" at end of line: /$/
602 \& MEOL no Same, assuming multiline: /$/m
603 \& EOS no Match "" at end of string: /\ez/
604 \&
605 \& # Match Start Anchors:
606 \& GPOS no Matches where last m//g left off.
607 \&
608 \& # Word Boundary Opcodes:
609 \& BOUND no Like BOUNDA for non\-utf8, otherwise like
610 \& BOUNDU
611 \& BOUNDL no Like BOUND/BOUNDU, but \ew and \eW are
612 \& defined by current locale
613 \& BOUNDU no Match "" at any boundary of a given type
614 \& using /u rules.
615 \& BOUNDA no Match "" at any boundary between \ew\eW or
616 \& \eW\ew, where \ew is [_a\-zA\-Z0\-9]
617 \& NBOUND no Like NBOUNDA for non\-utf8, otherwise like
618 \& BOUNDU
619 \& NBOUNDL no Like NBOUND/NBOUNDU, but \ew and \eW are
620 \& defined by current locale
621 \& NBOUNDU no Match "" at any non\-boundary of a given
622 \& type using using /u rules.
623 \& NBOUNDA no Match "" betweeen any \ew\ew or \eW\eW, where
624 \& \ew is [_a\-zA\-Z0\-9]
625 \&
626 \& # [Special] alternatives:
627 \& REG_ANY no Match any one character (except newline).
628 \& SANY no Match any one character.
629 \& ANYOF sv Match character in (or not in) this class,
630 \& charclass single char match only
631 \& ANYOFD sv Like ANYOF, but /d is in effect
632 \& charclass
633 \& ANYOFL sv Like ANYOF, but /l is in effect
634 \& charclass
635 \& ANYOFPOSIXL sv Like ANYOFL, but matches [[:posix:]]
636 \& charclass_ classes
637 \& posixl
638 \&
639 \& ANYOFH sv 1 Like ANYOF, but only has "High" matches,
640 \& none in the bitmap; the flags field
641 \& contains the lowest matchable UTF\-8 start
642 \& byte
643 \& ANYOFHb sv 1 Like ANYOFH, but all matches share the same
644 \& UTF\-8 start byte, given in the flags field
645 \& ANYOFHr sv 1 Like ANYOFH, but the flags field contains
646 \& packed bounds for all matchable UTF\-8 start
647 \& bytes.
648 \& ANYOFHs sv:str 1 Like ANYOFHb, but has a string field that
649 \& gives the leading matchable UTF\-8 bytes;
650 \& flags field is len
651 \& ANYOFR packed 1 Matches any character in the range given by
652 \& its packed args: upper 12 bits is the max
653 \& delta from the base lower 20; the flags
654 \& field contains the lowest matchable UTF\-8
655 \& start byte
656 \& ANYOFRb packed 1 Like ANYOFR, but all matches share the same
657 \& UTF\-8 start byte, given in the flags field
658 \&
659 \& ANYOFHbbm none bbm Like ANYOFHb, but only for 2\-byte UTF\-8
660 \& characters; uses a bitmap to match the
661 \& continuation byte
662 \&
663 \& ANYOFM byte 1 Like ANYOF, but matches an invariant byte
664 \& as determined by the mask and arg
665 \& NANYOFM byte 1 complement of ANYOFM
666 \&
667 \& # POSIX Character Classes:
668 \& POSIXD none Some [[:class:]] under /d; the FLAGS field
669 \& gives which one
670 \& POSIXL none Some [[:class:]] under /l; the FLAGS field
671 \& gives which one
672 \& POSIXU none Some [[:class:]] under /u; the FLAGS field
673 \& gives which one
674 \& POSIXA none Some [[:class:]] under /a; the FLAGS field
675 \& gives which one
676 \& NPOSIXD none complement of POSIXD, [[:^class:]]
677 \& NPOSIXL none complement of POSIXL, [[:^class:]]
678 \& NPOSIXU none complement of POSIXU, [[:^class:]]
679 \& NPOSIXA none complement of POSIXA, [[:^class:]]
680 \&
681 \& CLUMP no Match any extended grapheme cluster
682 \& sequence
683 \&
684 \& # Alternation
685 \&
686 \& # BRANCH The set of branches constituting a single choice are
687 \& # hooked together with their "next" pointers, since
688 \& # precedence prevents anything being concatenated to
689 \& # any individual branch. The "next" pointer of the last
690 \& # BRANCH in a choice points to the thing following the
691 \& # whole choice. This is also where the final "next"
692 \& # pointer of each individual branch points; each branch
693 \& # starts with the operand node of a BRANCH node.
694 \& #
695 \& BRANCH node 1 Match this alternative, or the next...
696 \&
697 \& # Literals
698 \&
699 \& EXACT str Match this string (flags field is the
700 \& length).
701 \&
702 \& # In a long string node, the U32 argument is the length, and is
703 \& # immediately followed by the string.
704 \& LEXACT len:str 1 Match this long string (preceded by length;
705 \& flags unused).
706 \& EXACTL str Like EXACT, but /l is in effect (used so
707 \& locale\-related warnings can be checked for)
708 \& EXACTF str Like EXACT, but match using /id rules;
709 \& (string not UTF\-8, ASCII folded; non\-ASCII
710 \& not)
711 \& EXACTFL str Like EXACT, but match using /il rules;
712 \& (string not likely to be folded)
713 \& EXACTFU str Like EXACT, but match using /iu rules;
714 \& (string folded)
715 \&
716 \& EXACTFAA str Like EXACT, but match using /iaa rules;
717 \& (string folded except MICRO in non\-UTF8
718 \& patterns; doesn\*(Aqt contain SHARP S unless
719 \& UTF\-8; folded length <= unfolded)
720 \& EXACTFAA_NO_TRIE str Like EXACTFAA, (string not UTF\-8, folded
721 \& except: MICRO, SHARP S; folded length <=
722 \& unfolded, not currently trie\-able)
723 \&
724 \& EXACTFUP str Like EXACT, but match using /iu rules;
725 \& (string not UTF\-8, folded except MICRO:
726 \& hence Problematic)
727 \&
728 \& EXACTFLU8 str Like EXACTFU, but use /il, UTF\-8, (string
729 \& is folded, and everything in it is above
730 \& 255
731 \& EXACT_REQ8 str Like EXACT, but only UTF\-8 encoded targets
732 \& can match
733 \& LEXACT_REQ8 len:str 1 Like LEXACT, but only UTF\-8 encoded targets
734 \& can match
735 \& EXACTFU_REQ8 str Like EXACTFU, but only UTF\-8 encoded
736 \& targets can match
737 \&
738 \& EXACTFU_S_EDGE str /di rules, but nothing in it precludes /ui,
739 \& except begins and/or ends with [Ss];
740 \& (string not UTF\-8; compile\-time only)
741 \&
742 \& # New charclass like patterns
743 \& LNBREAK none generic newline pattern
744 \&
745 \& # Trie Related
746 \&
747 \& # Behave the same as A|LIST|OF|WORDS would. The \*(Aq..C\*(Aq variants
748 \& # have inline charclass data (ascii only), the \*(AqC\*(Aq store it in the
749 \& # structure.
750 \&
751 \& TRIE trie 1 Match many EXACT(F[ALU]?)? at once.
752 \& flags==type
753 \& TRIEC trie Same as TRIE, but with embedded charclass
754 \& charclass data
755 \&
756 \& AHOCORASICK trie 1 Aho Corasick stclass. flags==type
757 \& AHOCORASICKC trie Same as AHOCORASICK, but with embedded
758 \& charclass charclass data
759 \&
760 \& # Do nothing types
761 \&
762 \& NOTHING no Match empty string.
763 \& # A variant of above which delimits a group, thus stops optimizations
764 \& TAIL no Match empty string. Can jump here from
765 \& outside.
766 \&
767 \& # Loops
768 \&
769 \& # STAR,PLUS \*(Aq?\*(Aq, and complex \*(Aq*\*(Aq and \*(Aq+\*(Aq, are implemented as
770 \& # circular BRANCH structures. Simple cases
771 \& # (one character per match) are implemented with STAR
772 \& # and PLUS for speed and to minimize recursive plunges.
773 \& #
774 \& STAR node Match this (simple) thing 0 or more times:
775 \& /A{0,}B/ where A is width 1 char
776 \& PLUS node Match this (simple) thing 1 or more times:
777 \& /A{1,}B/ where A is width 1 char
778 \&
779 \& CURLY sv 3 Match this (simple) thing {n,m} times:
780 \& /A{m,n}B/ where A is width 1 char
781 \& CURLYN no 3 Capture next\-after\-this simple thing:
782 \& /(A){m,n}B/ where A is width 1 char
783 \& CURLYM no 3 Capture this medium\-complex thing {n,m}
784 \& times: /(A){m,n}B/ where A is fixed\-length
785 \& CURLYX sv 3 Match/Capture this complex thing {n,m}
786 \& times.
787 \&
788 \& # This terminator creates a loop structure for CURLYX
789 \& WHILEM no Do curly processing and see if rest
790 \& matches.
791 \&
792 \& # Buffer related
793 \&
794 \& # OPEN,CLOSE,GROUPP ...are numbered at compile time.
795 \& OPEN num 1 Mark this point in input as start of #n.
796 \& CLOSE num 1 Close corresponding OPEN of #n.
797 \& SROPEN none Same as OPEN, but for script run
798 \& SRCLOSE none Close preceding SROPEN
799 \&
800 \& REF num 2 Match some already matched string
801 \& REFF num 2 Match already matched string, using /di
802 \& rules.
803 \& REFFL num 2 Match already matched string, using /li
804 \& rules.
805 \& REFFU num 2 Match already matched string, usng /ui.
806 \& REFFA num 2 Match already matched string, using /aai
807 \& rules.
808 \&
809 \& # Named references. Code in regcomp.c assumes that these all are after
810 \& # the numbered references
811 \& REFN no\-sv 2 Match some already matched string
812 \& REFFN no\-sv 2 Match already matched string, using /di
813 \& rules.
814 \& REFFLN no\-sv 2 Match already matched string, using /li
815 \& rules.
816 \& REFFUN num 2 Match already matched string, using /ui
817 \& rules.
818 \& REFFAN num 2 Match already matched string, using /aai
819 \& rules.
820 \&
821 \& # Support for long RE
822 \& LONGJMP off 1 1 Jump far away.
823 \& BRANCHJ off 2 1 BRANCH with long offset.
824 \&
825 \& # Special Case Regops
826 \& IFMATCH off 1 1 Succeeds if the following matches; non\-zero
827 \& flags "f", next_off "o" means lookbehind
828 \& assertion starting "f..(f\-o)" characters
829 \& before current
830 \& UNLESSM off 1 1 Fails if the following matches; non\-zero
831 \& flags "f", next_off "o" means lookbehind
832 \& assertion starting "f..(f\-o)" characters
833 \& before current
834 \& SUSPEND off 1 1 "Independent" sub\-RE.
835 \& IFTHEN off 1 1 Switch, should be preceded by switcher.
836 \& GROUPP num 1 Whether the group matched.
837 \&
838 \& # The heavy worker
839 \&
840 \& EVAL evl/flags Execute some Perl code.
841 \& 2
842 \&
843 \& # Modifiers
844 \&
845 \& MINMOD no Next operator is not greedy.
846 \& LOGICAL no Next opcode should set the flag only.
847 \&
848 \& # This is not used yet
849 \& RENUM off 1 1 Group with independently numbered parens.
850 \&
851 \& # Regex Subroutines
852 \& GOSUB num/ofs 2 recurse to paren arg1 at (signed) ofs arg2
853 \&
854 \& # Special conditionals
855 \& GROUPPN no\-sv 1 Whether the group matched.
856 \& INSUBP num 1 Whether we are in a specific recurse.
857 \& DEFINEP none 1 Never execute directly.
858 \&
859 \& # Backtracking Verbs
860 \& ENDLIKE none Used only for the type field of verbs
861 \& OPFAIL no\-sv 1 Same as (?!), but with verb arg
862 \& ACCEPT no\-sv/num Accepts the current matched string, with
863 \& 2 verbar
864 \&
865 \& # Verbs With Arguments
866 \& VERB no\-sv 1 Used only for the type field of verbs
867 \& PRUNE no\-sv 1 Pattern fails at this startpoint if no\-
868 \& backtracking through this
869 \& MARKPOINT no\-sv 1 Push the current location for rollback by
870 \& cut.
871 \& SKIP no\-sv 1 On failure skip forward (to the mark)
872 \& before retrying
873 \& COMMIT no\-sv 1 Pattern fails outright if backtracking
874 \& through this
875 \& CUTGROUP no\-sv 1 On failure go to the next alternation in
876 \& the group
877 \&
878 \& # Control what to keep in $&.
879 \& KEEPS no $& begins here.
880 \&
881 \& # Validate that lookbehind IFMATCH and UNLESSM end at the right place
882 \& LOOKBEHIND_END no Return from lookbehind (IFMATCH/UNLESSM)
883 \& and validate position
884 \&
885 \& # SPECIAL REGOPS
886 \&
887 \& # This is not really a node, but an optimized away piece of a "long"
888 \& # node. To simplify debugging output, we mark it as if it were a node
889 \& OPTIMIZED off Placeholder for dump.
890 \&
891 \& # Special opcode with the property that no opcode in a compiled program
892 \& # will ever be of this type. Thus it can be used as a flag value that
893 \& # no other opcode has been seen. END is used similarly, in that an END
894 \& # node cant be optimized. So END implies "unoptimizable" and PSEUDO
895 \& # mean "not seen anything to optimize yet".
896 \& PSEUDO off Pseudo opcode for internal use.
897 \&
898 \& REGEX_SET depth p Regex set, temporary node used in pre\-
899 \& optimization compilation
900 .Ve
901 .PP
902 Following the optimizer information is a dump of the offset/length
903 table, here split across several lines:
904 .PP
905 .Vb 5
906 \& Offsets: [45]
907 \& 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
908 \& 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
909 \& 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
910 \& 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
911 .Ve
912 .PP
913 The first line here indicates that the offset/length table contains 45
914 entries. Each entry is a pair of integers, denoted by \f(CW\*(C`offset[length]\*(C'\fR.
915 Entries are numbered starting with 1, so entry #1 here is \f(CW\*(C`1[4]\*(C'\fR and
916 entry #12 is \f(CW\*(C`5[1]\*(C'\fR. \f(CW\*(C`1[4]\*(C'\fR indicates that the node labeled \f(CW\*(C`1:\*(C'\fR
917 (the \f(CW\*(C`1: ANYOF[bc]\*(C'\fR) begins at character position 1 in the
918 pre-compiled form of the regex, and has a length of 4 characters.
919 \&\f(CW\*(C`5[1]\*(C'\fR in position 12
920 indicates that the node labeled \f(CW\*(C`12:\*(C'\fR
921 (the \f(CW\*(C`12: EXACT <d>\*(C'\fR) begins at character position 5 in the
922 pre-compiled form of the regex, and has a length of 1 character.
923 \&\f(CW\*(C`12[1]\*(C'\fR in position 14
924 indicates that the node labeled \f(CW\*(C`14:\*(C'\fR
925 (the \f(CW\*(C`14: CURLYX[0] {1,32767}\*(C'\fR) begins at character position 12 in the
926 pre-compiled form of the regex, and has a length of 1 character\-\-\-that
927 is, it corresponds to the \f(CW\*(C`+\*(C'\fR symbol in the precompiled regex.
928 .PP
929 \&\f(CW\*(C`0[0]\*(C'\fR items indicate that there is no corresponding node.
930 .SS "Run-time Output"
931 .IX Subsection "Run-time Output"
932 First of all, when doing a match, one may get no run-time output even
933 if debugging is enabled. This means that the regex engine was never
934 entered and that all of the job was therefore done by the optimizer.
935 .PP
936 If the regex engine was entered, the output may look like this:
937 .PP
938 .Vb 10
939 \& Matching \*(Aq[bc]d(ef*g)+h[ij]k$\*(Aq against \*(Aqabcdefg_\|_gh_\|_\*(Aq
940 \& Setting an EVAL scope, savestack=3
941 \& 2 <ab> <cdefg_\|_gh_> | 1: ANYOF
942 \& 3 <abc> <defg_\|_gh_> | 11: EXACT <d>
943 \& 4 <abcd> <efg_\|_gh_> | 13: CURLYX {1,32767}
944 \& 4 <abcd> <efg_\|_gh_> | 26: WHILEM
945 \& 0 out of 1..32767 cc=effff31c
946 \& 4 <abcd> <efg_\|_gh_> | 15: OPEN1
947 \& 4 <abcd> <efg_\|_gh_> | 17: EXACT <e>
948 \& 5 <abcde> <fg_\|_gh_> | 19: STAR
949 \& EXACT <f> can match 1 times out of 32767...
950 \& Setting an EVAL scope, savestack=3
951 \& 6 <bcdef> <g_\|_gh_\|_> | 22: EXACT <g>
952 \& 7 <bcdefg> <_\|_gh_\|_> | 24: CLOSE1
953 \& 7 <bcdefg> <_\|_gh_\|_> | 26: WHILEM
954 \& 1 out of 1..32767 cc=effff31c
955 \& Setting an EVAL scope, savestack=12
956 \& 7 <bcdefg> <_\|_gh_\|_> | 15: OPEN1
957 \& 7 <bcdefg> <_\|_gh_\|_> | 17: EXACT <e>
958 \& restoring \e1 to 4(4)..7
959 \& failed, try continuation...
960 \& 7 <bcdefg> <_\|_gh_\|_> | 27: NOTHING
961 \& 7 <bcdefg> <_\|_gh_\|_> | 28: EXACT <h>
962 \& failed...
963 \& failed...
964 .Ve
965 .PP
966 The most significant information in the output is about the particular \fInode\fR
967 of the compiled regex that is currently being tested against the target string.
968 The format of these lines is
969 .PP
970 \&\f(CW\*(C` \*(C'\fR\fISTRING-OFFSET\fR <\fIPRE-STRING\fR> <\fIPOST-STRING\fR> |\fIID\fR: \fITYPE\fR
971 .PP
972 The \fITYPE\fR info is indented with respect to the backtracking level.
973 Other incidental information appears interspersed within.
974 .SH "Debugging Perl Memory Usage"
975 .IX Header "Debugging Perl Memory Usage"
976 Perl is a profligate wastrel when it comes to memory use. There
977 is a saying that to estimate memory usage of Perl, assume a reasonable
978 algorithm for memory allocation, multiply that estimate by 10, and
979 while you still may miss the mark, at least you won't be quite so
980 astonished. This is not absolutely true, but may provide a good
981 grasp of what happens.
982 .PP
983 Assume that an integer cannot take less than 20 bytes of memory, a
984 float cannot take less than 24 bytes, a string cannot take less
985 than 32 bytes (all these examples assume 32\-bit architectures, the
986 result are quite a bit worse on 64\-bit architectures). If a variable
987 is accessed in two of three different ways (which require an integer,
988 a float, or a string), the memory footprint may increase yet another
989 20 bytes. A sloppy \fBmalloc\fR\|(3) implementation can inflate these
990 numbers dramatically.
991 .PP
992 On the opposite end of the scale, a declaration like
993 .PP
994 .Vb 1
995 \& sub foo;
996 .Ve
997 .PP
998 may take up to 500 bytes of memory, depending on which release of Perl
999 you're running.
1000 .PP
1001 Anecdotal estimates of source-to-compiled code bloat suggest an
1002 eightfold increase. This means that the compiled form of reasonable
1003 (normally commented, properly indented etc.) code will take
1004 about eight times more space in memory than the code took
1005 on disk.
1006 .PP
1007 The \fB\-DL\fR command-line switch is obsolete since circa Perl 5.6.0
1008 (it was available only if Perl was built with \f(CW\*(C`\-DDEBUGGING\*(C'\fR).
1009 The switch was used to track Perl's memory allocations and possible
1010 memory leaks. These days the use of malloc debugging tools like
1011 \&\fIPurify\fR or \fIvalgrind\fR is suggested instead. See also
1012 "PERL_MEM_LOG" in perlhacktips.
1013 .PP
1014 One way to find out how much memory is being used by Perl data
1015 structures is to install the Devel::Size module from CPAN: it gives
1016 you the minimum number of bytes required to store a particular data
1017 structure. Please be mindful of the difference between the \fBsize()\fR
1018 and \fBtotal_size()\fR.
1019 .PP
1020 If Perl has been compiled using Perl's malloc you can analyze Perl
1021 memory usage by setting \f(CW$ENV\fR{PERL_DEBUG_MSTATS}.
1022 .ie n .SS "Using $ENV{PERL_DEBUG_MSTATS}"
1023 .el .SS "Using \f(CW$ENV{PERL_DEBUG_MSTATS}\fP"
1024 .IX Subsection "Using $ENV{PERL_DEBUG_MSTATS}"
1025 If your perl is using Perl's \fBmalloc()\fR and was compiled with the
1026 necessary switches (this is the default), then it will print memory
1027 usage statistics after compiling your code when \f(CW\*(C`$ENV{PERL_DEBUG_MSTATS}
1028 > 1\*(C'\fR, and before termination of the program when \f(CW\*(C`$ENV{PERL_DEBUG_MSTATS} >= 1\*(C'\fR. The report format is similar to
1029 the following example:
1030 .PP
1031 .Vb 10
1032 \& $ PERL_DEBUG_MSTATS=2 perl \-e "require Carp"
1033 \& Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
1034 \& 14216 free: 130 117 28 7 9 0 2 2 1 0 0
1035 \& 437 61 36 0 5
1036 \& 60924 used: 125 137 161 55 7 8 6 16 2 0 1
1037 \& 74 109 304 84 20
1038 \& Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
1039 \& Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
1040 \& 30888 free: 245 78 85 13 6 2 1 3 2 0 1
1041 \& 315 162 39 42 11
1042 \& 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
1043 \& 196 178 1066 798 39
1044 \& Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
1045 .Ve
1046 .PP
1047 It is possible to ask for such a statistic at arbitrary points in
1048 your execution using the \fBmstat()\fR function out of the standard
1049 Devel::Peek module.
1050 .PP
1051 Here is some explanation of that format:
1052 .ie n .IP """buckets SMALLEST(APPROX)..GREATEST(APPROX)""" 4
1053 .el .IP "\f(CWbuckets SMALLEST(APPROX)..GREATEST(APPROX)\fR" 4
1054 .IX Item "buckets SMALLEST(APPROX)..GREATEST(APPROX)"
1055 Perl's \fBmalloc()\fR uses bucketed allocations. Every request is rounded
1056 up to the closest bucket size available, and a bucket is taken from
1057 the pool of buckets of that size.
1058 .Sp
1059 The line above describes the limits of buckets currently in use.
1060 Each bucket has two sizes: memory footprint and the maximal size
1061 of user data that can fit into this bucket. Suppose in the above
1062 example that the smallest bucket were size 4. The biggest bucket
1063 would have usable size 8188, and the memory footprint would be 8192.
1064 .Sp
1065 In a Perl built for debugging, some buckets may have negative usable
1066 size. This means that these buckets cannot (and will not) be used.
1067 For larger buckets, the memory footprint may be one page greater
1068 than a power of 2. If so, the corresponding power of two is
1069 printed in the \f(CW\*(C`APPROX\*(C'\fR field above.
1070 .IP Free/Used 4
1071 .IX Item "Free/Used"
1072 The 1 or 2 rows of numbers following that correspond to the number
1073 of buckets of each size between \f(CW\*(C`SMALLEST\*(C'\fR and \f(CW\*(C`GREATEST\*(C'\fR. In
1074 the first row, the sizes (memory footprints) of buckets are powers
1075 of two\-\-or possibly one page greater. In the second row, if present,
1076 the memory footprints of the buckets are between the memory footprints
1077 of two buckets "above".
1078 .Sp
1079 For example, suppose under the previous example, the memory footprints
1080 were
1081 .Sp
1082 .Vb 2
1083 \& free: 8 16 32 64 128 256 512 1024 2048 4096 8192
1084 \& 4 12 24 48 80
1085 .Ve
1086 .Sp
1087 With a non\-\f(CW\*(C`DEBUGGING\*(C'\fR perl, the buckets starting from \f(CW128\fR have
1088 a 4\-byte overhead, and thus an 8192\-long bucket may take up to
1089 8188\-byte allocations.
1090 .ie n .IP """Total sbrk(): SBRKed/SBRKs:CONTINUOUS""" 4
1091 .el .IP "\f(CWTotal sbrk(): SBRKed/SBRKs:CONTINUOUS\fR" 4
1092 .IX Item "Total sbrk(): SBRKed/SBRKs:CONTINUOUS"
1093 The first two fields give the total amount of memory perl \fBsbrk\fR\|(2)ed
1094 (ess-broken? :\-) and number of \fBsbrk\fR\|(2)s used. The third number is
1095 what perl thinks about continuity of returned chunks. So long as
1096 this number is positive, \fBmalloc()\fR will assume that it is probable
1097 that \fBsbrk\fR\|(2) will provide continuous memory.
1098 .Sp
1099 Memory allocated by external libraries is not counted.
1100 .ie n .IP """pad: 0""" 4
1101 .el .IP "\f(CWpad: 0\fR" 4
1102 .IX Item "pad: 0"
1103 The amount of \fBsbrk\fR\|(2)ed memory needed to keep buckets aligned.
1104 .ie n .IP """heads: 2192""" 4
1105 .el .IP "\f(CWheads: 2192\fR" 4
1106 .IX Item "heads: 2192"
1107 Although memory overhead of bigger buckets is kept inside the bucket, for
1108 smaller buckets, it is kept in separate areas. This field gives the
1109 total size of these areas.
1110 .ie n .IP """chain: 0""" 4
1111 .el .IP "\f(CWchain: 0\fR" 4
1112 .IX Item "chain: 0"
1113 \&\fBmalloc()\fR may want to subdivide a bigger bucket into smaller buckets.
1114 If only a part of the deceased bucket is left unsubdivided, the rest
1115 is kept as an element of a linked list. This field gives the total
1116 size of these chunks.
1117 .ie n .IP """tail: 6144""" 4
1118 .el .IP "\f(CWtail: 6144\fR" 4
1119 .IX Item "tail: 6144"
1120 To minimize the number of \fBsbrk\fR\|(2)s, \fBmalloc()\fR asks for more memory. This
1121 field gives the size of the yet unused part, which is \fBsbrk\fR\|(2)ed, but
1122 never touched.
1123 .SH "SEE ALSO"
1124 .IX Header "SEE ALSO"
1125 perldebug,
1126 perl5db.pl,
1127 perlguts,
1128 perlrun,
1129 re,
1130 and
1131 Devel::DProf.