| 1 |
_ _ ____ _ |
| 2 |
___| | | | _ \| | |
| 3 |
/ __| | | | |_) | | |
| 4 |
| (__| |_| | _ <| |___ |
| 5 |
\___|\___/|_| \_\_____| |
| 6 |
|
| 7 |
Known Bugs |
| 8 |
|
| 9 |
These are problems and bugs known to exist at the time of this release. Feel |
| 10 |
free to join in and help us correct one or more of these! Also be sure to |
| 11 |
check the changelog of the current development status, as one or more of these |
| 12 |
problems may have been fixed or changed somewhat since this was written! |
| 13 |
|
| 14 |
1. HTTP |
| 15 |
1.2 Multiple methods in a single WWW-Authenticate: header |
| 16 |
1.3 STARTTRANSFER time is wrong for HTTP POSTs |
| 17 |
1.4 multipart formposts file name encoding |
| 18 |
1.5 Expect-100 meets 417 |
| 19 |
1.6 Unnecessary close when 401 received waiting for 100 |
| 20 |
1.7 Deflate error after all content was received |
| 21 |
1.8 DoH isn't used for all name resolves when enabled |
| 22 |
1.9 HTTP/2 frames while in the connection pool kill reuse |
| 23 |
1.11 CURLOPT_SEEKFUNCTION not called with CURLFORM_STREAM |
| 24 |
|
| 25 |
2. TLS |
| 26 |
2.1 CURLINFO_SSL_VERIFYRESULT has limited support |
| 27 |
2.2 DER in keychain |
| 28 |
2.3 GnuTLS backend skips really long certificate fields |
| 29 |
2.4 DarwinSSL won't import PKCS#12 client certificates without a password |
| 30 |
2.5 Client cert handling with Issuer DN differs between backends |
| 31 |
2.6 CURL_GLOBAL_SSL |
| 32 |
2.7 Client cert (MTLS) issues with Schannel |
| 33 |
2.8 Schannel disable CURLOPT_SSL_VERIFYPEER and verify hostname |
| 34 |
2.9 TLS session cache doesn't work with TFO |
| 35 |
|
| 36 |
3. Email protocols |
| 37 |
3.1 IMAP SEARCH ALL truncated response |
| 38 |
3.2 No disconnect command |
| 39 |
3.3 POP3 expects "CRLF.CRLF" eob for some single-line responses |
| 40 |
3.4 AUTH PLAIN for SMTP is not working on all servers |
| 41 |
|
| 42 |
4. Command line |
| 43 |
4.1 -J and -O with %-encoded file names |
| 44 |
4.2 -J with -C - fails |
| 45 |
4.3 --retry and transfer timeouts |
| 46 |
4.4 --upload-file . hang if delay in STDIN |
| 47 |
4.5 Improve --data-urlencode space encoding |
| 48 |
|
| 49 |
5. Build and portability issues |
| 50 |
5.1 USE_UNIX_SOCKETS on Windows |
| 51 |
5.2 curl-config --libs contains private details |
| 52 |
5.3 curl compiled on OSX 10.13 failed to run on OSX 10.10 |
| 53 |
5.4 Cannot compile against a static build of OpenLDAP |
| 54 |
5.5 can't handle Unicode arguments in Windows |
| 55 |
5.6 cmake support gaps |
| 56 |
5.7 Visual Studio project gaps |
| 57 |
5.8 configure finding libs in wrong directory |
| 58 |
5.9 Utilize Requires.private directives in libcurl.pc |
| 59 |
5.10 IDN tests failing on Windows / MSYS2 |
| 60 |
|
| 61 |
6. Authentication |
| 62 |
6.1 NTLM authentication and unicode |
| 63 |
6.2 MIT Kerberos for Windows build |
| 64 |
6.3 NTLM in system context uses wrong name |
| 65 |
6.4 Negotiate and Kerberos V5 need a fake user name |
| 66 |
6.5 NTLM doesn't support password with § character |
| 67 |
6.6 libcurl can fail to try alternatives with --proxy-any |
| 68 |
6.7 Don't clear digest for single realm |
| 69 |
|
| 70 |
7. FTP |
| 71 |
7.1 FTP without or slow 220 response |
| 72 |
7.2 FTP with CONNECT and slow server |
| 73 |
7.3 FTP with NOBODY and FAILONERROR |
| 74 |
7.4 FTP with ACCT |
| 75 |
7.5 ASCII FTP |
| 76 |
7.6 FTP with NULs in URL parts |
| 77 |
7.7 FTP and empty path parts in the URL |
| 78 |
7.8 Premature transfer end but healthy control channel |
| 79 |
7.9 Passive transfer tries only one IP address |
| 80 |
7.10 Stick to same family over SOCKS proxy |
| 81 |
|
| 82 |
8. TELNET |
| 83 |
8.1 TELNET and time limitations don't work |
| 84 |
8.2 Microsoft telnet server |
| 85 |
|
| 86 |
9. SFTP and SCP |
| 87 |
9.1 SFTP doesn't do CURLOPT_POSTQUOTE correct |
| 88 |
|
| 89 |
10. SOCKS |
| 90 |
10.3 FTPS over SOCKS |
| 91 |
10.4 active FTP over a SOCKS |
| 92 |
|
| 93 |
11. Internals |
| 94 |
11.1 Curl leaks .onion hostnames in DNS |
| 95 |
11.2 error buffer not set if connection to multiple addresses fails |
| 96 |
11.3 c-ares deviates from stock resolver on http://1346569778 |
| 97 |
11.4 HTTP test server 'connection-monitor' problems |
| 98 |
11.5 Connection information when using TCP Fast Open |
| 99 |
11.6 slow connect to localhost on Windows |
| 100 |
11.7 signal-based resolver timeouts |
| 101 |
|
| 102 |
12. LDAP and OpenLDAP |
| 103 |
12.1 OpenLDAP hangs after returning results |
| 104 |
12.2 LDAP on Windows does authentication wrong? |
| 105 |
12.3 LDAP on Windows doesn't work |
| 106 |
|
| 107 |
13. TCP/IP |
| 108 |
13.1 --interface for ipv6 binds to unusable IP address |
| 109 |
|
| 110 |
14 DICT |
| 111 |
14.1 DICT responses show the underlying protocol |
| 112 |
|
| 113 |
============================================================================== |
| 114 |
|
| 115 |
1. HTTP |
| 116 |
|
| 117 |
1.2 Multiple methods in a single WWW-Authenticate: header |
| 118 |
|
| 119 |
The HTTP responses headers WWW-Authenticate: can provide information about |
| 120 |
multiple authentication methods as multiple headers or as several methods |
| 121 |
within a single header. The latter way, several methods in the same physical |
| 122 |
line, is not supported by libcurl's parser. (For no good reason.) |
| 123 |
|
| 124 |
1.3 STARTTRANSFER time is wrong for HTTP POSTs |
| 125 |
|
| 126 |
Wrong STARTTRANSFER timer accounting for POST requests Timer works fine with |
| 127 |
GET requests, but while using POST the time for CURLINFO_STARTTRANSFER_TIME |
| 128 |
is wrong. While using POST CURLINFO_STARTTRANSFER_TIME minus |
| 129 |
CURLINFO_PRETRANSFER_TIME is near to zero every time. |
| 130 |
|
| 131 |
https://github.com/curl/curl/issues/218 |
| 132 |
https://curl.haxx.se/bug/view.cgi?id=1213 |
| 133 |
|
| 134 |
1.4 multipart formposts file name encoding |
| 135 |
|
| 136 |
When creating multipart formposts. The file name part can be encoded with |
| 137 |
something beyond ascii but currently libcurl will only pass in the verbatim |
| 138 |
string the app provides. There are several browsers that already do this |
| 139 |
encoding. The key seems to be the updated draft to RFC2231: |
| 140 |
https://tools.ietf.org/html/draft-reschke-rfc2231-in-http-02 |
| 141 |
|
| 142 |
1.5 Expect-100 meets 417 |
| 143 |
|
| 144 |
If an upload using Expect: 100-continue receives an HTTP 417 response, it |
| 145 |
ought to be automatically resent without the Expect:. A workaround is for |
| 146 |
the client application to redo the transfer after disabling Expect:. |
| 147 |
https://curl.haxx.se/mail/archive-2008-02/0043.html |
| 148 |
|
| 149 |
1.6 Unnecessary close when 401 received waiting for 100 |
| 150 |
|
| 151 |
libcurl closes the connection if an HTTP 401 reply is received while it is |
| 152 |
waiting for the 100-continue response. |
| 153 |
https://curl.haxx.se/mail/lib-2008-08/0462.html |
| 154 |
|
| 155 |
1.7 Deflate error after all content was received |
| 156 |
|
| 157 |
There's a situation where we can get an error in a HTTP response that is |
| 158 |
compressed, when that error is detected after all the actual body contents |
| 159 |
have been received and delivered to the application. This is tricky, but is |
| 160 |
ultimately a broken server. |
| 161 |
|
| 162 |
See https://github.com/curl/curl/issues/2719 |
| 163 |
|
| 164 |
1.8 DoH isn't used for all name resolves when enabled |
| 165 |
|
| 166 |
Even if DoH is specified to be used, there are some name resolves that are |
| 167 |
done without it. This should be fixed. When the internal function |
| 168 |
`Curl_resolver_wait_resolv()` is called, it doesn't use DoH to complete the |
| 169 |
resolve as it otherwise should. |
| 170 |
|
| 171 |
See https://github.com/curl/curl/pull/3857 and |
| 172 |
https://github.com/curl/curl/pull/3850 |
| 173 |
|
| 174 |
1.9 HTTP/2 frames while in the connection pool kill reuse |
| 175 |
|
| 176 |
If the server sends HTTP/2 frames (like for example an HTTP/2 PING frame) to |
| 177 |
curl while the connection is held in curl's connection pool, the socket will |
| 178 |
be found readable when considered for reuse and that makes curl think it is |
| 179 |
dead and then it will be closed and a new connection gets created instead. |
| 180 |
|
| 181 |
This is *best* fixed by adding monitoring to connections while they are kept |
| 182 |
in the pool so that pings can be responded to appropriately. |
| 183 |
|
| 184 |
1.11 CURLOPT_SEEKFUNCTION not called with CURLFORM_STREAM |
| 185 |
|
| 186 |
I'm using libcurl to POST form data using a FILE* with the CURLFORM_STREAM |
| 187 |
option of curl_formadd(). I've noticed that if the connection drops at just |
| 188 |
the right time, the POST is reattempted without the data from the file. It |
| 189 |
seems like the file stream position isn't getting reset to the beginning of |
| 190 |
the file. I found the CURLOPT_SEEKFUNCTION option and set that with a |
| 191 |
function that performs an fseek() on the FILE*. However, setting that didn't |
| 192 |
seem to fix the issue or even get called. See |
| 193 |
https://github.com/curl/curl/issues/768 |
| 194 |
|
| 195 |
|
| 196 |
2. TLS |
| 197 |
|
| 198 |
2.1 CURLINFO_SSL_VERIFYRESULT has limited support |
| 199 |
|
| 200 |
CURLINFO_SSL_VERIFYRESULT is only implemented for the OpenSSL and NSS |
| 201 |
backends, so relying on this information in a generic app is flaky. |
| 202 |
|
| 203 |
2.2 DER in keychain |
| 204 |
|
| 205 |
Curl doesn't recognize certificates in DER format in keychain, but it works |
| 206 |
with PEM. https://curl.haxx.se/bug/view.cgi?id=1065 |
| 207 |
|
| 208 |
2.3 GnuTLS backend skips really long certificate fields |
| 209 |
|
| 210 |
libcurl calls gnutls_x509_crt_get_dn() with a fixed buffer size and if the |
| 211 |
field is too long in the cert, it'll just return an error and the field will |
| 212 |
be displayed blank. |
| 213 |
|
| 214 |
2.4 DarwinSSL won't import PKCS#12 client certificates without a password |
| 215 |
|
| 216 |
libcurl calls SecPKCS12Import with the PKCS#12 client certificate, but that |
| 217 |
function rejects certificates that do not have a password. |
| 218 |
https://github.com/curl/curl/issues/1308 |
| 219 |
|
| 220 |
2.5 Client cert handling with Issuer DN differs between backends |
| 221 |
|
| 222 |
When the specified client certificate doesn't match any of the |
| 223 |
server-specified DNs, the OpenSSL and GnuTLS backends behave differently. |
| 224 |
The github discussion may contain a solution. |
| 225 |
|
| 226 |
See https://github.com/curl/curl/issues/1411 |
| 227 |
|
| 228 |
2.6 CURL_GLOBAL_SSL |
| 229 |
|
| 230 |
Since libcurl 7.57.0, the flag CURL_GLOBAL_SSL is a no-op. The change was |
| 231 |
merged in https://github.com/curl/curl/commit/d661b0afb571a |
| 232 |
|
| 233 |
It was removed since it was |
| 234 |
|
| 235 |
A) never clear for applications on how to deal with init in the light of |
| 236 |
different SSL backends (the option was added back in the days when life |
| 237 |
was simpler) |
| 238 |
|
| 239 |
B) multissl introduced dynamic switching between SSL backends which |
| 240 |
emphasized (A) even more |
| 241 |
|
| 242 |
C) libcurl uses some TLS backend functionality even for non-TLS functions (to |
| 243 |
get "good" random) so applications trying to avoid the init for |
| 244 |
performance reasons would do wrong anyway |
| 245 |
|
| 246 |
D) never very carefully documented so all this mostly just happened to work |
| 247 |
for some users |
| 248 |
|
| 249 |
However, in spite of the problems with the feature, there were some users who |
| 250 |
apparently depended on this feature and who now claim libcurl is broken for |
| 251 |
them. The fix for this situation is not obvious as a downright revert of the |
| 252 |
patch is totally ruled out due to those reasons above. |
| 253 |
|
| 254 |
https://github.com/curl/curl/issues/2276 |
| 255 |
|
| 256 |
2.7 Client cert (MTLS) issues with Schannel |
| 257 |
|
| 258 |
See https://github.com/curl/curl/issues/3145 |
| 259 |
|
| 260 |
2.8 Schannel disable CURLOPT_SSL_VERIFYPEER and verify hostname |
| 261 |
|
| 262 |
This seems to be a limitation in the underlying Schannel API. |
| 263 |
|
| 264 |
https://github.com/curl/curl/issues/3284 |
| 265 |
|
| 266 |
2.9 TLS session cache doesn't work with TFO |
| 267 |
|
| 268 |
See https://github.com/curl/curl/issues/4301 |
| 269 |
|
| 270 |
3. Email protocols |
| 271 |
|
| 272 |
3.1 IMAP SEARCH ALL truncated response |
| 273 |
|
| 274 |
IMAP "SEARCH ALL" truncates output on large boxes. "A quick search of the |
| 275 |
code reveals that pingpong.c contains some truncation code, at line 408, when |
| 276 |
it deems the server response to be too large truncating it to 40 characters" |
| 277 |
https://curl.haxx.se/bug/view.cgi?id=1366 |
| 278 |
|
| 279 |
3.2 No disconnect command |
| 280 |
|
| 281 |
The disconnect commands (LOGOUT and QUIT) may not be sent by IMAP, POP3 and |
| 282 |
SMTP if a failure occurs during the authentication phase of a connection. |
| 283 |
|
| 284 |
3.3 POP3 expects "CRLF.CRLF" eob for some single-line responses |
| 285 |
|
| 286 |
You have to tell libcurl not to expect a body, when dealing with one line |
| 287 |
response commands. Please see the POP3 examples and test cases which show |
| 288 |
this for the NOOP and DELE commands. https://curl.haxx.se/bug/?i=740 |
| 289 |
|
| 290 |
3.4 AUTH PLAIN for SMTP is not working on all servers |
| 291 |
|
| 292 |
Specifying "--login-options AUTH=PLAIN" on the command line doesn't seem to |
| 293 |
work correctly. |
| 294 |
|
| 295 |
See https://github.com/curl/curl/issues/4080 |
| 296 |
|
| 297 |
4. Command line |
| 298 |
|
| 299 |
4.1 -J and -O with %-encoded file names |
| 300 |
|
| 301 |
-J/--remote-header-name doesn't decode %-encoded file names. RFC6266 details |
| 302 |
how it should be done. The can of worm is basically that we have no charset |
| 303 |
handling in curl and ascii >=128 is a challenge for us. Not to mention that |
| 304 |
decoding also means that we need to check for nastiness that is attempted, |
| 305 |
like "../" sequences and the like. Probably everything to the left of any |
| 306 |
embedded slashes should be cut off. |
| 307 |
https://curl.haxx.se/bug/view.cgi?id=1294 |
| 308 |
|
| 309 |
-O also doesn't decode %-encoded names, and while it has even less |
| 310 |
information about the charset involved the process is similar to the -J case. |
| 311 |
|
| 312 |
Note that we won't add decoding to -O without the user asking for it with |
| 313 |
some other means as well, since -O has always been documented to use the name |
| 314 |
exactly as specified in the URL. |
| 315 |
|
| 316 |
4.2 -J with -C - fails |
| 317 |
|
| 318 |
When using -J (with -O), automatically resumed downloading together with "-C |
| 319 |
-" fails. Without -J the same command line works! This happens because the |
| 320 |
resume logic is worked out before the target file name (and thus its |
| 321 |
pre-transfer size) has been figured out! |
| 322 |
https://curl.haxx.se/bug/view.cgi?id=1169 |
| 323 |
|
| 324 |
4.3 --retry and transfer timeouts |
| 325 |
|
| 326 |
If using --retry and the transfer timeouts (possibly due to using -m or |
| 327 |
-y/-Y) the next attempt doesn't resume the transfer properly from what was |
| 328 |
downloaded in the previous attempt but will truncate and restart at the |
| 329 |
original position where it was at before the previous failed attempt. See |
| 330 |
https://curl.haxx.se/mail/lib-2008-01/0080.html and Mandriva bug report |
| 331 |
https://qa.mandriva.com/show_bug.cgi?id=22565 |
| 332 |
|
| 333 |
4.4 --upload-file . hangs if delay in STDIN |
| 334 |
|
| 335 |
"(echo start; sleep 1; echo end) | curl --upload-file . http://mywebsite -vv" |
| 336 |
|
| 337 |
... causes a hang when it shouldn't. |
| 338 |
|
| 339 |
See https://github.com/curl/curl/issues/2051 |
| 340 |
|
| 341 |
4.5 Improve --data-urlencode space encoding |
| 342 |
|
| 343 |
ASCII space characters in --data-urlencode are currently encoded as %20 |
| 344 |
rather than +, which RFC 1866 says should be used. |
| 345 |
|
| 346 |
See https://github.com/curl/curl/issues/3229 |
| 347 |
|
| 348 |
5. Build and portability issues |
| 349 |
|
| 350 |
5.1 USE_UNIX_SOCKETS on Windows |
| 351 |
|
| 352 |
Due to incorrect CMake checks for the presence of the feature, it will never |
| 353 |
be enabled for windows in a cmake build. |
| 354 |
|
| 355 |
See https://github.com/curl/curl/issues/4040 |
| 356 |
|
| 357 |
5.2 curl-config --libs contains private details |
| 358 |
|
| 359 |
"curl-config --libs" will include details set in LDFLAGS when configure is |
| 360 |
run that might be needed only for building libcurl. Further, curl-config |
| 361 |
--cflags suffers from the same effects with CFLAGS/CPPFLAGS. |
| 362 |
|
| 363 |
5.3 curl compiled on OSX 10.13 failed to run on OSX 10.10 |
| 364 |
|
| 365 |
See https://github.com/curl/curl/issues/2905 |
| 366 |
|
| 367 |
5.4 Cannot compile against a static build of OpenLDAP |
| 368 |
|
| 369 |
See https://github.com/curl/curl/issues/2367 |
| 370 |
|
| 371 |
5.5 can't handle Unicode arguments in Windows |
| 372 |
|
| 373 |
If a URL or filename can't be encoded using the user's current codepage then |
| 374 |
it can only be encoded properly in the Unicode character set. Windows uses |
| 375 |
UTF-16 encoding for Unicode and stores it in wide characters, however curl |
| 376 |
and libcurl are not equipped for that at the moment. And, except for Cygwin, |
| 377 |
Windows can't use UTF-8 as a locale. |
| 378 |
|
| 379 |
https://curl.haxx.se/bug/?i=345 |
| 380 |
https://curl.haxx.se/bug/?i=731 |
| 381 |
|
| 382 |
5.6 cmake support gaps |
| 383 |
|
| 384 |
The cmake build setup lacks several features that the autoconf build |
| 385 |
offers. This includes: |
| 386 |
|
| 387 |
- use of correct soname for the shared library build |
| 388 |
|
| 389 |
- support for several TLS backends are missing |
| 390 |
|
| 391 |
- the unit tests cause link failures in regular non-static builds |
| 392 |
|
| 393 |
- no nghttp2 check |
| 394 |
|
| 395 |
- unusable tool_hugehelp.c with MinGW, see |
| 396 |
https://github.com/curl/curl/issues/3125 |
| 397 |
|
| 398 |
5.7 Visual Studio project gaps |
| 399 |
|
| 400 |
The Visual Studio projects lack some features that the autoconf and nmake |
| 401 |
builds offer, such as the following: |
| 402 |
|
| 403 |
- support for zlib and nghttp2 |
| 404 |
- use of static runtime libraries |
| 405 |
- add the test suite components |
| 406 |
|
| 407 |
In addition to this the following could be implemented: |
| 408 |
|
| 409 |
- support for other development IDEs |
| 410 |
- add PATH environment variables for third-party DLLs |
| 411 |
|
| 412 |
5.8 configure finding libs in wrong directory |
| 413 |
|
| 414 |
When the configure script checks for third-party libraries, it adds those |
| 415 |
directories to the LDFLAGS variable and then tries linking to see if it |
| 416 |
works. When successful, the found directory is kept in the LDFLAGS variable |
| 417 |
when the script continues to execute and do more tests and possibly check for |
| 418 |
more libraries. |
| 419 |
|
| 420 |
This can make subsequent checks for libraries wrongly detect another |
| 421 |
installation in a directory that was previously added to LDFLAGS by another |
| 422 |
library check! |
| 423 |
|
| 424 |
A possibly better way to do these checks would be to keep the pristine LDFLAGS |
| 425 |
even after successful checks and instead add those verified paths to a |
| 426 |
separate variable that only after all library checks have been performed gets |
| 427 |
appended to LDFLAGS. |
| 428 |
|
| 429 |
5.9 Utilize Requires.private directives in libcurl.pc |
| 430 |
|
| 431 |
https://github.com/curl/curl/issues/864 |
| 432 |
|
| 433 |
5.10 IDN tests failing on Windows / MSYS2 |
| 434 |
|
| 435 |
It seems like MSYS2 does some UTF-8-to-something-else conversion for Windows |
| 436 |
compatibility. |
| 437 |
|
| 438 |
https://github.com/curl/curl/issues/3747 |
| 439 |
|
| 440 |
6. Authentication |
| 441 |
|
| 442 |
6.1 NTLM authentication and unicode |
| 443 |
|
| 444 |
NTLM authentication involving unicode user name or password only works |
| 445 |
properly if built with UNICODE defined together with the WinSSL/Schannel |
| 446 |
backend. The original problem was mentioned in: |
| 447 |
https://curl.haxx.se/mail/lib-2009-10/0024.html |
| 448 |
https://curl.haxx.se/bug/view.cgi?id=896 |
| 449 |
|
| 450 |
The WinSSL/Schannel version verified to work as mentioned in |
| 451 |
https://curl.haxx.se/mail/lib-2012-07/0073.html |
| 452 |
|
| 453 |
6.2 MIT Kerberos for Windows build |
| 454 |
|
| 455 |
libcurl fails to build with MIT Kerberos for Windows (KfW) due to KfW's |
| 456 |
library header files exporting symbols/macros that should be kept private to |
| 457 |
the KfW library. See ticket #5601 at https://krbdev.mit.edu/rt/ |
| 458 |
|
| 459 |
6.3 NTLM in system context uses wrong name |
| 460 |
|
| 461 |
NTLM authentication using SSPI (on Windows) when (lib)curl is running in |
| 462 |
"system context" will make it use wrong(?) user name - at least when compared |
| 463 |
to what winhttp does. See https://curl.haxx.se/bug/view.cgi?id=535 |
| 464 |
|
| 465 |
6.4 Negotiate and Kerberos V5 need a fake user name |
| 466 |
|
| 467 |
In order to get Negotiate (SPNEGO) authentication to work in HTTP or Kerberos |
| 468 |
V5 in the e-mail protocols, you need to provide a (fake) user name (this |
| 469 |
concerns both curl and the lib) because the code wrongly only considers |
| 470 |
authentication if there's a user name provided by setting |
| 471 |
conn->bits.user_passwd in url.c https://curl.haxx.se/bug/view.cgi?id=440 How? |
| 472 |
https://curl.haxx.se/mail/lib-2004-08/0182.html A possible solution is to |
| 473 |
either modify this variable to be set or introduce a variable such as |
| 474 |
new conn->bits.want_authentication which is set when any of the authentication |
| 475 |
options are set. |
| 476 |
|
| 477 |
6.5 NTLM doesn't support password with § character |
| 478 |
|
| 479 |
https://github.com/curl/curl/issues/2120 |
| 480 |
|
| 481 |
6.6 libcurl can fail to try alternatives with --proxy-any |
| 482 |
|
| 483 |
When connecting via a proxy using --proxy-any, a failure to establish an |
| 484 |
authentication will cause libcurl to abort trying other options if the |
| 485 |
failed method has a higher preference than the alternatives. As an example, |
| 486 |
--proxy-any against a proxy which advertise Negotiate and NTLM, but which |
| 487 |
fails to set up Kerberos authentication won't proceed to try authentication |
| 488 |
using NTLM. |
| 489 |
|
| 490 |
https://github.com/curl/curl/issues/876 |
| 491 |
|
| 492 |
6.7 Don't clear digest for single realm |
| 493 |
|
| 494 |
https://github.com/curl/curl/issues/3267 |
| 495 |
|
| 496 |
7. FTP |
| 497 |
|
| 498 |
7.1 FTP without or slow 220 response |
| 499 |
|
| 500 |
If a connection is made to a FTP server but the server then just never sends |
| 501 |
the 220 response or otherwise is dead slow, libcurl will not acknowledge the |
| 502 |
connection timeout during that phase but only the "real" timeout - which may |
| 503 |
surprise users as it is probably considered to be the connect phase to most |
| 504 |
people. Brought up (and is being misunderstood) in: |
| 505 |
https://curl.haxx.se/bug/view.cgi?id=856 |
| 506 |
|
| 507 |
7.2 FTP with CONNECT and slow server |
| 508 |
|
| 509 |
When doing FTP over a socks proxy or CONNECT through HTTP proxy and the multi |
| 510 |
interface is used, libcurl will fail if the (passive) TCP connection for the |
| 511 |
data transfer isn't more or less instant as the code does not properly wait |
| 512 |
for the connect to be confirmed. See test case 564 for a first shot at a test |
| 513 |
case. |
| 514 |
|
| 515 |
7.3 FTP with NOBODY and FAILONERROR |
| 516 |
|
| 517 |
It seems sensible to be able to use CURLOPT_NOBODY and CURLOPT_FAILONERROR |
| 518 |
with FTP to detect if a file exists or not, but it is not working: |
| 519 |
https://curl.haxx.se/mail/lib-2008-07/0295.html |
| 520 |
|
| 521 |
7.4 FTP with ACCT |
| 522 |
|
| 523 |
When doing an operation over FTP that requires the ACCT command (but not when |
| 524 |
logging in), the operation will fail since libcurl doesn't detect this and |
| 525 |
thus fails to issue the correct command: |
| 526 |
https://curl.haxx.se/bug/view.cgi?id=635 |
| 527 |
|
| 528 |
7.5 ASCII FTP |
| 529 |
|
| 530 |
FTP ASCII transfers do not follow RFC959. They don't convert the data |
| 531 |
accordingly (not for sending nor for receiving). RFC 959 section 3.1.1.1 |
| 532 |
clearly describes how this should be done: |
| 533 |
|
| 534 |
The sender converts the data from an internal character representation to |
| 535 |
the standard 8-bit NVT-ASCII representation (see the Telnet |
| 536 |
specification). The receiver will convert the data from the standard |
| 537 |
form to his own internal form. |
| 538 |
|
| 539 |
Since 7.15.4 at least line endings are converted. |
| 540 |
|
| 541 |
7.6 FTP with NULs in URL parts |
| 542 |
|
| 543 |
FTP URLs passed to curl may contain NUL (0x00) in the RFC 1738 <user>, |
| 544 |
<password>, and <fpath> components, encoded as "%00". The problem is that |
| 545 |
curl_unescape does not detect this, but instead returns a shortened C string. |
| 546 |
From a strict FTP protocol standpoint, NUL is a valid character within RFC |
| 547 |
959 <string>, so the way to handle this correctly in curl would be to use a |
| 548 |
data structure other than a plain C string, one that can handle embedded NUL |
| 549 |
characters. From a practical standpoint, most FTP servers would not |
| 550 |
meaningfully support NUL characters within RFC 959 <string>, anyway (e.g., |
| 551 |
Unix pathnames may not contain NUL). |
| 552 |
|
| 553 |
7.7 FTP and empty path parts in the URL |
| 554 |
|
| 555 |
libcurl ignores empty path parts in FTP URLs, whereas RFC1738 states that |
| 556 |
such parts should be sent to the server as 'CWD ' (without an argument). The |
| 557 |
only exception to this rule, is that we knowingly break this if the empty |
| 558 |
part is first in the path, as then we use the double slashes to indicate that |
| 559 |
the user wants to reach the root dir (this exception SHALL remain even when |
| 560 |
this bug is fixed). |
| 561 |
|
| 562 |
7.8 Premature transfer end but healthy control channel |
| 563 |
|
| 564 |
When 'multi_done' is called before the transfer has been completed the normal |
| 565 |
way, it is considered a "premature" transfer end. In this situation, libcurl |
| 566 |
closes the connection assuming it doesn't know the state of the connection so |
| 567 |
it can't be reused for subsequent requests. |
| 568 |
|
| 569 |
With FTP however, this isn't necessarily true but there are a bunch of |
| 570 |
situations (listed in the ftp_done code) where it *could* keep the connection |
| 571 |
alive even in this situation - but the current code doesn't. Fixing this would |
| 572 |
allow libcurl to reuse FTP connections better. |
| 573 |
|
| 574 |
7.9 Passive transfer tries only one IP address |
| 575 |
|
| 576 |
When doing FTP operations through a proxy at localhost, the reported spotted |
| 577 |
that curl only tried to connect once to the proxy, while it had multiple |
| 578 |
addresses and a failed connect on one address should make it try the next. |
| 579 |
|
| 580 |
After switching to passive mode (EPSV), curl should try all IP addresses for |
| 581 |
"localhost". Currently it tries ::1, but it should also try 127.0.0.1. |
| 582 |
|
| 583 |
See https://github.com/curl/curl/issues/1508 |
| 584 |
|
| 585 |
7.10 Stick to same family over SOCKS proxy |
| 586 |
|
| 587 |
When asked to do FTP over a SOCKS proxy, it might connect to the proxy (and |
| 588 |
then subsequently to the remote server) using for example IPv4. When doing |
| 589 |
the second connection, curl should make sure that the second connection is |
| 590 |
using the same IP protocol version as the first connection did and not try |
| 591 |
others, since the remote server will only accept the same. |
| 592 |
|
| 593 |
See https://curl.haxx.se/mail/archive-2018-07/0000.html |
| 594 |
|
| 595 |
8. TELNET |
| 596 |
|
| 597 |
8.1 TELNET and time limitations don't work |
| 598 |
|
| 599 |
When using telnet, the time limitation options don't work. |
| 600 |
https://curl.haxx.se/bug/view.cgi?id=846 |
| 601 |
|
| 602 |
8.2 Microsoft telnet server |
| 603 |
|
| 604 |
There seems to be a problem when connecting to the Microsoft telnet server. |
| 605 |
https://curl.haxx.se/bug/view.cgi?id=649 |
| 606 |
|
| 607 |
|
| 608 |
9. SFTP and SCP |
| 609 |
|
| 610 |
9.1 SFTP doesn't do CURLOPT_POSTQUOTE correct |
| 611 |
|
| 612 |
When libcurl sends CURLOPT_POSTQUOTE commands when connected to a SFTP server |
| 613 |
using the multi interface, the commands are not being sent correctly and |
| 614 |
instead the connection is "cancelled" (the operation is considered done) |
| 615 |
prematurely. There is a half-baked (busy-looping) patch provided in the bug |
| 616 |
report but it cannot be accepted as-is. See |
| 617 |
https://curl.haxx.se/bug/view.cgi?id=748 |
| 618 |
|
| 619 |
|
| 620 |
10. SOCKS |
| 621 |
|
| 622 |
10.3 FTPS over SOCKS |
| 623 |
|
| 624 |
libcurl doesn't support FTPS over a SOCKS proxy. |
| 625 |
|
| 626 |
10.4 active FTP over a SOCKS |
| 627 |
|
| 628 |
libcurl doesn't support active FTP over a SOCKS proxy |
| 629 |
|
| 630 |
|
| 631 |
11. Internals |
| 632 |
|
| 633 |
11.1 Curl leaks .onion hostnames in DNS |
| 634 |
|
| 635 |
Curl sends DNS requests for hostnames with a .onion TLD. This leaks |
| 636 |
information about what the user is attempting to access, and violates this |
| 637 |
requirement of RFC7686: https://tools.ietf.org/html/rfc7686 |
| 638 |
|
| 639 |
Issue: https://github.com/curl/curl/issues/543 |
| 640 |
|
| 641 |
11.2 error buffer not set if connection to multiple addresses fails |
| 642 |
|
| 643 |
If you ask libcurl to resolve a hostname like example.com to IPv6 addresses |
| 644 |
only. But you only have IPv4 connectivity. libcurl will correctly fail with |
| 645 |
CURLE_COULDNT_CONNECT. But the error buffer set by CURLOPT_ERRORBUFFER |
| 646 |
remains empty. Issue: https://github.com/curl/curl/issues/544 |
| 647 |
|
| 648 |
11.3 c-ares deviates from stock resolver on http://1346569778 |
| 649 |
|
| 650 |
When using the socket resolvers, that URL becomes: |
| 651 |
|
| 652 |
* Rebuilt URL to: http://1346569778/ |
| 653 |
* Trying 80.67.6.50... |
| 654 |
|
| 655 |
but with c-ares it instead says "Could not resolve: 1346569778 (Domain name |
| 656 |
not found)" |
| 657 |
|
| 658 |
See https://github.com/curl/curl/issues/893 |
| 659 |
|
| 660 |
11.4 HTTP test server 'connection-monitor' problems |
| 661 |
|
| 662 |
The 'connection-monitor' feature of the sws HTTP test server doesn't work |
| 663 |
properly if some tests are run in unexpected order. Like 1509 and then 1525. |
| 664 |
|
| 665 |
See https://github.com/curl/curl/issues/868 |
| 666 |
|
| 667 |
11.5 Connection information when using TCP Fast Open |
| 668 |
|
| 669 |
CURLINFO_LOCAL_PORT (and possibly a few other) fails when TCP Fast Open is |
| 670 |
enabled. |
| 671 |
|
| 672 |
See https://github.com/curl/curl/issues/1332 and |
| 673 |
https://github.com/curl/curl/issues/4296 |
| 674 |
|
| 675 |
11.6 slow connect to localhost on Windows |
| 676 |
|
| 677 |
When connecting to "localhost" on Windows, curl will resolve the name for |
| 678 |
both ipv4 and ipv6 and try to connect to both happy eyeballs-style. Something |
| 679 |
in there does however make it take 200 milliseconds to succeed - which is the |
| 680 |
HAPPY_EYEBALLS_TIMEOUT define exactly. Lowering that define speeds up the |
| 681 |
connection, suggesting a problem in the HE handling. |
| 682 |
|
| 683 |
If we can *know* that we're talking to a local host, we should lower the |
| 684 |
happy eyeballs delay timeout for IPv6 (related: hardcode the "localhost" |
| 685 |
addresses, mentioned in TODO). Possibly we should reduce that delay for all. |
| 686 |
|
| 687 |
https://github.com/curl/curl/issues/2281 |
| 688 |
|
| 689 |
11.7 signal-based resolver timeouts |
| 690 |
|
| 691 |
libcurl built without an asynchronous resolver library uses alarm() to time |
| 692 |
out DNS lookups. When a timeout occurs, this causes libcurl to jump from the |
| 693 |
signal handler back into the library with a sigsetjmp, which effectively |
| 694 |
causes libcurl to continue running within the signal handler. This is |
| 695 |
non-portable and could cause problems on some platforms. A discussion on the |
| 696 |
problem is available at https://curl.haxx.se/mail/lib-2008-09/0197.html |
| 697 |
|
| 698 |
Also, alarm() provides timeout resolution only to the nearest second. alarm |
| 699 |
ought to be replaced by setitimer on systems that support it. |
| 700 |
|
| 701 |
|
| 702 |
12. LDAP and OpenLDAP |
| 703 |
|
| 704 |
12.1 OpenLDAP hangs after returning results |
| 705 |
|
| 706 |
By configuration defaults, openldap automatically chase referrals on |
| 707 |
secondary socket descriptors. The OpenLDAP backend is asynchronous and thus |
| 708 |
should monitor all socket descriptors involved. Currently, these secondary |
| 709 |
descriptors are not monitored, causing openldap library to never receive |
| 710 |
data from them. |
| 711 |
|
| 712 |
As a temporary workaround, disable referrals chasing by configuration. |
| 713 |
|
| 714 |
The fix is not easy: proper automatic referrals chasing requires a |
| 715 |
synchronous bind callback and monitoring an arbitrary number of socket |
| 716 |
descriptors for a single easy handle (currently limited to 5). |
| 717 |
|
| 718 |
Generic LDAP is synchronous: OK. |
| 719 |
|
| 720 |
See https://github.com/curl/curl/issues/622 and |
| 721 |
https://curl.haxx.se/mail/lib-2016-01/0101.html |
| 722 |
|
| 723 |
12.2 LDAP on Windows does authentication wrong? |
| 724 |
|
| 725 |
https://github.com/curl/curl/issues/3116 |
| 726 |
|
| 727 |
12.3 LDAP on Windows doesn't work |
| 728 |
|
| 729 |
A simple curl command line getting "ldap://ldap.forumsys.com" returns an |
| 730 |
error that says "no memory" ! |
| 731 |
|
| 732 |
https://github.com/curl/curl/issues/4261 |
| 733 |
|
| 734 |
13. TCP/IP |
| 735 |
|
| 736 |
13.1 --interface for ipv6 binds to unusable IP address |
| 737 |
|
| 738 |
Since IPv6 provides a lot of addresses with different scope, binding to an |
| 739 |
IPv6 address needs to take the proper care so that it doesn't bind to a |
| 740 |
locally scoped address as that is bound to fail. |
| 741 |
|
| 742 |
https://github.com/curl/curl/issues/686 |
| 743 |
|
| 744 |
14. DICT |
| 745 |
|
| 746 |
14.1 DICT responses show the underlying protocol |
| 747 |
|
| 748 |
When getting a DICT response, the protocol parts of DICT aren't stripped off |
| 749 |
from the output. |
| 750 |
|
| 751 |
https://github.com/curl/curl/issues/1809 |