5 \___|\___/|_| \_\_____|
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!
15 1.1 CURLFORM_CONTENTLEN in an array
16 1.2 Disabling HTTP Pipelining
17 1.3 STARTTRANSFER time is wrong for HTTP POSTs
18 1.4 multipart formposts file name encoding
19 1.5 Expect-100 meets 417
20 1.6 Unnecessary close when 401 received waiting for 100
21 1.9 HTTP/2 frames while in the connection pool kill reuse
22 1.10 Strips trailing dot from host name
23 1.11 CURLOPT_SEEKFUNCTION not called with CURLFORM_STREAM
26 2.1 CURLINFO_SSL_VERIFYRESULT has limited support
28 2.3 GnuTLS backend skips really long certificate fields
29 2.4 DarwinSSL won't import PKCS#12 client certificates without a password
32 3.1 IMAP SEARCH ALL truncated response
33 3.2 No disconnect command
34 3.3 SMTP to multiple recipients
35 3.4 POP3 expects "CRLF.CRLF" eob for some single-line responses
38 4.1 -J with %-encoded file nameas
39 4.2 -J with -C - fails
40 4.3 --retry and transfer timeouts
42 5. Build and portability issues
43 5.1 Windows Borland compiler
44 5.2 curl-config --libs contains private details
45 5.4 AIX shared build with c-ares fails
46 5.5 can't handle Unicode arguments in Windows
47 5.6 cmake support gaps
48 5.7 Visual Studio project gaps
49 5.8 configure finding libs in wrong directory
50 5.9 Utilize Requires.private directives in libcurl.pc
53 6.1 NTLM authentication and unicode
54 6.2 MIT Kerberos for Windows build
55 6.3 NTLM in system context uses wrong name
56 6.4 Negotiate and Kerberos V5 need a fake user name
59 7.1 FTP without or slow 220 response
60 7.2 FTP with CONNECT and slow server
61 7.3 FTP with NOBODY and FAILONERROR
64 7.6 FTP with NULs in URL parts
65 7.7 FTP and empty path parts in the URL
66 7.8 Premature transfer end but healthy control channel
69 8.1 TELNET and time limtiations don't work
70 8.2 Microsoft telnet server
73 9.1 SFTP doesn't do CURLOPT_POSTQUOTE correct
76 10.1 SOCKS proxy connections are done blocking
77 10.2 SOCKS don't support timeouts
79 10.4 active FTP over a SOCKS
82 11.1 Curl leaks .onion hostnames in DNS
83 11.2 error buffer not set if connection to multiple addresses fails
84 11.3 c-ares deviates from stock resolver on http://1346569778
85 11.4 HTTP test server 'connection-monitor' problems
88 12.1 OpenLDAP hangs after returning results
91 13.1 --interface for ipv6 binds to unusable IP address
94 ==============================================================================
98 1.1 CURLFORM_CONTENTLEN in an array
100 It is not possible to pass a 64-bit value using CURLFORM_CONTENTLEN with
101 CURLFORM_ARRAY, when compiled on 32-bit platforms that support 64-bit
102 integers. This is because the underlying structure 'curl_forms' uses a dual
103 purpose char* for storing these values in via casting. For more information
104 see the now closed related issue:
105 https://github.com/curl/curl/issues/608
107 1.2 Disabling HTTP Pipelining
109 Disabling HTTP Pipelining when there are ongoing transfers can lead to
110 heap corruption and crash. https://curl.haxx.se/bug/view.cgi?id=1411
112 1.3 STARTTRANSFER time is wrong for HTTP POSTs
114 Wrong STARTTRANSFER timer accounting for POST requests Timer works fine with
115 GET requests, but while using POST the time for CURLINFO_STARTTRANSFER_TIME
116 is wrong. While using POST CURLINFO_STARTTRANSFER_TIME minus
117 CURLINFO_PRETRANSFER_TIME is near to zero every time.
119 https://github.com/curl/curl/issues/218
120 https://curl.haxx.se/bug/view.cgi?id=1213
122 1.4 multipart formposts file name encoding
124 When creating multipart formposts. The file name part can be encoded with
125 something beyond ascii but currently libcurl will only pass in the verbatim
126 string the app provides. There are several browsers that already do this
127 encoding. The key seems to be the updated draft to RFC2231:
128 https://tools.ietf.org/html/draft-reschke-rfc2231-in-http-02
130 1.5 Expect-100 meets 417
132 If an upload using Expect: 100-continue receives an HTTP 417 response, it
133 ought to be automatically resent without the Expect:. A workaround is for
134 the client application to redo the transfer after disabling Expect:.
135 https://curl.haxx.se/mail/archive-2008-02/0043.html
137 1.6 Unnecessary close when 401 received waiting for 100
139 libcurl closes the connection if an HTTP 401 reply is received while it is
140 waiting for the the 100-continue response.
141 https://curl.haxx.se/mail/lib-2008-08/0462.html
143 1.9 HTTP/2 frames while in the connection pool kill reuse
145 If the server sends HTTP/2 frames (like for example an HTTP/2 PING frame) to
146 curl while the connection is held in curl's connection pool, the socket will
147 be found readable when considered for reuse and that makes curl think it is
148 dead and then it will be closed and a new connection gets created instead.
150 This is *best* fixed by adding monitoring to connections while they are kept
151 in the pool so that pings can be responded to appropriately.
153 1.10 Strips trailing dot from host name
155 When given a URL with a trailing dot for the host name part:
156 "https://example.com./", libcurl will strip off the dot and use the name
157 without a dot internally and send it dot-less in HTTP Host: headers and in
160 The HTTP part violates RFC 7230 section 5.4 but the SNI part is accordance
161 with RFC 6066 section 3.
163 URLs using these trailing dots are very rare in the wild and we have not seen
164 or gotten any real-world problems with such URLs reported. The popular
165 browsers seem to have stayed with not stripping the dot for both uses (thus
166 they violate RFC 6066 instead of RFC 7230).
168 Daniel took the discussion to the HTTPbis mailing list in March 2016:
169 https://lists.w3.org/Archives/Public/ietf-http-wg/2016JanMar/0430.html but
170 there was not major rush or interest to fix this. The impression I get is
171 that most HTTP people rather not rock the boat now and instead prioritize web
172 compatibility rather than to strictly adhere to these RFCs.
174 Our current approach allows a knowing client to send a custom HTTP header
177 It can also be noted that while adding a trailing dot to the host name in
178 most (all?) cases will make the name resolve to the same set of IP addresses,
179 many HTTP servers will not happily accept the trailing dot there unless that
180 has been specifically configured to be a fine virtual host.
182 If URLs with trailing dots for host names become more popular or even just
183 used more than for just plain fun experiments, I'm sure we will have reason
184 to go back and reconsider.
186 See https://github.com/curl/curl/issues/716 for the discussion.
188 1.11 CURLOPT_SEEKFUNCTION not called with CURLFORM_STREAM
190 I'm using libcurl to POST form data using a FILE* with the CURLFORM_STREAM
191 option of curl_formadd(). I've noticed that if the connection drops at just
192 the right time, the POST is reattempted without the data from the file. It
193 seems like the file stream position isn't getting reset to the beginning of
194 the file. I found the CURLOPT_SEEKFUNCTION option and set that with a
195 function that performs an fseek() on the FILE*. However, setting that didn't
196 seem to fix the issue or even get called. See
197 https://github.com/curl/curl/issues/768
202 2.1 CURLINFO_SSL_VERIFYRESULT has limited support
204 CURLINFO_SSL_VERIFYRESULT is only implemented for the OpenSSL and NSS
205 backends, so relying on this information in a generic app is flaky.
209 Curl doesn't recognize certificates in DER format in keychain, but it works
210 with PEM. https://curl.haxx.se/bug/view.cgi?id=1065
212 2.3 GnuTLS backend skips really long certificate fields
214 libcurl calls gnutls_x509_crt_get_dn() with a fixed buffer size and if the
215 field is too long in the cert, it'll just return an error and the field will
218 2.4 DarwinSSL won't import PKCS#12 client certificates without a password
220 libcurl calls SecPKCS12Import with the PKCS#12 client certificate, but that
221 function rejects certificates that do not have a password.
222 https://github.com/curl/curl/issues/1308
227 3.1 IMAP SEARCH ALL truncated response
229 IMAP "SEARCH ALL" truncates output on large boxes. "A quick search of the
230 code reveals that pingpong.c contains some truncation code, at line 408, when
231 it deems the server response to be too large truncating it to 40 characters"
232 https://curl.haxx.se/bug/view.cgi?id=1366
234 3.2 No disconnect command
236 The disconnect commands (LOGOUT and QUIT) may not be sent by IMAP, POP3 and
237 SMTP if a failure occurs during the authentication phase of a connection.
239 3.3 SMTP to multiple recipients
241 When sending data to multiple recipients, curl will abort and return failure
242 if one of the recipients indicate failure (on the "RCPT TO"
243 command). Ordinary mail programs would proceed and still send to the ones
244 that can receive data. This is subject for change in the future.
245 https://curl.haxx.se/bug/view.cgi?id=1116
247 3.4 POP3 expects "CRLF.CRLF" eob for some single-line responses
249 You have to tell libcurl not to expect a body, when dealing with one line
250 response commands. Please see the POP3 examples and test cases which show
251 this for the NOOP and DELE commands. https://curl.haxx.se/bug/?i=740
256 4.1 -J with %-encoded file nameas
258 -J/--remote-header-name doesn't decode %-encoded file names. RFC6266 details
259 how it should be done. The can of worm is basically that we have no charset
260 handling in curl and ascii >=128 is a challenge for us. Not to mention that
261 decoding also means that we need to check for nastiness that is attempted,
262 like "../" sequences and the like. Probably everything to the left of any
263 embedded slashes should be cut off.
264 https://curl.haxx.se/bug/view.cgi?id=1294
266 4.2 -J with -C - fails
268 When using -J (with -O), automatically resumed downloading together with "-C
269 -" fails. Without -J the same command line works! This happens because the
270 resume logic is worked out before the target file name (and thus its
271 pre-transfer size) has been figured out!
272 https://curl.haxx.se/bug/view.cgi?id=1169
274 4.3 --retry and transfer timeouts
276 If using --retry and the transfer timeouts (possibly due to using -m or
277 -y/-Y) the next attempt doesn't resume the transfer properly from what was
278 downloaded in the previous attempt but will truncate and restart at the
279 original position where it was at before the previous failed attempt. See
280 https://curl.haxx.se/mail/lib-2008-01/0080.html and Mandriva bug report
281 https://qa.mandriva.com/show_bug.cgi?id=22565
284 5. Build and portability issues
286 5.1 Windows Borland compiler
288 When building with the Windows Borland compiler, it fails because the "tlib"
289 tool doesn't support hyphens (minus signs) in file names and we have such in
290 the build. https://curl.haxx.se/bug/view.cgi?id=1222
292 5.2 curl-config --libs contains private details
294 "curl-config --libs" will include details set in LDFLAGS when configure is
295 run that might be needed only for building libcurl. Further, curl-config
296 --cflags suffers from the same effects with CFLAGS/CPPFLAGS.
298 5.4 AIX shared build with c-ares fails
300 curl version 7.12.2 fails on AIX if compiled with --enable-ares. The
301 workaround is to combine --enable-ares with --disable-shared
303 5.5 can't handle Unicode arguments in Windows
305 If a URL or filename can't be encoded using the user's current codepage then
306 it can only be encoded properly in the Unicode character set. Windows uses
307 UTF-16 encoding for Unicode and stores it in wide characters, however curl
308 and libcurl are not equipped for that at the moment. And, except for Cygwin,
309 Windows can't use UTF-8 as a locale.
311 https://curl.haxx.se/bug/?i=345
312 https://curl.haxx.se/bug/?i=731
314 5.6 cmake support gaps
316 The cmake build setup lacks several features that the autoconf build
317 offers. This includes:
319 - use of correct soname for the shared library build
320 - support for several TLS backends are missing
321 - the unit tests cause link failures in regular non-static builds
324 5.7 Visual Studio project gaps
326 The Visual Studio projects lack some features that the autoconf and nmake
327 builds offer, such as the following:
329 - support for zlib and nghttp2
330 - use of static runtime libraries
331 - add the test suite components
333 In addition to this the following could be implemented:
335 - support for other development IDEs
336 - add PATH environment variables for third-party DLLs
338 5.8 configure finding libs in wrong directory
340 When the configure script checks for third-party libraries, it adds those
341 directories to the LDFLAGS variable and then tries linking to see if it
342 works. When successful, the found directory is kept in the LDFLAGS variable
343 when the script continues to execute and do more tests and possibly check for
346 This can make subsequent checks for libraries wrongly detect another
347 installation in a directory that was previously added to LDFLAGS by another
350 A possibly better way to do these checks would be to keep the pristine LDFLAGS
351 even after successful checks and instead add those verified paths to a
352 separate variable that only after all library checks have been performed gets
355 5.9 Utilize Requires.private directives in libcurl.pc
357 https://github.com/curl/curl/issues/864
361 6.1 NTLM authentication and unicode
363 NTLM authentication involving unicode user name or password only works
364 properly if built with UNICODE defined together with the WinSSL/schannel
365 backend. The original problem was mentioned in:
366 https://curl.haxx.se/mail/lib-2009-10/0024.html
367 https://curl.haxx.se/bug/view.cgi?id=896
369 The WinSSL/schannel version verified to work as mentioned in
370 https://curl.haxx.se/mail/lib-2012-07/0073.html
372 6.2 MIT Kerberos for Windows build
374 libcurl fails to build with MIT Kerberos for Windows (KfW) due to KfW's
375 library header files exporting symbols/macros that should be kept private to
376 the KfW library. See ticket #5601 at https://krbdev.mit.edu/rt/
378 6.3 NTLM in system context uses wrong name
380 NTLM authentication using SSPI (on Windows) when (lib)curl is running in
381 "system context" will make it use wrong(?) user name - at least when compared
382 to what winhttp does. See https://curl.haxx.se/bug/view.cgi?id=535
384 6.4 Negotiate and Kerberos V5 need a fake user name
386 In order to get Negotiate (SPNEGO) authentication to work in HTTP or Kerberos
387 V5 in the e-mail protocols, you need to provide a (fake) user name (this
388 concerns both curl and the lib) because the code wrongly only considers
389 authentication if there's a user name provided by setting
390 conn->bits.user_passwd in url.c https://curl.haxx.se/bug/view.cgi?id=440 How?
391 https://curl.haxx.se/mail/lib-2004-08/0182.html A possible solution is to
392 either modify this variable to be set or introduce a variable such as
393 new conn->bits.want_authentication which is set when any of the authentication
399 7.1 FTP without or slow 220 response
401 If a connection is made to a FTP server but the server then just never sends
402 the 220 response or otherwise is dead slow, libcurl will not acknowledge the
403 connection timeout during that phase but only the "real" timeout - which may
404 surprise users as it is probably considered to be the connect phase to most
405 people. Brought up (and is being misunderstood) in:
406 https://curl.haxx.se/bug/view.cgi?id=856
408 7.2 FTP with CONNECT and slow server
410 When doing FTP over a socks proxy or CONNECT through HTTP proxy and the multi
411 interface is used, libcurl will fail if the (passive) TCP connection for the
412 data transfer isn't more or less instant as the code does not properly wait
413 for the connect to be confirmed. See test case 564 for a first shot at a test
416 7.3 FTP with NOBODY and FAILONERROR
418 It seems sensible to be able to use CURLOPT_NOBODY and CURLOPT_FAILONERROR
419 with FTP to detect if a file exists or not, but it is not working:
420 https://curl.haxx.se/mail/lib-2008-07/0295.html
424 When doing an operation over FTP that requires the ACCT command (but not when
425 logging in), the operation will fail since libcurl doesn't detect this and
426 thus fails to issue the correct command:
427 https://curl.haxx.se/bug/view.cgi?id=635
431 FTP ASCII transfers do not follow RFC959. They don't convert the data
432 accordingly (not for sending nor for receiving). RFC 959 section 3.1.1.1
433 clearly describes how this should be done:
435 The sender converts the data from an internal character representation to
436 the standard 8-bit NVT-ASCII representation (see the Telnet
437 specification). The receiver will convert the data from the standard
438 form to his own internal form.
440 Since 7.15.4 at least line endings are converted.
442 7.6 FTP with NULs in URL parts
444 FTP URLs passed to curl may contain NUL (0x00) in the RFC 1738 <user>,
445 <password>, and <fpath> components, encoded as "%00". The problem is that
446 curl_unescape does not detect this, but instead returns a shortened C string.
447 From a strict FTP protocol standpoint, NUL is a valid character within RFC
448 959 <string>, so the way to handle this correctly in curl would be to use a
449 data structure other than a plain C string, one that can handle embedded NUL
450 characters. From a practical standpoint, most FTP servers would not
451 meaningfully support NUL characters within RFC 959 <string>, anyway (e.g.,
452 Unix pathnames may not contain NUL).
454 7.7 FTP and empty path parts in the URL
456 libcurl ignores empty path parts in FTP URLs, whereas RFC1738 states that
457 such parts should be sent to the server as 'CWD ' (without an argument). The
458 only exception to this rule, is that we knowingly break this if the empty
459 part is first in the path, as then we use the double slashes to indicate that
460 the user wants to reach the root dir (this exception SHALL remain even when
463 7.8 Premature transfer end but healthy control channel
465 When 'multi_done' is called before the transfer has been completed the normal
466 way, it is considered a "premature" transfer end. In this situation, libcurl
467 closes the connection assuming it doesn't know the state of the connection so
468 it can't be reused for subsequent requests.
470 With FTP however, this isn't necessarily true but there are a bunch of
471 situations (listed in the ftp_done code) where it *could* keep the connection
472 alive even in this situation - but the current code doesn't. Fixing this would
473 allow libcurl to reuse FTP connections better.
477 8.1 TELNET and time limtiations don't work
479 When using telnet, the time limitation options don't work.
480 https://curl.haxx.se/bug/view.cgi?id=846
482 8.2 Microsoft telnet server
484 There seems to be a problem when connecting to the Microsoft telnet server.
485 https://curl.haxx.se/bug/view.cgi?id=649
490 9.1 SFTP doesn't do CURLOPT_POSTQUOTE correct
492 When libcurl sends CURLOPT_POSTQUOTE commands when connected to a SFTP server
493 using the multi interface, the commands are not being sent correctly and
494 instead the connection is "cancelled" (the operation is considered done)
495 prematurely. There is a half-baked (busy-looping) patch provided in the bug
496 report but it cannot be accepted as-is. See
497 https://curl.haxx.se/bug/view.cgi?id=748
502 10.1 SOCKS proxy connections are done blocking
504 Both SOCKS5 and SOCKS4 proxy connections are done blocking, which is very bad
505 when used with the multi interface.
507 10.2 SOCKS don't support timeouts
509 The SOCKS4 connection codes don't properly acknowledge (connect) timeouts.
510 According to bug #1556528, even the SOCKS5 connect code does not do it right:
511 https://curl.haxx.se/bug/view.cgi?id=604
513 When connecting to a SOCK proxy, the (connect) timeout is not properly
514 acknowledged after the actual TCP connect (during the SOCKS "negotiate"
519 libcurl doesn't support FTPS over a SOCKS proxy.
521 10.4 active FTP over a SOCKS
523 libcurl doesn't support active FTP over a SOCKS proxy
528 11.1 Curl leaks .onion hostnames in DNS
530 Curl sends DNS requests for hostnames with a .onion TLD. This leaks
531 information about what the user is attempting to access, and violates this
532 requirement of RFC7686: https://tools.ietf.org/html/rfc7686
534 Issue: https://github.com/curl/curl/issues/543
536 11.2 error buffer not set if connection to multiple addresses fails
538 If you ask libcurl to resolve a hostname like example.com to IPv6 addresses
539 only. But you only have IPv4 connectivity. libcurl will correctly fail with
540 CURLE_COULDNT_CONNECT. But the error buffer set by CURLOPT_ERRORBUFFER
541 remains empty. Issue: https://github.com/curl/curl/issues/544
543 11.3 c-ares deviates from stock resolver on http://1346569778
545 When using the socket resolvers, that URL becomes:
547 * Rebuilt URL to: http://1346569778/
548 * Trying 80.67.6.50...
550 but with c-ares it instead says "Could not resolve: 1346569778 (Domain name
553 See https://github.com/curl/curl/issues/893
555 11.4 HTTP test server 'connection-monitor' problems
557 The 'connection-monitor' feature of the sws HTTP test server doesn't work
558 properly if some tests are run in unexpected order. Like 1509 and then 1525.
560 See https://github.com/curl/curl/issues/868
563 12. LDAP and OpenLDAP
565 12.1 OpenLDAP hangs after returning results
567 By configuration defaults, openldap automatically chase referrals on
568 secondary socket descriptors. The OpenLDAP backend is asynchronous and thus
569 should monitor all socket descriptors involved. Currently, these secondary
570 descriptors are not monitored, causing openldap library to never receive
573 As a temporary workaround, disable referrals chasing by configuration.
575 The fix is not easy: proper automatic referrals chasing requires a
576 synchronous bind callback and monitoring an arbitrary number of socket
577 descriptors for a single easy handle (currently limited to 5).
579 Generic LDAP is synchronous: OK.
581 See https://github.com/curl/curl/issues/622 and
582 https://curl.haxx.se/mail/lib-2016-01/0101.html
587 13.1 --interface for ipv6 binds to unusable IP address
589 Since IPv6 provides a lot of addresses with different scope, binding to an
590 IPv6 address needs to take the proper care so that it doesn't bind to a
591 locally scoped address as that is bound to fail.
593 https://github.com/curl/curl/issues/686