curl/docs/KNOWN_BUGS

These are problems known to exist at the time of this release. Feel free to
join in and help us correct one or more of these! Also be sure to check the
changelog of the current development status, as one or more of these problems
may have been fixed since this was written!

27. "libcurl built with GNUTLS ignores the SSLCERT option" - Unlike
  Curl_ossl_connect(), the Curl_gtls_connect() function does not send the user
  certificate to the peer. In fact, it ignores the conn->data->set.cert field
  completely, it always uses the anonymous credentials. See
  http://curl.haxx.se/bug/view.cgi?id=1348930 

26. NTLM authentication using SSPI (on Windows) when (lib)curl is running in
  "system context" will make it use wrong(?) user name - at least when compared
  to what winhttp does. See http://curl.haxx.se/bug/view.cgi?id=1281867

25. When doing a CONNECT request with curl it doesn't properly handle if the
  proxy closes the connection within the authentication "negotiation phase".
  Like if you do HTTPS or similar over a proxy and you use perhaps
  --proxy-anyauth. There's work in progress on this problem, and a recent
  patch was posted here: http://curl.haxx.se/mail/lib-2005-08/0074.html

24. Harshal Pradhan's Use-after-free with libcurl+ares. This probably occurs
  because there is a pending ares callback that gets called after the
  connection struct has been freed in libcurl:
        http://curl.haxx.se/mail/lib-2005-08/0022.html
  Fixing this properly most likely requires a new c-ares function.

23. We don't support SOCKS for IPv6. We don't support FTPS over a SOCKS proxy.
  We don't have any test cases for SOCKS proxy. We probably have even more
  bugs and lack of features when a SOCKS proxy is used.

22. Sending files to a FTP server using curl on VMS, might lead to curl
  complaining on "unaligned file size" on completion. The problem is related
  to VMS file structures and the perceived file sizes stat() returns. A
  possible fix would involve sending a "STRU VMS" command.
  http://curl.haxx.se/bug/view.cgi?id=1156287
  
21. FTP ASCII transfers do not follow RFC959. They don't convert the data
   accordingly (not for sending nor for receiving). RFC 959 section 3.1.1.1
   clearly describes how this should be done:

     The sender converts the data from an internal character representation to
     the standard 8-bit NVT-ASCII representation (see the Telnet
     specification).  The receiver will convert the data from the standard
     form to his own internal form.

19. FTP 3rd party transfers with the multi interface doesn't work. Test:
  define CURL_MULTIEASY, rebuild curl, run test case 230 - 232.

18. test case 57 has </test> that should be </client> but when corrected, the
  test case fails!

16. FTP URLs passed to curl may contain NUL (0x00) in the RFC 1738 <user>,
  <password>, and <fpath> components, encoded as "%00".  The problem is that
  curl_unescape does not detect this, but instead returns a shortened C
  string.  From a strict FTP protocol standpoint, NUL is a valid character
  within RFC 959 <string>, so the way to handle this correctly in curl would
  be to use a data structure other than a plain C string, one that can handle
  embedded NUL characters.  From a practical standpoint, most FTP servers
  would not meaningfully support NUL characters within RFC 959 <string>,
  anyway (e.g., UNIX pathnames may not contain NUL).

14. Test case 165 might fail on system which has libidn present, but with an
  old iconv version (2.1.3 is a known bad version), since it doesn't recognize
  the charset when named ISO8859-1. Changing the name to ISO-8859-1 makes the
  test pass, but instead makes it fail on Solaris hosts that use its native
  iconv.

13. curl version 7.12.2 fails on AIX if compiled with --enable-ares.
  The workaround is to combine --enable-ares with --disable-shared

12. When connecting to a SOCKS proxy, the (connect) timeout is not properly
  acknowledged after the actual TCP connect (during the SOCKS "negotiate"
  phase). Pointed out by Lucas. Fix: need to select() and timeout properly.

11. Using configure --disable-[protocol] may cause 'make test' to fail for
  tests using the disabled protocol(s).

10. To get HTTP Negotiate authentication to work fine, you need to provide a
  (fake) user name (this concerns both curl and the lib) because the code
  wrongly only considers authentication if there's a user name provided.
  http://curl.haxx.se/bug/view.cgi?id=1004841. How?
  http://curl.haxx.se/mail/lib-2004-08/0182.html

9. --limit-rate using -d or -F does not work. This is because the limit logic
  is provided by the curl app in its read/write callbacks, and when doing
  -d/-F the callbacks aren't used! http://curl.haxx.se/bug/view.cgi?id=921395

8. Doing resumed upload over HTTP does not work with '-C -', because curl
  doesn't do a HEAD first to get the initial size. This needs to be done
  manually for HTTP PUT resume to work, and then '-C [index]'.

7. CURLOPT_USERPWD and CURLOPT_PROXYUSERPWD have no way of providing user names
  that contain a colon. This can't be fixed easily in a backwards compatible
  way without adding new options (and then, they should most probably allow
  setting user name and password separately).

6. libcurl ignores empty path parts in FTP URLs, whereas RFC1738 states that
  such parts should be sent to the server as 'CWD ' (without an argument).
  The only exception to this rule, is that we knowingly break this if the
  empty part is first in the path, as then we use the double slashes to
  indicate that the user wants to reach the root dir (this exception SHALL
  remain even when this bug is fixed).

5. libcurl doesn't treat the content-length of compressed data properly, as
  it seems HTTP servers send the *uncompressed* length in that header and
  libcurl thinks of it as the *compressed* length. Some explanations are here:
  http://curl.haxx.se/mail/lib-2003-06/0146.html

3. GOPHER transfers seem broken

2. If a HTTP server responds to a HEAD request and includes a body (thus
  violating the RFC2616), curl won't wait to read the response but just stop
  reading and return back. If a second request (let's assume a GET) is then
  immediately made to the same server again, the connection will be re-used
  fine of course, and the second request will be sent off but when the
  response is to get read, the previous response-body is what curl will read
  and havoc is what happens.
  More details on this is found in this libcurl mailing list thread:
  http://curl.haxx.se/mail/lib-2002-08/0000.html