1These are problems known to exist at the time of this release. Feel free to 2join in and help us correct one or more of these! Also be sure to check the 3changelog of the current development status, as one or more of these problems 4may have been fixed since this was written! 5 687. -J/--remote-header-name doesn't decode %-encoded file names. RFC6266 7 details how it should be done. The can of worm is basically that we have no 8 charset handling in curl and ascii >=128 is a challenge for us. Not to 9 mention that decoding also means that we need to check for nastiness that is 10 attempted, like "../" sequences and the like. Probably everything to the left 11 of any embedded slashes should be cut off. 12 http://curl.haxx.se/bug/view.cgi?id=1294 13 1486. The disconnect commands (LOGOUT and QUIT) may not be sent by IMAP, POP3 15 and SMTP if a failure occurs during the authentication phase of a 16 connection. 17 1885. Wrong STARTTRANSFER timer accounting for POST requests 19 Timer works fine with GET requests, but while using POST the time for 20 CURLINFO_STARTTRANSFER_TIME is wrong. While using POST 21 CURLINFO_STARTTRANSFER_TIME minus CURLINFO_PRETRANSFER_TIME is near to zero 22 every time. 23 http://curl.haxx.se/bug/view.cgi?id=1213 24 2584. CURLINFO_SSL_VERIFYRESULT is only implemented for the OpenSSL and NSS 26 backends, so relying on this information in a generic app is flaky. 27 2883. curl is unable to load non-default openssl engines, because openssl isn't 29 initialized properly. This seems to require OpenSSL_config() or 30 CONF_modules_load_file() to be used by libcurl but the first seems to not 31 work and we've gotten not reports from tests with the latter. Possibly we 32 need to discuss with OpenSSL developers how this is supposed to be done. We 33 need users with actual external openssl engines for testing to work on this. 34 http://curl.haxx.se/bug/view.cgi?id=1208 35 3682. When building with the Windows Borland compiler, it fails because the 37 "tlib" tool doesn't support hyphens (minus signs) in file names and we have 38 such in the build. 39 http://curl.haxx.se/bug/view.cgi?id=1222 40 4181. When using -J (with -O), automatically resumed downloading together with 42 "-C -" fails. Without -J the same command line works! This happens because 43 the resume logic is worked out before the target file name (and thus its 44 pre-transfer size) has been figured out! 45 http://curl.haxx.se/bug/view.cgi?id=1169 46 4780. Curl doesn't recognize certificates in DER format in keychain, but it 48 works with PEM. 49 http://curl.haxx.se/bug/view.cgi?id=1065 50 5179. SMTP. When sending data to multiple recipients, curl will abort and return 52 failure if one of the recipients indicate failure (on the "RCPT TO" 53 command). Ordinary mail programs would proceed and still send to the ones 54 that can receive data. This is subject for change in the future. 55 http://curl.haxx.se/bug/view.cgi?id=1116 56 5778. curl and libcurl don't always signal the client properly when "sending" 58 zero bytes files - it makes for example the command line client not creating 59 any file at all. Like when using FTP. 60 http://curl.haxx.se/bug/view.cgi?id=1063 61 6277. CURLOPT_FORBID_REUSE on a handle prevents NTLM from working since it 63 "abuses" the underlying connection re-use system and if connections are 64 forced to close they break the NTLM support. 65 6676. The SOCKET type in Win64 is 64 bits large (and thus so is curl_socket_t on 67 that platform), and long is only 32 bits. It makes it impossible for 68 curl_easy_getinfo() to return a socket properly with the CURLINFO_LASTSOCKET 69 option as for all other operating systems. 70 7175. NTLM authentication involving unicode user name or password only works 72 properly if built with UNICODE defined together with the WinSSL/schannel 73 backend. The original problem was mentioned in: 74 http://curl.haxx.se/mail/lib-2009-10/0024.html 75 http://curl.haxx.se/bug/view.cgi?id=896 76 77 The WinSSL/schannel version verified to work as mentioned in 78 http://curl.haxx.se/mail/lib-2012-07/0073.html 79 8073. if a connection is made to a FTP server but the server then just never 81 sends the 220 response or otherwise is dead slow, libcurl will not 82 acknowledge the connection timeout during that phase but only the "real" 83 timeout - which may surprise users as it is probably considered to be the 84 connect phase to most people. Brought up (and is being misunderstood) in: 85 http://curl.haxx.se/bug/view.cgi?id=856 86 8772. "Pausing pipeline problems." 88 http://curl.haxx.se/mail/lib-2009-07/0214.html 89 9070. Problem re-using easy handle after call to curl_multi_remove_handle 91 http://curl.haxx.se/mail/lib-2009-07/0249.html 92 9368. "More questions about ares behavior". 94 http://curl.haxx.se/mail/lib-2009-08/0012.html 95 9667. When creating multipart formposts. The file name part can be encoded with 97 something beyond ascii but currently libcurl will only pass in the verbatim 98 string the app provides. There are several browsers that already do this 99 encoding. The key seems to be the updated draft to RFC2231: 100 http://tools.ietf.org/html/draft-reschke-rfc2231-in-http-02 101 10266. When using telnet, the time limitation options don't work. 103 http://curl.haxx.se/bug/view.cgi?id=846 104 10565. When doing FTP over a socks proxy or CONNECT through HTTP proxy and the 106 multi interface is used, libcurl will fail if the (passive) TCP connection 107 for the data transfer isn't more or less instant as the code does not 108 properly wait for the connect to be confirmed. See test case 564 for a first 109 shot at a test case. 110 11163. When CURLOPT_CONNECT_ONLY is used, the handle cannot reliably be re-used 112 for any further requests or transfers. The work-around is then to close that 113 handle with curl_easy_cleanup() and create a new. Some more details: 114 http://curl.haxx.se/mail/lib-2009-04/0300.html 115 11661. If an upload using Expect: 100-continue receives an HTTP 417 response, 117 it ought to be automatically resent without the Expect:. A workaround is 118 for the client application to redo the transfer after disabling Expect:. 119 http://curl.haxx.se/mail/archive-2008-02/0043.html 120 12160. libcurl closes the connection if an HTTP 401 reply is received while it 122 is waiting for the the 100-continue response. 123 http://curl.haxx.se/mail/lib-2008-08/0462.html 124 12558. It seems sensible to be able to use CURLOPT_NOBODY and 126 CURLOPT_FAILONERROR with FTP to detect if a file exists or not, but it is 127 not working: http://curl.haxx.se/mail/lib-2008-07/0295.html 128 12956. When libcurl sends CURLOPT_POSTQUOTE commands when connected to a SFTP 130 server using the multi interface, the commands are not being sent correctly 131 and instead the connection is "cancelled" (the operation is considered done) 132 prematurely. There is a half-baked (busy-looping) patch provided in the bug 133 report but it cannot be accepted as-is. See 134 http://curl.haxx.se/bug/view.cgi?id=748 135 13655. libcurl fails to build with MIT Kerberos for Windows (KfW) due to KfW's 137 library header files exporting symbols/macros that should be kept private 138 to the KfW library. See ticket #5601 at http://krbdev.mit.edu/rt/ 139 14052. Gautam Kachroo's issue that identifies a problem with the multi interface 141 where a connection can be re-used without actually being properly 142 SSL-negotiated: 143 http://curl.haxx.se/mail/lib-2008-01/0277.html 144 14549. If using --retry and the transfer timeouts (possibly due to using -m or 146 -y/-Y) the next attempt doesn't resume the transfer properly from what was 147 downloaded in the previous attempt but will truncate and restart at the 148 original position where it was at before the previous failed attempt. See 149 http://curl.haxx.se/mail/lib-2008-01/0080.html and Mandriva bug report 150 https://qa.mandriva.com/show_bug.cgi?id=22565 151 15248. If a CONNECT response-headers are larger than BUFSIZE (16KB) when the 153 connection is meant to be kept alive (like for NTLM proxy auth), the 154 function will return prematurely and will confuse the rest of the HTTP 155 protocol code. This should be very rare. 156 15743. There seems to be a problem when connecting to the Microsoft telnet server. 158 http://curl.haxx.se/bug/view.cgi?id=649 159 16041. When doing an operation over FTP that requires the ACCT command (but not 161 when logging in), the operation will fail since libcurl doesn't detect this 162 and thus fails to issue the correct command: 163 http://curl.haxx.se/bug/view.cgi?id=635 164 16539. Steffen Rumler's Race Condition in Curl_proxyCONNECT: 166 http://curl.haxx.se/mail/lib-2007-01/0045.html 167 16838. Kumar Swamy Bhatt's problem in ftp/ssl "LIST" operation: 169 http://curl.haxx.se/mail/lib-2007-01/0103.html 170 17135. Both SOCKS5 and SOCKS4 proxy connections are done blocking, which is very 172 bad when used with the multi interface. 173 17434. The SOCKS4 connection codes don't properly acknowledge (connect) timeouts. 175 Also see #12. According to bug #1556528, even the SOCKS5 connect code does 176 not do it right: http://curl.haxx.se/bug/view.cgi?id=604 177 17831. "curl-config --libs" will include details set in LDFLAGS when configure is 179 run that might be needed only for building libcurl. Further, curl-config 180 --cflags suffers from the same effects with CFLAGS/CPPFLAGS. 181 18230. You need to use -g to the command line tool in order to use RFC2732-style 183 IPv6 numerical addresses in URLs. 184 18529. IPv6 URLs with zone ID is not nicely supported. 186 http://www.ietf.org/internet-drafts/draft-fenner-literal-zone-02.txt (expired) 187 specifies the use of a plus sign instead of a percent when specifying zone 188 IDs in URLs to get around the problem of percent signs being 189 special. According to the reporter, Firefox deals with the URL _with_ a 190 percent letter (which seems like a blatant URL spec violation). 191 libcurl supports zone IDs where the percent sign is URL-escaped (i.e. %25): 192 http://curl.haxx.se/bug/view.cgi?id=555 193 19426. NTLM authentication using SSPI (on Windows) when (lib)curl is running in 195 "system context" will make it use wrong(?) user name - at least when compared 196 to what winhttp does. See http://curl.haxx.se/bug/view.cgi?id=535 197 19823. SOCKS-related problems: 199 B) libcurl doesn't support FTPS over a SOCKS proxy. 200 E) libcurl doesn't support active FTP over a SOCKS proxy 201 202 We probably have even more bugs and lack of features when a SOCKS proxy is 203 used. 204 20521. FTP ASCII transfers do not follow RFC959. They don't convert the data 206 accordingly (not for sending nor for receiving). RFC 959 section 3.1.1.1 207 clearly describes how this should be done: 208 209 The sender converts the data from an internal character representation to 210 the standard 8-bit NVT-ASCII representation (see the Telnet 211 specification). The receiver will convert the data from the standard 212 form to his own internal form. 213 214 Since 7.15.4 at least line endings are converted. 215 21616. FTP URLs passed to curl may contain NUL (0x00) in the RFC 1738 <user>, 217 <password>, and <fpath> components, encoded as "%00". The problem is that 218 curl_unescape does not detect this, but instead returns a shortened C 219 string. From a strict FTP protocol standpoint, NUL is a valid character 220 within RFC 959 <string>, so the way to handle this correctly in curl would 221 be to use a data structure other than a plain C string, one that can handle 222 embedded NUL characters. From a practical standpoint, most FTP servers 223 would not meaningfully support NUL characters within RFC 959 <string>, 224 anyway (e.g., UNIX pathnames may not contain NUL). 225 22614. Test case 165 might fail on a system which has libidn present, but with an 227 old iconv version (2.1.3 is a known bad version), since it doesn't recognize 228 the charset when named ISO8859-1. Changing the name to ISO-8859-1 makes the 229 test pass, but instead makes it fail on Solaris hosts that use its native 230 iconv. 231 23213. curl version 7.12.2 fails on AIX if compiled with --enable-ares. 233 The workaround is to combine --enable-ares with --disable-shared 234 23512. When connecting to a SOCKS proxy, the (connect) timeout is not properly 236 acknowledged after the actual TCP connect (during the SOCKS "negotiate" 237 phase). 238 23910. To get HTTP Negotiate authentication to work fine, you need to provide a 240 (fake) user name (this concerns both curl and the lib) because the code 241 wrongly only considers authentication if there's a user name provided. 242 http://curl.haxx.se/bug/view.cgi?id=440 How? 243 http://curl.haxx.se/mail/lib-2004-08/0182.html 244 2458. Doing resumed upload over HTTP does not work with '-C -', because curl 246 doesn't do a HEAD first to get the initial size. This needs to be done 247 manually for HTTP PUT resume to work, and then '-C [index]'. 248 2496. libcurl ignores empty path parts in FTP URLs, whereas RFC1738 states that 250 such parts should be sent to the server as 'CWD ' (without an argument). 251 The only exception to this rule, is that we knowingly break this if the 252 empty part is first in the path, as then we use the double slashes to 253 indicate that the user wants to reach the root dir (this exception SHALL 254 remain even when this bug is fixed). 255 2565. libcurl doesn't treat the content-length of compressed data properly, as 257 it seems HTTP servers send the *uncompressed* length in that header and 258 libcurl thinks of it as the *compressed* length. Some explanations are here: 259 http://curl.haxx.se/mail/lib-2003-06/0146.html 260 2612. If a HTTP server responds to a HEAD request and includes a body (thus 262 violating the RFC2616), curl won't wait to read the response but just stop 263 reading and return back. If a second request (let's assume a GET) is then 264 immediately made to the same server again, the connection will be re-used 265 fine of course, and the second request will be sent off but when the 266 response is to get read, the previous response-body is what curl will read 267 and havoc is what happens. 268 More details on this is found in this libcurl mailing list thread: 269 http://curl.haxx.se/mail/lib-2002-08/0000.html 270