platform/upstream/curl.git
14 years agoUpdate email address.
Gisle Vanem [Mon, 7 Sep 2009 10:59:54 +0000 (10:59 +0000)]
Update email address.

14 years agoUpdate my email address. Add ares_config.h as dependency for 'make depend'.
Gisle Vanem [Mon, 7 Sep 2009 10:01:17 +0000 (10:01 +0000)]
Update my email address. Add ares_config.h as dependency for 'make depend'.

14 years agoT_SRV portability check
Yang Tse [Sun, 6 Sep 2009 23:46:07 +0000 (23:46 +0000)]
T_SRV portability check

14 years agoadded casts to silent compiler warning on 64bit systems.
Gunter Knauf [Sun, 6 Sep 2009 20:16:59 +0000 (20:16 +0000)]
added casts to silent compiler warning on 64bit systems.

14 years agouse our define struct_stat to be compatible with largefile support.
Gunter Knauf [Sun, 6 Sep 2009 19:45:08 +0000 (19:45 +0000)]
use our define struct_stat to be compatible with largefile support.

14 years agoadded base64.h include to silent warnings about missing prototype for ATOB_ConvertAsc...
Gunter Knauf [Sun, 6 Sep 2009 17:59:46 +0000 (17:59 +0000)]
added base64.h include to silent warnings about missing prototype for ATOB_ConvertAsciiToItem.

14 years agolanguage! / take2
Gunter Knauf [Sun, 6 Sep 2009 00:00:19 +0000 (00:00 +0000)]
language! / take2

14 years agochanged includes to match style how we do with all other *.c files.
Gunter Knauf [Sat, 5 Sep 2009 20:11:32 +0000 (20:11 +0000)]
changed includes to match style how we do with all other *.c files.

14 years agoadd ftpgetinfo
Daniel Stenberg [Sat, 5 Sep 2009 17:54:30 +0000 (17:54 +0000)]
add ftpgetinfo

14 years agoMake CURLOPT_FTPPORT support an additional port number
Daniel Stenberg [Sat, 5 Sep 2009 17:49:07 +0000 (17:49 +0000)]
Make CURLOPT_FTPPORT support an additional port number

done!

14 years agoadded ftpgetinfo sample since users asked frequently for such a sample.
Gunter Knauf [Sat, 5 Sep 2009 15:23:37 +0000 (15:23 +0000)]
added ftpgetinfo sample since users asked frequently for such a sample.

14 years agochanged u_int16_t to unsigned short because it is the only place within ares and...
Gunter Knauf [Sat, 5 Sep 2009 15:12:30 +0000 (15:12 +0000)]
changed u_int16_t to unsigned short because it is the only place within ares and curl where such a type would be used;
also it broke many autobuilds. We should probably introduce an ares_port_t if we want to use a type here.

14 years agoReplace 'uint16_t' with 'u_int16_t' since the latter is used in ares.h.
Gisle Vanem [Sat, 5 Sep 2009 12:33:08 +0000 (12:33 +0000)]
Replace 'uint16_t' with 'u_int16_t' since the latter is used in ares.h.

14 years agoAdded 'ares_parse_srv_reply.obj'. Added definition of 'u_int16_t'. This is I don...
Gisle Vanem [Sat, 5 Sep 2009 12:29:35 +0000 (12:29 +0000)]
Added 'ares_parse_srv_reply.obj'. Added definition of 'u_int16_t'. This is I don't like; we should not depend on such non-universal types in a public header. But this is just a quick fix.

14 years ago- Jakub Hrozek added ares_parse_srv_reply() for SRV parsing
Daniel Stenberg [Fri, 4 Sep 2009 21:56:14 +0000 (21:56 +0000)]
- Jakub Hrozek added ares_parse_srv_reply() for SRV parsing

14 years agoadded --with-nss to configure output.
Gunter Knauf [Fri, 4 Sep 2009 17:46:59 +0000 (17:46 +0000)]
added --with-nss to configure output.

14 years agoadded MSVC6 caveats.
Gunter Knauf [Fri, 4 Sep 2009 16:59:10 +0000 (16:59 +0000)]
added MSVC6 caveats.

14 years agoadded chkspeed to samples.
Gunter Knauf [Thu, 3 Sep 2009 17:54:02 +0000 (17:54 +0000)]
added chkspeed to samples.

14 years agoupdated MingW32 makefile for recent external libs.
Gunter Knauf [Thu, 3 Sep 2009 17:53:21 +0000 (17:53 +0000)]
updated MingW32 makefile for recent external libs.

14 years agofix copyright year
Daniel Stenberg [Thu, 3 Sep 2009 08:30:28 +0000 (08:30 +0000)]
fix copyright year

14 years agoprovide and export Curl_parsedate() as a library-wide internal function
Daniel Stenberg [Thu, 3 Sep 2009 08:13:32 +0000 (08:13 +0000)]
provide and export Curl_parsedate() as a library-wide internal function
for a better API to date parsing than the external API is

14 years agolanguage!
Daniel Stenberg [Wed, 2 Sep 2009 21:13:24 +0000 (21:13 +0000)]
language!

14 years ago- Daniel Johnson founded a flaw in the code convering sftp-errors to libcurl
Daniel Stenberg [Wed, 2 Sep 2009 21:05:47 +0000 (21:05 +0000)]
- Daniel Johnson founded a flaw in the code convering sftp-errors to libcurl
  errors.

14 years agoextended CURL_CHECK_PKGCONFIG to check for a host-specific version of the
Daniel Stenberg [Wed, 2 Sep 2009 17:48:26 +0000 (17:48 +0000)]
extended CURL_CHECK_PKGCONFIG to check for a host-specific version of the
pkg-config first before the "normal" one (if cross-compiling)

14 years agospell out that when running_handles is zero, there is no longer any transfers
Daniel Stenberg [Wed, 2 Sep 2009 14:57:05 +0000 (14:57 +0000)]
spell out that when running_handles is zero, there is no longer any transfers
going on

14 years agocomment the "stunnel exited with [num]" stuff since this is now happen
Daniel Stenberg [Tue, 1 Sep 2009 17:05:24 +0000 (17:05 +0000)]
comment the "stunnel exited with [num]" stuff since this is now happen
in normal tests and is annoying

14 years ago- Peter Sylvester made a debug featuer for Curl_resolv() that now will force
Daniel Stenberg [Tue, 1 Sep 2009 14:27:01 +0000 (14:27 +0000)]
- Peter Sylvester made a debug featuer for Curl_resolv() that now will force
  libcurl to resolve 'localhost' whatever name you use in the URL *if* you set
  the --interface option to (exactly) "LocalHost". This will enable us to
  write tests for custom hosts names but still use a local host server.

14 years agoa failed libssh2 check must not pollute the variables
Daniel Stenberg [Tue, 1 Sep 2009 08:39:39 +0000 (08:39 +0000)]
a failed libssh2 check must not pollute the variables

14 years agoclarify: building outside of the source tree is not special for the
Daniel Stenberg [Tue, 1 Sep 2009 07:02:37 +0000 (07:02 +0000)]
clarify: building outside of the source tree is not special for the
test script

14 years ago- configure now tries to use pkg-config for a number of sub-dependencies even
Daniel Stenberg [Tue, 1 Sep 2009 06:53:01 +0000 (06:53 +0000)]
- configure now tries to use pkg-config for a number of sub-dependencies even
  when cross-compiling. The key to success is then you properly setup
  PKG_CONFIG_PATH before invoking configure.

  I also improved how NSS is detected by trying nss-config if pkg-config isn't
  present, and as a last resort just use the lib name and force the user to
  setup the LIBS/LDFLAGS/CFLAGS etc properly. The previous last resort would
  add a range of various libs that would almost never be quite correct.

14 years ago73. if a connection is made to a FTP server but the server then just never
Daniel Stenberg [Mon, 31 Aug 2009 21:57:24 +0000 (21:57 +0000)]
73. if a connection is made to a FTP server but the server then just never
  sends the 220 response or otherwise is dead slow, libcurl will not
  acknowledge the connection timeout during that phase but only the "real"
  timeout - which may surprise users as it is probably considered to be the
  connect phase to most people. Brought up (and is being misunderstood) in:
  http://curl.haxx.se/bug/view.cgi?id=2844077

14 years ago- When using the multi interface with FTP and you asked for NOBODY, you did no
Daniel Stenberg [Mon, 31 Aug 2009 20:49:30 +0000 (20:49 +0000)]
- When using the multi interface with FTP and you asked for NOBODY, you did no
  QUOTE commands and the request used the same path as the connection had
  already changed to, it would decide that no commands would be necessary for
  the "DO" action and that was not handled properly but libcurl would instead
  hang.

14 years agoAdded a warning comment about curlbuild.h inclusion leading path.
Patrick Monnerat [Mon, 31 Aug 2009 15:48:52 +0000 (15:48 +0000)]
Added a warning comment about curlbuild.h inclusion leading path.

14 years agoOS400: adjust inclusion paths in include files to avoid "include file not found"...
Patrick Monnerat [Mon, 31 Aug 2009 15:22:16 +0000 (15:22 +0000)]
OS400: adjust inclusion paths in include files to avoid "include file not found" in DB2 mode.

14 years agoadded curl prefix for curlbuild.h back until we find a better solution;
Gunter Knauf [Mon, 31 Aug 2009 13:58:47 +0000 (13:58 +0000)]
added curl prefix for curlbuild.h back until we find a better solution;
it breaks almost all autobuilds.

14 years agoadded simple chkspeed sample.
Gunter Knauf [Mon, 31 Aug 2009 02:06:19 +0000 (02:06 +0000)]
added simple chkspeed sample.

14 years agoremoved curl prefix for curlbuild.h as Patrick suggested.
Gunter Knauf [Mon, 31 Aug 2009 02:05:04 +0000 (02:05 +0000)]
removed curl prefix for curlbuild.h as Patrick suggested.

14 years agoadd casts to silent compiler warnings with 64bit systems.
Gunter Knauf [Sun, 30 Aug 2009 01:36:01 +0000 (01:36 +0000)]
add casts to silent compiler warnings with 64bit systems.

14 years agoadd cast to silent compiler warning with 64bit systems.
Gunter Knauf [Sat, 29 Aug 2009 15:44:31 +0000 (15:44 +0000)]
add cast to silent compiler warning with 64bit systems.

14 years agoadd cast to silent compiler warning with 64bit systems.
Gunter Knauf [Sat, 29 Aug 2009 04:34:44 +0000 (04:34 +0000)]
add cast to silent compiler warning with 64bit systems.

14 years agofix shadow definition of outp.
Gunter Knauf [Sat, 29 Aug 2009 04:26:02 +0000 (04:26 +0000)]
fix shadow definition of outp.

14 years agoadd cast to silent compiler warning with 64bit systems.
Gunter Knauf [Sat, 29 Aug 2009 04:12:51 +0000 (04:12 +0000)]
add cast to silent compiler warning with 64bit systems.

14 years agoremoved obsolete casts.
Gunter Knauf [Sat, 29 Aug 2009 03:57:28 +0000 (03:57 +0000)]
removed obsolete casts.

14 years agoadd casts to silent compiler warnings with 64bit systems.
Gunter Knauf [Sat, 29 Aug 2009 03:53:09 +0000 (03:53 +0000)]
add casts to silent compiler warnings with 64bit systems.

14 years agoadd cast to silient compiler warning with 64bit systems.
Gunter Knauf [Sat, 29 Aug 2009 03:42:13 +0000 (03:42 +0000)]
add cast to silient compiler warning with 64bit systems.

14 years ago- Improved error message for not matching certificate subject name in
Kamil Dudka [Fri, 28 Aug 2009 12:06:51 +0000 (12:06 +0000)]
- Improved error message for not matching certificate subject name in
  libcurl-NSS. Originally reported at:
  https://bugzilla.redhat.com/show_bug.cgi?id=516056#c9

14 years agodisable the tests that differ between 32 and 64 bit
Daniel Stenberg [Thu, 27 Aug 2009 18:15:38 +0000 (18:15 +0000)]
disable the tests that differ between 32 and 64 bit

14 years agoSupport lookup of IPv4 literals in ares_gethostbyname(), even when the address family...
Steinar H. Gunderson [Thu, 27 Aug 2009 09:53:55 +0000 (09:53 +0000)]
Support lookup of IPv4 literals in ares_gethostbyname(), even when the address family is set to AF_INET6.

14 years ago- Introduced a SYST-based test to properly set-up name format when dealing with...
Patrick Monnerat [Mon, 24 Aug 2009 12:57:25 +0000 (12:57 +0000)]
- Introduced a SYST-based test to properly set-up name format when dealing  with the OS/400 FTP server.
- Fixed an ftp_readresp() bug preventing detection of failing control socket  and causing FTP client to loop forever.

14 years ago- Marc de Bruin pointed out that configure --with-gnutls=PATH didn't work
Daniel Stenberg [Mon, 24 Aug 2009 11:38:59 +0000 (11:38 +0000)]
- Marc de Bruin pointed out that configure --with-gnutls=PATH didn't work
  properly and provided a fix. http://curl.haxx.se/bug/view.cgi?id=2843008

14 years ago- Eric Wong introduced support for the new option -T. (dot) that makes curl
Daniel Stenberg [Mon, 24 Aug 2009 10:57:17 +0000 (10:57 +0000)]
- Eric Wong introduced support for the new option -T. (dot) that makes curl
  read stdin in a non-blocking fashion. This also brings back -T- (minus) to
  the previous blocking behavior since it could break stuff for people at
  times.

14 years agoclarify the code by initing newurl to NULL
Daniel Stenberg [Mon, 24 Aug 2009 08:41:17 +0000 (08:41 +0000)]
clarify the code by initing newurl to NULL

14 years agoWith CURLOPT_PROXY_TRANSFER_MODE, avoid sending invalid URLs like
Michal Marek [Fri, 21 Aug 2009 12:15:35 +0000 (12:15 +0000)]
With CURLOPT_PROXY_TRANSFER_MODE, avoid sending invalid URLs like
ftp://example.com;type=i if the user specified ftp://example.com without the
slash.

14 years ago- Andre Guibert de Bruet pointed out a missing return code check for a
Daniel Stenberg [Fri, 21 Aug 2009 12:01:36 +0000 (12:01 +0000)]
- Andre Guibert de Bruet pointed out a missing return code check for a
  strdup() that could lead to segfault if it returned NULL. I extended his
  suggest patch to now have Curl_retry_request() return a regular return code
  and better check that.

14 years ago- Lots of good work by Krister Johansen, mostly related to pipelining:
Daniel Stenberg [Fri, 21 Aug 2009 07:11:20 +0000 (07:11 +0000)]
- Lots of good work by Krister Johansen, mostly related to pipelining:

  Fix SIGSEGV on free'd easy_conn when pipe unexpectedly breaks
  Fix data corruption issue with re-connected transfers
  Fix use after free if we're completed but easy_conn not NULL

14 years ago5.3 Sort outgoing cookies
Daniel Stenberg [Wed, 19 Aug 2009 07:09:13 +0000 (07:09 +0000)]
5.3 Sort outgoing cookies
5.4 Rearrange request header order

Things to play with when you're bored

14 years agothe mozilla browser is called Firefox...
Daniel Stenberg [Tue, 18 Aug 2009 19:51:18 +0000 (19:51 +0000)]
the mozilla browser is called Firefox...

14 years agoPull the certificate files from the source directory. Ensure that the
Dan Fandrich [Tue, 18 Aug 2009 00:18:54 +0000 (00:18 +0000)]
Pull the certificate files from the source directory. Ensure that the
certificate tests only run on a localhost-hosted test server since the
host name is explicitly checked.

14 years agoadded missing curl_easy_pause to export list.
Gunter Knauf [Sun, 16 Aug 2009 15:03:58 +0000 (15:03 +0000)]
added missing curl_easy_pause to export list.

14 years agoI think it's worth clarifying that curl DOES NOT validate a given URL more
Daniel Stenberg [Fri, 14 Aug 2009 18:09:42 +0000 (18:09 +0000)]
I think it's worth clarifying that curl DOES NOT validate a given URL more
than what's absolutely necessary:

curl will do its best to use what you pass to it as a URL. It is not trying to
validate it as a syntactically correct URL by any means but is instead
VERY liberal with what it accepts.

14 years ago- Changed NSS code to not ignore the value of ssl.verifyhost and produce more
Kamil Dudka [Thu, 13 Aug 2009 16:04:51 +0000 (16:04 +0000)]
- Changed NSS code to not ignore the value of ssl.verifyhost and produce more
  verbose error messages. Originally reported at:
  https://bugzilla.redhat.com/show_bug.cgi?id=516056

14 years agomention yesterday's changes
Daniel Stenberg [Thu, 13 Aug 2009 08:51:45 +0000 (08:51 +0000)]
mention yesterday's changes

14 years agoadd missing file, as pointed out by Karl M
Daniel Stenberg [Wed, 12 Aug 2009 17:25:16 +0000 (17:25 +0000)]
add missing file, as pointed out by Karl M

14 years agostart over fresh again towards 7.19.7
Daniel Stenberg [Wed, 12 Aug 2009 11:24:52 +0000 (11:24 +0000)]
start over fresh again towards 7.19.7

14 years agoimported names from the 7.19.6 RELEASE-NOTES
Daniel Stenberg [Wed, 12 Aug 2009 11:24:09 +0000 (11:24 +0000)]
imported names from the 7.19.6 RELEASE-NOTES

14 years agoAdded a range of new fun date strings to try. This set of dates come from a
Daniel Stenberg [Wed, 12 Aug 2009 11:18:55 +0000 (11:18 +0000)]
Added a range of new fun date strings to try. This set of dates come from a
mail posted to the http-state mailing list, from Adam Barth, and is said to be
the set of date formats the Chrome browser code is tested against:
http://www.ietf.org/mail-archive/web/http-state/current/msg00129.html

libcurl parses most of them identically, but not all of them.

14 years ago7.19.6
Daniel Stenberg [Wed, 12 Aug 2009 08:59:30 +0000 (08:59 +0000)]
7.19.6

14 years ago- Carsten Lange reported a bug and provided a patch for TFTP upload and the
Daniel Stenberg [Wed, 12 Aug 2009 08:19:39 +0000 (08:19 +0000)]
- Carsten Lange reported a bug and provided a patch for TFTP upload and the
  sending of the TSIZE option. I don't like fixing bugs just hours before
  a release, but since it was broken and the patch fixes this for him I decided
  to get it in anyway.

14 years agouse --insecure to allow non-matching known hosts for SSH-based protocols
Daniel Stenberg [Wed, 12 Aug 2009 08:14:53 +0000 (08:14 +0000)]
use --insecure to allow non-matching known hosts for SSH-based protocols

14 years agopasted here (and renumbered) from the TODO-RELEASE since they are in fact
Daniel Stenberg [Wed, 12 Aug 2009 08:02:42 +0000 (08:02 +0000)]
pasted here (and renumbered) from the TODO-RELEASE since they are in fact
bugs we know about that will appear in the next release (too)

14 years ago- Peter Sylvester made the HTTPS test server use specific certificates for
Daniel Stenberg [Tue, 11 Aug 2009 21:48:58 +0000 (21:48 +0000)]
- Peter Sylvester made the HTTPS test server use specific certificates for
  each test, so that the test suite can now be used to actually test the
  verification of cert names etc. This made an error show up in the OpenSSL-
  specific code where it would attempt to match the CN field even if a
  subjectAltName exists that doesn't match. This is now fixed and verified
  in test 311.

14 years agocredit
Daniel Stenberg [Tue, 11 Aug 2009 20:57:15 +0000 (20:57 +0000)]
credit

14 years ago- Benbuck Nason posted the bug report #2835196
Daniel Stenberg [Tue, 11 Aug 2009 20:43:12 +0000 (20:43 +0000)]
- Benbuck Nason posted the bug report #2835196
  (http://curl.haxx.se/bug/view.cgi?id=2835196), fixing a few compiler
  warnings when mixing ints and bools.

14 years agoInclude the Android make file in the source package even though the
Dan Fandrich [Tue, 11 Aug 2009 18:11:40 +0000 (18:11 +0000)]
Include the Android make file in the source package even though the
config.h issue hasn't been completely solved.  This will save some effort
for someone desperate to use curl on Android.

14 years agoFix definition of CURLOPT_SOCKS5_GSSAPI_SERVICE from LONG to OBJECTPOINT
Patrick Monnerat [Tue, 11 Aug 2009 14:07:08 +0000 (14:07 +0000)]
Fix definition of CURLOPT_SOCKS5_GSSAPI_SERVICE from LONG to OBJECTPOINT
Fix OS400 makefile for tests to use the new Makefile.inc in libtest
Update the OS400 wrappers and RPG binding according to the current CVS source state

14 years agoAdded links to more details on most issues. Moved all these issues to 7.19.7
Daniel Stenberg [Tue, 11 Aug 2009 07:56:16 +0000 (07:56 +0000)]
Added links to more details on most issues. Moved all these issues to 7.19.7
now since we won't manage to get them done for 7.19.6.

14 years agoFixed a memory leak in the FTP code and an off-by-one heap buffer overflow.
Dan Fandrich [Tue, 11 Aug 2009 02:30:53 +0000 (02:30 +0000)]
Fixed a memory leak in the FTP code and an off-by-one heap buffer overflow.

14 years agoFixed some memory leaks in the command-line tool that caused most of the
Dan Fandrich [Sun, 9 Aug 2009 23:15:20 +0000 (23:15 +0000)]
Fixed some memory leaks in the command-line tool that caused most of the
torture tests to fail.

14 years agofix cast for some systems which are broken due to absense of socklen_t, therefore...
Gunter Knauf [Fri, 7 Aug 2009 23:32:38 +0000 (23:32 +0000)]
fix cast for some systems which are broken due to absense of socklen_t, therefore now use curl_socklen_t.

14 years agoadded a cast to silent compiler warning with 64bit systems.
Gunter Knauf [Thu, 6 Aug 2009 13:23:00 +0000 (13:23 +0000)]
added a cast to silent compiler warning with 64bit systems.

14 years agofixed cast added with last commit.
Gunter Knauf [Thu, 6 Aug 2009 12:02:00 +0000 (12:02 +0000)]
fixed cast added with last commit.

14 years agocast to fix 64bit build warnings. From manpage:
Gunter Knauf [Thu, 6 Aug 2009 11:10:30 +0000 (11:10 +0000)]
cast to fix 64bit build warnings. From manpage:
POSIX.1-2001. Note that RFC 2553 defines a prototype where the last parameter cnt is of type size_t.
Many systems follow RFC 2553. Glibc 2.0 and 2.1 have size_t, but 2.2 has socklen_t.

14 years agoRFC1867 was updated by RFC2388
Daniel Stenberg [Tue, 4 Aug 2009 12:02:27 +0000 (12:02 +0000)]
RFC1867 was updated by RFC2388

14 years agoavoid possible NULL dereference caused by my previous fix
Daniel Stenberg [Mon, 3 Aug 2009 14:07:57 +0000 (14:07 +0000)]
avoid possible NULL dereference caused by my previous fix

14 years agoRemove call to LoadLibrary(). (leftover from debugging).
Gisle Vanem [Mon, 3 Aug 2009 12:20:03 +0000 (12:20 +0000)]
Remove call to LoadLibrary(). (leftover from debugging).

14 years agoFix bad sentence.
Gisle Vanem [Mon, 3 Aug 2009 12:16:00 +0000 (12:16 +0000)]
Fix bad sentence.

14 years ago- Timo Teras changed the reason code used in the resolve callback done when
Daniel Stenberg [Mon, 3 Aug 2009 11:51:06 +0000 (11:51 +0000)]
- Timo Teras changed the reason code used in the resolve callback done when
  ares_cancel() is used, to be ARES_ECANCELLED instead of ARES_ETIMEOUT to
  better allow the callback to know what's happening.

14 years ago256 - "More questions about ares behavior"
Daniel Stenberg [Mon, 3 Aug 2009 11:39:43 +0000 (11:39 +0000)]
256 - "More questions about ares behavior"
yet another issue not yet sorted out

14 years agoindentation fixes only
Daniel Stenberg [Mon, 3 Aug 2009 11:32:55 +0000 (11:32 +0000)]
indentation fixes only

14 years ago- Joshua Kwan fixed the init routine to fill in the defaults for stuff that
Daniel Stenberg [Mon, 3 Aug 2009 11:29:17 +0000 (11:29 +0000)]
- Joshua Kwan fixed the init routine to fill in the defaults for stuff that
  fails to get inited by other means. This fixes a case of when the c-ares
  init fails when internet access is fone.

14 years agorespect error code from ftruncate(), mentioned by Peter Sylvester
Daniel Stenberg [Mon, 3 Aug 2009 09:06:35 +0000 (09:06 +0000)]
respect error code from ftruncate(), mentioned by Peter Sylvester

14 years agoReverted the zero-byte-in-name check to instead rely on the fact that strlen
Daniel Stenberg [Mon, 3 Aug 2009 08:45:19 +0000 (08:45 +0000)]
Reverted the zero-byte-in-name check to instead rely on the fact that strlen
and the name length differ in those cases and thus leave the matching function
unmodified from before, as the matching functions never have to bother with
the zero bytes in legitimate cases. Peter Sylvester helped me realize that
this fix is slightly better as it leaves more code unmodified and makes the
detection a bit more obvious in the code.

14 years agoclarified configure detection of GnuTLS
Daniel Stenberg [Sun, 2 Aug 2009 22:38:15 +0000 (22:38 +0000)]
clarified configure detection of GnuTLS

14 years agoExtended my embedded-zero-in-cert-name fix based on a comment from Scott
Daniel Stenberg [Sun, 2 Aug 2009 22:34:00 +0000 (22:34 +0000)]
Extended my embedded-zero-in-cert-name fix based on a comment from Scott
Cantor. My previous attempt was half-baked and didn't cover the normal CN
case.

14 years agomention two crashing bugs we'd like fixed
Daniel Stenberg [Sun, 2 Aug 2009 21:43:04 +0000 (21:43 +0000)]
mention two crashing bugs we'd like fixed

14 years agoclarify the description of the null byte in cert name fix
Daniel Stenberg [Sat, 1 Aug 2009 22:18:37 +0000 (22:18 +0000)]
clarify the description of the null byte in cert name fix

14 years ago- Curt Bogmine reported a problem with SNI enabled on a particular server. We
Daniel Stenberg [Sat, 1 Aug 2009 22:11:58 +0000 (22:11 +0000)]
- Curt Bogmine reported a problem with SNI enabled on a particular server. We
  should introduce an option to disable SNI, but as we're in feature freeze
  now I've addressed the obvious bug here (pointed out by Peter Sylvester): we
  shouldn't try to enable SNI when SSLv2 or SSLv3 is explicitly selected.
  Code for OpenSSL and GnuTLS was fixed. NSS doesn't seem to have a particular
  option for SNI, or are we simply not using it?

14 years ago- Scott Cantor posted the bug report #2829955
Daniel Stenberg [Sat, 1 Aug 2009 21:56:59 +0000 (21:56 +0000)]
- Scott Cantor posted the bug report #2829955
  (http://curl.haxx.se/bug/view.cgi?id=2829955) mentioning the recent SSL cert
  verification flaw found and exploited by Moxie Marlinspike. The presentation
  he did at Black Hat is available here:
  https://www.blackhat.com/html/bh-usa-09/bh-usa-09-archives.html#Marlinspike

  Apparently at least one CA allowed a subjectAltName or CN that contain a
  zero byte, and thus clients that assumed they would never have zero bytes
  were exploited to OK a certificate that didn't actually match the site. Like
  if the name in the cert was "example.com\0theatualsite.com", libcurl would
  happily verify that cert for example.com.

  libcurl now better use the length of the extracted name, not assuming it is
  zero terminated.

14 years ago- Tanguy Fautre pointed out that OpenSSL's function RAND_screen() (present
Daniel Stenberg [Sat, 1 Aug 2009 11:09:02 +0000 (11:09 +0000)]
- Tanguy Fautre pointed out that OpenSSL's function RAND_screen() (present
  only in some OpenSSL installs - like on Windows) isn't thread-safe and we
  agreed that moving it to the global_init() function is a decent way to deal
  with this situation.

14 years ago- Alexander Beedie provided the patch for a noproxy problem: If I have set
Daniel Stenberg [Sat, 1 Aug 2009 11:02:10 +0000 (11:02 +0000)]
- Alexander Beedie provided the patch for a noproxy problem: If I have set
  CURLOPT_NOPROXY to "*", or to a host that should not use a proxy, I actually
  could still end up using a proxy if a proxy environment variable was set.