platform/upstream/curl.git
11 years agobuild: prevent global LIBS from influencing libtest build targets
Yang Tse [Fri, 30 Nov 2012 13:34:48 +0000 (14:34 +0100)]
build: prevent global LIBS from influencing libtest build targets

11 years agobuild: prevent global LIBS from influencing test server build targets
Yang Tse [Thu, 29 Nov 2012 20:56:48 +0000 (21:56 +0100)]
build: prevent global LIBS from influencing test server build targets

11 years agobuild: fix Windows build targets damaged since commit 550e403f00
Yang Tse [Wed, 28 Nov 2012 19:51:13 +0000 (20:51 +0100)]
build: fix Windows build targets damaged since commit 550e403f00

11 years agobuild: avoid linkage of directly unused libraries
Yang Tse [Wed, 28 Nov 2012 17:31:23 +0000 (18:31 +0100)]
build: avoid linkage of directly unused libraries

11 years agodd missing NTLM feature for tests 2025, and 2028 to 2032
Yang Tse [Tue, 27 Nov 2012 17:01:55 +0000 (18:01 +0100)]
dd missing NTLM feature for tests 2025, and 2028 to 2032

11 years agoavoid mixing of enumerated type with another type
Yang Tse [Mon, 26 Nov 2012 15:23:02 +0000 (16:23 +0100)]
avoid mixing of enumerated type with another type

11 years agomulti.c: disambiguate precedence of bitwise and relational operation
Yang Tse [Mon, 26 Nov 2012 15:20:53 +0000 (16:20 +0100)]
multi.c: disambiguate precedence of bitwise and relational operation

11 years agoRemove stray CRLF in chunk-encoded content-free request bodies
Fabian Keil [Sun, 25 Nov 2012 17:34:01 +0000 (18:34 +0100)]
Remove stray CRLF in chunk-encoded content-free request bodies

.. that are sent when auth-negotiating before a chunked
upload or when setting the 'Transfer-Encoding: chunked'
header and intentionally sending no content.

Adjust test565 and test1333 accordingly.

11 years agoFAQ: clarify the 3.4 section
Daniel Stenberg [Sun, 25 Nov 2012 18:44:49 +0000 (19:44 +0100)]
FAQ: clarify the 3.4 section

You can do custom commands to FTP without sending anything by using the
CURLOPT_NOBODY, which -I sets.

11 years agoexamples: Updated asiohiper.cpp to remove connect from opensocket
Lijo Antony [Sun, 25 Nov 2012 06:00:58 +0000 (10:00 +0400)]
examples: Updated asiohiper.cpp to remove connect from opensocket

Blocking connect on the socket has been removed from opensocket
callback. opensocket just opens a new socket and gives it back to
libcurl and libcurl will take care of the connect. sockopt_callback has
also been removed, as it is no longer required.

11 years agobuild: fix AIX compilation and usage
Yang Tse [Fri, 23 Nov 2012 16:57:00 +0000 (17:57 +0100)]
build: fix AIX compilation and usage

AIX sys/poll.h header file defines 'events' and 'revents' as C
preprocessor macros. Usage of these literals in libcurl's external
API was introduced in commit de24d7bd4c causing AIX build failures.
Appropriate inclusion of sys/poll.h by libcurl's external interface
fixes AIX build and usage issues while avoiding a SONAME bump.

11 years agoDOCS: Updated CURLOPT_CONNECT_ONLY to reflect usage in other protocols
Steve Holme [Fri, 23 Nov 2012 13:43:53 +0000 (13:43 +0000)]
DOCS: Updated CURLOPT_CONNECT_ONLY to reflect usage in other protocols

11 years agotest: offer "automake" output and check for perl better
Daniel Stenberg [Fri, 23 Nov 2012 10:56:26 +0000 (11:56 +0100)]
test: offer "automake" output and check for perl better

runtests.pl -am now uses the "PASS/FAIL: [desc]" output for each
executed test. You can run 'make test-am' in the root build directory to
invoke that. The reason for this output style is to better allow generic
test suite parsers to also grok our test output.

The test Makefile now also tests that perl was indeed found and that the
PERL variable points to an executable before it tries to run the main
test perl script runtests.pl,

11 years agoTest 206: Use a Content-Length header for the 407 response
Fabian Keil [Wed, 21 Nov 2012 15:01:59 +0000 (16:01 +0100)]
Test 206: Use a Content-Length header for the 407 response

Otherwise curl would have to guess where the body ends.

11 years agoTest 206: Don't respond to a succesful CONNECT request with a body
Fabian Keil [Wed, 21 Nov 2012 14:42:00 +0000 (15:42 +0100)]
Test 206: Don't respond to a succesful CONNECT request with a body

It's against the spec and caused test failures when header
and response were read from the network separately in which
case bug #39 wasn't triggered.

11 years agohtmltitle: use .cpp extension for C++ examples
Daniel Stenberg [Wed, 21 Nov 2012 17:26:42 +0000 (18:26 +0100)]
htmltitle: use .cpp extension for C++ examples

11 years agoexamples: Added a c++ example of using multi with boost::asio
Lijo Antony [Wed, 21 Nov 2012 10:19:45 +0000 (14:19 +0400)]
examples: Added a c++ example of using multi with boost::asio

Added an example for demonstrating the usage of curl multi interface
with boost::asio in c++

11 years agoVC Makefiles: add missing hostcheck
Daniel Stenberg [Wed, 21 Nov 2012 15:17:45 +0000 (16:17 +0100)]
VC Makefiles: add missing hostcheck

the newly introduced hostcheck.h/c is missing in the Visual Studio
Makefiles as obj file.

Bug: http://curl.haxx.se/mail/lib-2012-11/0176.html

11 years agocompiler warning fixes
Daniel Stenberg [Tue, 20 Nov 2012 19:57:18 +0000 (20:57 +0100)]
compiler warning fixes

The conversions from ssize_t to int need to be typecasted.

11 years agobump: start working on 7.28.2
Daniel Stenberg [Tue, 20 Nov 2012 09:27:36 +0000 (10:27 +0100)]
bump: start working on 7.28.2

11 years agoTHANKS: added 14 contributors from the 7.28.1 release
Daniel Stenberg [Tue, 20 Nov 2012 09:25:25 +0000 (10:25 +0100)]
THANKS: added 14 contributors from the 7.28.1 release

11 years agoRELEASE-NOTES: synced with 52af6e69f079 / 7.28.1
Daniel Stenberg [Mon, 19 Nov 2012 21:58:05 +0000 (22:58 +0100)]
RELEASE-NOTES: synced with 52af6e69f079 / 7.28.1

11 years agoRELEASE-NOTES: NSS can be used for metalink hashing
Anthony Bryan [Mon, 19 Nov 2012 23:13:21 +0000 (00:13 +0100)]
RELEASE-NOTES: NSS can be used for metalink hashing

11 years agoGet test 2032 working when using valgrind
Fabian Keil [Sun, 18 Nov 2012 21:42:04 +0000 (22:42 +0100)]
Get test 2032 working when using valgrind

If curl_multi_fdset() sets maxfd to -1, the socket detection
loop is skipped and thus !found_new_socket is no cause for alarm.

11 years agotest2032: spurious failure caused by premature termination
Kamil Dudka [Mon, 15 Oct 2012 19:42:33 +0000 (21:42 +0200)]
test2032: spurious failure caused by premature termination

Bug: http://curl.haxx.se/mail/lib-2012-11/0095.html

11 years agoFix comment typos in test 517
Fabian Keil [Wed, 14 Nov 2012 11:53:47 +0000 (12:53 +0100)]
Fix comment typos in test 517

11 years agoTest 92 and 194: normalize spaces in the Server headers
Fabian Keil [Tue, 13 Nov 2012 12:58:14 +0000 (13:58 +0100)]
Test 92 and 194: normalize spaces in the Server headers

It makes no difference from curl's point of view but
makes it more convenient to use the tests with a
lws-normalizing proxy between curl and the test server.

11 years agoAdd a HOSTIP precheck for tests 31 and 1105
Fabian Keil [Sun, 18 Nov 2012 13:55:29 +0000 (14:55 +0100)]
Add a HOSTIP precheck for tests 31 and 1105

They currently only work for 127.0.0.1 which
is hardcoded and can't be easily changed.

11 years agoLet test 8 work as long as %HOSTIP ends with ".0.0.1"
Fabian Keil [Thu, 15 Nov 2012 13:02:21 +0000 (14:02 +0100)]
Let test 8 work as long as %HOSTIP ends with ".0.0.1"

.. and add a precheck to skip the test otherwise.

11 years agoAdd --resolve to the keywords and name of test 1318
Fabian Keil [Wed, 14 Nov 2012 18:44:11 +0000 (19:44 +0100)]
Add --resolve to the keywords and name of test 1318

This makes it easier to skip it automatically when
the test suite is used with external proxies.

11 years agoAdd FTP keywords for a couple of currently keyword-less FTP tests
Fabian Keil [Fri, 16 Nov 2012 11:42:51 +0000 (12:42 +0100)]
Add FTP keywords for a couple of currently keyword-less FTP tests

11 years agoAdd keywords for a couple of currently keyword-less HTTP tests
Fabian Keil [Tue, 13 Nov 2012 13:02:42 +0000 (14:02 +0100)]
Add keywords for a couple of currently keyword-less HTTP tests

11 years agoUse carriage returns in all headers in test 31
Fabian Keil [Fri, 16 Nov 2012 12:54:59 +0000 (13:54 +0100)]
Use carriage returns in all headers in test 31

Trailing spaces were left unmodifed, assuming they were intentional.

11 years agoDo not mix CRLF and LF header endings in a couple of HTTP tests
Fabian Keil [Wed, 14 Nov 2012 18:49:12 +0000 (19:49 +0100)]
Do not mix CRLF and LF header endings in a couple of HTTP tests

Consistently use CRLF instead. The mixed endings weren't
documented so I assume they were unintentional.

This change doesn't matter for curl itself but makes using
the tests with a proxy between curl and the test server
more convenient.

Tests that consistently use no carriage returns were
left unmodified as one can easily work around this.

11 years agofixed memory leak: CURLOPT_RESOLVE with multi interface
Daniel Stenberg [Sun, 18 Nov 2012 15:17:37 +0000 (16:17 +0100)]
fixed memory leak: CURLOPT_RESOLVE with multi interface

DNS cache entries populated with CURLOPT_RESOLVE were not properly freed
again when done using the multi interface.

Test case 1502 added to verify.

Bug: http://curl.haxx.se/bug/view.cgi?id=3575448
Reported by: Alex Gruz

11 years agoRELEASE-NOTES: synced with ee588fe08807778
Daniel Stenberg [Sat, 17 Nov 2012 13:23:41 +0000 (14:23 +0100)]
RELEASE-NOTES: synced with ee588fe08807778

4 more bug fixes and 4 more contributors

11 years agomem-include-scan: verify memory #includes
Daniel Stenberg [Fri, 16 Nov 2012 23:59:42 +0000 (00:59 +0100)]
mem-include-scan: verify memory #includes

If we use memory functions (malloc, free, strdup etc) in C sources in
libcurl and we fail to include curl_memory.h or memdebug.h we either
fail to properly support user-provided memory callbacks or the memory
leak system of the test suite fails.

After Ajit's report of a failure in the first category in http_proxy.c,
I spotted a few in the second category as well. These problems are now
tested for by test 1132 which runs a perl program that scans for and
attempts to check that we use the correct include files if a memory
related function is used in the source code.

Reported by: Ajit Dhumale
Bug: http://curl.haxx.se/mail/lib-2012-11/0125.html

11 years agotftp_rx: code style cleanup
Daniel Stenberg [Fri, 16 Nov 2012 21:00:17 +0000 (22:00 +0100)]
tftp_rx: code style cleanup

Fixed checksrc warnings

11 years agoFix the libauthretry changes from 7c0cbcf2f61
Fabian Keil [Fri, 16 Nov 2012 18:22:12 +0000 (19:22 +0100)]
Fix the libauthretry changes from 7c0cbcf2f61

They broke the NTLM tests from 2023 to 2031.

11 years agotftp_rx: handle resends
Christian Vogt [Fri, 16 Nov 2012 14:30:52 +0000 (15:30 +0100)]
tftp_rx: handle resends

Re-send ACK for block X in case we receive block X data again while
waiting for block X+1.

Based on an earlier patch by Marcin Adamski.

11 years agoautoconf: don't force-disable compiler debug option
Daniel Stenberg [Fri, 16 Nov 2012 12:06:49 +0000 (13:06 +0100)]
autoconf: don't force-disable compiler debug option

When nothing is told to configure, we should not enforce switching off
debug options with -g0 (or similar). We instead don't use -g at all in
that situaion and therefore allow the user's CFLAGS settings possibly
dictate what to do.

11 years agowinbuild: Fix PDB file output
Mark Snelling [Wed, 14 Nov 2012 22:19:32 +0000 (23:19 +0100)]
winbuild: Fix PDB file output

And fix some newlines to be proper CRLF

Bug: http://curl.haxx.se/bug/view.cgi?id=3586741

11 years agoRELEASE-NOTES: synced with fa1ae0abcde
Daniel Stenberg [Wed, 14 Nov 2012 21:32:19 +0000 (22:32 +0100)]
RELEASE-NOTES: synced with fa1ae0abcde

11 years agoOpenSSL: Disable SSL/TLS compression
Cristian Rodríguez [Mon, 12 Nov 2012 15:41:58 +0000 (16:41 +0100)]
OpenSSL: Disable SSL/TLS compression

It either causes increased memory usage or exposes users
to the "CRIME attack" (CVE-2012-4929)

11 years agoFILE: Make upload-writes unbuffered by not using FILE streams
Sebastian Rasmussen [Mon, 8 Oct 2012 09:49:47 +0000 (11:49 +0200)]
FILE: Make upload-writes unbuffered by not using FILE streams

11 years agotool_metalink: fix error detection of hash alg initialization
Kamil Dudka [Tue, 13 Nov 2012 12:09:43 +0000 (13:09 +0100)]
tool_metalink: fix error detection of hash alg initialization

The {MD5,SHA1,SHA256}_Init functions from OpenSSL are called directly
without any wrappers and they return 1 for success, 0 otherwise.  Hence,
we have to use the same approach in all the wrapper functions that are
used for the other crypto libraries.

This commit fixes a regression introduced in commit dca8ae5f.

11 years agoRELEASE-NOTES: synced with 7c0cbcf2f617b
Daniel Stenberg [Tue, 13 Nov 2012 12:03:38 +0000 (13:03 +0100)]
RELEASE-NOTES: synced with 7c0cbcf2f617b

11 years agofixed Visual Studio 2010 compilation
Sergei Nikulov [Fri, 9 Nov 2012 13:29:02 +0000 (17:29 +0400)]
fixed Visual Studio 2010 compilation

11 years agoftp: EPSV-disable fix over SOCKS
Anton Malov [Mon, 12 Nov 2012 12:06:31 +0000 (16:06 +0400)]
ftp: EPSV-disable fix over SOCKS

Bug: http://curl.haxx.se/bug/view.cgi?id=3586338

11 years agoMerge branch 'master' of github.com:bagder/curl
Patrick Monnerat [Mon, 12 Nov 2012 13:27:43 +0000 (14:27 +0100)]
Merge branch 'master' of github.com:bagder/curl

11 years agoOS400: upgrade wrappers for the 7.28.1 release.
Patrick Monnerat [Mon, 12 Nov 2012 13:26:16 +0000 (14:26 +0100)]
OS400: upgrade wrappers for the 7.28.1 release.

11 years agoruntests: limit execessive logging/output
Daniel Stenberg [Mon, 12 Nov 2012 12:50:00 +0000 (13:50 +0100)]
runtests: limit execessive logging/output

11 years agoDigst: Add microseconds into nounce calculation
Gabriel Sjoberg [Fri, 9 Nov 2012 23:19:46 +0000 (17:19 -0600)]
Digst: Add microseconds into nounce calculation

When using only 1 second precision, curl doesn't create new cnonce
values quickly enough for all uses.

For example, issuing the following command multiple times to a recent
Tomcat causes authentication failures:

curl --digest -utest:test http://tomcat.test.com:8080/manager/list

This is because curl uses the same cnonce for several seconds, but
doesn't increment the nonce counter.  Tomcat correctly interprets
this as a replay attack and rejects the request.

When microsecond-precision is available, this commit causes curl to
change cnonce values much more frequently.

With microsecond resolution, increasing the nounce length used in the
headers to 32 was made to further reduce the risk of duplication.

11 years agoSCP/SFTP: improve error code used for send failures
Daniel Stenberg [Mon, 12 Nov 2012 09:04:44 +0000 (10:04 +0100)]
SCP/SFTP: improve error code used for send failures

Instead of relying on the generic CURLE error for SCP or SFTP send
failures, try passing back a more suitable error if possible.

11 years agoCurl_write: remove unneeded typecast
Daniel Stenberg [Mon, 12 Nov 2012 09:04:19 +0000 (10:04 +0100)]
Curl_write: remove unneeded typecast

11 years agotool_metalink: allow to use hash algorithms provided by NSS
Kamil Dudka [Tue, 30 Oct 2012 13:59:48 +0000 (14:59 +0100)]
tool_metalink: allow to use hash algorithms provided by NSS

Fixes bug #3578163:
http://sourceforge.net/tracker/?func=detail&atid=100976&aid=3578163&group_id=976

11 years agotool_metalink: allow to handle failure of hash alg initialization
Kamil Dudka [Tue, 30 Oct 2012 13:21:54 +0000 (14:21 +0100)]
tool_metalink: allow to handle failure of hash alg initialization

11 years agotool_metalink: introduce metalink_cleanup() in the internal API
Kamil Dudka [Wed, 31 Oct 2012 09:43:36 +0000 (10:43 +0100)]
tool_metalink: introduce metalink_cleanup() in the internal API

... to release resources allocated at global scope

11 years agohostcheck: only build for the actual users
Daniel Stenberg [Thu, 8 Nov 2012 21:37:53 +0000 (22:37 +0100)]
hostcheck: only build for the actual users

and make local function static

11 years agoSSL: Several SSL-backend related fixes
Oscar Koeroo [Sat, 3 Nov 2012 01:06:51 +0000 (02:06 +0100)]
SSL: Several SSL-backend related fixes

axTLS:

This will make the axTLS backend perform the RFC2818 checks, honoring
the VERIFYHOST setting similar to the OpenSSL backend.

Generic for OpenSSL and axTLS:

Move the hostcheck and cert_hostcheck functions from the lib/ssluse.c
files to make them genericly available for both the OpenSSL, axTLS and
other SSL backends. They are now in the new lib/hostcheck.c file.

CyaSSL:

CyaSSL now also has the RFC2818 checks enabled by default. There is a
limitation that the verifyhost can not be enabled exclusively on the
Subject CN field comparison. This SSL backend will thus behave like the
NSS and the GnuTLS (meaning: RFC2818 ok, or bust). In other words:
setting verifyhost to 0 or 1 will disable the Subject Alt Names checks
too.

Schannel:

Updated the schannel information messages: Split the IP address usage
message from the verifyhost setting and changed the message about
disabling SNI (Server Name Indication, used in HTTP virtual hosting)
into a message stating that the Subject Alternative Names checks are
being disabled when verifyhost is set to 0 or 1. As a side effect of
switching off the RFC2818 related servername checks with
SCH_CRED_NO_SERVERNAME_CHECK
(http://msdn.microsoft.com/en-us/library/aa923430.aspx) the SNI feature
is being disabled. This effect is not documented in MSDN, but Wireshark
output clearly shows the effect (details on the libcurl maillist).

PolarSSL:

Fix the prototype change in PolarSSL of ssl_set_session() and the move
of the peer_cert from the ssl_context to the ssl_session. Found this
change in the PolarSSL SVN between r1316 and r1317 where the
POLARSSL_VERSION_NUMBER was at 0x01010100. But to accommodate the Ubuntu
PolarSSL version 1.1.4 the check is to discriminate between lower then
PolarSSL version 1.2.0 and 1.2.0 and higher. Note: The PolarSSL SVN
trunk jumped from version 1.1.1 to 1.2.0.

Generic:

All the SSL backends are fixed and checked to work with the
ssl.verifyhost as a boolean, which is an internal API change.

11 years agolibcurl: VERSIONINFO update
Daniel Stenberg [Thu, 8 Nov 2012 19:25:08 +0000 (20:25 +0100)]
libcurl: VERSIONINFO update

Since we added the curl_multi_wait function, the VERSIONINFO needed
updating.

Reported by: Patrick Monnerat

11 years agoAdded .def file to output.
Guenter Knauf [Thu, 8 Nov 2012 17:50:48 +0000 (18:50 +0100)]
Added .def file to output.

Requested by Johnny Luong on the libcurl list.

11 years agoAdded deps for static metalink-aware MinGW builds.
Guenter Knauf [Thu, 8 Nov 2012 17:41:59 +0000 (18:41 +0100)]
Added deps for static metalink-aware MinGW builds.

11 years agoFix compilation of lib1501
Fabian Keil [Thu, 8 Nov 2012 15:36:55 +0000 (16:36 +0100)]
Fix compilation of lib1501

11 years agoCurl_readwrite: remove debug output
Daniel Stenberg [Thu, 8 Nov 2012 09:47:11 +0000 (10:47 +0100)]
Curl_readwrite: remove debug output

The text "additional stuff not fine" text was added for debug purposes a
while ago, but it isn't really helping anyone and for some reason some
Linux distributions provide their libcurls built with debug info still
present and thus (far too many) users get to read this info.

11 years agoRELEASE-NOTES: synced with 487538e87a3d5e
Daniel Stenberg [Wed, 7 Nov 2012 22:21:55 +0000 (23:21 +0100)]
RELEASE-NOTES: synced with 487538e87a3d5e

6 new bugfixes and 3 more contributors...

11 years agohttp_perhapsrewind: consider NTLM over proxy too
Daniel Stenberg [Wed, 7 Nov 2012 22:08:29 +0000 (23:08 +0100)]
http_perhapsrewind: consider NTLM over proxy too

The logic previously checked for a started NTLM negotiation only for
host and not also with proxy, leading to problems doing POSTs over a
proxy NTLM that are larger than 2000 bytes. Now it includes proxy in the
check.

Bug: http://curl.haxx.se/bug/view.cgi?id=3582321
Reported by: John Suprock

11 years agoCurl_connecthost: friendlier "couldn't connect" message
Lars Buitinck [Wed, 31 Oct 2012 15:58:53 +0000 (16:58 +0100)]
Curl_connecthost: friendlier "couldn't connect" message

11 years agotest1413: verify redirects to URLs with fragments
Daniel Stenberg [Tue, 6 Nov 2012 22:25:52 +0000 (23:25 +0100)]
test1413: verify redirects to URLs with fragments

The bug report claimed it didn't work. This problem was probably fixed
in 473003fbdf.

Bug: http://curl.haxx.se/bug/view.cgi?id=3581898

11 years agoURL parser: cut off '#' fragments from URLs (better)
Daniel Stenberg [Tue, 6 Nov 2012 22:17:57 +0000 (23:17 +0100)]
URL parser: cut off '#' fragments from URLs (better)

The existing logic only cut off the fragment from the separate 'path'
buffer which is used when sending HTTP to hosts. The buffer that held
the full URL used for proxies were not dealt with. It is now.

Test case 5 was updated to use a fragment on a URL over a proxy.

Bug: http://curl.haxx.se/bug/view.cgi?id=3579813

11 years agoOpenSSL/servercert: use correct buffer size, not size of pointer
Daniel Stenberg [Tue, 6 Nov 2012 21:55:22 +0000 (22:55 +0100)]
OpenSSL/servercert: use correct buffer size, not size of pointer

Bug: http://curl.haxx.se/bug/view.cgi?id=3579286

11 years agocurl: set CURLOPT_SSL_VERIFYHOST to 0 to disable
Daniel Stenberg [Tue, 6 Nov 2012 21:27:25 +0000 (22:27 +0100)]
curl: set CURLOPT_SSL_VERIFYHOST to 0 to disable

11 years agotest 2027/2030: take duplicate Digest requests into account
Daniel Stenberg [Mon, 5 Nov 2012 22:58:31 +0000 (23:58 +0100)]
test 2027/2030: take duplicate Digest requests into account

With the reversion of ce8311c7e49eca and the new clear logic, this flaw
is present and we allow it.

11 years agoCurl_pretransfer: clear out unwanted auth methods
Daniel Stenberg [Mon, 5 Nov 2012 22:31:24 +0000 (23:31 +0100)]
Curl_pretransfer: clear out unwanted auth methods

As a handle can be re-used after having done HTTP auth in a previous
request, it must make sure to clear out the HTTP types that aren't
wanted in this new request.

11 years agotest1412: verify Digest with repeated URLs
Daniel Stenberg [Sun, 4 Nov 2012 22:48:52 +0000 (23:48 +0100)]
test1412: verify Digest with repeated URLs

This test case verifies that bug 3582718 is fixed.

Bug: http://curl.haxx.se/bug/view.cgi?id=3582718
Reported by: Nick Zitzmann (originally)

11 years agoRevert "Zero out auth structs before transfer"
Daniel Stenberg [Sun, 4 Nov 2012 22:47:06 +0000 (23:47 +0100)]
Revert "Zero out auth structs before transfer"

This reverts commit ce8311c7e49eca93c136b58efa6763853541ec97.

The commit made test 2024 work but caused a regression with repeated
Digest authentication. We need to fix this differently.

11 years agoCURLOPT_SSL_VERIFYHOST: stop supporting the 1 value
Daniel Stenberg [Sat, 27 Oct 2012 10:31:39 +0000 (12:31 +0200)]
CURLOPT_SSL_VERIFYHOST: stop supporting the 1 value

After a research team wrote a document[1] that found several live source
codes out there in the wild that misused the CURLOPT_SSL_VERIFYHOST
option thinking it was a boolean, this change now bans 1 as a value and
will make libcurl return error for it.

1 was never a sensible value to use in production but was introduced
back in the days to help debugging. It was always documented clearly
this way.

1 was never supported by all SSL backends in libcurl, so this cleanup
makes the treatment of it unified.

The report's list of mistakes for this option were all PHP code and
while there's a binding layer between libcurl and PHP, the PHP team has
decided that they have an as thin layer as possible on top of libcurl so
they will not alter or specifically filter a 'TRUE' value for this
particular option. I sympathize with that position.

[1] = http://daniel.haxx.se/blog/2012/10/25/libcurl-claimed-to-be-dangerous/

11 years agognutls: fix compiler warnings
Daniel Stenberg [Tue, 6 Nov 2012 18:45:51 +0000 (19:45 +0100)]
gnutls: fix compiler warnings

11 years agognutls: print alerts during handshake
Alessandro Ghedini [Wed, 24 Oct 2012 12:34:00 +0000 (14:34 +0200)]
gnutls: print alerts during handshake

11 years agognutls: fix the error_is_fatal logic
Alessandro Ghedini [Wed, 24 Oct 2012 09:47:32 +0000 (11:47 +0200)]
gnutls: fix the error_is_fatal logic

11 years agoRELEASE-NOTES: synced with fa6d78829fd30ad
Daniel Stenberg [Tue, 6 Nov 2012 11:03:29 +0000 (12:03 +0100)]
RELEASE-NOTES: synced with fa6d78829fd30ad

11 years agohttpcustomheader.c: free the headers after use
Daniel Stenberg [Tue, 6 Nov 2012 10:50:50 +0000 (11:50 +0100)]
httpcustomheader.c: free the headers after use

11 years agouniformly use AM_CPPFLAGS, avoid deprecated INCLUDES
Dave Reisner [Wed, 10 Oct 2012 08:05:02 +0000 (10:05 +0200)]
uniformly use AM_CPPFLAGS, avoid deprecated INCLUDES

Since automake 1.12.4, the warnings are issued on running automake:

  warning: 'INCLUDES' is the old name for 'AM_CPPFLAGS' (or '*_CPPFLAGS')

Avoid INCLUDES and roll these flags into AM_CPPFLAGS.

Compile tested on:
  Ubuntu 10.04 (automake 1:1.11.1-1)
  Ubuntu 12.04 (automake 1:1.11.3-1ubuntu2)
  Arch Linux (automake 1.12.4)

11 years agolibauthretry.c: shorten lines to fit within 80 cols
Daniel Stenberg [Mon, 5 Nov 2012 21:02:56 +0000 (22:02 +0100)]
libauthretry.c: shorten lines to fit within 80 cols

11 years agoftp_readresp: fix build without krb4 support
Daniel Stenberg [Mon, 5 Nov 2012 12:01:48 +0000 (13:01 +0100)]
ftp_readresp: fix build without krb4 support

Oops, my previous commit broke builds with krb support.

11 years agotest/README: mention the 1500 test number range
Daniel Stenberg [Sun, 4 Nov 2012 22:18:20 +0000 (23:18 +0100)]
test/README: mention the 1500 test number range

11 years agoFTP: prevent the multi interface from blocking
Daniel Stenberg [Sun, 4 Nov 2012 17:22:48 +0000 (18:22 +0100)]
FTP: prevent the multi interface from blocking

As pointed out in Bug report #3579064, curl_multi_perform() would
wrongly use a blocking mechanism internally for some commands which
could lead to for example a very long block if the LIST response never
showed.

The solution was to make sure to properly continue to use the multi
interface non-blocking state machine.

The new test 1501 verifies the fix.

Bug: http://curl.haxx.se/bug/view.cgi?id=3579064
Reported by: Guido Berhoerster

11 years agowinbuild: Use machine type of development environment
Marc Hoersken [Thu, 1 Nov 2012 21:23:05 +0000 (22:23 +0100)]
winbuild: Use machine type of development environment

This patch restores the original behavior instead of always
falling back to x86 if no MACHINE-type was specified.

11 years agowinbuild: Additional clean up
Marc Hoersken [Thu, 1 Nov 2012 21:16:47 +0000 (22:16 +0100)]
winbuild: Additional clean up

11 years agoEven more winbuild refactoring
Sapien2 [Thu, 25 Oct 2012 10:54:07 +0000 (14:54 +0400)]
Even more winbuild refactoring

11 years agoMinor winbuild refactoring
Sapien2 [Thu, 25 Oct 2012 10:24:41 +0000 (14:24 +0400)]
Minor winbuild refactoring

11 years agoArchitecture selection for winbuild and minor makefiles refactoring
Sapien2 [Thu, 25 Oct 2012 08:19:50 +0000 (12:19 +0400)]
Architecture selection for winbuild and minor makefiles refactoring

11 years agoBUGS: fix the bug tracker URL
Daniel Stenberg [Thu, 1 Nov 2012 20:33:37 +0000 (21:33 +0100)]
BUGS: fix the bug tracker URL

The URL we used before is the one that goes directly to 'add' a bug
report, but since you can only do that after first having logged in to
sourceforge, the link often doesn't work for visitors.

Bug: http://curl.haxx.se/bug/view.cgi?id=3582408
Reported by: Oscar Norlander

11 years agoevhiperfifo: fix the pointer passed to WRITEDATA
Daniel Stenberg [Thu, 1 Nov 2012 13:19:21 +0000 (14:19 +0100)]
evhiperfifo: fix the pointer passed to WRITEDATA

Bug: http://curl.haxx.se/bug/view.cgi?id=3582407
Reported by: Oscar Norlander

11 years agoFixed MSVC libssh2 static build.
Guenter Knauf [Thu, 1 Nov 2012 00:03:12 +0000 (01:03 +0100)]
Fixed MSVC libssh2 static build.

Since libssh2 supports now agent stuff it also depends on user32.lib.
Posted to the list by Jan Ehrhardt.

11 years agotlsauthtype: deal with the string case insensitively
Daniel Stenberg [Tue, 23 Oct 2012 21:12:58 +0000 (23:12 +0200)]
tlsauthtype: deal with the string case insensitively

When given a string as 'srp' it didn't work, but required 'SRP'.
Starting now, the check disregards casing.

Bug: http://curl.haxx.se/bug/view.cgi?id=3578418
Reported by: Jeff Connelly

11 years agoasyn-ares: restore working with c-ares < 1.6.1
Daniel Stenberg [Tue, 23 Oct 2012 21:06:38 +0000 (23:06 +0200)]
asyn-ares: restore working with c-ares < 1.6.1

Back in those days the public ares.h header didn't include the
ares_version.h header so it needs to be included here.

Bug: http://curl.haxx.se/bug/view.cgi?id=3577710

11 years agometalink/md5: Use CommonCrypto on Apple operating systems
Nick Zitzmann [Tue, 16 Oct 2012 17:48:55 +0000 (11:48 -0600)]
metalink/md5: Use CommonCrypto on Apple operating systems

Previously the Metalink code used Apple's CommonCrypto library only if
curl was built using the --with-darwinssl option. Now we use CommonCrypto
on all Apple operating systems including Tiger or later, or iOS 5 or
later, so you don't need to build --with-darwinssl anymore. Also rolled
out this change to libcurl's md5 code.

11 years agohref_extractor.c: fix the URL
Daniel Stenberg [Thu, 18 Oct 2012 17:42:31 +0000 (19:42 +0200)]
href_extractor.c: fix the URL

11 years agohref_extractor: example code extracting href elements
Michał Kowalczyk [Thu, 18 Oct 2012 14:45:51 +0000 (16:45 +0200)]
href_extractor: example code extracting href elements

It does so in a streaming manner using the "Streaming HTML parser".