]> granicus.if.org Git - curl/log
curl
10 years agoFILEFORMAT: Added test requires feature not present information
Steve Holme [Fri, 21 Nov 2014 18:42:40 +0000 (18:42 +0000)]
FILEFORMAT: Added test requires feature not present information

Such as !SSPI as we do for the NTLM and Digest tests.

10 years agohttp.c: log if it notices HTTP 1.1 after a upgrade to http2
Daniel Stenberg [Fri, 5 Sep 2014 22:05:03 +0000 (00:05 +0200)]
http.c: log if it notices HTTP 1.1 after a upgrade to http2

10 years agotest1801: first real http2 test case
Daniel Stenberg [Sat, 2 Aug 2014 22:47:21 +0000 (00:47 +0200)]
test1801: first real http2 test case

10 years agosws: initial tiny steps toward http2 support
Daniel Stenberg [Sat, 2 Aug 2014 22:46:39 +0000 (00:46 +0200)]
sws: initial tiny steps toward http2 support

10 years agoFILEFORMAT: mention the new upgrade support
Daniel Stenberg [Sat, 2 Aug 2014 22:46:16 +0000 (00:46 +0200)]
FILEFORMAT: mention the new upgrade support

10 years agotest1800: first plain-text http2 test case
Daniel Stenberg [Sat, 2 Aug 2014 21:51:46 +0000 (23:51 +0200)]
test1800: first plain-text http2 test case

Verifies the upgrade request, but gets a plain 1.1 response

10 years agohttp: Disable pipelining for HTTP/2 and upgraded connections
Tatsuhiro Tsujikawa [Thu, 20 Nov 2014 12:44:00 +0000 (21:44 +0900)]
http: Disable pipelining for HTTP/2 and upgraded connections

This commit disables pipelining for HTTP/2 or upgraded connections.  For
HTTP/2, we do not support multiplexing.  In general, requests cannot be
pipelined in an upgraded connection, since it is now different protocol.

10 years agoCURLOPT_POSTFIELDS.3: mention the COPYPOSTFIELDS option
Brad Harder [Thu, 20 Nov 2014 09:27:09 +0000 (10:27 +0100)]
CURLOPT_POSTFIELDS.3: mention the COPYPOSTFIELDS option

10 years agomulti-uv.c: Updated for curl coding standards
Steve Holme [Wed, 19 Nov 2014 20:10:27 +0000 (20:10 +0000)]
multi-uv.c: Updated for curl coding standards

10 years agoconncache: Fixed specifiers in infof() for long and size_t variables
Steve Holme [Wed, 19 Nov 2014 20:04:21 +0000 (20:04 +0000)]
conncache: Fixed specifiers in infof() for long and size_t variables

10 years agocmake: add Kerberos to the supported features
Peter Wu [Wed, 19 Nov 2014 15:53:22 +0000 (16:53 +0100)]
cmake: add Kerberos to the supported features

Updated following commit eda919f and a4b7f71.

Acked-by: Brad King <brad.king@kitware.com>
Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agocmake: fix NTLM detection when CURL_DISABLE_HTTP defined
Peter Wu [Wed, 19 Nov 2014 15:48:23 +0000 (16:48 +0100)]
cmake: fix NTLM detection when CURL_DISABLE_HTTP defined

Updated following changes in commit f0d860d.

Acked-by: Brad King <brad.king@kitware.com>
Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agoRELEASE-NOTES: synced with cb13fad733e
Daniel Stenberg [Wed, 19 Nov 2014 13:06:07 +0000 (14:06 +0100)]
RELEASE-NOTES: synced with cb13fad733e

10 years agoexamples: Wait recommended 100ms when no file descriptors are ready
Jay Satiro [Mon, 17 Nov 2014 07:26:03 +0000 (02:26 -0500)]
examples: Wait recommended 100ms when no file descriptors are ready

Prior to this change when no file descriptors were ready on platforms
other than Windows the multi examples would sleep whatever was in
timeout, which may or may not have been less than the minimum
recommended value [1] of 100ms.

[1]: http://curl.haxx.se/libcurl/c/curl_multi_fdset.html

10 years agomulti-uv.c: close the file handle after download
Waldek Kozba [Tue, 7 Oct 2014 10:53:59 +0000 (12:53 +0200)]
multi-uv.c: close the file handle after download

10 years agomulti: inform about closed sockets before they are closed
Jon Spencer [Fri, 14 Nov 2014 18:32:47 +0000 (10:32 -0800)]
multi: inform about closed sockets before they are closed

When the connection code decides to close a socket it informs the multi
system via the Curl_multi_closed function. The multi system may, in
turn, invoke the CURLMOPT_SOCKETFUNCTION function with
CURL_POLL_REMOVE. This happens after the socket has already been
closed. Reorder the code so that CURL_POLL_REMOVE is called before the
socket is closed.

10 years agobuild: in Makefile.m32 moved target autodetection.
Guenter Knauf [Wed, 19 Nov 2014 11:39:20 +0000 (12:39 +0100)]
build: in Makefile.m32 moved target autodetection.

Moved target autodetection block after defining CC macro.

10 years agobuild: in Makefile.m32 simplify platform flags.
Guenter Knauf [Wed, 19 Nov 2014 10:46:02 +0000 (11:46 +0100)]
build: in Makefile.m32 simplify platform flags.

10 years agobuild: in Makefile.m32 try to detect 64bit target.
Guenter Knauf [Wed, 19 Nov 2014 10:38:37 +0000 (11:38 +0100)]
build: in Makefile.m32 try to detect 64bit target.

10 years agoCMake: Simplify if() conditions on check result variables
Brad King [Fri, 14 Nov 2014 21:33:14 +0000 (22:33 +0100)]
CMake: Simplify if() conditions on check result variables

Remove use of an old hack that takes advantage of the auto-dereference
behavior of the if() command to detect if a variable is defined.  The
hack has the form:

 if("${VAR} MATCHES "^${VAR}$")

where "${VAR}" is a macro argument reference.  Use if(DEFINED) instead.
This also avoids warnings for CMake Policy CMP0054 in CMake 3.1.

10 years agoTODO-RELEASE: removed
Daniel Stenberg [Wed, 19 Nov 2014 10:17:45 +0000 (11:17 +0100)]
TODO-RELEASE: removed

10 years agodebug: added new connection cache output, plus fixups
Carlo Wood [Tue, 18 Nov 2014 22:02:40 +0000 (23:02 +0100)]
debug: added new connection cache output, plus fixups

Debug output 'typo' fix.

Don't print an extra "0x" in
  * Pipe broke: handle 0x0x2546d88, url = /

Add debug output.
Print the number of connections in the connection cache when
  adding one, and not only when one is removed.

Fix typos in comments.

10 years agomulti: move the ending condition into the loop as well
Daniel Stenberg [Tue, 18 Nov 2014 19:51:54 +0000 (20:51 +0100)]
multi: move the ending condition into the loop as well

... as it was before I changed the loop in commit e04ccbd50. It caused
test 2030 and 2032 to fail.

10 years agomulti: Prefer we don't use CURLE_OK and NULL in comparisons
Steve Holme [Tue, 18 Nov 2014 20:25:05 +0000 (20:25 +0000)]
multi: Prefer we don't use CURLE_OK and NULL in comparisons

10 years agomulti_runsingle: use 'result' for local CURLcode storage
Daniel Stenberg [Tue, 18 Nov 2014 14:18:21 +0000 (15:18 +0100)]
multi_runsingle: use 'result' for local CURLcode storage

... and assign data->result only at the end. Makes the code more compact
(easier to read) and more similar to other code.

10 years agomulti_runsingle: rename result to rc
Daniel Stenberg [Tue, 18 Nov 2014 13:37:57 +0000 (14:37 +0100)]
multi_runsingle: rename result to rc

save 'result' for CURLcode types

10 years agomulti: make multi_runsingle loop internally
Daniel Stenberg [Tue, 18 Nov 2014 13:34:55 +0000 (14:34 +0100)]
multi: make multi_runsingle loop internally

simplifies the use of this function at little cost.

10 years agomulti: when leaving for timeout, close accordingly
Carlo Wood [Sun, 16 Nov 2014 20:53:56 +0000 (21:53 +0100)]
multi: when leaving for timeout, close accordingly

Fixes the problem when a transfer in a pipeline times out.

10 years agobuild: in Makefile.m32 add -m32 flag for 32bit.
Guenter Knauf [Tue, 18 Nov 2014 13:25:41 +0000 (14:25 +0100)]
build: in Makefile.m32 add -m32 flag for 32bit.

10 years agomk-ca-bundle.vbs: update copyright year.
Guenter Knauf [Tue, 18 Nov 2014 12:47:13 +0000 (13:47 +0100)]
mk-ca-bundle.vbs: update copyright year.

10 years agobuild: in Makefile.m32 pass -F flag to windres.
Guenter Knauf [Tue, 18 Nov 2014 12:44:03 +0000 (13:44 +0100)]
build: in Makefile.m32 pass -F flag to windres.

10 years agoconfig-win32: Fixed build targets for the VS2012+ Windows XP toolset
Steve Holme [Tue, 18 Nov 2014 00:01:27 +0000 (00:01 +0000)]
config-win32: Fixed build targets for the VS2012+ Windows XP toolset

Even though commit 23e70e1cc6 mentioned the v110_xp toolset, I had
forgotten to include the relevant pre-processor definitions.

10 years agosasl_sspi: Removed note about the NTLM functions being a wrapper
Steve Holme [Sun, 16 Nov 2014 22:44:58 +0000 (22:44 +0000)]
sasl_sspi: Removed note about the NTLM functions being a wrapper

10 years agoconnect.c: Fixed compilation warning when no verbose string support
Steve Holme [Sun, 16 Nov 2014 18:19:58 +0000 (18:19 +0000)]
connect.c: Fixed compilation warning when no verbose string support

warning: unused parameter 'reason'

10 years agoeasy.c: Fixed compilation warning when no verbose string support
Steve Holme [Sun, 16 Nov 2014 18:16:27 +0000 (18:16 +0000)]
easy.c: Fixed compilation warning when no verbose string support

warning: unused parameter 'easy'

10 years agowin32: Updated some legacy APIs to use the newer extended versions
Steve Holme [Sun, 16 Nov 2014 17:27:44 +0000 (17:27 +0000)]
win32: Updated some legacy APIs to use the newer extended versions

Updated the usage of some legacy APIs, that are preventing curl from
compiling for Windows Store and Windows Phone build targets.

Suggested-by: Stefan Neis
Feature: http://sourceforge.net/p/curl/feature-requests/82/

10 years agoconfig-win32: Introduce build targets for VS2012+
Steve Holme [Sun, 16 Nov 2014 17:18:19 +0000 (17:18 +0000)]
config-win32: Introduce build targets for VS2012+

Visual Studio 2012 introduced support for Windows Store apps as well as
supporting Windows Phone 8. Introduced build targets that allow more
modern APIs to be used as certain legacy ones are not available on these
new platforms.

10 years agosasl_sspi: Fixed compilation warnings when no verbose string support
Steve Holme [Sun, 16 Nov 2014 14:39:26 +0000 (14:39 +0000)]
sasl_sspi: Fixed compilation warnings when no verbose string support

10 years agosasl_sspi: Added base64 decoding debug failure messages
Steve Holme [Sun, 16 Nov 2014 13:27:03 +0000 (13:27 +0000)]
sasl_sspi: Added base64 decoding debug failure messages

Just like in the NTLM code, added infof() failure messages for
DIGEST-MD5 and GSSAPI authentication when base64 decoding fails.

10 years agontlm: Moved the SSPI based Type-3 message generation into the SASL module
Steve Holme [Sun, 16 Nov 2014 13:18:26 +0000 (13:18 +0000)]
ntlm: Moved the SSPI based Type-3 message generation into the SASL module

10 years agontlm: Moved the SSPI based Type-2 message decoding into the SASL module
Steve Holme [Sun, 16 Nov 2014 12:58:04 +0000 (12:58 +0000)]
ntlm: Moved the SSPI based Type-2 message decoding into the SASL module

10 years agontlm: Moved the SSPI based Type-1 message generation into the SASL module
Steve Holme [Sun, 16 Nov 2014 12:45:30 +0000 (12:45 +0000)]
ntlm: Moved the SSPI based Type-1 message generation into the SASL module

10 years agokerberos: Use symbol qualified with _KERBEROS5
Michael Osipov [Sun, 16 Nov 2014 09:50:06 +0000 (09:50 +0000)]
kerberos: Use symbol qualified with _KERBEROS5

For consistency renamed USE_KRB5 to USE_KERBEROS5.

10 years agoexamples: Don't call select() to sleep on windows
Jay Satiro [Fri, 14 Nov 2014 08:59:46 +0000 (03:59 -0500)]
examples: Don't call select() to sleep on windows

Windows does not support using select() for sleeping without a dummy
socket. Instead use Windows' Sleep() and sleep for 100ms which is the
minimum suggested value in the curl_multi_fdset() doc.

Prior to this change the multi examples would exit prematurely since
select() would error instead of sleeping when called without an fd.

Reported-by: Johan Lantz
Bug: http://curl.haxx.se/mail/lib-2014-11/0221.html

10 years agohttp2: Don't send Upgrade headers when we already do HTTP/2
Tatsuhiro Tsujikawa [Fri, 14 Nov 2014 15:38:29 +0000 (00:38 +0900)]
http2: Don't send Upgrade headers when we already do HTTP/2

10 years agosasl: Corrected Curl_sasl_build_spn() function description
Steve Holme [Sat, 15 Nov 2014 18:35:40 +0000 (18:35 +0000)]
sasl: Corrected Curl_sasl_build_spn() function description

There was a mismatch in function parameter names.

10 years agotool: Removed krb4 from the supported features
Steve Holme [Fri, 14 Nov 2014 14:54:21 +0000 (14:54 +0000)]
tool: Removed krb4 from the supported features

Although libcurl would never return CURL_VERSION_KERBEROS4 after 7.33,
so would not be output with --version, removed krb4 from the supported
features output.

10 years agotool: Use Kerberos for supported features
Michael Osipov [Sat, 15 Nov 2014 14:08:43 +0000 (14:08 +0000)]
tool: Use Kerberos for supported features

10 years agourldata: Don't define sec_complete when no GSS-API support present
Steve Holme [Sat, 15 Nov 2014 13:02:31 +0000 (13:02 +0000)]
urldata: Don't define sec_complete when no GSS-API support present

This variable is only used with HAVE_GSSAPI is defined by the FTP code
so let's place the definition with the other GSS-API based variables.

10 years agodocs: Use consistent naming for Kerberos
Michael Osipov [Sat, 15 Nov 2014 11:10:29 +0000 (11:10 +0000)]
docs: Use consistent naming for Kerberos

10 years agoTODO: Lets support QOP options in GSSAPI authentication
Steve Holme [Sat, 15 Nov 2014 00:27:53 +0000 (00:27 +0000)]
TODO: Lets support QOP options in GSSAPI authentication

10 years agosasl_sspi: Corrected a couple of comment typos
Steve Holme [Fri, 14 Nov 2014 22:20:28 +0000 (22:20 +0000)]
sasl_sspi: Corrected a couple of comment typos

10 years agosasl: Moved Curl_sasl_gssapi_cleanup() definition into header file
Steve Holme [Fri, 14 Nov 2014 21:16:08 +0000 (21:16 +0000)]
sasl: Moved Curl_sasl_gssapi_cleanup() definition into header file

Rather than define the function as extern in the source files that use
it, moved the function declaration into the SASL header file just like
the Digest and NTLM clean-up functions.

Additionally, added a function description comment block.

10 years agosasl_sspi: Added missing RFC reference for HTTP Digest authentication
Steve Holme [Fri, 14 Nov 2014 21:05:38 +0000 (21:05 +0000)]
sasl_sspi: Added missing RFC reference for HTTP Digest authentication

10 years agontlm: Clean-up and standardisation of base64 decoding
Steve Holme [Fri, 14 Nov 2014 21:02:50 +0000 (21:02 +0000)]
ntlm: Clean-up and standardisation of base64 decoding

10 years agontlm: We prefer 'CURLcode result'
Steve Holme [Fri, 14 Nov 2014 20:43:15 +0000 (20:43 +0000)]
ntlm: We prefer 'CURLcode result'

10 years agoCMake: Restore order-dependent library checks
Brad King [Mon, 10 Nov 2014 18:56:10 +0000 (19:56 +0100)]
CMake: Restore order-dependent library checks

Revert commit 2257deb502 (Cmake: Avoid cycle directory dependencies,
2014-08-22) and add a comment explaining the purpose of the original
code.

The check_library_exists_concat macro is intended to be called multiple
times on a sequence of possibly dependent libraries.  Later libraries
may depend on earlier libraries when they are static.  They cannot be
safely linked in reverse order on some platforms.

Signed-off-by: Brad King <brad.king@kitware.com>
10 years agoCMake: Restore order-dependent header checks
Brad King [Mon, 10 Nov 2014 18:56:09 +0000 (19:56 +0100)]
CMake: Restore order-dependent header checks

Revert commit 1269df2e3b (Cmake: Don't check for all headers each
time, 2014-08-15) and add a comment explaining the purpose of the
original code.

The check_include_file_concat macro is intended to be called multiple
times on a sequence of possibly dependent headers.  Later headers
may depend on earlier headers to provide declarations.  They cannot
be safely included independently on some platforms.

For example, many POSIX APIs document including sys/types.h before some
other headers.  Also on some OS X versions sys/socket.h must be included
before net/if.h or the check for the latter will fail.

Signed-off-by: Brad King <brad.king@kitware.com>
10 years agotest22: expand a backtick command
Peter Wu [Thu, 13 Nov 2014 22:06:44 +0000 (23:06 +0100)]
test22: expand a backtick command

This is the only user of the backtick operator in the command. As the
commands will soon not be executed by a shell anymore (but by perl),
replace the command with its output.

Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agoRELEASE-NOTES: synced with 2ee3c63b13
Daniel Stenberg [Thu, 13 Nov 2014 15:02:52 +0000 (16:02 +0100)]
RELEASE-NOTES: synced with 2ee3c63b13

10 years agohttp2: fix switched macro when http2 is not enabled
Daniel Stenberg [Thu, 13 Nov 2014 14:39:15 +0000 (15:39 +0100)]
http2: fix switched macro when http2 is not enabled

10 years agohttp2: Deal with HTTP/2 data inside response header buffer
Tatsuhiro Tsujikawa [Wed, 12 Nov 2014 17:07:24 +0000 (02:07 +0900)]
http2: Deal with HTTP/2 data inside response header buffer

Previously if HTTP/2 traffic is appended to HTTP Upgrade response header
(thus they are in the same buffer), the trailing HTTP/2 traffic is not
processed and lost.  The appended data is most likely SETTINGS frame.
If it is lost, nghttp2 library complains server does not obey the HTTP/2
protocol and issues GOAWAY frame and curl eventually drops connection.
This commit fixes this problem and now trailing data is processed.

10 years agoconfigure: Fixed inclusion of krb5 when CURL_DISABLE_CRYPTO_AUTH is defined
Steve Holme [Tue, 11 Nov 2014 00:04:05 +0000 (00:04 +0000)]
configure: Fixed inclusion of krb5 when CURL_DISABLE_CRYPTO_AUTH is defined

Commit fe0f8967bf fixed a problem with krb5 not being defined as a
supported feature when HAVE_GSSAPI is defined, however, it should
only be included if CURL_DISABLE_CRYPTO_AUTH is not set, like when
SPNEGO is listed as a feature.

10 years agomulti: removed Curl_multi_set_easy_connection
Daniel Stenberg [Mon, 10 Nov 2014 09:09:40 +0000 (10:09 +0100)]
multi: removed Curl_multi_set_easy_connection

It isn't used anywhere!

Reported-by: Carlo Wood
10 years agosymbol-scan.pl: do not require autotools
Peter Wu [Thu, 6 Nov 2014 00:32:46 +0000 (01:32 +0100)]
symbol-scan.pl: do not require autotools

Makes test1119 pass when building with cmake.

configurehelp.pm is generated by configure (autotools). As cmake does
not provide a separate variable for the C preprocessor, default to cpp.
Before commit ef24ecde68a5f577a7f0f423a767620f09a0ab16 ("symbol-scan:
use configure script knowledge about how to run the C preprocessor"),
this tool would also use 'cpp'.

Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agocmake: add ENABLE_THREADED_RESOLVER, rename ARES
Peter Wu [Thu, 6 Nov 2014 00:32:45 +0000 (01:32 +0100)]
cmake: add ENABLE_THREADED_RESOLVER, rename ARES

Fix detection of the AsynchDNS feature which not just depends on
pthreads support, but also on whether USE_POSIX_THREADS is set or not.
Caught by test 1014.

This patch adds a new ENABLE_THREADED_RESOLVER option (corresponding to
--enable-threaded-resolver of autotools) which also needs a check for
HAVE_PTHREAD_H.

For symmetry with autotools, CURL_USE_ARES is renamed to ENABLE_ARES
(--enable-ares). Checks that test for the availability actually use
USE_ARES instead as that is the result of whether a-res is available or
not (in practice this does not matter as CARES is marked as required
package, but nevertheless it is better to write the intent).

Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agocmake: build libhostname for test suite
Peter Wu [Thu, 6 Nov 2014 00:32:44 +0000 (01:32 +0100)]
cmake: build libhostname for test suite

Used by some test cases via LD_PRELOAD in order to fake the host name.

Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agocmake: fix HAVE_GETHOSTNAME definition
Peter Wu [Thu, 6 Nov 2014 00:32:43 +0000 (01:32 +0100)]
cmake: fix HAVE_GETHOSTNAME definition

Otherwise Curl_gethostname always fails. Windows has gethostname
since Vista according to
http://msdn.microsoft.com/en-us/library/ms738527%28VS.85%29.aspx, but
accordings to byte_bucket's VC 2005 documentation, it is available even
in Windows 95. (possibly after installing a Platform SDK, the
Windows Server 2003 SP1 Platform SDK should be sufficient).

Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agotests: fix libhostname visibility
Peter Wu [Thu, 6 Nov 2014 00:32:42 +0000 (01:32 +0100)]
tests: fix libhostname visibility

I noticed that a patched cmake build would pass tests with a fake local
hostname, but the autotools build skips them:

    got unexpected host name back, LD_PRELOAD failed

It turns out that -fvisibility=hidden hides the symbol, and since the
tests are not part of libcurl, it fails too. Just remove the LIBCURL
guard.

Broken since cURL 7.30 (commit 83a42ee20ea7fc25abb61c0b7ef56ebe712d7093,
"curl.h: stricter CURL_EXTERN linkage decorations logic").

Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agotests: fix memleak in server/resolve.c
Peter Wu [Thu, 6 Nov 2014 00:32:41 +0000 (01:32 +0100)]
tests: fix memleak in server/resolve.c

This makes LeakSanitizer happy.

Signed-off-by: Peter Wu <peter@lekensteyn.nl>
10 years agoconfigure: assume krb5 when gss-api works
Daniel Stenberg [Mon, 10 Nov 2014 08:05:56 +0000 (09:05 +0100)]
configure: assume krb5 when gss-api works

To please test 1014 while we work out if this is truly the a correct
assumption.

10 years agovtls.h: Fixed compiler warning when compiled without SSL
Steve Holme [Sun, 9 Nov 2014 18:09:58 +0000 (18:09 +0000)]
vtls.h: Fixed compiler warning when compiled without SSL

vtls.c:185:46: warning: unused parameter 'data'

10 years agoRELEASE-NOTES: Synced with 2fbf23875f
Steve Holme [Sun, 9 Nov 2014 15:30:01 +0000 (15:30 +0000)]
RELEASE-NOTES: Synced with 2fbf23875f

10 years agontlm: Added separate SSPI based functions
Steve Holme [Sun, 9 Nov 2014 15:12:56 +0000 (15:12 +0000)]
ntlm: Added separate SSPI based functions

In preparation for moving the NTLM message code into the SASL module,
and separating the native code from the SSPI code, added functions that
simply call the functions in curl_ntlm_msg.c.

10 years agohttp_ntlm: Use the SASL functions instead
Steve Holme [Sun, 9 Nov 2014 14:55:41 +0000 (14:55 +0000)]
http_ntlm: Use the SASL functions instead

In preparation for moving the NTLM message code into the SASL module
use the SASL functions in the HTTP code instead.

10 years agolibssh2: detect features based on version, not configure checks
Daniel Stenberg [Sun, 9 Nov 2014 14:42:12 +0000 (15:42 +0100)]
libssh2: detect features based on version, not configure checks

... so that non-configure builds get the correct functions too based on
the libssh2 version used.

10 years agoSSH: use the port number as well for known_known checks
Nobuhiro Ban [Sun, 9 Nov 2014 14:30:06 +0000 (15:30 +0100)]
SSH: use the port number as well for known_known checks

... if the libssh2 version is new enough.

Bug: http://curl.haxx.se/bug/view.cgi?id=1448

10 years agoINSTALL: Updated pre-processor references to the old VC6 project files
Steve Holme [Sun, 9 Nov 2014 14:19:50 +0000 (14:19 +0000)]
INSTALL: Updated pre-processor references to the old VC6 project files

Reworked the two sections that discuss modifying the Visual Studio pre-
processor settings, and vc6libcurl.dsw/vc6libcurl.dsp, to remove the
project files references as they have been superseded by a more thorough
set of project files for VC6 through VC12, but to also give the correct
reference to this setting in later versions of Visual Studio.

10 years agoINSTALL: Added email protocols to the "Disabling in Win32 builds" section
Steve Holme [Sun, 9 Nov 2014 13:35:08 +0000 (13:35 +0000)]
INSTALL: Added email protocols to the "Disabling in Win32 builds" section

10 years agoconfigure: Fixed NTLM missing from features when CURL_DISABLE_HTTP defined
Steve Holme [Sun, 9 Nov 2014 12:58:07 +0000 (12:58 +0000)]
configure: Fixed NTLM missing from features when CURL_DISABLE_HTTP defined

10 years agobuild: Fixed no NTLM support for email when CURL_DISABLE_HTTP is defined
Steve Holme [Sun, 9 Nov 2014 12:46:00 +0000 (12:46 +0000)]
build: Fixed no NTLM support for email when CURL_DISABLE_HTTP is defined

USE_NTLM would only be defined if: HTTP support was enabled, NTLM and
cryptography weren't disabled, and either a supporting cryptography
library or Windows SSPI was being compiled against.

This means it was not possible to build libcurl without HTTP support
and use NTLM for other protocols such as IMAP, POP3 and SMTP. Rather
than introduce a new SASL pre-processor definition, removed the HTTP
prerequisite just like USE_SPNEGO and USE_KRB5.

Note: Winbind support still needs to be dependent on CURL_DISABLE_HTTP
as it is only available to HTTP at present.

This bug dates back to August 2011 when I started to add support for
NTLM to SMTP.

10 years agontlm: Removed an unnecessary free of native Target Info
Steve Holme [Sun, 9 Nov 2014 11:34:36 +0000 (11:34 +0000)]
ntlm: Removed an unnecessary free of native Target Info

Due to commit 40ee1ba0dc the free in Curl_ntlm_decode_type2_target() is
longer required.

10 years agontlm: Moved the native Target Info clean-up from HTTP specific function
Steve Holme [Sun, 9 Nov 2014 11:25:10 +0000 (11:25 +0000)]
ntlm: Moved the native Target Info clean-up from HTTP specific function

10 years agontlm: Moved SSPI clean-up code into SASL module
Steve Holme [Thu, 9 Oct 2014 10:03:14 +0000 (11:03 +0100)]
ntlm: Moved SSPI clean-up code into SASL module

10 years agoMakefile.dist: Added support for WinIDN
Steve Holme [Sat, 8 Nov 2014 18:27:54 +0000 (18:27 +0000)]
Makefile.dist: Added support for WinIDN

10 years agoMakefile.vc6: Added support for WinIDN
Steve Holme [Sat, 8 Nov 2014 17:47:08 +0000 (17:47 +0000)]
Makefile.vc6: Added support for WinIDN

10 years agoMakefile.dist: Added some missing SSPI configurations
Steve Holme [Sat, 8 Nov 2014 17:07:45 +0000 (17:07 +0000)]
Makefile.dist: Added some missing SSPI configurations

10 years agoMakefile.dist: Separated the groups of SSL configurations from each other
Steve Holme [Sat, 8 Nov 2014 16:49:35 +0000 (16:49 +0000)]
Makefile.dist: Separated the groups of SSL configurations from each other

10 years agoMakefile.dist: Grouped the x64 configurations next to their x86 counterparts
Steve Holme [Sat, 8 Nov 2014 16:37:31 +0000 (16:37 +0000)]
Makefile.dist: Grouped the x64 configurations next to their x86 counterparts

10 years agocurl.h: Tidy up of CURL_VERSION_* flags
Steve Holme [Fri, 7 Nov 2014 11:26:06 +0000 (11:26 +0000)]
curl.h: Tidy up of CURL_VERSION_* flags

As the list has gotten a little messy and hard to read, especially with
the introduction of deprecated items, aligned the values and comments
into clean columns and reworked some of the comments in the process.

10 years agocurl_tool: Added krb5 to the supported features
Steve Holme [Fri, 7 Nov 2014 10:44:00 +0000 (10:44 +0000)]
curl_tool: Added krb5 to the supported features

10 years agoconfigure: Added krb5 to the supported features
Steve Holme [Fri, 7 Nov 2014 10:42:54 +0000 (10:42 +0000)]
configure: Added krb5 to the supported features

10 years agoversion info: Added Kerberos V5 to the supported features
Steve Holme [Fri, 7 Nov 2014 10:40:01 +0000 (10:40 +0000)]
version info: Added Kerberos V5 to the supported features

10 years agomk-ca-bundle.vbs: switch to new certdata.txt url.
Guenter Knauf [Fri, 7 Nov 2014 09:27:26 +0000 (10:27 +0100)]
mk-ca-bundle.vbs: switch to new certdata.txt url.

10 years agoRELEASE-NOTES: Synced with dcad09e125
Steve Holme [Fri, 7 Nov 2014 00:08:16 +0000 (00:08 +0000)]
RELEASE-NOTES: Synced with dcad09e125

10 years agohttp_digest: Fixed some memory leaks introduced in commit 6f8d8131b1
Steve Holme [Fri, 7 Nov 2014 00:03:11 +0000 (00:03 +0000)]
http_digest: Fixed some memory leaks introduced in commit 6f8d8131b1

Fixed a couple of memory leaks as a result of moving code that used to
populate allocuserpwd and relied on it's clean up.

10 years agodocs: Updated following the addition of SSPI based HTTP digest auth
Steve Holme [Thu, 6 Nov 2014 23:47:18 +0000 (23:47 +0000)]
docs: Updated following the addition of SSPI based HTTP digest auth

10 years agosasl_sspi: Tidy up of the existing digest code
Steve Holme [Thu, 6 Nov 2014 23:12:18 +0000 (23:12 +0000)]
sasl_sspi: Tidy up of the existing digest code

Following the addition of SSPI support for HTTP digest, synchronised
elements of the email digest code with that of the new HTTP code.

10 years agohttp_digest: Post SSPI support tidy up
Steve Holme [Thu, 6 Nov 2014 23:01:36 +0000 (23:01 +0000)]
http_digest: Post SSPI support tidy up

Post tidy up to ensure commonality of code style and variable names.

10 years agotest552: Don't run HTTP digest tests for SSPI based builds
Dan Fandrich [Thu, 6 Nov 2014 21:28:48 +0000 (22:28 +0100)]
test552: Don't run HTTP digest tests for SSPI based builds

Technical difficulties prevented this from going into the
previous commit.