platform/upstream/curl.git
10 years agotests: Removed test 807 as it has been superseded by tests 815 and 816
Steve Holme [Sun, 9 Feb 2014 20:37:44 +0000 (20:37 +0000)]
tests: Removed test 807 as it has been superseded by tests 815 and 816

10 years agotests: Updated the titles of tests 815 and 816
Steve Holme [Sun, 9 Feb 2014 20:32:44 +0000 (20:32 +0000)]
tests: Updated the titles of tests 815 and 816

10 years agotool_metalink: fix compiler warning when built without metalink
Daniel Stenberg [Sun, 9 Feb 2014 19:08:13 +0000 (20:08 +0100)]
tool_metalink: fix compiler warning when built without metalink

10 years agotool_operate: Move the trace and error file closure to tool_cfgable
Steve Holme [Sun, 9 Feb 2014 17:16:15 +0000 (17:16 +0000)]
tool_operate: Move the trace and error file closure to tool_cfgable

10 years agoTODO: Removed url-specific options
Steve Holme [Sun, 9 Feb 2014 13:29:00 +0000 (13:29 +0000)]
TODO: Removed url-specific options

10 years agotests: Re-enabled IMAP tests that require URL specific option support
Steve Holme [Sun, 9 Feb 2014 13:18:29 +0000 (13:18 +0000)]
tests: Re-enabled IMAP tests that require URL specific option support

10 years agoRELEASE-NOTES: Synced with 8e62f7a6503a
Steve Holme [Sun, 9 Feb 2014 13:04:33 +0000 (13:04 +0000)]
RELEASE-NOTES: Synced with 8e62f7a6503a

10 years agosecureserver: FIPS option is only supported since stunnel 5.00
Marc Hoersken [Sun, 9 Feb 2014 12:36:11 +0000 (13:36 +0100)]
secureserver: FIPS option is only supported since stunnel 5.00

10 years agotool_operate: Added support for performing URL specific operations
Steve Holme [Sun, 9 Feb 2014 11:01:36 +0000 (11:01 +0000)]
tool_operate: Added support for performing URL specific operations

10 years agotool_operate: Let curl handle cleanup take place in config_free()
Steve Holme [Fri, 7 Feb 2014 21:32:29 +0000 (21:32 +0000)]
tool_operate: Let curl handle cleanup take place in config_free()

10 years agoformdata: Must use Curl_safefree instead of free
Dan Fandrich [Sat, 8 Feb 2014 12:59:40 +0000 (13:59 +0100)]
formdata: Must use Curl_safefree instead of free

10 years agotest96: updated according to recent changes
Daniel Stenberg [Sat, 8 Feb 2014 22:20:10 +0000 (23:20 +0100)]
test96: updated according to recent changes

10 years agoruntests: allow <strippart> to remove lines
Daniel Stenberg [Sat, 8 Feb 2014 22:19:10 +0000 (23:19 +0100)]
runtests: allow <strippart> to remove lines

For verify file, if the strippart condition removes the line completely
it is now removed from the array.

10 years agotool_getparam: Added support for parsing of specific URL options
Steve Holme [Fri, 7 Feb 2014 21:14:43 +0000 (21:14 +0000)]
tool_getparam: Added support for parsing of specific URL options

10 years agosecureserver: Disable FIPS mode for stunnel
Dan Fandrich [Sat, 8 Feb 2014 10:51:28 +0000 (11:51 +0100)]
secureserver: Disable FIPS mode for stunnel

It's unnecessary for curl testing, and it can otherwise cause
stunnel to fail to start if OpenSSL doesn't support FIPS mode.

10 years agoformdata: Fixed memory leak on OOM condition
Dan Fandrich [Sat, 8 Feb 2014 10:33:43 +0000 (11:33 +0100)]
formdata: Fixed memory leak on OOM condition

10 years agoruntests: Disable valgrind when debugging
Dan Fandrich [Fri, 7 Feb 2014 21:15:02 +0000 (22:15 +0100)]
runtests: Disable valgrind when debugging

This was already mostly being done, except that analysis after the
test still assumed that the valgrind log files would be available. An
alternative way to handle the valgrind + gdb combination could be to
enable one of the valgrind debugger hooks.

10 years agotool_cfgable: For consistency renamed init_config() to config_init()
Steve Holme [Fri, 7 Feb 2014 19:40:45 +0000 (19:40 +0000)]
tool_cfgable: For consistency renamed init_config() to config_init()

10 years agotool_cfgable: Introduced config_free() function
Steve Holme [Fri, 7 Feb 2014 19:29:46 +0000 (19:29 +0000)]
tool_cfgable: Introduced config_free() function

10 years ago--help: add missing --tlsv1.x options
Daniel Stenberg [Fri, 7 Feb 2014 19:28:53 +0000 (20:28 +0100)]
--help: add missing --tlsv1.x options

10 years agolib1515.c: Fixed various compilation warnings
Steve Holme [Fri, 7 Feb 2014 15:00:51 +0000 (15:00 +0000)]
lib1515.c: Fixed various compilation warnings

lib1515.c:38:26 warning: unused parameter 'curl'
lib1515.c:38:81 warning: unused parameter 'ptr'
lib1515.c:38:5 warning: no previous prototype for 'debug_callback'
lib1515.c:46:5 warning: no previous prototype for 'do_one_request'
lib1515.c:120:3  warning: ISO C90 forbids mixed declarations and code

As well as some code policing such as white space and braces.

10 years agohttp2: updated README after NSS addition
Daniel Stenberg [Fri, 7 Feb 2014 14:50:31 +0000 (15:50 +0100)]
http2: updated README after NSS addition

Changed the support to a little matrix and added brief explanation of
what ALPN and NPN are for.

10 years agonss: support pre-ALPN versions
Daniel Stenberg [Fri, 7 Feb 2014 14:38:45 +0000 (15:38 +0100)]
nss: support pre-ALPN versions

10 years agonss: ALPN and NPN support
Fabian Frank [Fri, 7 Feb 2014 09:48:34 +0000 (01:48 -0800)]
nss: ALPN and NPN support

Add ALPN and NPN support for NSS. This allows cURL to negotiate
HTTP/2.0 connections when built with NSS.

10 years agoformpost: use semicolon in multipart/mixed
Daniel Stenberg [Fri, 7 Feb 2014 08:39:50 +0000 (09:39 +0100)]
formpost: use semicolon in multipart/mixed

Not comma, which is an inconsistency and a mistake probably inherited
from the examples section of RFC1867.

This bug has been present since the day curl started to support
multipart formposts, back in the 90s.

Reported-by: Rob Davies
Bug: http://curl.haxx.se/bug/view.cgi?id=1333

10 years agotests: Document use of the MEMDEBUG_LOG_SYNC macro
Dan Fandrich [Thu, 6 Feb 2014 22:55:25 +0000 (23:55 +0100)]
tests: Document use of the MEMDEBUG_LOG_SYNC macro

10 years agossh: Fixed a NULL pointer dereference on OOM condition
Dan Fandrich [Thu, 6 Feb 2014 22:53:48 +0000 (23:53 +0100)]
ssh: Fixed a NULL pointer dereference on OOM condition

10 years agonss: Updated copyright year for recent edits
Steve Holme [Thu, 6 Feb 2014 22:32:56 +0000 (22:32 +0000)]
nss: Updated copyright year for recent edits

10 years ago100-continue: fix timeout condition
Remi Gacogne [Thu, 6 Feb 2014 22:16:37 +0000 (23:16 +0100)]
100-continue: fix timeout condition

When using the multi socket interface, libcurl calls the
curl_multi_timer_callback asking to be woken up after
CURL_TIMEOUT_EXPECT_100 milliseconds.

After the timeout has expired, calling curl_multi_socket_action with
CURL_SOCKET_TIMEOUT as sockfd leads libcurl to check expired
timeouts. When handling the 100-continue one, the following check in
Curl_readwrite() fails if exactly CURL_TIMEOUT_EXPECT_100 milliseconds
passed since the timeout has been set!

It seems logical to consider that having waited for exactly
CURL_TIMEOUT_EXPECT_100 ms is enough.

Bug: http://curl.haxx.se/bug/view.cgi?id=1334

10 years agonss: prefer highest available TLS version
Fabian Frank [Thu, 6 Feb 2014 08:41:53 +0000 (00:41 -0800)]
nss: prefer highest available TLS version

Offer TLSv1.0 to 1.2 by default, still fall back to SSLv3
if --tlsv1[.N] was not specified on the command line.

10 years agotests: add test for bug #1327 (dns cache timeout)
Romulo A. Ceccon [Mon, 3 Feb 2014 19:09:33 +0000 (17:09 -0200)]
tests: add test for bug #1327 (dns cache timeout)

Fix for bug #1303 (030a2b8cb) was not complete.
libcurl still pruned DNS entries added manually
after detecting a dead connection. This test
checks such behavior.

10 years agotests: add test for bug #1303 (dns cache timeout)
Romulo A. Ceccon [Fri, 31 Jan 2014 19:03:13 +0000 (17:03 -0200)]
tests: add test for bug #1303 (dns cache timeout)

Test-case 1515 reproduces bug #1303, where libcurl
would incorrectly prune DNS entries added via
CURLOPT_RESOLVE after the DNS_CACHE_TIMEOUT had
expired.

10 years agohttp2: spell fixed README and added version requirement
Daniel Stenberg [Thu, 6 Feb 2014 21:26:47 +0000 (22:26 +0100)]
http2: spell fixed README and added version requirement

10 years agotool_operate: Removed unused argument parameters from operate_do()
Steve Holme [Thu, 6 Feb 2014 19:51:18 +0000 (19:51 +0000)]
tool_operate: Removed unused argument parameters from operate_do()

10 years agotool_operate: Moved list SSL engines code into operate()
Steve Holme [Thu, 6 Feb 2014 19:43:50 +0000 (19:43 +0000)]
tool_operate: Moved list SSL engines code into operate()

10 years agotool_operate: Moved argument parsing into operate()
Steve Holme [Thu, 6 Feb 2014 19:31:44 +0000 (19:31 +0000)]
tool_operate: Moved argument parsing into operate()

10 years agoruntests: add suppression generator help
Daniel Stenberg [Wed, 5 Feb 2014 22:48:44 +0000 (23:48 +0100)]
runtests: add suppression generator help

Leave the valgrind --gen-suppressions option in there, commented, to
make it easier for next update.

10 years agovalgrind: updated suppressions file
Daniel Stenberg [Wed, 5 Feb 2014 22:46:31 +0000 (23:46 +0100)]
valgrind: updated suppressions file

The call stack was modified in 2dc7ad23 so the supressions didn't work
anymore.

10 years agoruntests: detect 'ares' better
Daniel Stenberg [Wed, 5 Feb 2014 22:36:16 +0000 (23:36 +0100)]
runtests: detect 'ares' better

... caused false detections of the threaded resolver otherwise

10 years agotool_operate: Moved .curlrc parsing code into operate()
Steve Holme [Wed, 5 Feb 2014 20:28:36 +0000 (20:28 +0000)]
tool_operate: Moved .curlrc parsing code into operate()

10 years agotool_operate: Moved locale setup code into operate_init()
Steve Holme [Wed, 5 Feb 2014 20:20:45 +0000 (20:20 +0000)]
tool_operate: Moved locale setup code into operate_init()

10 years agohttp2: minor update of the README
Daniel Stenberg [Wed, 5 Feb 2014 14:31:29 +0000 (15:31 +0100)]
http2: minor update of the README

10 years agohttp2: rely on content-encoding header
Fabian Frank [Wed, 5 Feb 2014 08:21:16 +0000 (00:21 -0800)]
http2: rely on content-encoding header

A server might respond with a content-encoding header and a response
that was encoded accordingly in HTTP-draft-09/2.0 mode, even if the
client did not send an accept-encoding header earlier. The server might
not send a content-encoding header if the identity encoding was used to
encode the response.

See:
http://tools.ietf.org/html/draft-ietf-httpbis-http2-09#section-9.3

10 years agotool_operate: shortened too-long source line
Dan Fandrich [Tue, 4 Feb 2014 20:49:10 +0000 (21:49 +0100)]
tool_operate: shortened too-long source line

10 years agotool_operate: Introduced operate_free() function
Steve Holme [Sun, 2 Feb 2014 15:37:28 +0000 (15:37 +0000)]
tool_operate: Introduced operate_free() function

10 years agotool_operate: Introduced operate_init() function
Steve Holme [Sun, 2 Feb 2014 15:28:25 +0000 (15:28 +0000)]
tool_operate: Introduced operate_init() function

10 years agotool_operate: Introduced new operate() function
Steve Holme [Sun, 2 Feb 2014 15:20:37 +0000 (15:20 +0000)]
tool_operate: Introduced new operate() function

10 years agohttp2: enforce gzip auto-decompress
Daniel Stenberg [Tue, 4 Feb 2014 14:07:08 +0000 (15:07 +0100)]
http2: enforce gzip auto-decompress

As this is mandated by the http2 spec draft-09

10 years agohttp2: handle incoming data larger than remaining buffer
Tatsuhiro Tsujikawa [Tue, 4 Feb 2014 13:57:29 +0000 (14:57 +0100)]
http2: handle incoming data larger than remaining buffer

10 years agohttp2: Check stream ID we are interested in
Tatsuhiro Tsujikawa [Sun, 2 Feb 2014 12:31:13 +0000 (21:31 +0900)]
http2: Check stream ID we are interested in

10 years agohttp2: store response header in temporary buffer
Tatsuhiro Tsujikawa [Tue, 4 Feb 2014 13:54:42 +0000 (14:54 +0100)]
http2: store response header in temporary buffer

10 years agoHTTP2: add layer between existing http and socket(TLS) layer
Tatsuhiro Tsujikawa [Fri, 31 Jan 2014 15:51:24 +0000 (00:51 +0900)]
HTTP2: add layer between existing http and socket(TLS) layer

This patch chooses different approach to integrate HTTP2 into HTTP curl
stack. The idea is that we insert HTTP2 layer between HTTP code and
socket(TLS) layer. When HTTP2 is initialized (either in NPN or Upgrade),
we replace the Curl_recv/Curl_send callbacks with HTTP2's, but keep the
original callbacks in http_conn struct. When sending serialized data by
nghttp2, we use original Curl_send callback. Likewise, when reading data
from network, we use original Curl_recv callback. In this way we can
treat both TLS and non-TLS connections.

With this patch, one can transfer contents from https://twitter.com and
from nghttp2 test server in plain HTTP as well.

The code still has rough edges. The notable one is I could not figure
out how to call nghttp2_session_send() when underlying socket is
writable.

10 years agogtls: add ALPN support
Fabian Frank [Tue, 4 Feb 2014 08:10:37 +0000 (00:10 -0800)]
gtls: add ALPN support

Add ALPN support when using GnuTLS >= 3.2.0. This allows
libcurl to negotiate HTTP/2.0 for https connections when
built with GnuTLS.

See:
http://www.gnutls.org/manual/gnutls.html#Application-Layer-Protocol-Negotiation-_0028ALPN_0029
http://tools.ietf.org/html/draft-ietf-tls-applayerprotoneg-04

10 years agotool_operate: Moved libcurl information gathering to tool_main
Steve Holme [Sun, 2 Feb 2014 15:11:22 +0000 (15:11 +0000)]
tool_operate: Moved libcurl information gathering to tool_main

10 years agoopenssl: add ALPN support
Fabian Frank [Mon, 3 Feb 2014 02:33:28 +0000 (18:33 -0800)]
openssl: add ALPN support

Add ALPN support when using OpenSSL. This will offer ALPN and NPN to the
server, who can respond with either one or none of the two. OpenSSL >=
1.0.2 is required, which means as of today obtaining a snapshot from
ftp://ftp.openssl.org/snapshot/.

See:
http://tools.ietf.org/html/draft-ietf-tls-applayerprotoneg-04
https://github.com/openssl/openssl/blob/ba168244a14bbd056e502d7daa04cae4aabe9d0d/ssl/ssl_lib.c#L1787

10 years agotool_operate: Moved command line argument parsing into separate function
Steve Holme [Sun, 2 Feb 2014 14:58:25 +0000 (14:58 +0000)]
tool_operate: Moved command line argument parsing into separate function

10 years agotool_operate: Simplified parse .curlrc decision logic
Steve Holme [Sun, 2 Feb 2014 14:29:54 +0000 (14:29 +0000)]
tool_operate: Simplified parse .curlrc decision logic

10 years agotool_operate: Moved main initialisation and cleanup code into tool_main
Steve Holme [Sun, 2 Feb 2014 14:21:53 +0000 (14:21 +0000)]
tool_operate: Moved main initialisation and cleanup code into tool_main

10 years agotool_main: Fixed compilation warning from commit 0104678c79
Steve Holme [Mon, 3 Feb 2014 12:16:22 +0000 (12:16 +0000)]
tool_main: Fixed compilation warning from commit 0104678c79

no previous prototype for function 'memory_tracking_init'

10 years agotool_main: Changed stack based config struct to be heap based
Steve Holme [Sun, 2 Feb 2014 13:58:35 +0000 (13:58 +0000)]
tool_main: Changed stack based config struct to be heap based

10 years agotests: Moved some comments so the test data files parse as XML
Dan Fandrich [Sun, 2 Feb 2014 23:23:34 +0000 (00:23 +0100)]
tests: Moved some comments so the test data files parse as XML

10 years agotool_operate: Moved memory tracking initialisation into tool_main
Steve Holme [Sun, 2 Feb 2014 13:45:35 +0000 (13:45 +0000)]
tool_operate: Moved memory tracking initialisation into tool_main

10 years agotests: Fixed test172 cookie expiry
Steve Holme [Sun, 2 Feb 2014 11:01:10 +0000 (11:01 +0000)]
tests: Fixed test172 cookie expiry

The test contains a cookie jar file where one of the cookies has an
expiry date of 1391252187 -- Sat, 1 Feb 2014 10:56:27 GMT which has
now expired. Updated to Wed, 14 Oct 2037 16:36:33 GMT as per test
179.

Reported-by: Adam Sampson
Bug: http://curl.haxx.se/bug/view.cgi?id=1330

10 years agotool_operate: Moved initial config setup into new init_config() function
Steve Holme [Sat, 1 Feb 2014 13:36:48 +0000 (13:36 +0000)]
tool_operate: Moved initial config setup into new init_config() function

10 years agotool_main: Moved config struct initialisation into a separate function
Steve Holme [Sat, 1 Feb 2014 13:26:09 +0000 (13:26 +0000)]
tool_main: Moved config struct initialisation into a separate function

In preparation for adding URL specific options moved the initialisation
of the Configurable structure into a separate function in tool_cfgable.

10 years agotest 500: workaround low timer resolution on Windows
Marc Hoersken [Sat, 1 Feb 2014 00:01:10 +0000 (01:01 +0100)]
test 500: workaround low timer resolution on Windows

Since the timer resolution is lower, there are actually cases that
the compared values are equal. Therefore we check for previous
timestamps being greater than the current one instead.

10 years agotest suite: stop conversion of valid output to CRLF on Windows
Marc Hoersken [Sat, 1 Feb 2014 12:49:58 +0000 (13:49 +0100)]
test suite: stop conversion of valid output to CRLF on Windows

Since the output isn't actually being written in text-mode and it
was rather used as a workaround, disable text-mode for these tests.

10 years agoHTTP tests: use CRLF as header seperator according to RFC 2616
Marc Hoersken [Sat, 1 Feb 2014 12:47:00 +0000 (13:47 +0100)]
HTTP tests: use CRLF as header seperator according to RFC 2616

10 years agoFTP tests: enable text-mode for more datacheck sections
Marc Hoersken [Sat, 1 Feb 2014 12:43:19 +0000 (13:43 +0100)]
FTP tests: enable text-mode for more datacheck sections

10 years agoFTP tests: enable text-mode for data and datacheck sections
Marc Hoersken [Sun, 26 Jan 2014 10:32:44 +0000 (11:32 +0100)]
FTP tests: enable text-mode for data and datacheck sections

10 years agoruntests.pl: added support for text-mode within datacheck section
Marc Hoersken [Sun, 26 Jan 2014 10:31:52 +0000 (11:31 +0100)]
runtests.pl: added support for text-mode within datacheck section

10 years agoftpserver.pl: directory LISTings use [CR][LF] for ASCII transfer
Marc Hoersken [Sun, 26 Jan 2014 10:30:41 +0000 (11:30 +0100)]
ftpserver.pl: directory LISTings use [CR][LF] for ASCII transfer

According to section 2.2 of RFC959 the End-of-Line is defined as:
 The end-of-line sequence defines the separation of printing
 lines.  The sequence is Carriage Return, followed by Line Feed.

Verified by sniffing traffic between a Windows FTP client (FileZilla)
and Unix-hosted FTP server (ProFTPD).

10 years agoruntests.pl: reverse line-ending conversion on Windows
Marc Hoersken [Sun, 26 Jan 2014 10:23:11 +0000 (11:23 +0100)]
runtests.pl: reverse line-ending conversion on Windows

It makes more sense to convert the expected output to [CR][LF] on
Windows than to force the actual, probably correct, output to [LF].

This way it is actually possible to see if curl outputs the correct
line-ending excepted by a text-aware test case.

10 years agowinssl: improved default SSL/TLS protocol selection
Marc Hoersken [Sun, 19 Jan 2014 13:13:21 +0000 (14:13 +0100)]
winssl: improved default SSL/TLS protocol selection

For some reason Windows 7 SP1 chooses TLS 1.0 instead of TLS 1.2
if it is not explicitly enabled within grbitEnabledProtocols.

More information can be found on MSDN:
http://msdn.microsoft.com/library/windows/desktop/aa379810.aspx

10 years agoINSTALL: Corrected mentioned version number as release 7.34.1 became 7.35.0
Steve Holme [Fri, 31 Jan 2014 00:25:11 +0000 (00:25 +0000)]
INSTALL: Corrected mentioned version number as release 7.34.1 became 7.35.0

10 years agoRELEASE-NOTES: Synced with 0f213fdca1
Steve Holme [Fri, 31 Jan 2014 00:20:32 +0000 (00:20 +0000)]
RELEASE-NOTES: Synced with 0f213fdca1

10 years agopipeline: Fixed a NULL pointer dereference on OOM
Dan Fandrich [Thu, 30 Jan 2014 23:05:36 +0000 (00:05 +0100)]
pipeline: Fixed a NULL pointer dereference on OOM

10 years agotests: make the authorization retry tests pass the torture tests
Dan Fandrich [Thu, 30 Jan 2014 22:18:20 +0000 (23:18 +0100)]
tests: make the authorization retry tests pass the torture tests

10 years agoftp: fixed a memory leak on wildcard error path
Dan Fandrich [Wed, 29 Jan 2014 21:40:57 +0000 (22:40 +0100)]
ftp: fixed a memory leak on wildcard error path

10 years agonetrc: Fixed a memory leak in an OOM condition
Dan Fandrich [Wed, 29 Jan 2014 07:10:26 +0000 (08:10 +0100)]
netrc: Fixed a memory leak in an OOM condition

10 years agontlm: Fixed a memory leak when using NTLM with a proxy server
Steve Holme [Thu, 30 Jan 2014 20:59:26 +0000 (20:59 +0000)]
ntlm: Fixed a memory leak when using NTLM with a proxy server

10 years agotests: Missed updating a type-3 message in commit 1c9aaa0bac
Steve Holme [Thu, 30 Jan 2014 20:51:34 +0000 (20:51 +0000)]
tests: Missed updating a type-3 message in commit 1c9aaa0bac

10 years agohttp2: fix size check in on_data_chunk_recv
Daniel Stenberg [Thu, 30 Jan 2014 19:22:54 +0000 (20:22 +0100)]
http2: fix size check in on_data_chunk_recv

10 years agohttp2: add CRLF when first data arrives
Daniel Stenberg [Thu, 30 Jan 2014 17:21:57 +0000 (18:21 +0100)]
http2: add CRLF when first data arrives

10 years agotests: Updated NTLM tests for NTLMv2 type-3 message
Steve Holme [Thu, 30 Jan 2014 19:11:21 +0000 (19:11 +0000)]
tests: Updated NTLM tests for NTLMv2 type-3 message

10 years agohttp2_recv: Return written length on CURLE_AGAIN
Tatsuhiro Tsujikawa [Thu, 30 Jan 2014 16:49:35 +0000 (17:49 +0100)]
http2_recv: Return written length on CURLE_AGAIN

10 years agohttp2: Use nghttp2_session_mem_recv and nghttp2_session_upgrade
Tatsuhiro Tsujikawa [Thu, 30 Jan 2014 14:26:20 +0000 (23:26 +0900)]
http2: Use nghttp2_session_mem_recv and nghttp2_session_upgrade

10 years agohttp2: call it "HTTP 2" and not 2.0
Daniel Stenberg [Thu, 30 Jan 2014 15:32:03 +0000 (16:32 +0100)]
http2: call it "HTTP 2" and not 2.0

The minor version will be dropped for HTTP 2 so it will make sense to
avoid using it in option names etc.

10 years agohttp2: basic version of receiving DATA
Daniel Stenberg [Thu, 30 Jan 2014 15:09:36 +0000 (16:09 +0100)]
http2: basic version of receiving DATA

10 years agohttp2: convert HEADER frames to HTTP1-like headers
Daniel Stenberg [Thu, 30 Jan 2014 14:58:07 +0000 (15:58 +0100)]
http2: convert HEADER frames to HTTP1-like headers

... and then go through the "normal" HTTP engine.

10 years agohttp2: fix EWOULDBLOCK in recv_callback()
Daniel Stenberg [Thu, 30 Jan 2014 13:31:05 +0000 (14:31 +0100)]
http2: fix EWOULDBLOCK in recv_callback()

10 years agohttp2: do the POST Upgrade dance properly
Daniel Stenberg [Thu, 30 Jan 2014 13:26:00 +0000 (14:26 +0100)]
http2: do the POST Upgrade dance properly

10 years agontlm: Use static client nonce for the test suite
Steve Holme [Thu, 30 Jan 2014 11:38:30 +0000 (11:38 +0000)]
ntlm: Use static client nonce for the test suite

10 years agohttp2.h: provide empty macros for non-http2 builds
Daniel Stenberg [Thu, 30 Jan 2014 10:46:59 +0000 (11:46 +0100)]
http2.h: provide empty macros for non-http2 builds

10 years agohttp2: switch into http2 mode if NPN indicates
Fabian Frank [Thu, 30 Jan 2014 05:28:50 +0000 (21:28 -0800)]
http2: switch into http2 mode if NPN indicates

Check the NPN result before preparing an HTTP request and switch into
HTTP/2.0 mode if necessary. This is a work in progress, the actual code
to prepare and send the request using nghttp2 is still missing from
Curl_http2_send_request().

10 years agohttp2: s/Curl_http2_request/Curl_http2_request_upgrade
Daniel Stenberg [Thu, 30 Jan 2014 10:32:04 +0000 (11:32 +0100)]
http2: s/Curl_http2_request/Curl_http2_request_upgrade

To better reflect its purpose

10 years agohttp2-openssl: verify that NPN functionality is present
Daniel Stenberg [Thu, 30 Jan 2014 10:24:15 +0000 (11:24 +0100)]
http2-openssl: verify that NPN functionality is present

10 years agoopenssl: set up hooks with to perform NPN
Fabian Frank [Thu, 30 Jan 2014 05:18:03 +0000 (21:18 -0800)]
openssl: set up hooks with to perform NPN

NPN is what is available in the wild today to negotiate SPDY or HTTP/2.0
connections. It is expected to be replaced by ALPN in the future. If
HTTP/2.0 is negotiated, this is indicated for the entire connection and
http.c is expected to initialize itself for HTTP/2.0 instead of
HTTP/1.1.

see:
http://technotes.googlecode.com/git/nextprotoneg.html
http://tools.ietf.org/html/draft-ietf-tls-applayerprotoneg-04

10 years agohttp2: added stubs for all nghttp2 callbacks
Daniel Stenberg [Wed, 29 Jan 2014 23:11:56 +0000 (00:11 +0100)]
http2: added stubs for all nghttp2 callbacks

This makes it easier to trace what's happening.

10 years agohttp2: use FIRSTSOCKET instead of 0 to index the sockets array
Daniel Stenberg [Wed, 29 Jan 2014 22:47:24 +0000 (23:47 +0100)]
http2: use FIRSTSOCKET instead of 0 to index the sockets array