platform/upstream/curl.git
10 years agobase64: Fixed compilation warnings when using Curl_base64_decode()
Steve Holme [Sun, 3 Nov 2013 12:25:07 +0000 (12:25 +0000)]
base64: Fixed compilation warnings when using Curl_base64_decode()

curl_sasl.c:294: warning: dereferencing type-punned pointer will break
strict-aliasing rules

getpart.c:201: warning: dereferencing type-punned pointer will break
strict-aliasing rules

10 years agoconnect: Fixed "Whut?" no server connection failures
Steve Holme [Sun, 3 Nov 2013 11:27:12 +0000 (11:27 +0000)]
connect: Fixed "Whut?" no server connection failures

Introduced in commit 7d7df831981fee curl would loop displaying "Whut?"
if it was trying to connect to an address and port that didn't have
anything listening on it.

10 years agohttp: Post base64 decoding tidy up
Steve Holme [Sun, 3 Nov 2013 10:17:26 +0000 (10:17 +0000)]
http: Post base64 decoding tidy up

Renamed copy_header_value() to Curl_copy_header_value() as this
function is now non static.

Simplified proxy flag in Curl_http_input_auth() when calling
sub-functions.

Removed unnecessary white space removal when using negotiate as it had
been missed in commit cdccb422671aeb.

10 years agoglob_range: pass the closing bracket for a-z ranges
Daniel Stenberg [Sun, 3 Nov 2013 09:08:10 +0000 (10:08 +0100)]
glob_range: pass the closing bracket for a-z ranges

Regression since commit 5ca96cb844102 (release in 7.33.0)

Reported-by: Marcin Gryszkalis
10 years agogetpart: Fixed base64 encoded parts following commit e17c1b25bc33eb
Steve Holme [Sat, 2 Nov 2013 17:00:00 +0000 (17:00 +0000)]
getpart: Fixed base64 encoded parts following commit e17c1b25bc33eb

10 years agohttp: Added proxy tunnel authentication message header value extraction
Steve Holme [Sat, 2 Nov 2013 11:18:39 +0000 (11:18 +0000)]
http: Added proxy tunnel authentication message header value extraction

...following recent changes to Curl_base64_decode() rather than trying
to parse a header line for the authentication mechanisms which is CRLF
terminated and inline zero terminate it.

10 years agohttp: Added authentication message header value extraction
Steve Holme [Wed, 30 Oct 2013 21:33:28 +0000 (21:33 +0000)]
http: Added authentication message header value extraction

...following recent changes to Curl_base64_decode() rather than trying
to parse a header line for the authentication mechanisms which is CRLF
terminated and inline zero terminate it.

10 years agocurl_multi_wait: accept 0 from multi_timeout() as valid timeout
Daniel Stenberg [Wed, 30 Oct 2013 22:48:08 +0000 (23:48 +0100)]
curl_multi_wait: accept 0 from multi_timeout() as valid timeout

The code rejected 0 as a valid timeout while in fact the function could
indeed legitimately return that and it should be respected.

Reported-by: Bjorn Stenberg
10 years agoemail: Corrected a couple of typos from commit aa0eaef4838ccd
Steve Holme [Wed, 30 Oct 2013 21:25:15 +0000 (21:25 +0000)]
email: Corrected a couple of typos from commit aa0eaef4838ccd

10 years agoTODO: Removed the 'Graceful base64 decoding failure' sections
Steve Holme [Wed, 30 Oct 2013 20:59:11 +0000 (20:59 +0000)]
TODO: Removed the 'Graceful base64 decoding failure' sections

Updated following the recent changes to support graceful failures
during the authentication phrase.

10 years agoemail: Post graceful SASL authentication cancellation tidy up
Steve Holme [Wed, 30 Oct 2013 20:56:38 +0000 (20:56 +0000)]
email: Post graceful SASL authentication cancellation tidy up

10 years agotests: use proper padding in NTLM responses
Kamil Dudka [Wed, 30 Oct 2013 16:52:19 +0000 (17:52 +0100)]
tests: use proper padding in NTLM responses

10 years agoNSS: support for CERTINFO feature
Patrick Monnerat [Wed, 30 Oct 2013 10:12:06 +0000 (11:12 +0100)]
NSS: support for CERTINFO feature

10 years agobase64: removed trailing white space
Daniel Stenberg [Wed, 30 Oct 2013 08:11:10 +0000 (09:11 +0100)]
base64: removed trailing white space

and updated copyright year

10 years agobase64: Added basic validation to base64 input string when decoding
Steve Holme [Wed, 30 Oct 2013 07:31:22 +0000 (07:31 +0000)]
base64: Added basic validation to base64 input string when decoding

A base64 string should be a multiple of 4 characters in length, not
contain any more than 2 padding characters and only contain padding
characters at the end of string. For example: Y3VybA==

Strings such as the following are considered invalid:

Y=   - Invalid length
Y==  - Invalid length
Y=== - More than two padding characters
Y=x= - Padding character contained within string

10 years agoRELEASE-NOTES: synced with 255826c40f9316
Daniel Stenberg [Tue, 29 Oct 2013 22:06:04 +0000 (23:06 +0100)]
RELEASE-NOTES: synced with 255826c40f9316

10 years agobugfix: Don't block waiting for socket1 connect.
Björn Stenberg [Tue, 29 Oct 2013 13:43:01 +0000 (14:43 +0100)]
bugfix: Don't block waiting for socket1 connect.

This patch fixes a bug in Happy Eyeballs where curl would wait for a
connect response from socket1 before checking socket2.

Also, it updates error messages for failed connections, showing the ip
addresses that failed rather than just the host name repeatedly.

Bug: http://curl.haxx.se/mail/lib-2013-10/0236.html
Reported-by: Paul Marks
10 years agosasl: Updated create_digest_md5_message() to use a dynamic buffer
Steve Holme [Sun, 27 Oct 2013 22:53:07 +0000 (22:53 +0000)]
sasl: Updated create_digest_md5_message() to use a dynamic buffer

10 years agoSECURITY: "curl security for developers"
Daniel Stenberg [Mon, 28 Oct 2013 22:19:55 +0000 (23:19 +0100)]
SECURITY: "curl security for developers"

Describes our security process from a project and curl developer's
perspective.

10 years agoOS400: coding style standards
Patrick Monnerat [Mon, 28 Oct 2013 11:00:22 +0000 (12:00 +0100)]
OS400: coding style standards

10 years agoemail: Added support for cancelling NTLM authentication
Steve Holme [Sun, 27 Oct 2013 22:20:18 +0000 (22:20 +0000)]
email: Added support for cancelling NTLM authentication

10 years agosasl: Removed unused variables from commit b87ba2c94217c0
Steve Holme [Sun, 27 Oct 2013 17:04:56 +0000 (17:04 +0000)]
sasl: Removed unused variables from commit b87ba2c94217c0

10 years agoemail: Added support for cancelling DIGEST-MD5 authentication
Steve Holme [Sun, 27 Oct 2013 16:27:38 +0000 (16:27 +0000)]
email: Added support for cancelling DIGEST-MD5 authentication

10 years agoemail: Corrected a couple of typos from 1e39b95682781f
Steve Holme [Sun, 27 Oct 2013 16:24:03 +0000 (16:24 +0000)]
email: Corrected a couple of typos from 1e39b95682781f

10 years agodocs/examples/httpput.c: fix build for MSVC
Gisle Vanem [Thu, 24 Oct 2013 13:21:16 +0000 (15:21 +0200)]
docs/examples/httpput.c: fix build for MSVC

"Dan Fandrich" <dan@coneharvesters.com> wrote:

>> But I'm not sure <unistd.h> is needed at all.
>
> It's needed for close(2). But the only reason that's needed is because fstat
> is used instead of stat(2); if you fix that, then you could remove that
> include altogether.

Okay. I've tested the following with MSVC and MingW. htttput.c now
simply uses stat():

10 years agoemail: Added support for canceling CRAM-MD5 authentication
Steve Holme [Sun, 27 Oct 2013 12:34:56 +0000 (12:34 +0000)]
email: Added support for canceling CRAM-MD5 authentication

10 years agoTypo fix in trynextip().
Björn Stenberg [Sun, 27 Oct 2013 11:27:52 +0000 (12:27 +0100)]
Typo fix in trynextip().

10 years agoTODO: remove "Happy Eyeball dual stack connect"
Daniel Stenberg [Sun, 27 Oct 2013 10:34:09 +0000 (11:34 +0100)]
TODO: remove "Happy Eyeball dual stack connect"

... as it was just merged in commit 7d7df

10 years agoAdd "Happy Eyeballs" for IPv4/IPv6.
Björn Stenberg [Sat, 26 Oct 2013 12:17:33 +0000 (14:17 +0200)]
Add "Happy Eyeballs" for IPv4/IPv6.

This patch invokes two socket connect()s nearly simultaneously, and
the socket that is first connected "wins" and is subsequently used for
the connection. The other is terminated.

There is a very slight IPv4 preference, in that if both sockets connect
simultaneously IPv4 is checked first and thus will win.

10 years agoemail: Added initial support for cancelling authentication
Steve Holme [Sun, 27 Oct 2013 09:10:38 +0000 (09:10 +0000)]
email: Added initial support for cancelling authentication

Should a client application fail to decode an authentication message
received from a server, or not support any of the parameters given by
the server in the message, then the authentication phrase should be
cancelled gracefully by the client rather than simply terminating the
connection.

The authentication phrase should be cancelled by simply sending a '*'
to the server, in response to erroneous data being received, as per
RFC-3501, RFC-4954 and RFC-5034.

This patch adds the necessary state machine constants and appropriate
response handlers in order to add this functionality for the CRAM-MD5,
DIGEST-MD5 and NTLM authentication mechanisms.

10 years agoemail: Moved authentication message parsing into a separate function
Steve Holme [Sun, 27 Oct 2013 09:00:41 +0000 (09:00 +0000)]
email: Moved authentication message parsing into a separate function

...in preparation for upcoming modifications.

10 years agoftp: Fixed compiler warning
Steve Holme [Sat, 26 Oct 2013 23:00:01 +0000 (00:00 +0100)]
ftp: Fixed compiler warning

warning: 'result' may be used uninitialized in this function

10 years agoFTP: make the data connection work when going through proxy
Daniel Stenberg [Sat, 26 Oct 2013 18:19:27 +0000 (20:19 +0200)]
FTP: make the data connection work when going through proxy

This is a regression since the switch to always-multi internally
c43127414d89c.

Test 1316 was modified since we now clearly call the Curl_client_write()
function when doing the LIST transfer part and then the
handler->protocol says FTP and ftpc.transfertype is 'A' which implies
text converting even though that the response is initially a HTTP
CONNECT response in this case.

10 years agotool_help: Added login options to --user description
Steve Holme [Sat, 26 Oct 2013 19:11:48 +0000 (20:11 +0100)]
tool_help: Added login options to --user description

10 years agoemail: Added references to SASL LOGIN authentication draft proposal
Steve Holme [Sat, 26 Oct 2013 13:35:27 +0000 (14:35 +0100)]
email: Added references to SASL LOGIN authentication draft proposal

10 years agotests: Tidy up of SMTP and POP3 tests
Steve Holme [Sat, 26 Oct 2013 12:44:31 +0000 (13:44 +0100)]
tests: Tidy up of SMTP and POP3 tests

Corrected line endings, RFC references and standardised on user names
and passwords used in the tests.

10 years agotool_help: Added clarity to the --oauth2-bearer option
Steve Holme [Sat, 26 Oct 2013 12:01:33 +0000 (13:01 +0100)]
tool_help: Added clarity to the --oauth2-bearer option

...as XOAUTH2 is the extended (or non-standard) SASL identifier and
OAuth 2 is the protocol name (and version).

10 years agosmtp: Fixed response code parsing for bad AUTH continuation responses
Steve Holme [Sat, 26 Oct 2013 10:09:31 +0000 (11:09 +0100)]
smtp: Fixed response code parsing for bad AUTH continuation responses

This workaround had been previously been implemented for IMAP and POP3
but not SMTP. Some of the recent test case additions implemented this
behaviour to emulate a bad server and the SMTP code didn't cope with it.

10 years agogskit.c: Code policing following commit 2cc9246477285d
Steve Holme [Fri, 25 Oct 2013 23:57:45 +0000 (00:57 +0100)]
gskit.c: Code policing following commit 2cc9246477285d

Corrected 80 character line length error and pointer declarations (some
of which were previously incorrect)

10 years agotest907: Corrected DIGEST-MD5 response given in commit 820ed48a0088cd
Steve Holme [Fri, 25 Oct 2013 23:24:45 +0000 (00:24 +0100)]
test907: Corrected DIGEST-MD5 response given in commit 820ed48a0088cd

As the URI, which is contained within the DIGEST-MD5 response, is
constructed from the service and realm, the encoded message differs
from that generated under POP3.

10 years agoRELEASE-NOTES: Synced with d24b7953c2132a
Steve Holme [Fri, 25 Oct 2013 19:00:00 +0000 (20:00 +0100)]
RELEASE-NOTES: Synced with d24b7953c2132a

10 years agotests: Added SMTP OAUTH2 authentication with initial response test
Steve Holme [Fri, 25 Oct 2013 18:48:48 +0000 (19:48 +0100)]
tests: Added SMTP OAUTH2 authentication with initial response test

10 years agotests: Added SMTP NTLM authentication with initial response test
Steve Holme [Fri, 25 Oct 2013 18:40:40 +0000 (19:40 +0100)]
tests: Added SMTP NTLM authentication with initial response test

10 years agotests: Added SMTP OAUTH2 authentication test
Steve Holme [Fri, 25 Oct 2013 18:31:30 +0000 (19:31 +0100)]
tests: Added SMTP OAUTH2 authentication test

10 years agotests: Added SMTP DIGEST-MD5 authentication test
Steve Holme [Fri, 25 Oct 2013 18:22:30 +0000 (19:22 +0100)]
tests: Added SMTP DIGEST-MD5 authentication test

10 years agotests: Regrouped SMTP authentication tests
Steve Holme [Fri, 25 Oct 2013 18:17:15 +0000 (19:17 +0100)]
tests: Regrouped SMTP authentication tests

10 years agoOS400: sync RPG wrapper, zlib support, fix header file names, ...
Patrick Monnerat [Fri, 25 Oct 2013 16:37:37 +0000 (18:37 +0200)]
OS400: sync RPG wrapper, zlib support, fix header file names, ...
IFS compilation support, SSL GSKit backend by default, TLSv1.[12] support in
  GSKit for OS400 >= V7R1, no more tabs in make scripts.

10 years agosasl: Fixed memory leak in OAUTH2 message creation
Steve Holme [Wed, 23 Oct 2013 23:16:59 +0000 (00:16 +0100)]
sasl: Fixed memory leak in OAUTH2 message creation

10 years agoftpserver.pl: Added support for empty pop3 authentication data
Steve Holme [Wed, 23 Oct 2013 21:19:42 +0000 (22:19 +0100)]
ftpserver.pl: Added support for empty pop3 authentication data

10 years agoCURLOPT_RESOLVE: mention they don't time-out
Daniel Stenberg [Wed, 23 Oct 2013 21:14:17 +0000 (23:14 +0200)]
CURLOPT_RESOLVE: mention they don't time-out

Clarify in the documentation that DNS entries added with CURLOPT_RESOLVE
won't time-out.

Bug: http://curl.haxx.se/mail/lib-2013-10/0062.html
Reported-by: Romulo Ceccon
10 years agotests: Added POP3 OAUTH2 authentication test
Steve Holme [Wed, 23 Oct 2013 21:10:10 +0000 (22:10 +0100)]
tests: Added POP3 OAUTH2 authentication test

10 years agotests: Added empty response support to custom replies
Steve Holme [Wed, 23 Oct 2013 21:05:22 +0000 (22:05 +0100)]
tests: Added empty response support to custom replies

...and fixed up test869 as DIGEST-MD transcript is as follows:

S: Challenge
C: Authentication String
S: Continue Response
C: Empty String

10 years agosasl: fix compiler warning
Daniel Stenberg [Wed, 23 Oct 2013 20:28:13 +0000 (22:28 +0200)]
sasl: fix compiler warning

error: unused variable 'table16'

10 years agotests: Added POP3 DIGEST-MD5 authentication test
Steve Holme [Wed, 23 Oct 2013 17:29:29 +0000 (18:29 +0100)]
tests: Added POP3 DIGEST-MD5 authentication test

10 years agoconfigure: check for long long when building with cyassl
Daniel Stenberg [Tue, 22 Oct 2013 20:55:48 +0000 (22:55 +0200)]
configure: check for long long when building with cyassl

cyassl/ctaocrypt/types.h needs SIZEOF_LONG_LONG

Reported-by: Chris Conlon
10 years agotest1240: verify 867b52a7ac52 (glob ranges with text to the right)
Daniel Stenberg [Mon, 21 Oct 2013 22:10:16 +0000 (00:10 +0200)]
test1240: verify 867b52a7ac52 (glob ranges with text to the right)

10 years agoglob: fix regression from commit 5ca96cb844
Daniel Stenberg [Mon, 21 Oct 2013 22:01:17 +0000 (00:01 +0200)]
glob: fix regression from commit 5ca96cb844

Plain strings after glob ranges/lists weren't treated correctly but
caused broken URLs to get used.

Reported-by: Javier Barroso
10 years agoAdding a .travis.yml file to use the travis-ci.org
Rémy Léone [Thu, 17 Oct 2013 13:10:18 +0000 (15:10 +0200)]
Adding a .travis.yml file to use the travis-ci.org

From wikipedia:

Travis CI is a hosted, distributed continuous integration service used
to build and test projects hosted at GitHub.

Travis CI is configured by adding a file named .travis.yml, which is a
YAML format text file, to the root directory of the GitHub repository.

Travis CI automatically detects when a commit has been made and pushed
to a GitHub repository that is using Travis CI, and each time this
happens, it will try to build the project and run tests. This includes
commits to all branches, not just to the master branch. When that
process has completed, it will notify a developer in the way it has been
configured to do so — for example, by sending an email containing the
test results (showing success or failure), or by posting a message on an
IRC channel. It can be configured to run the tests on a range of
different machines, with different software installed (such as older
versions of a programming language, to test for compatibility).

10 years agossh: initialize per-handle data in ssh_connect()
Kamil Dudka [Mon, 21 Oct 2013 16:47:54 +0000 (18:47 +0200)]
ssh: initialize per-handle data in ssh_connect()

... if not already initialized.  This fixes a regression introduced by
commit 4ad8e142da463ab208d5b5565e53291c8e5ef038, which caused test619
to intermittently fail on certain machines (namely Fedora build hosts).

10 years agocurl.1: add missing exit-code
Gisle Vanem [Wed, 16 Oct 2013 20:31:51 +0000 (22:31 +0200)]
curl.1: add missing exit-code

I noted a missing text for exit-code 89 in docs/curl.1

10 years agocmake: unbreak for non-Windows platforms
Daniel Stenberg [Sun, 20 Oct 2013 15:12:56 +0000 (17:12 +0200)]
cmake: unbreak for non-Windows platforms

Patch-by: Oliver Kuckertz
Bug: http://curl.haxx.se/bug/view.cgi?id=1292

10 years agoftpserver.pl: Fixed syntax error from commit 5b31b38c27bb7a
Steve Holme [Sat, 19 Oct 2013 23:30:03 +0000 (00:30 +0100)]
ftpserver.pl: Fixed syntax error from commit 5b31b38c27bb7a

10 years agotest866: Fixed user response from commit 7f7fbe7fbdb449
Steve Holme [Sat, 19 Oct 2013 20:43:20 +0000 (21:43 +0100)]
test866: Fixed user response from commit 7f7fbe7fbdb449

10 years agoftpserver.pl: Fixed processing of POP3 authentication strings
Steve Holme [Sat, 19 Oct 2013 19:39:18 +0000 (20:39 +0100)]
ftpserver.pl: Fixed processing of POP3 authentication strings

...and corrected response when check fails from 500 to -ERR.

10 years agotests: Added POP3 NTLM authentication test
Steve Holme [Sat, 19 Oct 2013 12:17:25 +0000 (13:17 +0100)]
tests: Added POP3 NTLM authentication test

10 years agotests: Added POP3 CRAM-MD5 authentication test
Steve Holme [Sat, 19 Oct 2013 11:57:25 +0000 (12:57 +0100)]
tests: Added POP3 CRAM-MD5 authentication test

10 years agotests: Added POP3 login authentication test
Steve Holme [Sat, 19 Oct 2013 11:42:25 +0000 (12:42 +0100)]
tests: Added POP3 login authentication test

10 years agotests: Added POP3 plain authentication test
Steve Holme [Sat, 19 Oct 2013 11:31:55 +0000 (12:31 +0100)]
tests: Added POP3 plain authentication test

10 years agotests: Added POP3 APOP authentication test
Steve Holme [Sat, 19 Oct 2013 11:20:00 +0000 (12:20 +0100)]
tests: Added POP3 APOP authentication test

10 years agoftpserver.pl: Added support for APOP POP3 authentication
Steve Holme [Sat, 19 Oct 2013 09:38:19 +0000 (10:38 +0100)]
ftpserver.pl: Added support for APOP POP3 authentication

10 years agotests: Added POP3 RSET test
Steve Holme [Fri, 18 Oct 2013 23:10:29 +0000 (00:10 +0100)]
tests: Added POP3 RSET test

10 years agoRELEASE-NOTES: Synced with ce61510127ea60
Steve Holme [Fri, 18 Oct 2013 18:54:30 +0000 (19:54 +0100)]
RELEASE-NOTES: Synced with ce61510127ea60

10 years agoemail: Fixed QUIT / LOGOUT being sent when SSL connect fails
Steve Holme [Fri, 18 Oct 2013 18:28:20 +0000 (19:28 +0100)]
email: Fixed QUIT / LOGOUT being sent when SSL connect fails

10 years agocurl_sasl: initialize NSS before using crypto
Kamil Dudka [Fri, 18 Oct 2013 13:37:18 +0000 (15:37 +0200)]
curl_sasl: initialize NSS before using crypto

10 years agoSSL: Follow up work to commits 6a1363128f1107 and 87861c9b0e8155
Steve Holme [Thu, 17 Oct 2013 18:57:26 +0000 (19:57 +0100)]
SSL: Follow up work to commits 6a1363128f1107 and 87861c9b0e8155

Changed the failure code when TLS v1.1 and v1.2 is requested but not
supported by older OpenSSL versions, following review from libcurl
peers, and reduced the number of required preprocessor if statements.

10 years agoSSL: Added unsupported cipher version check for OpenSSL
Steve Holme [Wed, 16 Oct 2013 19:48:24 +0000 (20:48 +0100)]
SSL: Added unsupported cipher version check for OpenSSL

...with the use of CURL_SSLVERSION_TLSv1_1 and CURL_SSLVERSION_TLSv1_2
being conditional on OpenSSL v1.0.1 as the appropriate flags are not
supported under earlier versions.

10 years agoDOCS: Added libcurl version number to CURLOPT_SSLVERSION
Steve Holme [Wed, 16 Oct 2013 19:18:15 +0000 (20:18 +0100)]
DOCS: Added libcurl version number to CURLOPT_SSLVERSION

10 years agoSSL: Corrected version number for new symbols from commit ad34a2d5c87c7f
Steve Holme [Wed, 16 Oct 2013 19:06:23 +0000 (20:06 +0100)]
SSL: Corrected version number for new symbols from commit ad34a2d5c87c7f

10 years agoSSL: Corrected typo from commit 87861c9b0e8155
Steve Holme [Tue, 15 Oct 2013 23:57:01 +0000 (00:57 +0100)]
SSL: Corrected typo from commit 87861c9b0e8155

10 years agoSSL: Fixed OpenSSL builds prior to v1.0.1
Steve Holme [Tue, 15 Oct 2013 23:49:23 +0000 (00:49 +0100)]
SSL: Fixed OpenSSL builds prior to v1.0.1

Commit ad34a2d5c87c7f relies on definitions that are only present in
OpenSSL v1.0.1 and up. This quick fix allows the builds that use
older versions of OpenSSL to continue building.

10 years agotest906: Fixed failing test on some platforms
Steve Holme [Tue, 15 Oct 2013 20:31:14 +0000 (21:31 +0100)]
test906: Fixed failing test on some platforms

Bug: http://sourceforge.net/p/curl/bugs/1291
Reported-by: David Walser
10 years agoNSS: acknowledge the --no-sessionid/CURLOPT_SSL_SESSIONID_CACHE option
Paul Donohue [Tue, 15 Oct 2013 19:36:32 +0000 (21:36 +0200)]
NSS: acknowledge the --no-sessionid/CURLOPT_SSL_SESSIONID_CACHE option

10 years agossh: Handle successful SSH_USERAUTH_NONE
Tyler Hall [Mon, 14 Oct 2013 20:24:17 +0000 (22:24 +0200)]
ssh: Handle successful SSH_USERAUTH_NONE

According to the documentation for libssh2_userauth_list(), a NULL
return value is not necessarily an error. You must call
libssh2_userauth_authenticated() to determine if the SSH_USERAUTH_NONE
request was successful.

This fixes a segv when using sftp on a server that allows logins with an
empty password. When NULL was interpreted as an error, it would
free the session but not flag an error since the libssh2 errno would be
clear. This resulted in dereferencing a NULL session pointer.

Signed-off-by: Tyler Hall <tylerwhall@gmail.com>
10 years agousercertinmem: fix memory leaks
Ishan SinghLevett [Tue, 15 Oct 2013 18:48:22 +0000 (20:48 +0200)]
usercertinmem: fix memory leaks

10 years agobuild: distribute and install libcurl.m4 by default
Dave Reisner [Sun, 13 Oct 2013 17:39:42 +0000 (19:39 +0200)]
build: distribute and install libcurl.m4 by default

10 years agotool: use XFERFUNCTION to save some casts
Dave Reisner [Sun, 13 Oct 2013 17:39:41 +0000 (19:39 +0200)]
tool: use XFERFUNCTION to save some casts

10 years agocurl.1: fix typo conjuction -> conjunction
Alessandro Ghedini [Mon, 14 Oct 2013 18:49:54 +0000 (20:49 +0200)]
curl.1: fix typo conjuction -> conjunction

10 years agocurl: document the new --tlsv1.[012] options
Daniel Stenberg [Tue, 15 Oct 2013 18:31:04 +0000 (20:31 +0200)]
curl: document the new --tlsv1.[012] options

10 years agoSSL: protocol version can be specified more precisely
Gergely Nagy [Thu, 19 Sep 2013 13:17:13 +0000 (15:17 +0200)]
SSL: protocol version can be specified more precisely

CURL_SSLVERSION_TLSv1_0, CURL_SSLVERSION_TLSv1_1,
CURL_SSLVERSION_TLSv1_2 enum values are added to force exact TLS version
(CURL_SSLVERSION_TLSv1 means TLS 1.x).

axTLS:
axTLS only supports TLS 1.0 and 1.1 but it cannot be set that only one
of these should be used, so we don't allow the new enum values.

darwinssl:
Added support for the new enum values.

SChannel:
Added support for the new enum values.

CyaSSL:
Added support for the new enum values.
Bug: The original CURL_SSLVERSION_TLSv1 value enables only TLS 1.0 (it
did the same before this commit), because CyaSSL cannot be configured to
use TLS 1.0-1.2.

GSKit:
GSKit doesn't seem to support TLS 1.1 and TLS 1.2, so we do not allow
those values.
Bugfix: There was a typo that caused wrong SSL versions to be passed to
GSKit.

NSS:
TLS minor version cannot be set, so we don't allow the new enum values.

QsoSSL:
TLS minor version cannot be set, so we don't allow the new enum values.

OpenSSL:
Added support for the new enum values.
Bugfix: The original CURL_SSLVERSION_TLSv1 value enabled only TLS 1.0,
now it enables 1.0-1.2.

Command-line tool:
Added command line options for the new values.

10 years agodarwinssl: un-break iOS build after PKCS#12 feature added
Nick Zitzmann [Mon, 14 Oct 2013 23:03:32 +0000 (18:03 -0500)]
darwinssl: un-break iOS build after PKCS#12 feature added

SecPKCS12Import() returns a few errors that are enumerated in OS X's
headers but not in iOS' headers for some reason.

10 years agobump: start working on 7.33.1
Daniel Stenberg [Mon, 14 Oct 2013 21:11:20 +0000 (23:11 +0200)]
bump: start working on 7.33.1

10 years agoTHANKS: added contributors from the 7.33.0 announcement
Daniel Stenberg [Mon, 14 Oct 2013 21:10:35 +0000 (23:10 +0200)]
THANKS: added contributors from the 7.33.0 announcement

10 years agoRELEASE-NOTES: synced with 92cf6141ed0de
Daniel Stenberg [Sun, 13 Oct 2013 21:24:21 +0000 (23:24 +0200)]
RELEASE-NOTES: synced with 92cf6141ed0de

10 years agocurl: fix --oauth2-bearer in the --help output
Daniel Stenberg [Sun, 13 Oct 2013 21:21:12 +0000 (23:21 +0200)]
curl: fix --oauth2-bearer in the --help output

After the option rename in 5df04bfafd1

10 years agoOpenSSL: improve the grammar of the language in 39beaa5ffbcc
Daniel Stenberg [Sun, 13 Oct 2013 21:07:44 +0000 (23:07 +0200)]
OpenSSL: improve the grammar of the language in 39beaa5ffbcc

Reported-by: Petr Pisar
10 years agoOpenSSL: use failf() when subjectAltName mismatches
Andrej E Baranov [Sat, 12 Oct 2013 23:02:03 +0000 (01:02 +0200)]
OpenSSL: use failf() when subjectAltName mismatches

Write to CURLOPT_ERRORBUFFER information about mismatch alternative
certificate subject names.

Signed-off-by: Andrej E Baranov <admin@andrej-andb.ru>
10 years agocurl: rename --bearer to --oauth2-bearer
Daniel Stenberg [Sat, 12 Oct 2013 21:26:38 +0000 (23:26 +0200)]
curl: rename --bearer to --oauth2-bearer

The option '--bearer' might be slightly ambiguous in name. It doesn't
create any conflict that I am aware of at the moment, however, OAUTH v2
is not the only authentication mechanism which uses "bearer" tokens.

Reported-by: Kyle L. Huff
URL: http://curl.haxx.se/mail/lib-2013-10/0064.html

10 years agossh: improve the logic for detecting blocking direction
Kamil Dudka [Mon, 7 Oct 2013 14:07:50 +0000 (16:07 +0200)]
ssh: improve the logic for detecting blocking direction

This fixes a regression introduced by commit 0feeab78 limiting the speed
of SCP upload to 16384 B/s on a fast connection (such as localhost).

10 years agoFixed typo in Makefile.inc that left http2.h out of the tar ball
Dan Fandrich [Sat, 12 Oct 2013 19:09:17 +0000 (21:09 +0200)]
Fixed typo in Makefile.inc that left http2.h out of the tar ball

10 years agominor fix in doc
Heinrich Schaefer [Thu, 10 Oct 2013 21:08:47 +0000 (23:08 +0200)]
minor fix in doc