Merge pull request #46 from nekopsykose/fix-asyncns-addrinfo fix macro define for freeaddrinfo with asyncns committer: GitHub <noreply@github.com>

fix macro define for freeaddrinfo with asyncns previously, this define was only made in lm-asyncns-resolver.c, which does not propagate it to lm-resolver.c, where freeaddrinfo() is used. this means that when asyncns support is used, the standard libc freeaddrinfo() is called on an addrinfo acquired from asyncns_getaddrinfo, which is undefined behaviour, and leads to a segfault on musl libc.

Merge pull request #43 from ryandesign/patch-1 Fix typo: GnuTSL -> GnuTLS committer: GitHub <noreply@github.com>

Fix typo: GnuTSL -> GnuTLS committer: GitHub <noreply@github.com>

update tags

Bump version to 1.5.4

g_type_class_add_private is deprecated since glib 2.58

Fix getaddrinfo() handling in blocking resolver If getaddrinfo() fails, the first call setting the result will already free the resolver. Trying to access it afterwards will lead to a warning to the console from glib. getaddrinfo() shouldn't return NULL for the result list if it returns successful.

Handle G_IO_HUP and G_IO_NVAL when connecting

Removed the last g_print calls The calls g_print() in lm-debug.c don't count, because the log handler is normally provided by the application.

Always signal a disconnect after connection_do_close()

Don't use compression API for GnuTLS >= 3.6.0 GnuTLS removed compression completely.

lm-resolver.c: limit conditional include to OS X Explicitly including arpa/nameser_compat.h is only necessary for OS X. It's unneeded on other OSes. Some, e.g. Linux and NetBSD, include nameser_compat.h from nameser.h anyway, so at minimum, it's redundant. In the case of NetBSD, this was causing build breakages on some architectures due to legacy code in nameser_compat.h being exposed before the inclusion of nameser.h, which sets dependent macros. committer: mcabber <lm@freakysoft.de>

Fix compilation without deprecated OpenSSL 1.1 APIs committer: mcabber <lm@freakysoft.de>

Merge pull request #36 from kempniu/fix-falling-back-to-another-resolver-result-when-connecting Fix falling back to another resolver result when connecting committer: GitHub <noreply@github.com>

Fix falling back to another resolver result when connecting

Bump minimum GLib version See PR #35.

Merge pull request #35 from kempniu/move-away-from-g_type_class_add_private Move away from g_type_class_add_private() committer: GitHub <noreply@github.com>

Merge pull request #34 from kempniu/support-connecting-to-ipv6-servers Support connecting to IPv6 servers committer: GitHub <noreply@github.com>

Support connecting to IPv6 servers

Move away from g_type_class_add_private() g_type_class_add_private() will be deprecated in GLib 2.58. Replace: - g_type_class_add_private() with G_DEFINE_TYPE_WITH_PRIVATE() - G_TYPE_INSTANCE_GET_PRIVATE() with *_get_instance_private()

Merge pull request #33 from kempniu/GIOCondition-is-a-bit-field GIOCondition is a bit field committer: GitHub <noreply@github.com>

GIOCondition is a bit field Use bitwise AND instead of equality checks when processing a GIOCondition value passed to a GIOChannel callback since it may be a bitwise combination of multiple enum values (e.g. G_IO_OUT | G_IO_ERR).

Merge pull request #31 from tmp170422/patch-2 An apparent typo in lm-sasl.c:532 committer: GitHub <noreply@github.com>

An apparent typo committer: GitHub <noreply@github.com>

Merge pull request #29 from tmp170422/patch-1 Fix domain names with wildcard CN committer: GitHub <noreply@github.com>

Fix a typo committer: GitHub <noreply@github.com>

Fix an issue with wildcard domain names in CN When CN contains wildcards, the SAN list would be checked against corrupt server name. committer: GitHub <noreply@github.com>

Merge pull request #28 from MikePetullo/master Allow explicit disable of libidn at configure time committer: GitHub <noreply@github.com>

Allow explicit disable of libidn at configure time Signed-off-by: W. Michael Petullo <mike@flyn.org>

retry gnutls handshake on GNUTLS_E_INTERRUPTED

retry gnu tls handshake on GNUTLS_E_AGAIN committer: Frank Zschockelt <lm@freakysoft.de>

Return the correct enum lm_message_get_sub_type

lm-asyncns-resolver: Fix resolver uninitialization * Hopefully fix crash, when resolver CB gets called on non-existing object, if there's some problem with DNS and library user destroys LM instance before DNS timeout. * This introduces freeze, if underlying libasyncns uses pthreads instead of forking, but that is a bug in libasyncns, that will be hopefully fixed by upstream. Not reported yet.

Remove dummy files from sources list

LmMessageNode: Expose attributes member * New typedef: LmMessageNodeAttribute

autogen: Fix option name in help

update tags

Merge pull request #25 from mbiebl/dist-autogen Include autogen.sh in dist tarball

Include autogen.sh in dist tarball It is convenient to have autogen.sh in the dist tarball in case one needs to regenerate the build system.

Bumped version to 1.5.3

Deprecate MD5 fingerprints for SHA256

Fix coding style

Load system certs only if user doesn't specify trusted certs

Update required GnuTLS version number

Use portable gnutls function to trust system certs GnuTLS supports that function since 3.0.20.

Use "dnl" for configure.ac-only comments

Add --with-pkgconfigdir ./configure option

Fix "TEST_PROGS multiply defined" warning

Make gtk-doc.m4 optional

Update .gitignore

Fix automake warning about requiring AM_PROG_AR in configure.ac

Fix "AM_CPPFLAGS multiply defined" warnings

Add empty m4 directory

Fix libtoolize warnings

Update the postal address of the FSF

be as precise as possible when handling return values

update tags

Merge pull request #23 from mbiebl/drop-dolt Stop using dolt

Stop using dolt Dolt is no longer maintained upstream [1] and libtool has gotten much better. The permformance difference is neglible nowadays. [1] http://dolt.freedesktop.org/

Bumped version to 1.5.2

Check for __res_query in libresolv Changeset 74d6f2447f59a08c breaks builds here (Debian Jessie), because __res_query is not found...

Merge pull request #22 from moench-tegeder/build_20160111 assorted minor build fixes

openssl: always fill the fingerprint field of LmSSL

remove rcsid as loudmouth is maintained in git, the whole concept of rcsid does not really apply anymore - but we get warnings from -Wunused which are promptly upgraded to errors (-Werror)

try generic cc before testing for gcc FreeBSD has clang as it's primary C compiler (and often no gcc, these days). To avoid using the "wrong" compiler (gcc installed via ports if available), test for cc first.

remove libresolv relict As far as I can tell, libresolv is not needed anymore. To avoid build breakage on systems where the resolver functions live in libc, just remove -lresolv

include the right gssapi.h some time ago gssapi.h moved to gssapi/gssapi.h, and on some systems (e.g. FreeBSD) gssapi.h carries a deprecation #warning - which becomes a fatal error since we're compiling with -Werror

include netinet/in.h in resolver On some platforms (notably FreeBSD) netinet/in.h is required for the struct in_addr type and IPPROTO_* constants. As this header and both required definitions are documented in IEEE Std 1003.1, this should be safe on all systems.

Add libasyncns to Libs.private

Add the new API to the reference documentation

update tags

Update README wrt new release

Update release information files Change the maintainers.

Bumped version to 1.5.1

Fix a few memory leaks

Merge pull request #18 from tyll/libidn Do not overlink libidn

Do not overlink libidn References: https://bugzilla.redhat.com/show_bug.cgi?id=836761 committer: Till Maas <opensource@till.name>

Merge pull request #19 from hph86/master Link to TLS-secured issue tracker at GitHub.

Link to TLS-secured issue tracker at GitHub.

Merge pull request #17 from tyll/whitespace Remove trailing whitespace and mixed indenting

Remove trailing whitespace and mixed indenting

configure: make it compatible with current glib/gtk Signed-off-by: Paul Fertser <fercerpav@gmail.com> Signed-off-by: Mikael Berthe <mikael@lilotux.net> committer: Mikael Berthe <mikael@lilotux.net>

Merge pull request #15 from tuxad-loudmouth/master SSL/TLS protocol usage; use SANs of CERTs

CERTs: added checking of SAN(s) when CN does not match; fix CN wildcard handling

activate TLS 1.1 and 1.2; deactivate SSL 2 and 3

Merge pull request #14 from languitar/fix-compile-warnings Fix compilation warnings

fix compilation warnings * Make LmLogLevelFlags a typedef to GLogLevelFlags to avoid implicit casting * Remove a useless comparison of an unsigned int < 0

Merge pull request #11 from languitar/subdir-obje Use subdir-objects option for autoconf

Use subdir-objects option for autoconf This option is required for recent autoconf versions. Otherwise the build fails.

Merge pull request #12 from weiss/define-dir-type Make opendir(3) prototypes available

Make opendir(3) prototypes available

coding style fix

Fix TCP keepalives on OS X 10.9 Mac OS X 10.9 defines TCP_KEEPCNT, which triggers USE_TCP_KEEPALIVES. However, it defines TCP_KEEPALIVE rather than TCP_KEEPIDLE.

Implemented lm_ssl_set_ca This enables the application to choose additional certificates to trust

Implemented lm_ssl_set_cipher_list This enables the application to choose the cipher suites which are used by openssl or gnutls

Use the default cipher suite of GnuTLS and disable compression According to Nikos from GnuTLS, deflate can only cause harm: http://lists.gnutls.org/pipermail/gnutls-devel/2013-August/006428.html

Disable g_type_init for glib >= 2.35

Fix resolver record parsing * Compensate position for removed GETSHORT statement in previous commit 19a4b61c6b475ec67effe686e9037ccaad6e5f83 * Closes mcabber bug #111 * Thanks to Christoph Moench-Tegeder

update tags

Change mapping of openssl errors to lm ssl errors

Use connection_call_auth_cb in one more place

Minor documentation fixes

Merge pull request #7 from isbear/master Fix mcabber room name

Add COMP-NULL to prioritystring for gnutls

Fix MUC room address in README

Merge pull request #6 from isbear/master Fix remaining deprecated symbols

Update contacts in README

[autogen] always fail, if not in source root

gnutls fixes part 2

Merge pull request #5 from isbear/master Building fixes

Non-doc scenario for autogen.sh

Gnutls fixes part1

Add _unregister_reply_handler to gtkdoc sections

Update gitignore (add *~ and setup-build.stamp)

s/INCLUDES/AM_CPPFLAGS/

Update gtkdoc.m4

Fix compilation warnings

Merge pull request #4 from tetromino/master Fix building with glib-2.32

Fix building with glib-2.32 In >=glib-2.31, most glib subheaders cannot be included directly. See http://git.gnome.org/browse/glib/commit/?id=7455dd370eb37ce3b0b409ff6120501f37b50569

Revert mistake in previous changeset

Fix some gcc warnings

Workaround for parser to support prefixed tag names

Protect GMarkup parser against invalid unicode characters

Give up unneeded ping request on connection timeout

Fix segfault in asyncns resolver

Revert "Code looks for an auth element below the returned streams:feature" This reverts commit b94583bb7db80803c48add517a5559d45b6f3a58. Do not try to hack around bugs of the server...

Ignore case when comparing the CN with the domain

Removed an internal asyncns relict

Remove internal asyncns

Fix a double-free triggered by failed connection attempts

Replaced every g_warning() with g_log() This fixes display errors when using ncurses. committer: Frank Zschockelt <lm@freakysoft.de>

Refine "Allow replies to non-iq stanzas" (c8e0edc96a65cc29abc8fd6eeb9dc8a62d12f0f1) The old check was useful for iq-stanzas, because otherwise iq requests to yourself will break. Therefore replies to iq-stanzas should only be forwarded to the id-handlers if it is a result or error sub type.

lm_connection_get_keep_alive_rate was missing from loudmouth.sym

Replace g_print with g_log in asyncns-resolver: I'll probably drop the internal asyncns, if nobody complains loud enough.

Removed duplicated declarations

Added lm_connection_unregister_reply_handler()

Code looks for an auth element below the returned streams:feature element which afaics from rfc3920 does not go here. This causes the connection to just hang. committer: Frank Zschockelt <lm@freakysoft.de>

Some style fixes

asyncns resolver: fail when the server doesn't have a SRV field. Otherwise old_socket_resolver_srv_cb() does not try to connect to the domain from the jid instead.

lm_old_socket_close() leaks a file descriptor if connect() has not succeeded. socket_close_io_channel() is not called on the socket file descriptor if lm_old_socket_close() is called between when connect() has been called and when it has succeeded/failed. The patch came from the user emezeske on github.

There were two assignments in assertions, reported by Tommi Rantala at http://loudmouth.lighthouseapp.com/projects/17276/tickets/43-asyncnsc-assignment-in-assert committer: Frank Zschockelt <lm@freakysoft.de>

A patch is required for loudmouth configure script to detect gnutls (by using pkg-config instead of libgnutls-config). The patch was collected from the gentoo ebuild, which was borrowed from Mandriva.

[PATCH] Drop stanzas when failing to convert them to LmMessages. When a stanza comes in that for some reason can't be parsed into an LmMessage, just drop them on the floor instead of blocking the parser. I've seen this issue happen in practise because some (buggy?) client sending an iq with a prefix e.g. <client:iq xmlns:client="jabber:client" ... /> committer: Frank Zschockelt <lm@freakysoft.de>

Theoretically, getsockopt could fail and leave 'error' unset. While I don't think this will happen in practice, the failure can be handled by returning the value of errno. committer: Frank Zschockelt <lm@freakysoft.de>

The Digest URI in the MD5 SASL Response wasn't set correctly when using SRV records. committer: Frank Zschockelt <lm@freakysoft.de>

Fix g_log issue when configured with --enable-debug=no.

Merge commit 'origin/loudmouth-1-5'

Applied parallel make and asyncns patches from the mailing list.

Removed Dangerous Instructions

Add new loudmouth-ruby files.

Updated tarball

Update release information files. * Combined AUTHORS and CONTRIBUTORS. * Added some more folks to the AUTHORS file based on commit history. * Added Makefile recipe to dynamically generate ChangeLog from last annotated tag or a specified ref. * Added ignores for tarballs and ChangeLog.

Added 1.5.0 tarball

Created gh-pages branch

Added File Listing Known Bugs

Made Tests Optional to Allow Older glib Debian Stable had old Glib, only used for gtest, so made it optional.

Logging Adjustments

Fixed Retry Behavior, Fixed Crash on IPv6 Addresses, Log and Exit On No Host

Removed Stray, Old Assert

Fixed Resolver Memory Management Issues (by creating less severe issues)

Explicitly Request IPv4 Names

The order of enums in the lm-message.h and lm-message.c for the stream:error and stream:features message types was transposed, leading to inaccurate identification of those two types.

Gut autogen.sh and replace most of it with autoreconf.

Applied ksmith's DNS patch. * Use SRV records when available. * Fallback to A records. * Use port specified in SRV record, else default.

Merge branch 'wyhaines' Conflicts: loudmouth/lm-data-objects.c

Added a newline at the end of the file, so that gcc does not complain.

Some temporary logging additions to aid in debugging.

Added a newline at the end of the file.

Removed some whitespace insertions that were altering the structure of the generated XML.

Don't check for sync dns problems when using asyncns [#33] lm_socket_create() checks for sync DNS failure, but the check is executed even if Loudmouth is using asyncns, in which case Loudmouth crashes. committer: Mikael Hallendal <micke@imendio.com>

fixed the coding style

Don't free connection internals before it's closed. [#34] When the connection freed up internal states (in this case the handler lists) before closing itself it segfaulted when trying to unregister the SASL handlers. Reported by Julien Puydt.

Don't free connection internals before it's closed. [#34] When the connection freed up internal states (in this case the handler lists) before closing itself it segfaulted when trying to unregister the SASL handlers. Reported by Julien Puydt.

Fixed compile when GSSAPI support is enabled. [#20] GSSAPI support required this fix after applying the patch. committer: Mikael Hallendal <micke@imendio.com>

Added interface dummy code for copying

Removed unused LmTcpSocket

Removed unmaintained spec-file Realized we still have an old and unmaintained spec-file for building rpms. The file was last updated 2003 and the distributions handle this themselves anyway.

Silence chdir by catching return value

Use option NO_TICKET for OpenSSL [#28] Google Talk seems to have a problem connecting with clients using OpenSSL and TLS. According to a bug report on Twister it was tracked down to a bug in Java. Setting the NO_TICKET option on the SSL context seemed to work around the problem.

Use option NO_TICKET for OpenSSL [#28] Google Talk seems to have a problem connecting with clients using OpenSSL and TLS. According to a bug report on Twister it was tracked down to a bug in Java. Setting the NO_TICKET option on the SSL context seemed to work around the problem.

Style fixes for GSSAPI changes [#20].

Applied patch to add GSSAPI support to Loudmouth [#20] This patch adds optional suppor for SASL support through GSSAPI mechanism. Patch from Jelmer Vernooij. committer: Mikael Hallendal <micke@imendio.com>

Applied patch to check for GSSAPI during configure [#20] Patch adds support for the GSSAPI (Kerberos) SASL mechanism. committer: Mikael Hallendal <micke@imendio.com>

Removed a left over debug printout

Merge branch 'master' of git@github.com:hallski/loudmouth

fixed compile error in asyncns.c

Don't append the 'id'-tag in opening stream headers. [#30] According to RFC 3920 (section 4.4) the id attribute SHOULD NOT be added by the initiating party. [#30 responsible:Hallski state:resolved milestone:Loudmouth 1.4.4 tagged:committed] committer: Mikael Hallendal <micke@imendio.com>

Don't append the 'id'-tag in opening stream headers. [#30] According to RFC 3920 (section 4.4) the id attribute SHOULD NOT be added by the initiating party. [#30 responsible:Hallski state:resolved milestone:Loudmouth 1.4.4 tagged:committed] committer: Mikael Hallendal <micke@imendio.com>

Bumped version to 1.4.3

Fixed up the documentation for 1.4.3 [#27] Went over the documentation and made sure that the API docs are back to 100% documented. [#27 responsible:Hallski state:resolved]

Removed the check for IO condition in socket_in_event [#23] A simplified version of the fix for [#23] since socket will never be set in blocking mode here.

Always set the socket to be non-blocking. Managed to remove the code that actually made the socket blocking when I removed the code to set it depending on whether the blocking or non-blocking LmConnection API was used or not.

Only check for IO Condition before trying another read when the socket is blocking [#23] In commit 79deb183d we added a check to see if there was more data to be read before trying again. This to avoid having a blocking socket hang on that call until some data became available. However, when using OpenSSL it sometimes reads everything off the TCP socket and buffers it internally which means that the TCP socket will tell us that there is no more data available which leads to us not trying to read from the OpenSSL layer again. In this commit I added a check to see if the socket was blocking before doing this. The long term fix will be that the socket is always going to be in async mode and the synchronous APIs will be handled at a higher level. [#23 responsible:Hallski milestone:Loudmouth 1.4.3 state:resolved]

Removed lm-soup-socket.[ch] These unused files was a test to implement the new network backend using the SoupSocket from libsoup. However, I'm not inclined to add libsoup as a dependency to Loudmouth so they will not be used.

Make the new libasyncns compile on mac os x

Updated asyncns.[ch]

Added some verbose debug output

Improved the resolvers a bit. LmOldSocket currently creates a new resolver when doing a host lookup after a successful srv lookup.

Removed optional definition of _lm_sock_set_keepalive. Always define the method but make the implementation noop if TCP keep alives are not available.

Removed the async_connect_waiting and blocking states from LmConnection. As a first step of cleaning up the network code underlying LmConnection I want to make the blocking and non-blocking calls be as similar as possible. The socket should be able to connect in asynchronous manner even if we want to block on the open call since the mainloop iteration will handle the "block" for lm_connection_open_and_block.

Minor style fix

Removed unused variable in lm_connection_open_and_block

Removed unused use_srv variable in LmConnection. It was just being set but not used anymore.

Minor style fixes

Switch to use LmAuthParameters instead of passing around username and password. This cleans up the code and reduces the number of strings that need to be copied and synced.

Made type property on LmResolver settable after creation. Currently the type of the resolver is changed after doing the initial SRV lookup so the property can't be creation only.

Silence git from generated files

Added support for Dolt in the build process. This transparently reduces the build time on systems where it is supported.

Added test code for lm-data-objects

Added lm-data-objects.[ch] for parameter objects

Minor style fix

Added some TODO items and minor style fixes

Style fix in lm_connection_new

Removed unneeded initializers in lm_connection_new

More styling

More style fixes

Style fixes

Fixed root Makefile

Fixed root Makefile

Moved Makefile.decl to build sub directory and created a Makefile.am.lm All Makefile's should include the new Makefile.am.lm which in turns includes common and test Makefiles.

More cleaning up in Makefiles

More cleaning up in Makefile

Further cleanup of Makefile.am

Updated the build system to use a build directory and a common marshal Makefile. Inspired by Richards commit to Devhelp in order to clean up the build system a bit. Makefile setup for generating signal marshal are now defined in build/Makefile.am.marshal which is included in the loudmough/Makefile.am file.

Removed old .cvsignore file

Removed unused defines in lm-connection.c

Went over all of the files and made sure they all were indented the same way. This means from now on use 4 space instead of tabs everywhere for indentation.

Make the emacs header changes to tests and examples

Change the tab width to 4 steps in the emacs headers

Change the emacs format string to use space instead of tab for indentation

Allow replies to non-iq stanzas. [#22] This changes so that any stanza type will check if there is a reply handler waiting for the incoming ID. When implementing for example XEP-0184 (Message Receipts) there will be returned message stanzas that needs to be given to a reply handler. committer: Mikael Hallendal <micke@imendio.com>

Allow replies to non-iq stanzas. [#22] This changes so that any stanza type will check if there is a reply handler waiting for the incoming ID. When implementing for example XEP-0184 (Message Receipts) there will be returned message stanzas that needs to be given to a reply handler. [#22 tagged:committed responsible:Hallski milestone:none state:resolved] committer: Mikael Hallendal <micke@imendio.com>

Updated copyright years

Use SASL before old style auth if both are supported. [#19] Fixes an issue when server provides both old style and SASL authentication mechanisms. Loudmouth currently only checked if old style auth were supported and assumed that SASL weren't in those cases. [#19 tagged:committed responsible:Hallski milestone:none state:resolved] committer: Mikael Hallendal <micke@imendio.com>

Use SASL before old style auth if both are supported. [#19] Fixes an issue when server provides both old style and SASL authentication mechanisms. Loudmouth currently only checked if old style auth were supported and assumed that SASL weren't in those cases. [#19 tagged:committed responsible:Hallski milestone:none state:resolved] committer: Mikael Hallendal <micke@imendio.com>

Fixed a crash LmFeaturePing. Missed sending the LmFeaturePing object as user data to the message handler.

Cleaned up the moved XMPP Ping code from LmConnection and hooked up the LmFeaturePing object. Removed the code from LmConnection and instead use the new LmFeaturePing object.

Added the code from LmConnection to send and receive the XMPP Ping to LmFeaturePing. Copied and refactored the code to fit into LmFeaturePing instead of LmConnection.

Added properties, signals and start/stop calls to LmFeaturePing.