platform/upstream/glib.git
10 years agoconfig.h.win32.in: Drop unneeded item
Chun-wei Fan [Fri, 16 Aug 2013 02:35:19 +0000 (10:35 +0800)]
config.h.win32.in: Drop unneeded item

...We no longer have the iconv cache code around.

10 years agoUpdate config.h.win32.in
Chun-wei Fan [Fri, 16 Aug 2013 02:29:41 +0000 (10:29 +0800)]
Update config.h.win32.in

Make its entries match the items that are being checked by the autotools
builds in config.h.in.

10 years agogiochannel: Add an out annotation for g_io_channel_read_unichar()
Colin Walters [Thu, 15 Aug 2013 22:24:45 +0000 (18:24 -0400)]
giochannel: Add an out annotation for g_io_channel_read_unichar()

Since it's an output variable.

https://bugzilla.gnome.org/show_bug.cgi?id=704165

10 years agogio/tests/task: fix a race condition in test_run_in_thread()
Dan Winship [Thu, 15 Aug 2013 15:30:59 +0000 (11:30 -0400)]
gio/tests/task: fix a race condition in test_run_in_thread()

When running a task in a thread, GTask may still be internally holding
a ref on the task in that thread even after the callback is called in
the original thread (depending on thread scheduling). Fix the test to
handle that by using a weak notify that signals a GCond, and wait for
that GCond from the main thread. (And add a corresponding check to
test_return_on_cancel().)

https://bugzilla.gnome.org/show_bug.cgi?id=705152

10 years agogobject: Handle ref_count==0 in notify_by_pspec
Nick Schermer [Wed, 7 Aug 2013 19:01:00 +0000 (21:01 +0200)]
gobject: Handle ref_count==0 in notify_by_pspec

Just like g_object_notify, check for a zero ref_count in
g_object_notify_by_pspec and leave if it is 0.

This allows using functions in ->finalize() that possibly also
notify a property change on the object.  Previously,
this resulted in an error from g_object_ref.

https://bugzilla.gnome.org/show_bug.cgi?id=705570

10 years agoUpdated Thai translation.
Theppitak Karoonboonyanan [Wed, 14 Aug 2013 10:07:38 +0000 (17:07 +0700)]
Updated Thai translation.

10 years agoGSimpleActionGroup: Deprecated redundant API
Matthias Clasen [Tue, 13 Aug 2013 20:48:51 +0000 (16:48 -0400)]
GSimpleActionGroup: Deprecated redundant API

See https://bugzilla.gnome.org/show_bug.cgi?id=705600

10 years agothemedicon: correctly fallback to symbolic icons
Cosimo Cecchi [Tue, 13 Aug 2013 13:16:48 +0000 (15:16 +0200)]
themedicon: correctly fallback to symbolic icons

When an icon is requested as symbolic, our generic fallback algorithm
uses fullcolor icons when the specified icon name is not found, treating
the "-symbolic" suffix as another component of the icon name.

Change the algorithm to check beforehand if the icon is symbolic, remove
the suffix if so, and re-add it at the end for all the generated icon
names.

https://bugzilla.gnome.org/show_bug.cgi?id=680926

10 years agoUpdated Lithuanian translation
Aurimas Černius [Sun, 11 Aug 2013 18:01:50 +0000 (21:01 +0300)]
Updated Lithuanian translation

10 years agoUpdated Norwegian bokmål translation
Kjartan Maraas [Thu, 8 Aug 2013 20:14:04 +0000 (22:14 +0200)]
Updated Norwegian bokmål translation

10 years agoUpdated Spanish translation
Daniel Mustieles [Thu, 8 Aug 2013 09:14:31 +0000 (11:14 +0200)]
Updated Spanish translation

10 years agog_get_tmp_dir(): Clean up envars
Ryan Lortie [Mon, 29 Jul 2013 14:00:20 +0000 (10:00 -0400)]
g_get_tmp_dir(): Clean up envars

On UNIX, we should only ever be looking at TMPDIR.

On Windows, we should only ever look at TEMP.

Also, clean up the documentation to better describe what is actually
happening.  The previous docs may have left someone confused about why
this function returns "/var/tmp" on Solaris, even with no TMPDIR set.

https://bugzilla.gnome.org/show_bug.cgi?id=705075

10 years agoUpdated Hebrew translation.
Yaron Shahrabani [Sun, 4 Aug 2013 07:31:19 +0000 (10:31 +0300)]
Updated Hebrew translation.

10 years agogtype: Fix typo in g_type_class_add_private() error message
Emanuele Aina [Sat, 3 Aug 2013 09:51:00 +0000 (11:51 +0200)]
gtype: Fix typo in g_type_class_add_private() error message

https://bugzilla.gnome.org/show_bug.cgi?id=705398

10 years agoUpdated Russian translation
Yuri Myasoedov [Sun, 4 Aug 2013 05:44:21 +0000 (09:44 +0400)]
Updated Russian translation

10 years agoUpdated Marathi Translations
Sandeep Sheshrao Shedmake [Sun, 4 Aug 2013 03:27:37 +0000 (08:57 +0530)]
Updated Marathi Translations

10 years agoUpdated Czech translation
Marek Černocký [Sat, 3 Aug 2013 18:05:34 +0000 (20:05 +0200)]
Updated Czech translation

10 years agoUpdated Traditional Chinese translation(Hong Kong and Taiwan)
Chao-Hsiung Liao [Sat, 3 Aug 2013 13:06:46 +0000 (21:06 +0800)]
Updated Traditional Chinese translation(Hong Kong and Taiwan)

10 years agoUpdated Brazilian Portuguese translation proofread by Enrico Nicoletto
Rafael Ferreira [Sat, 3 Aug 2013 02:15:58 +0000 (23:15 -0300)]
Updated Brazilian Portuguese translation proofread by Enrico Nicoletto

10 years agoUpdated Galician translations
Fran Diéguez [Fri, 2 Aug 2013 10:16:30 +0000 (12:16 +0200)]
Updated Galician translations

10 years agoGMenuModel: Fix typo in annotation
Rico Tzschichholz [Thu, 1 Aug 2013 18:41:50 +0000 (20:41 +0200)]
GMenuModel: Fix typo in annotation

10 years agoUpdated Gujarati Translations
Sweta Kothari [Thu, 1 Aug 2013 08:16:52 +0000 (13:46 +0530)]
Updated Gujarati Translations

10 years agogdesktopappinfo: Add missing return value
Daniel Svensson [Tue, 30 Jul 2013 22:20:44 +0000 (18:20 -0400)]
gdesktopappinfo: Add missing return value

10 years agoAssamese translation updated
Nilamdyuti Goswami [Tue, 30 Jul 2013 17:10:49 +0000 (22:40 +0530)]
Assamese translation updated

10 years agogsocket: fix a cut-and-pasted error message
Dan Winship [Tue, 30 Jul 2013 13:12:48 +0000 (09:12 -0400)]
gsocket: fix a cut-and-pasted error message

10 years agopost-release version bump
Matthias Clasen [Mon, 29 Jul 2013 22:07:32 +0000 (18:07 -0400)]
post-release version bump

10 years ago2.37.5 2.37.5
Matthias Clasen [Mon, 29 Jul 2013 21:18:06 +0000 (17:18 -0400)]
2.37.5

10 years agoGMenuModel: add annotations to virtual functions
Ryan Lortie [Mon, 29 Jul 2013 21:15:19 +0000 (17:15 -0400)]
GMenuModel: add annotations to virtual functions

Patch from Ted Gould.

https://bugzilla.gnome.org/show_bug.cgi?id=704931

10 years agoFix G_ADD_PRIVATE_DYNAMIC()
Sébastien Wilmet [Mon, 29 Jul 2013 15:29:05 +0000 (17:29 +0200)]
Fix G_ADD_PRIVATE_DYNAMIC()

10 years agoAdd MSVC implementations of G_GNUC_*_IGNORE_DEPRECATIONS
Chun-wei Fan [Fri, 19 Jul 2013 11:10:41 +0000 (19:10 +0800)]
Add MSVC implementations of G_GNUC_*_IGNORE_DEPRECATIONS

As Visual Studio 2008 and later have support for the __pragma keyword,
where the compiler pragmas can be used in a macro, we can support
G_GNUC_BEGIN_IGNORE_DEPRECATIONS and G_GNUC_END_IGNORE_DEPRECATIONS
for Visual Studio 2008 and later, so many deprecation (C4996) warnings
can be suppressed when using these compilers when we use these macros
in the code.

https://bugzilla.gnome.org/show_bug.cgi?id=704543

10 years agoTrivial typo fix
Matthias Clasen [Mon, 29 Jul 2013 02:06:57 +0000 (22:06 -0400)]
Trivial typo fix

10 years agoUse GLIB_PRIVATE_CALL macro
Matthias Clasen [Sun, 28 Jul 2013 22:44:41 +0000 (18:44 -0400)]
Use GLIB_PRIVATE_CALL macro

I accidentally pushed the fix for bug 704873 before
applying this cleanup.

10 years agoUse better assertion macros
Matthias Clasen [Sun, 28 Jul 2013 22:39:51 +0000 (18:39 -0400)]
Use better assertion macros

Use g_assert_cmpint and friends in the convert test, to get
more telling messages when they fail.

10 years agoAvoid an unused variable warning
Matthias Clasen [Sun, 28 Jul 2013 22:39:42 +0000 (18:39 -0400)]
Avoid an unused variable warning

10 years agoDrop iconv caching code
Colin Walters [Sat, 27 Jul 2013 23:21:00 +0000 (00:21 +0100)]
Drop iconv caching code

This was introduced for Solaris performance theoretically;
we have never been able to use it on Linux/glibc because
the UTF-16 BOM state isn't reset.

We have no data about Solaris performance; were some to
still exist, we could reintroduce the code with an explicit
check for Solaris, not a check for glibc.

https://bugzilla.gnome.org/show_bug.cgi?id=704999

10 years agoinotify: don't assume mainloop is running
Ryan Lortie [Thu, 25 Jul 2013 14:28:42 +0000 (10:28 -0400)]
inotify: don't assume mainloop is running

GFileMonitor takes great care to sample the thread-default main context
at the time that it is created in order that events can be dispatched to
the correct thread when they come in.

The inotify GFileMonitor implementation uses a global file descriptor
shared between all watches.  It has to poll this file descriptor from
somewhere so it arbitrarily picks the default main context.

The problem with that is that the user might not be running it.

Let's use the GLib worker thread for this instead.  It's guaranteed to
be running if you need it, and this is exactly the sort of problem it
was meant to solve.

https://bugzilla.gnome.org/show_bug.cgi?id=704873

10 years agoFix build with statvfs
Matthias Clasen [Sun, 28 Jul 2013 21:04:56 +0000 (17:04 -0400)]
Fix build with statvfs

As pointed out in bug 704587, the maze of ifdefs was missing the
variable declaration for statfs_result in one case.

https://bugzilla.gnome.org/show_bug.cgi?id=704587

10 years agoFix a typo
Rico Tzschichholz [Sat, 27 Jul 2013 07:02:58 +0000 (09:02 +0200)]
Fix a typo

Introduced by 3b01cbe8a8a9ee2d1169f072624069877b91097c

10 years agoAdd macros and symbols for private data on dynamic types
Emmanuele Bassi [Fri, 26 Jul 2013 15:45:48 +0000 (16:45 +0100)]
Add macros and symbols for private data on dynamic types

We need a TypeName_private_offset variable defined by the macros used to
register dynamic types. We also need to call the adjust_private_offset()
function inside class_init(). G_ADD_PRIVATE_DYNAMIC only sets the size
of the private data structure, and relies on the behaviour of the
g_type_class_adjuset_private_offset() function to register the private
data structure at class init time if passed a value greater than zero.

This allows using G_PRIVATE_OFFSET with dynamic types.

10 years agoconfigure: Use AC_USE_SYSTEM_EXTENSIONS
Colin Walters [Fri, 26 Jul 2013 11:39:51 +0000 (12:39 +0100)]
configure: Use AC_USE_SYSTEM_EXTENSIONS

...instead of detecting glibc and using _GNU_SOURCE manually.  This
should fix the build when using glibc-emulating libraries; we can
defer portability work to autoconf.

https://bugzilla.gnome.org/show_bug.cgi?id=684123

10 years agoDocument restrictions on private names
Matthias Clasen [Fri, 26 Jul 2013 18:40:10 +0000 (14:40 -0400)]
Document restrictions on private names

With the new machinery, private structs must be named
TypeNamePrivate if TypeName is the name of the instance
struct.

10 years agoupdated Gujarati translations
Sweta Kothari [Fri, 26 Jul 2013 09:20:22 +0000 (14:50 +0530)]
updated Gujarati translations

10 years agoAssamese translation updated
Nilamdyuti Goswami [Thu, 25 Jul 2013 07:48:58 +0000 (13:18 +0530)]
Assamese translation updated

10 years agomainloop: redo child source tests to not use timeouts
Dan Winship [Wed, 24 Jul 2013 14:02:38 +0000 (10:02 -0400)]
mainloop: redo child source tests to not use timeouts

The timeout-based tests could fail on slow or heavily-loaded machines.
Change them to use a counter-based "timeout" source rather than a
time-based one, so they no longer depend on wall time.

https://bugzilla.gnome.org/show_bug.cgi?id=700460

10 years agognetworkaddress: tweak to make Solaris happy
Dan Winship [Wed, 24 Jul 2013 14:45:36 +0000 (10:45 -0400)]
gnetworkaddress: tweak to make Solaris happy

https://bugzilla.gnome.org/show_bug.cgi?id=704704

10 years agoTamil Translation Updated
Shantha kumar [Wed, 24 Jul 2013 09:27:30 +0000 (14:57 +0530)]
Tamil Translation Updated

10 years agotests/unix: Don't error immediately on timeout
Colin Walters [Mon, 22 Jul 2013 18:31:35 +0000 (19:31 +0100)]
tests/unix: Don't error immediately on timeout

On a heavily loaded system, it's possible that both our normal
condition *and* the timeout occurred.  In that case we can just ignore
the timeout.

While I did add a "sig_timeout" boolean, we don't need to add any
assertions around whether or not it was reached - the assertions
covering the non-timeout case are sufficient.  The sig_timeout boolean
is mainly for later debugging.

https://bugzilla.gnome.org/show_bug.cgi?id=700460

10 years agogmain: Reset signal handlers to default when source is destroyed
Colin Walters [Mon, 22 Jul 2013 18:06:31 +0000 (19:06 +0100)]
gmain: Reset signal handlers to default when source is destroyed

If someone creates a unix signal source for e.g. SIGINT, and then
removes it, reset the handler to SIG_DFL.

Not doing this was the source of race conditions in the
glib/tests/unix test, but this will also just make us a "good citizen"
by cleaning up.

For example, if a project temporarily creates a handler for SIGTERM,
and then later removes it, they almost certainly want SIGTERM to
revert to the default of terminating the process, rather than doing
nothing.

https://bugzilla.gnome.org/show_bug.cgi?id=704699

10 years agoGRegex: add g_regex_get_max_lookbehind()
Sébastien Wilmet [Mon, 15 Jul 2013 11:52:14 +0000 (13:52 +0200)]
GRegex: add g_regex_get_max_lookbehind()

It is useful for multi-segment regex matching.

A unit test is included.

https://bugzilla.gnome.org/show_bug.cgi?id=689794

10 years agotests/closure: Send SIGUSR1 only to our pid
Colin Walters [Mon, 22 Jul 2013 14:11:13 +0000 (15:11 +0100)]
tests/closure: Send SIGUSR1 only to our pid

Otherwise in e.g. the gnome-ostree integrationtest system, we
end up sending SIGUSR1 to the *entire session*, which triggers
various badness in untested debugging paths from gnome-session.

10 years agotests/closure: Use correct prototype for signal callback
Colin Walters [Mon, 22 Jul 2013 11:51:00 +0000 (12:51 +0100)]
tests/closure: Use correct prototype for signal callback

This test worked when compiled without optimization, but fails with
-O2.  Presumably we just happened to find the GMainLoop off the stack
somewhere.

https://bugzilla.gnome.org/show_bug.cgi?id=704267

10 years agoUpdated Spanish translation
Daniel Mustieles [Mon, 22 Jul 2013 11:48:58 +0000 (13:48 +0200)]
Updated Spanish translation

10 years agog_thread_create_full() can dereference NULL pointer
Milan Crha [Mon, 22 Jul 2013 07:54:33 +0000 (09:54 +0200)]
g_thread_create_full() can dereference NULL pointer

In case of the thread creation failure, and the thread was not created
as joinable, the g_thread_create_full() could dereference a NULL pointer.

https://bugzilla.gnome.org/show_bug.cgi?id=704523

10 years agogsourceclosure.c: Fix build on Windows
Chun-wei Fan [Thu, 18 Jul 2013 09:22:53 +0000 (17:22 +0800)]
gsourceclosure.c: Fix build on Windows

GPid is a HANDLE (aka void *) on Windows, not an int, so treat pid
accordingly on Windows, as using pid as a gulong directly would likely be
undesirable on Windows

https://bugzilla.gnome.org/show_bug.cgi?id=704447

10 years agogobject/tests/closure: fix on win32
Dan Winship [Thu, 18 Jul 2013 12:11:35 +0000 (08:11 -0400)]
gobject/tests/closure: fix on win32

(The g_closure_unref() was wrong, but was not causing errors on linux
for some reason.)

https://bugzilla.gnome.org/show_bug.cgi?id=704447

10 years agotests/closure: Fix invalid unref
Colin Walters [Sun, 21 Jul 2013 18:42:34 +0000 (19:42 +0100)]
tests/closure: Fix invalid unref

The closure is floating, g_source_set_closure() will claim the
ref, so we do not need to unref it.

10 years agoUpdated Czech translation
Marek Černocký [Sun, 21 Jul 2013 09:00:38 +0000 (11:00 +0200)]
Updated Czech translation

10 years agoGRegex: update documentation for partial matching
Sébastien Wilmet [Fri, 12 Jul 2013 14:44:13 +0000 (16:44 +0200)]
GRegex: update documentation for partial matching

The restrictions on partial matching no longer apply with PCRE >= 8.00.

The pcrepartial manpage contains the "FORMERLY RESTRICTED PATTERNS"
section:

"For  releases  of PCRE prior to 8.00, because of the way certain
internal optimizations were implemented in the pcre_exec() function, the
PCRE_PARTIAL option (predecessor of PCRE_PARTIAL_SOFT) could not be used
with all patterns. From release 8.00 onwards, the restrictions no
longer  apply,  and  partial  matching  with  can  be requested for any
pattern."

https://bugzilla.gnome.org/show_bug.cgi?id=704250

10 years agoImprove documentation
Sébastien Wilmet [Fri, 5 Jul 2013 11:23:11 +0000 (13:23 +0200)]
Improve documentation

- Mention G_SOURCE_CONTINUE and G_SOURCE_REMOVE in the GSourceFunc doc;
- Mention G_PARAM_READWRITE and G_PARAM_STATIC_STRINGS in the
  GParamFlags doc;
- Fix "Since:" version for G_DEFINE_ABSTRACT_TYPE_WITH_PRIVATE;
- Fix typo.

https://bugzilla.gnome.org/show_bug.cgi?id=704250

10 years agogsettings tool: report failure to write
Matthias Clasen [Fri, 19 Jul 2013 23:41:24 +0000 (19:41 -0400)]
gsettings tool: report failure to write

If a key is locked down, we should report an error if we
fail to write it.

https://bugzilla.gnome.org/show_bug.cgi?id=704424

10 years agotests/gvariant: Canonicalize to lower case to work on Illumos (OpenSolaris)
Igor Pashev [Fri, 19 Jul 2013 22:23:29 +0000 (18:23 -0400)]
tests/gvariant: Canonicalize to lower case to work on Illumos (OpenSolaris)

Solaris libc gives lowercase strings.

https://bugzilla.gnome.org/show_bug.cgi?id=704585

10 years agogdbusnameowning: Don't spew an error if we're releasing a name due to closing
Colin Walters [Fri, 19 Jul 2013 15:50:35 +0000 (11:50 -0400)]
gdbusnameowning: Don't spew an error if we're releasing a name due to closing

The default GNOME 3.10 login process right now has gdm spawn a session
for the login screen, retaining the X server, but closing the session
bus.  Right now in this scenario many GNOME components such as
gnome-settings-daemon attempt to "clean up" on shutdown by releasing
their owned names.

But they're shutting down because the session bus went away, so
releasing the name is pointless, and presently spews an error into the
journal.

This patch avoids that error spew, which helps system administrators
find *real* problems.

https://bugzilla.gnome.org/show_bug.cgi?id=704567

10 years agoglib-unix: fix handling of multiple signal source for the same signal
Giovanni Campagna [Tue, 16 Jul 2013 13:26:02 +0000 (15:26 +0200)]
glib-unix: fix handling of multiple signal source for the same signal

We can't reset the pending flag for a signal until we've traversed
the whole list, as the documentation clearly says that in case multiple
sources they all get invoked.
This is still racy if you get a signal after checking the flag
but before resetting it, but it was the same before. The correct
fix would be to use sigwait() or signalfd(), but that would mean
blocking all signals in all threads, which is not compatible
with existing applications.

https://bugzilla.gnome.org/show_bug.cgi?id=704322

10 years agogsourceclosure: fix idle/timeout/signal closures, add child watch support
Dan Winship [Wed, 17 Jul 2013 22:58:18 +0000 (18:58 -0400)]
gsourceclosure: fix idle/timeout/signal closures, add child watch support

And add a test for all source types.

https://bugzilla.gnome.org/show_bug.cgi?id=704267

10 years agoVisual Studio Builds: Update property sheets
Chun-wei Fan [Wed, 17 Jul 2013 08:54:35 +0000 (16:54 +0800)]
Visual Studio Builds: Update property sheets

"Install" the newly-introduced gpropertyaction.h header file, which is
required for GIO.

10 years agoVisual Studio Builds: "Install" gdbus-codegen
Chun-wei Fan [Wed, 17 Jul 2013 08:45:50 +0000 (16:45 +0800)]
Visual Studio Builds: "Install" gdbus-codegen

10 years agogsettings tool: Improve documentation
Matthias Clasen [Tue, 16 Jul 2013 23:21:31 +0000 (19:21 -0400)]
gsettings tool: Improve documentation

Mention that gsettings needs a session bus to make changes.

10 years agoGApplication: reply to ActivateAction
Ryan Lortie [Sun, 14 Jul 2013 21:11:59 +0000 (17:11 -0400)]
GApplication: reply to ActivateAction

Send a reply for org.freedesktop.Application.ActivateAction calls.  We
missed this before because nobody was waiting for it.

10 years agoMake a test desktop file a little less invalid
Matthias Clasen [Sun, 14 Jul 2013 15:35:56 +0000 (11:35 -0400)]
Make a test desktop file a little less invalid

Add a missing Name entry, and add a terminal ; to the Actions
entry in org.gtk.test.dbusappinfo.desktop. desktop-file-validate
still contains about the DBusActivatable entry and about the
missing Exec entries. The former will go away when desktop-file-validate
gets updated for the latest spec revision.

10 years agotests/mainloop: Fix assertions to really avoid race conditions
Colin Walters [Thu, 11 Jul 2013 15:07:17 +0000 (11:07 -0400)]
tests/mainloop: Fix assertions to really avoid race conditions

As the comment says, we may be delayed an arbitrary amount of time on
non-idle systems; update the assertions to reflect this.

This should fix periodic failures in the gnome-ostree continuous
integration system.

https://bugzilla.gnome.org/700460

10 years agoUpdated German translation
Christian Kirbach [Sun, 14 Jul 2013 10:24:07 +0000 (12:24 +0200)]
Updated German translation

10 years agogcancellable: update GCancellableSource, fix a race condition
Dan Winship [Mon, 3 Jun 2013 11:13:50 +0000 (08:13 -0300)]
gcancellable: update GCancellableSource, fix a race condition

Update GCancellableSource to call g_source_set_ready_time() when its
cancellable is cancelled, rather than manually checking the state of
the cancellable from prepare() and check().

This means that we now need to use g_cancellable_connect() rather than
g_signal_connect() at construction time, to avoid the connect/cancel
race condition. Likewise, use g_cancellable_disconnect() to avoid the
disconnect/cancel race condition when freeing the source. (In fact,
that was necessary in the earlier code as well, and might have
occasionally caused spurious criticals or worse.)

https://bugzilla.gnome.org/show_bug.cgi?id=701511

10 years agoRemove some unnecessary source prepare/check functions
Dan Winship [Mon, 3 Jun 2013 11:10:31 +0000 (08:10 -0300)]
Remove some unnecessary source prepare/check functions

GPollableSource and GSocket's "broken" source never trigger on their
own, so with the changes to GSources in the last cycle, their check
and prepare functions are unnecessary (and undesired).

https://bugzilla.gnome.org/show_bug.cgi?id=701511

10 years agogio: port unix streams to GUnixFDSource
Dan Winship [Sat, 1 Jun 2013 18:24:13 +0000 (15:24 -0300)]
gio: port unix streams to GUnixFDSource

The unix input/output streams were using a gio-only source type that
was mostly identical to GUnixFDSource. Get rid of that source type
and just use GUnixFDSource instead.

https://bugzilla.gnome.org/show_bug.cgi?id=701511

10 years agogsourceclosure: Add support for GUnixSignalWatchSource and GUnixFDSource
Dan Winship [Sat, 1 Jun 2013 18:23:15 +0000 (15:23 -0300)]
gsourceclosure: Add support for GUnixSignalWatchSource and GUnixFDSource

https://bugzilla.gnome.org/show_bug.cgi?id=701511

10 years agogsourceclosure: use g_cclosure_marshal_generic
Dan Winship [Sun, 2 Jun 2013 22:59:23 +0000 (19:59 -0300)]
gsourceclosure: use g_cclosure_marshal_generic

For the glib-defined source types, and any source type that defines a
closure callback but not a closure marshal, use
g_cclosure_marshal_generic. And then remove all the other remaining
source closure marshals.

https://bugzilla.gnome.org/show_bug.cgi?id=701511

10 years agoFix g_return_if_fail vs. return_val_if_fail
Ryan Lortie [Sat, 13 Jul 2013 19:48:39 +0000 (15:48 -0400)]
Fix g_return_if_fail vs. return_val_if_fail

10 years agoGAction: add function for printing detailed names
Ryan Lortie [Thu, 4 Jul 2013 20:24:38 +0000 (16:24 -0400)]
GAction: add function for printing detailed names

A counterpart for parsing of detailed actions into (name, target) pairs,
this new function prints them back.

We also add a new function to check for validity of action names.  Only
valid action names are allowed when printing.  Parsing accepts _some_
invalid names for backwards compatibility.

https://bugzilla.gnome.org/show_bug.cgi?id=704157

10 years agognetworkmonitornetlink: handle default route via device
Dan Winship [Tue, 4 Jun 2013 20:29:55 +0000 (17:29 -0300)]
gnetworkmonitornetlink: handle default route via device

If the default route is via a device rather than a particular IP
address, then neither RTA_DST nor RTA_GATEWAY will be present in the
RTM_NEWROUTE message, and so GNetworkMonitorNetlink would ignore it,
and then think there was no default route. (This could happen with
certain kinds of VPNs, if they were set to route all traffic through
the VPN.)

Fix this by recognizing routes that specify RTA_OIF ("output
interface") instead of RTA_GATEWAY.

https://bugzilla.gnome.org/show_bug.cgi?id=701609

10 years agoFix a typo in a filename in Makefile.am
Ryan Lortie [Thu, 11 Jul 2013 19:57:05 +0000 (15:57 -0400)]
Fix a typo in a filename in Makefile.am

10 years agoAdd a testcase for DBusActivatable=true
Ryan Lortie [Sat, 8 Jun 2013 21:31:09 +0000 (17:31 -0400)]
Add a testcase for DBusActivatable=true

Add a fairly realistic testcase that ensures that GDesktopAppInfo with
DBusActivatable=true can successfully talk to GApplication for a variety
of purposes.

https://bugzilla.gnome.org/show_bug.cgi?id=699259

10 years agoUpdated POTFILES.in
Piotr Drąg [Thu, 11 Jul 2013 19:09:15 +0000 (21:09 +0200)]
Updated POTFILES.in

10 years agoDeal with new automake test stuff
Ryan Lortie [Thu, 11 Jul 2013 17:54:47 +0000 (13:54 -0400)]
Deal with new automake test stuff

Add *.trs and *.log to .gitignore along with test-driver

Add *.trs and *.log to CLEANFILES in glib.mk.

10 years agoAdd missing file
Ryan Lortie [Thu, 11 Jul 2013 17:47:45 +0000 (13:47 -0400)]
Add missing file

This got removed during the OnlyShowIn shuffle.

10 years agoGDesktopAppInfo: remove note about OnlyShowIn
Ryan Lortie [Thu, 11 Jul 2013 17:04:23 +0000 (13:04 -0400)]
GDesktopAppInfo: remove note about OnlyShowIn

A previous version of the patch had OnlyShowIn support for desktop file
actions.  This was removed from the spec and the patch rewritten, but
this bit of documentation slipped through.  Remove it.

10 years agoImplement the Desktop Action specification
Ryan Lortie [Thu, 11 Jul 2013 16:46:59 +0000 (12:46 -0400)]
Implement the Desktop Action specification

For some time, the desktop file specification has supported "additional
application actions".  This is intended to allow for additional methods
of starting an app, such as a mail client having a "Compose New Message"
action that brings up the compose window instead of the folder list.

This patch adds support for this with a relatively minimal API.

In the case that the application is a GApplication and DBusActivatable,
desktop actions are translated into GActions that have been added to the
application with g_action_map_add_action().  This more or less closes
the loop on being able to activate an application with an action
invocation (instead of 'activate').

https://bugzilla.gnome.org/show_bug.cgi?id=664444

10 years agoadd GPropertyAction
Ryan Lortie [Fri, 28 Jun 2013 15:41:43 +0000 (11:41 -0400)]
add GPropertyAction

Add a new type of GAction that represents the value of a property on an
object.  As an example, this might be used on the "visible-child-name"
property of a GtkStack.

https://bugzilla.gnome.org/show_bug.cgi?id=703270

10 years agoGVariant: fix two small/theoretical leaks
Ryan Lortie [Thu, 11 Jul 2013 16:26:33 +0000 (12:26 -0400)]
GVariant: fix two small/theoretical leaks

Fix some leaks that turned up while valgrinding the GVariant testcase.

These leaks are small and only occur when there is already an error in
the program: they are leaks of temp strings used when formatting a
critical message.

These show up as leaks again the testcase under the new "expect
messages" approach.  Previously, we fork()ed and these caused the
subprocess to abort, which is why this was not noticed before.

10 years agoAdd g_variant_new_printf
Ryan Lortie [Thu, 11 Jul 2013 16:16:44 +0000 (12:16 -0400)]
Add g_variant_new_printf

Add a quick way to get a floating GVariant from printf formatting.

10 years agobump version
Matthias Clasen [Tue, 9 Jul 2013 18:28:19 +0000 (14:28 -0400)]
bump version

10 years ago2.37.4 2.37.4
Matthias Clasen [Tue, 9 Jul 2013 16:54:00 +0000 (12:54 -0400)]
2.37.4

10 years agoAdd G_TYPE_CHECKSUM to the docs
Matthias Clasen [Tue, 9 Jul 2013 17:30:49 +0000 (13:30 -0400)]
Add G_TYPE_CHECKSUM to the docs

10 years agoDocument g_simple_proxy_resolver_set_ignore_hosts
Matthias Clasen [Tue, 9 Jul 2013 17:28:07 +0000 (13:28 -0400)]
Document g_simple_proxy_resolver_set_ignore_hosts

10 years agoMinor doc tweak
Matthias Clasen [Tue, 9 Jul 2013 17:23:12 +0000 (13:23 -0400)]
Minor doc tweak

10 years ago[l10n] Update Catalan translation
Gil Forcada [Fri, 5 Jul 2013 00:23:48 +0000 (02:23 +0200)]
[l10n] Update Catalan translation

10 years agogthread: Use pthread_cond_timedwait_monotonic() if available
Sebastian Dröge [Thu, 4 Jul 2013 08:27:02 +0000 (10:27 +0200)]
gthread: Use pthread_cond_timedwait_monotonic() if available

Otherwise we have to rely on pthread_cond_timedwait() actually using
the monotonic clock, which might be true or not. On Android at least
it is using the realtime clock, no pthread_condattr_setclock() is available
but instead pthread_cond_timedwait_monotonic() can be used.

10 years agoEnsure that MAX_ALLOWED keeps working with the type macros
Emmanuele Bassi [Thu, 27 Jun 2013 14:15:40 +0000 (15:15 +0100)]
Ensure that MAX_ALLOWED keeps working with the type macros

When using the GLIB_VERSION_MAX_ALLOWED macro to define the upper bound
of allowed API the G_DEFINE_TYPE_EXTENDED starts warning about the newly
added g_type_class_adjust_private_offset() function; this effectively
introduces a run-time dependency on GLib 2.38 even if we don't use its
API explicitly.

https://bugzilla.gnome.org/show_bug.cgi?id=703191

10 years agoAdd missing G_BEGIN/END_DECLS to gsettingsschema.h.
Martin Ejdestig [Tue, 2 Jul 2013 17:53:04 +0000 (19:53 +0200)]
Add missing G_BEGIN/END_DECLS to gsettingsschema.h.

Without this fix, using any of the enclosed functions when building with a
C++ compiler will result in undefined references.

https://bugzilla.gnome.org/show_bug.cgi?id=703478

10 years agoGDBusConnection: be more careful with async GetAll
Ryan Lortie [Mon, 1 Jul 2013 21:42:43 +0000 (17:42 -0400)]
GDBusConnection: be more careful with async GetAll

It's possible to get a org.freedesktop.Properties.GetAll call even if we
have no readable properties in the introspection, in which case we
should return the empty list in the usual way.

We should certainly _not_ be dispatching to the method call handler of
an interface which has no properties (since it will not be expecting
this).

Add a check to make sure that there is at least one readable property
before assuming that a NULL get_property handler implies that we want to
handle properties asynchronously.

Add a testcase that was failing before the change and works after it.

https://bugzilla.gnome.org/show_bug.cgi?id=703437

10 years agoRevert two errant commits to GMutex, etc.
Ryan Lortie [Tue, 2 Jul 2013 03:30:31 +0000 (23:30 -0400)]
Revert two errant commits to GMutex, etc.

This reverts commits dfbac178bdcbcb042be928dcbeaa87742b511836 and
56348210f38209a11902e54b2db41833aab36414.

These two commits introduce undesirable behaviour and were made with no
apparent approval from anybody at all, and without reference to a bug or
mailing list discussion.