platform/upstream/gstreamer.git
7 years agotracer/rusage: fix format string args
Stefan Sauer [Sat, 8 Oct 2016 11:20:58 +0000 (13:20 +0200)]
tracer/rusage: fix format string args

The format string contains a process id, but we did not provice one. This
caused us to log garbage since all args got shifted.

7 years agomessage: Fix typo in gst_message_new_progress() docs
Sebastian Dröge [Sat, 1 Oct 2016 13:47:05 +0000 (16:47 +0300)]
message: Fix typo in gst_message_new_progress() docs

test -> text

7 years agomeson: Setup pre-commit hooks when configuring
Thibault Saunier [Fri, 30 Sep 2016 12:57:57 +0000 (09:57 -0300)]
meson: Setup pre-commit hooks when configuring

7 years agomeson: update version
Tim-Philipp Müller [Fri, 30 Sep 2016 10:38:37 +0000 (11:38 +0100)]
meson: update version

7 years agoRelease 1.9.90
Sebastian Dröge [Fri, 30 Sep 2016 10:01:17 +0000 (13:01 +0300)]
Release 1.9.90

7 years agoUpdate .po files
Sebastian Dröge [Fri, 30 Sep 2016 09:08:52 +0000 (12:08 +0300)]
Update .po files

7 years agopo: Update translations
Sebastian Dröge [Fri, 30 Sep 2016 08:41:14 +0000 (11:41 +0300)]
po: Update translations

7 years agomeson: tracers: signal availability of libunwind and backtrace() to code
Tim-Philipp Müller [Tue, 27 Sep 2016 17:00:47 +0000 (18:00 +0100)]
meson: tracers: signal availability of libunwind and backtrace() to code

Not setting cdata here on purpose because of .. complications.

7 years agomeson: Build tracers
Thibault Saunier [Mon, 26 Sep 2016 21:21:19 +0000 (18:21 -0300)]
meson: Build tracers

7 years agomeson: Fix gtkdoc using new meson features
Thibault Saunier [Fri, 23 Sep 2016 23:40:39 +0000 (20:40 -0300)]
meson: Fix gtkdoc using new meson features

7 years agotests: remove unused valgrind stuff
Tim-Philipp Müller [Mon, 26 Sep 2016 11:14:14 +0000 (12:14 +0100)]
tests: remove unused valgrind stuff

Code was also checking the wrong define anyway.

7 years agotests: parse-launch: looks clean nowadays, so re-enable for valgrind
Tim-Philipp Müller [Mon, 26 Sep 2016 11:12:12 +0000 (12:12 +0100)]
tests: parse-launch: looks clean nowadays, so re-enable for valgrind

Also, the valgrind bits weren't hooked up properly anyway,
checking the wrong define.

7 years agomeson: remove incorrect and unneeded check for ptrdiff_t
Tim-Philipp Müller [Sat, 24 Sep 2016 17:22:26 +0000 (18:22 +0100)]
meson: remove incorrect and unneeded check for ptrdiff_t

Need to include stddef.h for it, so this would've worked:
if cc.has_type('ptrdiff_t', prefix: '#include <stddef.h>')

7 years agomeson: fix internal printf for %ll format modifier on 32-bit systems
Tim-Philipp Müller [Sat, 24 Sep 2016 17:06:31 +0000 (18:06 +0100)]
meson: fix internal printf for %ll format modifier on 32-bit systems

gst/gstprintf unit test would fail on 32-bit x86 with:
gstprintf.c:83:printf_I32_I64:0: 'str' (64-bit x value = b5a6978f) is not equal to '"64-bit x value = f1e2d3c4b5a6978f"'

7 years agoRevert "meson: Force gstenum_h to be built when using gst_dep"
Nirbheek Chauhan [Thu, 22 Sep 2016 22:49:47 +0000 (04:19 +0530)]
Revert "meson: Force gstenum_h to be built when using gst_dep"

This reverts commit cfc565e2d88a8e7d656b68c5c2a1b7acb08cdb7f.

The commit was redundant since gst_gen_sources already contains
gstenum_h. We're still investigating why some people are still seeing
a racy build failure.

7 years agomeson: Force gstenum_h to be built when using gst_dep
Nirbheek Chauhan [Thu, 22 Sep 2016 18:58:53 +0000 (00:28 +0530)]
meson: Force gstenum_h to be built when using gst_dep

This forces gstenumtypes.h to be built whenever something uses gst_dep
as a subproject dependency. This is needed since gst/gst.h includes
gstenumtypes.h

Closes https://github.com/mesonbuild/meson/issues/714 which is not
actually a Meson bug.

7 years agobin: Add forgotten "git commit --amend" for last commit
Sebastian Dröge [Mon, 19 Sep 2016 14:07:51 +0000 (10:07 -0400)]
bin: Add forgotten "git commit --amend" for last commit

Need to cast away the const as g_queue_foreach() takes a non-const GQueue*

7 years agobin: When copying the sort iterator, also copy its internal queue
Kouhei Sutou [Mon, 19 Sep 2016 14:04:55 +0000 (10:04 -0400)]
bin: When copying the sort iterator, also copy its internal queue

Otherwise both iterators share the same references, the second one
usually resulting in a crash when being freed.

https://bugzilla.gnome.org/show_bug.cgi?id=771649

7 years agoqueue2: Fix watermark test
Carlos Rafael Giani [Sun, 11 Sep 2016 13:28:43 +0000 (15:28 +0200)]
queue2: Fix watermark test

This carries over code for a similar test from multiqueue to ensure full
control over the dataflow while testing. (The previous attempt was racy
since the fill level changed without any thread sync with the test code.)

https://bugzilla.gnome.org/show_bug.cgi?id=771210

7 years agoqueue2: Update buffering if its enabled and low/high watermarks are changed
Carlos Rafael Giani [Sun, 11 Sep 2016 13:26:26 +0000 (15:26 +0200)]
queue2: Update buffering if its enabled and low/high watermarks are changed

https://bugzilla.gnome.org/show_bug.cgi?id=771210

7 years agomessage: Fix documentation for gst_message_new_duration()
Arun Raghavan [Thu, 15 Sep 2016 12:08:49 +0000 (17:38 +0530)]
message: Fix documentation for gst_message_new_duration()

Seems like there was some documentation left over from when this was
gst_message_new_duration().

7 years agobasesink: Remove unused fields and always use the buffer timestamp difference for...
Sebastian Dröge [Mon, 12 Sep 2016 15:41:16 +0000 (17:41 +0200)]
basesink: Remove unused fields and always use the buffer timestamp difference for calculating the QoS proportion

The buffer timestamps are only hints and more often than not have
nothing to do with reality.

https://bugzilla.gnome.org/show_bug.cgi?id=771306

7 years agogstconfig: Use __declspec when built with MinGW and linking with MSVC
Nirbheek Chauhan [Thu, 8 Sep 2016 07:28:54 +0000 (12:58 +0530)]
gstconfig: Use __declspec when built with MinGW and linking with MSVC

Earlier we were only using __declspec(dllexport/import) when we were
built with MSVC because when built with MinGW and linking with MinGW we
don't need it (and we get linker errors because of it).

However, when we're built with MinGW and someone wants to link to us
with MSVC, we still need the prototypes to have __declspec(dllimport)
since MSVC cannot do auto-import like GCC can.

https://bugzilla.gnome.org/show_bug.cgi?id=771029

7 years agobin: Also don't *unset* element flags if they're in the suppressed flags
Sebastian Dröge [Mon, 12 Sep 2016 15:07:09 +0000 (17:07 +0200)]
bin: Also don't *unset* element flags if they're in the suppressed flags

Otherwise our bin might lose various flags that were explicitly set on
it at arbitrary times.

7 years agotests: gstbin: add tests for suppressed flags
Thiago Santos [Sat, 10 Sep 2016 14:59:11 +0000 (11:59 -0300)]
tests: gstbin: add tests for suppressed flags

Some simple tests to make sure it keeps working

7 years agobin: Add setter and getter to suppress element flags
Wonchul Lee [Fri, 2 Sep 2016 08:39:17 +0000 (17:39 +0900)]
bin: Add setter and getter to suppress element flags

Suppress-flags is for preventing propagation of child element's specific
flag when it is added to the bin.

https://bugzilla.gnome.org/show_bug.cgi?id=770627

7 years agoAutomatic update of common submodule
Jan Schmidt [Sat, 10 Sep 2016 10:50:48 +0000 (20:50 +1000)]
Automatic update of common submodule

From b18d820 to f980fd9

7 years agogst: Ensure gst_value is initialised before gst_tag
Graham Leggett [Fri, 13 Nov 2015 16:00:02 +0000 (16:00 +0000)]
gst: Ensure gst_value is initialised before gst_tag

Otherwise GST_TYPE_FRACTION will work correctly in tags.

https://bugzilla.gnome.org/show_bug.cgi?id=753922

7 years agoelement: Reset the stop position to NONE in seek_simple()
Sebastian Dröge [Fri, 9 Sep 2016 08:46:11 +0000 (11:46 +0300)]
element: Reset the stop position to NONE in seek_simple()

When using seek_simple() in combination with other kinds of seeks, this
becomes problematic. seek_simple() does not reset the stop position to
GST_CLOCK_TIME_NONE but keeps whatever a previous seek did. So for example
when doing a seek_simple() after a rate=-1 seek, we would usually get
assertions that start>stop (and stop being the old stop from the rate=1 seek).

https://bugzilla.gnome.org/show_bug.cgi?id=771104

7 years agoAutomatic update of common submodule
Jan Schmidt [Fri, 9 Sep 2016 23:53:42 +0000 (09:53 +1000)]
Automatic update of common submodule

From f49c55e to b18d820

7 years agomeson:tests: Bump timeout to 3 minutes
Thibault Saunier [Fri, 9 Sep 2016 12:36:40 +0000 (09:36 -0300)]
meson:tests: Bump timeout to 3 minutes

Basically we already have each test with a 20sec timeout,
and testsuite can last more than the default 30secs from
meson. 3 minutes is another arbitrary timeout but should
be good enough.

7 years agobasesink: Use the average durations based on timestamps for the QoS proportion when...
Sebastian Dröge [Thu, 8 Sep 2016 12:19:38 +0000 (15:19 +0300)]
basesink: Use the average durations based on timestamps for the QoS proportion when doing trickmodes

The durations of the buffers are (usually) assuming that no frames are being
dropped and are just the durations coming from the stream. However if we do
trickmodes, frames are being dropped regularly especially if only key units
are supposed to be played.

Fixes completely bogus QoS proportion values in the above case.

7 years agomeson: Fix building with meson 0.34
Thibault Saunier [Mon, 5 Sep 2016 21:07:49 +0000 (18:07 -0300)]
meson: Fix building with meson 0.34

7 years agomeson: Allow others to build GIR files when using GStreamer as subproject
Thibault Saunier [Fri, 26 Aug 2016 23:06:59 +0000 (20:06 -0300)]
meson: Allow others to build GIR files when using GStreamer as subproject

And add a way to disable the introspection and bump version to 1.9.2

7 years agogstconfig.h.in: Add another version of the SH4 #define and S390x
Sebastian Dröge [Mon, 5 Sep 2016 08:11:29 +0000 (11:11 +0300)]
gstconfig.h.in: Add another version of the SH4 #define and S390x

https://bugzilla.gnome.org/show_bug.cgi?id=770731

7 years agoconfig: newer gcc doesn't know __ppc__ and __ppc64__ anymore
Wim Taymans [Mon, 5 Sep 2016 07:50:17 +0000 (09:50 +0200)]
config: newer gcc doesn't know __ppc__ and __ppc64__ anymore

__ppc__ and __ppc64__ are non-standard defines, we should use
__powerpc__ and __powerpc64__ instead because newer gcc doesn't know
them anymore.

7 years agotools: gst-inspect: add * for pointer signal arguments where needed
Tim-Philipp Müller [Sun, 4 Sep 2016 19:39:31 +0000 (20:39 +0100)]
tools: gst-inspect: add * for pointer signal arguments where needed

Print GObject argument properly with pointer marker:

  "client-added" :  void user_function (GstElement* object,
                                        GObject* arg0,
                                        gpointer user_data);

instead of

  "client-added" :  void user_function (GstElement* object,
                                        GObject arg0,
                                        gpointer user_data);

for gst-inspect-1.0 tcpserversink.

7 years agotests: don't build misc subdir if both examples and benchmarks have been disabled
Tim-Philipp Müller [Fri, 2 Sep 2016 22:22:17 +0000 (23:22 +0100)]
tests: don't build misc subdir if both examples and benchmarks have been disabled

https://bugzilla.gnome.org/show_bug.cgi?id=770740

7 years agoconfig: support System z
Wim Taymans [Thu, 1 Sep 2016 12:13:40 +0000 (14:13 +0200)]
config: support System z

7 years agoBack to development
Sebastian Dröge [Thu, 1 Sep 2016 09:25:23 +0000 (12:25 +0300)]
Back to development

7 years agoRelease 1.9.2
Sebastian Dröge [Thu, 1 Sep 2016 09:24:45 +0000 (12:24 +0300)]
Release 1.9.2

7 years agopo: Update translations
Sebastian Dröge [Thu, 1 Sep 2016 08:22:45 +0000 (11:22 +0300)]
po: Update translations

7 years agomultiqueue: Add higher-resolution low/high-watermark properties
Carlos Rafael Giani [Wed, 31 Aug 2016 07:49:03 +0000 (09:49 +0200)]
multiqueue: Add higher-resolution low/high-watermark properties

low/high-watermark are of type double, and given in range 0.0-1.0. This
makes it possible to set low/high watermarks with greater resolution,
which is useful with large multiqueue max sizes and watermarks like 0.5%.

Also adding a test to check the fill and watermark level behavior.

https://bugzilla.gnome.org/show_bug.cgi?id=770628

7 years agomultiqueue: Distinguish between buffering percentage and buffering level
Carlos Rafael Giani [Wed, 31 Aug 2016 07:48:53 +0000 (09:48 +0200)]
multiqueue: Distinguish between buffering percentage and buffering level

To make the code clearer, and to facilitate future improvements, introduce
a distinction between the buffering level and the buffering percentage.

Buffering level: the queue's current fill level. The low/high watermarks
are in this range.

Buffering percentage: percentage relative to the low/high watermarks
(0% = low watermark, 100% = high watermark).

To that end, get_percentage() is renamed to get_buffering_level(). Also,
low/high_percent are renamed to low/high_watermark to avoid confusion.
mq->buffering_percent values are now normalized in the 0..100 range for
buffering messages inside update_buffering(), and not just before sending
the buffering message. Finally the buffering level range is parameterized
by adding a new constant called MAX_BUFFERING_LEVEL.

https://bugzilla.gnome.org/show_bug.cgi?id=770628

7 years agomultiqueue: Rename percent/percent_changed to buffering_percent(_changed)
Carlos Rafael Giani [Wed, 31 Aug 2016 07:48:38 +0000 (09:48 +0200)]
multiqueue: Rename percent/percent_changed to buffering_percent(_changed)

This is a prerequisite for subsequent commits, and makes queue2 and
multiqueue code a little more consistent.

https://bugzilla.gnome.org/show_bug.cgi?id=770628

7 years agomultiqueue: Fix high_time wakeup logic
Edward Hervey [Tue, 23 Aug 2016 05:57:33 +0000 (14:57 +0900)]
multiqueue: Fix high_time wakeup logic

When calculating the high_time, cache the group value in each singlequeue.

This fixes the issue by which wake_up_next_non_linked() would use the global
high-time to decide whether to wake-up a waiting thread, instead of the group
one, resulting in those threads constantly spinning.

Tidy up a bit the waiting logic while we're at it.

With this patch, we go from 212% playing a 8 audio / 8 video file down to less
than 10% (most of it being the video decoding).

https://bugzilla.gnome.org/show_bug.cgi?id=770225

7 years agotools: gst-inspect: don't print internal pad request function name
Tim-Philipp Müller [Sun, 28 Aug 2016 15:02:14 +0000 (16:02 +0100)]
tools: gst-inspect: don't print internal pad request function name

This just confuses people, they look at it and try to call it
directly by name, instead of using the public GstElement API.
It stands to reason that it goes without saying that when an
element provides request pads that they can actually be
requested using the standard API, and there's no point in
printing internal implementation details of the element.

7 years agoMake use of the new GST_ELEMENT_FLOW_ERROR API all around.
Thibault Saunier [Tue, 23 Aug 2016 16:27:58 +0000 (13:27 -0300)]
Make use of the new GST_ELEMENT_FLOW_ERROR API all around.

https://bugzilla.gnome.org/show_bug.cgi?id=770158

7 years agoelement: Add API to more easily post messages about flowing issues
Thibault Saunier [Tue, 23 Aug 2016 16:27:20 +0000 (13:27 -0300)]
element: Add API to more easily post messages about flowing issues

In many parts of the code we raise streaming error when the flow
goes wrong, and each time we create more or less similare error
message. Also that message does not let the application know what
has actually gone wrong. In the new API we add a "flow-return" detail
field inside the GstMessage so that the application has all the information
if it needs it.

API:
    GST_ELEMENT_FLOW_ERROR

https://bugzilla.gnome.org/show_bug.cgi?id=770158

7 years agogstconfig: Decide GST_EXPORT declaration style at build time
Nirbheek Chauhan [Fri, 26 Aug 2016 13:57:22 +0000 (19:27 +0530)]
gstconfig: Decide GST_EXPORT declaration style at build time

We only use GST_EXPORT consistently when building with MSVC by using the
visual studio definitions files (win32/common/*.def), so always disable
it when building with Autotools and only enable it with Meson when
building with MSVC.

This allows you to use MinGW to link to a GStreamer built with MSVC and
get the correct function prototypes to find functions and variables in
DLLs.

7 years agodocs: fix typo in stream selection docs
Wonchul Lee [Fri, 26 Aug 2016 07:21:30 +0000 (16:21 +0900)]
docs: fix typo in stream selection docs

https://bugzilla.gnome.org//show_bug.cgi?id=770428

7 years agoelement: rename gst_element_message_new_details() to gst_make_element_message_details()
Tim-Philipp Müller [Fri, 26 Aug 2016 11:55:04 +0000 (12:55 +0100)]
element: rename gst_element_message_new_details() to gst_make_element_message_details()

Fixes g-i warning "Gst: Constructor return type mismatch
symbol='gst_element_message_new_details' constructed='Gst.Element'
return='Gst.Structure'".

This is a newly-added function in git that has not been in a stable
release yet, so it's fine to rename it. It's also only used indirectly
via macros.

7 years agodocs: fix various gtk-doc warnings
Tim-Philipp Müller [Fri, 26 Aug 2016 11:35:23 +0000 (12:35 +0100)]
docs: fix various gtk-doc warnings

e.g. "warning: multi-line since docs found"

7 years agog-i: info: allow passing NULL to gst_debug_remove_log_function()
Tim-Philipp Müller [Fri, 26 Aug 2016 11:04:33 +0000 (12:04 +0100)]
g-i: info: allow passing NULL to gst_debug_remove_log_function()

Useful for removing the default handler from bindings.

7 years agomeson: doc: Fix building documentation when using subprojects
Thibault Saunier [Thu, 25 Aug 2016 18:04:06 +0000 (15:04 -0300)]
meson: doc: Fix building documentation when using subprojects

and check the presence of gtk-doc before building the documentation

7 years agovalue: Implement can_intersect for GstFlagSet types
Jan Schmidt [Thu, 25 Aug 2016 17:17:41 +0000 (03:17 +1000)]
value: Implement can_intersect for GstFlagSet types

Make sure that gst_value_can_intersect returns TRUE
for GstFlagSet combinations that can successfully
intersect

7 years agoqueue2: Add higher-resolution low/high-watermark properties
Carlos Rafael Giani [Wed, 3 Aug 2016 13:20:20 +0000 (15:20 +0200)]
queue2: Add higher-resolution low/high-watermark properties

low/high-watermark are of type double, and given in range 0.0-1.0. This
makes it possible to set low/high watermarks with greater resolution,
which is useful with large queue2 max sizes and watermarks like 0.5%.

Also adding a test to check the fill and watermark level behavior.

https://bugzilla.gnome.org/show_bug.cgi?id=769449

7 years agoqueue2: Distinguish between buffering percentage and buffering level
Carlos Rafael Giani [Wed, 3 Aug 2016 13:27:40 +0000 (15:27 +0200)]
queue2: Distinguish between buffering percentage and buffering level

To make the code clearer, and to facilitate future improvements, introduce
a distinction between the buffering level and the buffering percentage.

Buffering level: the queue's current fill level. The low/high watermarks
are in this range.

Buffering percentage: percentage relative to the low/high watermarks
(0% = low watermark, 100% = high watermark).

To that end, get_buffering_percent() is renamed to get_buffering_level(),
and the code at the end that transforms to the buffering percentage is
factored out into a new convert_to_buffering_percent() function. Also,
the buffering level range is parameterized by adding a new constant called
MAX_BUFFERING_LEVEL.

https://bugzilla.gnome.org/show_bug.cgi?id=769449

7 years agodocs: release: add tag signing command
Tim-Philipp Müller [Tue, 23 Aug 2016 09:52:32 +0000 (10:52 +0100)]
docs: release: add tag signing command

7 years agobuffer: add explicit getters and setters for buffer flags
Arjen Veenhuizen [Thu, 7 Jul 2016 06:01:24 +0000 (08:01 +0200)]
buffer: add explicit getters and setters for buffer flags

These can be used from bindings.

https://bugzilla.gnome.org/show_bug.cgi?id=768301

7 years agomeson: install libgstcheck-1.0 and add api version to name
Tim-Philipp Müller [Sun, 21 Aug 2016 23:01:46 +0000 (00:01 +0100)]
meson: install libgstcheck-1.0 and add api version to name

7 years agostreams: update and emit notify signal only if taglist actually changed
Hoonhee Lee [Fri, 19 Aug 2016 23:54:27 +0000 (08:54 +0900)]
streams: update and emit notify signal only if taglist actually changed

https://bugzilla.gnome.org/show_bug.cgi?id=770161

7 years agoAdd support for Meson as alternative/parallel build system
Nirbheek Chauhan [Fri, 12 Aug 2016 14:55:17 +0000 (20:25 +0530)]
Add support for Meson as alternative/parallel build system

https://github.com/mesonbuild/meson

With contributions from:

Tim-Philipp Müller <tim@centricular.com>
Mathieu Duponchelle <mathieu.duponchelle@opencreed.com>
Jussi Pakkanen <jpakkane@gmail.com> (original port)

Highlights of the features provided are:
* Faster builds on Linux (~40-50% faster)
* The ability to build with MSVC on Windows
* Generate Visual Studio project files
* Generate XCode project files
* Much faster builds on Windows (on-par with Linux)
* Seriously fast configure and building on embedded

... and many more. For more details see:

http://blog.nirbheek.in/2016/05/gstreamer-and-meson-new-hope.html
http://blog.nirbheek.in/2016/07/building-and-developing-gstreamer-using.html

Building with Meson should work on both Linux and Windows, but may
need a few more tweaks on other operating systems.

7 years agoMove gstconfig.h back to normal include dir
Tim-Philipp Müller [Sat, 13 Aug 2016 12:55:15 +0000 (13:55 +0100)]
Move gstconfig.h back to normal include dir

Now that it's arch-independent again. Will need fixes in cerbero too.

7 years agogstconfig.h: Detect unaligned access support at compile-time
Nirbheek Chauhan [Tue, 21 Jun 2016 13:29:49 +0000 (18:59 +0530)]
gstconfig.h: Detect unaligned access support at compile-time

This makes gstconfig.h completely arch-independent. Should cover all
compilers that gstreamer is known to build on, and all architectures
that I could find information on. People are encouraged to file bugs if
their platform/arch is missing.

7 years agoRemove old alloc tracing code now that we have a GstTracer-based replacement
Tim-Philipp Müller [Sat, 13 Aug 2016 08:55:46 +0000 (09:55 +0100)]
Remove old alloc tracing code now that we have a GstTracer-based replacement

It's been internal API only in 1.x.

7 years agoqueue2: Post buffering messages earlier in ringbuffer mode
Edward Hervey [Fri, 12 Aug 2016 14:15:25 +0000 (16:15 +0200)]
queue2: Post buffering messages earlier in ringbuffer mode

In ringbuffer mode we need to make sure we post buffering messages *before*
blocking to wait for data to be drained.

Without this, we would end up in situations like this:
* pipeline is pre-rolling
* Downstream demuxer/decoder has pushed data to all sinks, and demuxer thread
  is blocking downstream (i.e. not pulling from upstream/queue2).
* Therefore pipeline has pre-rolled ...
* ... but queue2 hasn't filled up yet, therefore the application waits for
  the buffering 100% messages before setting the pipeline to PLAYING
* But queue2 can't post that message, since the 100% message will be posted
  *after* there is room available for that last buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=769802

7 years agomultiqueue: removed redundant call to g_thread_self
Josep Torra [Mon, 8 Aug 2016 14:42:06 +0000 (16:42 +0200)]
multiqueue: removed redundant call to g_thread_self

Remove an unneeded call to g_thread_self and minor coding style fix.

7 years agoinputselector: Handle stream-group-done
Jan Schmidt [Wed, 16 Mar 2016 07:00:15 +0000 (18:00 +1100)]
inputselector: Handle stream-group-done

Handle the new stream-group-done message to unblock pads which
are waiting for the running time to advance on that group.

https://bugzilla.gnome.org/show_bug.cgi?id=768995

7 years agoevents: Implement the stream-group-done event
Jan Schmidt [Mon, 14 Mar 2016 03:20:42 +0000 (14:20 +1100)]
events: Implement the stream-group-done event

A new event which precedes EOS in situations where we
need downstream to unblock any pads waiting on a stream
before we can send EOS. E.g, decodebin draining a chain
so it can switch pads.

https://bugzilla.gnome.org/show_bug.cgi?id=768995

7 years agomessage: Add redirect message
Carlos Rafael Giani [Mon, 25 Jul 2016 09:22:36 +0000 (11:22 +0200)]
message: Add redirect message

Redirection messages are already used in fragmented sources and in
uridecodebin, so it makes sense to introduce these as an official message
type.

https://bugzilla.gnome.org/show_bug.cgi?id=631673

7 years agoinputselector: Wake other pads when selected goes EOS
Jan Schmidt [Mon, 25 Jul 2016 09:15:15 +0000 (19:15 +1000)]
inputselector: Wake other pads when selected goes EOS

Other pads that are waiting for the stream on the selected
pad to advance before they finish waiting themselves
should be given the chance to do so when the selected pad
goes EOS. Fixes problems where input streams can end up
waiting forever if the active stream goes EOS earlier than
their own end time.

7 years agoelement: fix GST_ELEMENT_ERROR() error code expansion
Tim-Philipp Müller [Sun, 24 Jul 2016 00:35:41 +0000 (01:35 +0100)]
element: fix GST_ELEMENT_ERROR() error code expansion

In some corner cases, the error 'code' part passed to
GST_ELEMENT_ERROR() is a valid define as well, in which
case it won't survive two levels of macro expansion, but
only one. Fixes:

oss4-sink.c: In function ‘gst_oss4_sink_open’:
error: ‘GST_RESOURCE_ERROR_0x00000002’ undeclared (first use in this function)
GST_ ## domain ## _ERROR_ ## code, __txt, __dbg, __FILE__,

which is from GST_ELEMENT_ERROR(el,RESOURCE,OPEN_WRITE,..)
and OPEN_WRITE happens to be defined to 2 here.

https://bugzilla.gnome.org/show_bug.cgi?id=756806
https://bugzilla.gnome.org/show_bug.cgi?id=769117

7 years agomessage: fix some nonsensical annotations
Tim-Philipp Müller [Fri, 22 Jul 2016 16:32:33 +0000 (17:32 +0100)]
message: fix some nonsensical annotations

7 years agodocs: add GST_ELEMENT_*_WITH_DETAILS to doc list
Vincent Penquerc'h [Fri, 22 Jul 2016 14:25:09 +0000 (15:25 +0100)]
docs: add GST_ELEMENT_*_WITH_DETAILS to doc list

7 years agodocs: list new message details apis where they seem to belong
Vincent Penquerc'h [Fri, 22 Jul 2016 14:04:58 +0000 (15:04 +0100)]
docs: list new message details apis where they seem to belong

7 years agolibgstreamer.def: fix mentions of new message details api
Vincent Penquerc'h [Fri, 22 Jul 2016 13:59:56 +0000 (14:59 +0100)]
libgstreamer.def: fix mentions of new message details api

I had not updated it after the review changes

7 years agomessage: new API for additional custom data to error messages
Vincent Penquerc'h [Wed, 2 Mar 2016 11:22:23 +0000 (11:22 +0000)]
message: new API for additional custom data to error messages

https://bugzilla.gnome.org/show_bug.cgi?id=756806

7 years agotracers: leaks: update type filter later for unknown types
Guillaume Desmottes [Wed, 20 Jul 2016 10:22:10 +0000 (12:22 +0200)]
tracers: leaks: update type filter later for unknown types

This allow us to filter using an object type which is implemented
by a plugin like, say, GstGtkGLSink.

https://bugzilla.gnome.org/show_bug.cgi?id=768989

7 years agocollectpads: add g-i transfer annotations to peek/pop/read_buffer/take_buffer functions
Aurélien Zanelli [Tue, 19 Jul 2016 12:45:53 +0000 (14:45 +0200)]
collectpads: add g-i transfer annotations to peek/pop/read_buffer/take_buffer functions

https://bugzilla.gnome.org/show_bug.cgi?id=768948

7 years agotests: Use gst_clock_wait_for_sync () for net client clock
Jan Schmidt [Tue, 19 Jul 2016 13:18:24 +0000 (23:18 +1000)]
tests: Use gst_clock_wait_for_sync () for net client clock

Instead of looping, use the gst_clock_wait_for_sync() function
to give clocks up to 1 second to synchronise

7 years agomessage: fix annotation of parse_stream_{collection,streams_selected}
Guillaume Desmottes [Wed, 13 Jul 2016 13:43:21 +0000 (15:43 +0200)]
message: fix annotation of parse_stream_{collection,streams_selected}

gst_structure_id_get() returns a new reference so the returned object is
actually (transfer full).

The unit tests was already unreffing the objects.

https://bugzilla.gnome.org/show_bug.cgi?id=768776

7 years agomessage: fix annotation of parse_device_{added,removed}
Guillaume Desmottes [Wed, 13 Jul 2016 13:43:21 +0000 (15:43 +0200)]
message: fix annotation of parse_device_{added,removed}

gst_structure_id_get() returns a new reference so the returned device is
actually (transfer full).

The code using this API was already correct but the code example in
comments was not.

https://bugzilla.gnome.org/show_bug.cgi?id=768776

7 years agopad: add g-i transfer annotatation to _store_sticky_event()
Guillaume Desmottes [Thu, 14 Jul 2016 14:40:22 +0000 (16:40 +0200)]
pad: add g-i transfer annotatation to _store_sticky_event()

For clarity.

https://bugzilla.gnome.org/show_bug.cgi?id=768810

7 years agobasesrc: Fix automatic-eos=false mode if a segment.stop is given
Sebastian Dröge [Tue, 12 Jul 2016 09:32:56 +0000 (12:32 +0300)]
basesrc: Fix automatic-eos=false mode if a segment.stop is given

If segment.stop was given, and the subclass provides a size that might be
smaller than segment.stop and also smaller than the actual size, we would
already stop there.

Instead try reading up to segment.stop, the goal is to ignore the (possibly
inaccurate) size the subclass gives and finish until segment.stop or when the
subclass tells us to stop.

7 years agoAutomatic update of common submodule
Stefan Sauer [Mon, 11 Jul 2016 19:13:28 +0000 (21:13 +0200)]
Automatic update of common submodule

From f363b32 to f49c55e

7 years agobenchmarks: Fix potential stack corruption in capsnego test
Nirbheek Chauhan [Mon, 11 Jul 2016 13:15:49 +0000 (18:45 +0530)]
benchmarks: Fix potential stack corruption in capsnego test

flavour_str is a non-const pointer that will be written to if the -f
option is passed

7 years agoleaks: check return values of libunwind calls
Guillaume Desmottes [Mon, 11 Jul 2016 09:34:02 +0000 (11:34 +0200)]
leaks: check return values of libunwind calls

7 years agoqueue2: Fix average input rate calculation on small input range
Edward Hervey [Mon, 11 Jul 2016 07:58:47 +0000 (09:58 +0200)]
queue2: Fix average input rate calculation on small input range

When dealing with small-ish input data coming into queue2, such as
adaptivedemux fragments, we would never take into account the last
<200ms of data coming in.

The problem is that usually on TCP connection the download rate
gradually increases (i.e. the rate is lower at the beginning of a
download than it is later on). Combined with small download time (less
than a second) we would end up with a computed average input rate
which was sometimes up to 30-50% off from the *actual* average input
rate for that fragment.

In order to fix this, force the average input rate calculation when
we receive an EOS so that we take into account that final window
of data.

https://bugzilla.gnome.org/show_bug.cgi?id=768649

7 years agominiobject: weak_unref: display the pointer of the object if failing
Guillaume Desmottes [Fri, 8 Jul 2016 14:31:56 +0000 (16:31 +0200)]
miniobject: weak_unref: display the pointer of the object if failing

That's generally the most useful information to help debugging the
problem.

https://bugzilla.gnome.org/show_bug.cgi?id=768579

7 years agobin: properly display the type of the removed message
Guillaume Desmottes [Fri, 8 Jul 2016 14:29:38 +0000 (16:29 +0200)]
bin: properly display the type of the removed message

Makes debugging easier.

https://bugzilla.gnome.org/show_bug.cgi?id=768579

7 years agoelement: re-create threadpool after cleaning up tasks
Tim-Philipp Müller [Fri, 8 Jul 2016 16:46:06 +0000 (17:46 +0100)]
element: re-create threadpool after cleaning up tasks

We don't free this from gst_deinit() but from gst_task_cleanup_all(),
so more GStreamer API may be called. In particular makes unit tests
work again with CK_FORK=no.

https://bugzilla.gnome.org/show_bug.cgi?id=768577

7 years agoleaks: warn if object is destroyed while the tracer is disposing
Guillaume Desmottes [Fri, 8 Jul 2016 14:53:51 +0000 (16:53 +0200)]
leaks: warn if object is destroyed while the tracer is disposing

This should not happen and generally means some thread is still running.

https://bugzilla.gnome.org/show_bug.cgi?id=768578

7 years agoelement: clean up thread pool from gst_task_cleanup_all()
Guillaume Desmottes [Fri, 8 Jul 2016 14:36:01 +0000 (16:36 +0200)]
element: clean up thread pool from gst_task_cleanup_all()

This ensures that all async operations (started from gst_element_call_async())
have been completed and so there is no extra thread running.

Fix races when checking for leaks on unit tests as some of those
operations were still running when the leaks tracer was checking for
leaked objects.

https://bugzilla.gnome.org/show_bug.cgi?id=768577

7 years agoleaks tracer: use G_OS_UNIX to check for signal support
Guillaume Desmottes [Fri, 8 Jul 2016 09:15:06 +0000 (11:15 +0200)]
leaks tracer: use G_OS_UNIX to check for signal support

Checking for signal.h is not good enough as it's present in Windows.
Those signals are UNIX specific anyway.

https://bugzilla.gnome.org/show_bug.cgi?id=767857

7 years agoleaks tracer: add creation stack trace support
Guillaume Desmottes [Wed, 22 Jun 2016 14:25:16 +0000 (16:25 +0200)]
leaks tracer: add creation stack trace support

This allow us to provide the trace of leaked objects making it easier
to debug.

https://bugzilla.gnome.org/show_bug.cgi?id=767862

7 years agoleaks tracer: add checkpoint support using SIGUSR2
Guillaume Desmottes [Wed, 1 Jun 2016 09:08:39 +0000 (11:08 +0200)]
leaks tracer: add checkpoint support using SIGUSR2

https://bugzilla.gnome.org/show_bug.cgi?id=767857

7 years agoleaks tracer: log alive objects when receiving SIGUSR1
Guillaume Desmottes [Tue, 31 May 2016 14:56:26 +0000 (16:56 +0200)]
leaks tracer: log alive objects when receiving SIGUSR1

We don't want to automatically catch signals so use an env variable to
enable this feature.

https://bugzilla.gnome.org/show_bug.cgi?id=767857

7 years agofunnel: Only forward sticky events on GAP events if needed
Sebastian Dröge [Thu, 7 Jul 2016 10:15:51 +0000 (13:15 +0300)]
funnel: Only forward sticky events on GAP events if needed

That is, if the active pad changed and if forwarding of sticky events is
requested at all. We otherwise forward events too often.

7 years agopad: check query caps answered and caps not NULL
Miguel París Díaz [Tue, 5 Jul 2016 14:50:16 +0000 (16:50 +0200)]
pad: check query caps answered and caps not NULL

https://bugzilla.gnome.org/show_bug.cgi?id=768450