platform/upstream/gstreamer.git
9 years agoxvimagesink: fix pool leak
Vineeth T M [Mon, 18 May 2015 07:16:10 +0000 (16:16 +0900)]
xvimagesink: fix pool leak

During set caps when config fails, the referenced newpool
is not unref ed.

https://bugzilla.gnome.org/show_bug.cgi?id=749530

9 years agoplaybin: check the flags before set again
eunhae choi [Mon, 18 May 2015 06:45:01 +0000 (15:45 +0900)]
playbin: check the flags before set again

check the previous flags of playsink to avoid the reconfigure of playsink repeatedly

https://bugzilla.gnome.org/show_bug.cgi?id=749528

9 years agodoc: Workaround gtkdoc issue
Nicolas Dufresne [Sun, 17 May 2015 03:33:55 +0000 (23:33 -0400)]
doc: Workaround gtkdoc issue

With gtkdoc 1.22, the XML generator fails when a itemizedlist is
followed by a refsect2. Workaround the issue by wrapping the refsect2
into para.

9 years agoplayback: use the new gst_object api
Stefan Sauer [Fri, 15 May 2015 12:49:47 +0000 (14:49 +0200)]
playback: use the new gst_object api

Use gst_object_has_as_anchestor instead of the now deprecated _has_ancestor.

9 years agodocs: fix up example pipeline
Tim-Philipp Müller [Sun, 10 May 2015 10:42:21 +0000 (11:42 +0100)]
docs: fix up example pipeline

9 years agodocs: update element example pipelines
Tim-Philipp Müller [Sat, 9 May 2015 21:33:26 +0000 (22:33 +0100)]
docs: update element example pipelines

- gst-launch -> gst-launch-1.0
- use autoaudiosink and audiovideosink more often
- review pipeline examples and descriptions

9 years agovideo: Update win32 exports for new libgstvideo API
Jan Schmidt [Sun, 10 May 2015 00:51:09 +0000 (10:51 +1000)]
video: Update win32 exports for new libgstvideo API

9 years agovideoconvert: Expose some properties from the videoconverter API
Vivia Nikolaidou [Fri, 8 May 2015 12:21:16 +0000 (15:21 +0300)]
videoconvert: Expose some properties from the videoconverter API

Expose chroma resampler, alpha mode, alpha value, chroma mode, matrix mode,
gamma mode and primaries mode from the videoconverter API.

https://bugzilla.gnome.org/show_bug.cgi?id=749105

9 years agovideo-converter: Change some implicit string enums to real enums
Vivia Nikolaidou [Fri, 8 May 2015 11:57:03 +0000 (14:57 +0300)]
video-converter: Change some implicit string enums to real enums

GST_VIDEO_CONVERTER_OPT_ALPHA_MODE, GST_VIDEO_CONVERTER_OPT_CHROMA_MODE,
GST_VIDEO_CONVERTER_OPT_MATRIX_MODE, GST_VIDEO_CONVERTER_OPT_GAMMA_MODE and
GST_VIDEO_CONVERTER_OPT_PRIMARIES_MODE were G_TYPE_STRING with only a few valid
options. Changed those to real enums.

https://bugzilla.gnome.org/show_bug.cgi?id=749104

9 years agoaudiodecoder: Also negotiate with downstream if needed before handling a GAP event
Sebastian Dröge [Fri, 8 May 2015 13:06:34 +0000 (15:06 +0200)]
audiodecoder: Also negotiate with downstream if needed before handling a GAP event

9 years agovideodecoder: Also negotiate with downstream if needed before handling a GAP event
Sebastian Dröge [Fri, 8 May 2015 13:02:48 +0000 (15:02 +0200)]
videodecoder: Also negotiate with downstream if needed before handling a GAP event

9 years agovideodecoder: Try to be smarter when clipping buffers without duration/framerate...
Sebastian Dröge [Wed, 6 May 2015 10:40:48 +0000 (12:40 +0200)]
videodecoder: Try to be smarter when clipping buffers without duration/framerate to the segment

2 second frame duration is rather unlikely... but if we don't clip
away buffers that far before the segment we can cause the pipeline to
lockup. This can happen if audio is properly clipped, and thus the
audio sink does not preroll yet but the video sink prerolls because
we already outputted a buffer here... and then queues run full.

In the worst case we will clip one buffer too many here now if no
framerate is given, no buffer duration is given and the actual
framerate is less than 0.5fps.

Fixes seeking on HLS/DASH streams, when seeking into the middle of
fragments and having no framerate/buffer duration.

9 years agoxvimagesink: fix navigation event leak when early returning
Guillaume Desmottes [Mon, 4 May 2015 15:59:30 +0000 (17:59 +0200)]
xvimagesink: fix navigation event leak when early returning

Create the event *after* the early return check so it's not leaked.

https://bugzilla.gnome.org/show_bug.cgi?id=748903

9 years agoxvimagesink: fix navigation event leak when not handled
Guillaume Desmottes [Mon, 4 May 2015 16:00:18 +0000 (18:00 +0200)]
xvimagesink: fix navigation event leak when not handled

gst_navigation_message_new_event() is *not* consuming the event so we should
always drop our extra reference.

https://bugzilla.gnome.org/show_bug.cgi?id=748903

9 years agonavigation: fix structure leak if subclass doesn't implement send_event()
Guillaume Desmottes [Mon, 4 May 2015 15:58:38 +0000 (17:58 +0200)]
navigation: fix structure leak if subclass doesn't implement send_event()

The send_event() implementation is supposed to consume @structure.

https://bugzilla.gnome.org/show_bug.cgi?id=748903

9 years agostreamsynchronizer: Don't override segment.base from upstream with 0
Sebastian Dröge [Tue, 5 May 2015 13:35:46 +0000 (15:35 +0200)]
streamsynchronizer: Don't override segment.base from upstream with 0

Upstream might want to use it to properly map timestamps to running/stream
times, if we just override it with 0 synchronization will be just wrong.

For this we remove some old 0.10 code related to segment accumulation, and
remove some more code that is useless now, and accumulate the group start time
(aka segment.base offset) manually now.

https://bugzilla.gnome.org/show_bug.cgi?id=635701

9 years agortpbasedepayload: Add some debug output
Sebastian Dröge [Tue, 5 May 2015 11:14:12 +0000 (13:14 +0200)]
rtpbasedepayload: Add some debug output

9 years agovideo: add NV61 format support
Aurélien Zanelli [Thu, 19 Mar 2015 09:50:22 +0000 (10:50 +0100)]
video: add NV61 format support

https://bugzilla.gnome.org/show_bug.cgi?id=746466

9 years agodocs: add new video API to docs
Tim-Philipp Müller [Mon, 4 May 2015 19:33:23 +0000 (20:33 +0100)]
docs: add new video API to docs

9 years agovideo: check colorimetry and chroma_site equality in gst_video_info_is_equal()
Jan Schmidt [Sun, 3 May 2015 16:18:22 +0000 (02:18 +1000)]
video: check colorimetry and chroma_site equality in gst_video_info_is_equal()

Add VideoInfo accessors for colorimetry and chroma_site and use them
when checking the equality of two GstVideoInfo

9 years agovideo-color: Add gst_video_colorimetry_is_equal()
Jan Schmidt [Sun, 3 May 2015 16:10:17 +0000 (02:10 +1000)]
video-color: Add gst_video_colorimetry_is_equal()

Add a function for comparing the equality of 2 colorimetry
structures.

9 years agooggdemux: remove unused code
Young Han Lee [Fri, 10 Apr 2015 07:05:45 +0000 (16:05 +0900)]
oggdemux: remove unused code

These lines have done nothing for about 10 years.

https://bugzilla.gnome.org/show_bug.cgi?id=748820

9 years agopbutils: Use more strict profile checking for hevc
Sreerenj Balachandran [Fri, 10 Apr 2015 12:24:28 +0000 (15:24 +0300)]
pbutils: Use more strict profile checking for hevc

Use the profile_idc value to set the profile string in caps.
Don't use compatibility flags for this purpose.

https://bugzilla.gnome.org/show_bug.cgi?id=747613

9 years agovideo-converter: Remove unused macro
Ravi Kiran K N [Thu, 30 Apr 2015 09:25:14 +0000 (14:55 +0530)]
video-converter: Remove unused macro

Remove unused macro GET_TMP_LINE

https://bugzilla.gnome.org/show_bug.cgi?id=748687

9 years agogst-play: add some more key navigation mappings
Tim-Philipp Müller [Wed, 29 Apr 2015 14:44:59 +0000 (15:44 +0100)]
gst-play: add some more key navigation mappings

And don't feed multi-character key descriptors to the
event handler, it won't be what it expects.

9 years agonavigation: sprinkle some since markers and add new API to .def file
Tim-Philipp Müller [Wed, 29 Apr 2015 14:30:02 +0000 (15:30 +0100)]
navigation: sprinkle some since markers and add new API to .def file

https://bugzilla.gnome.org/show_bug.cgi?id=747245

9 years agotools: Add mouse/keyboard handling from messages
Edward Hervey [Thu, 2 Apr 2015 14:16:58 +0000 (16:16 +0200)]
tools: Add mouse/keyboard handling from messages

Allows the user to control playback with the window in focus

https://bugzilla.gnome.org/show_bug.cgi?id=747245

9 years agoxvimagesink: Post unhandled navigation events on the bus
Edward Hervey [Thu, 2 Apr 2015 14:10:32 +0000 (16:10 +0200)]
xvimagesink: Post unhandled navigation events on the bus

https://bugzilla.gnome.org/show_bug.cgi?id=747245

9 years agovideo: Add a new "event" navigation message type
Edward Hervey [Thu, 2 Apr 2015 14:09:13 +0000 (16:09 +0200)]
video: Add a new "event" navigation message type

This will be useful for elements that wish to post unhandled navigation
events on the bus to give the application a chance to do something with
it

https://bugzilla.gnome.org/show_bug.cgi?id=747245

9 years agovideo-info: expose InterlaceMode conversion to/from string
Wim Taymans [Tue, 28 Apr 2015 10:01:02 +0000 (12:01 +0200)]
video-info: expose InterlaceMode conversion to/from string

Expose the methods used to convert a GstVideoInterlaceMode to and
from a string.

9 years agoRename property enums from ARG_ to PROP_
Luis de Bethencourt [Mon, 27 Apr 2015 10:26:10 +0000 (11:26 +0100)]
Rename property enums from ARG_ to PROP_

Property enum items should be named PROP_ for consistency and readability.

9 years agovideoconvert: Keep colorimetry and chroma-site fields if passthrough
Matthieu Bouron [Mon, 27 Apr 2015 09:06:58 +0000 (11:06 +0200)]
videoconvert: Keep colorimetry and chroma-site fields if passthrough

https://bugzilla.gnome.org/show_bug.cgi?id=748141

9 years agoaudio: Change the remaining "samples" in the ::delay() vfunc docs to "frames"
Sebastian Dröge [Mon, 27 Apr 2015 08:08:17 +0000 (10:08 +0200)]
audio: Change the remaining "samples" in the ::delay() vfunc docs to "frames"

https://bugzilla.gnome.org/show_bug.cgi?id=748289

9 years agotests: tcp: remove SOCK_CLOEXEC which causes build problems on OS/X
Tim-Philipp Müller [Sun, 26 Apr 2015 19:13:01 +0000 (20:13 +0100)]
tests: tcp: remove SOCK_CLOEXEC which causes build problems on OS/X

It's not needed here.

https://bugzilla.gnome.org/show_bug.cgi?id=747692

9 years agoaudio: The delay vfunc returns the number of frames, not samples
Sebastian Dröge [Sun, 26 Apr 2015 19:08:14 +0000 (21:08 +0200)]
audio: The delay vfunc returns the number of frames, not samples

https://bugzilla.gnome.org/show_bug.cgi?id=748289

9 years agoRemove obsolete Android build cruft
Tim-Philipp Müller [Sun, 26 Apr 2015 16:49:33 +0000 (17:49 +0100)]
Remove obsolete Android build cruft

This is not needed any longer.

9 years agotests: typefindfunctions: add test for UTF-16 MSS manifest typefinding
Tim-Philipp Müller [Sun, 26 Apr 2015 13:37:56 +0000 (14:37 +0100)]
tests: typefindfunctions: add test for UTF-16 MSS manifest typefinding

9 years agotypefinding: don't read more data than needed in MSS typefinder
Tim-Philipp Müller [Sun, 26 Apr 2015 13:44:33 +0000 (14:44 +0100)]
typefinding: don't read more data than needed in MSS typefinder

9 years agotypefinding: detect MSS manifests without using g_convert()
Tim-Philipp Müller [Sun, 26 Apr 2015 13:27:30 +0000 (14:27 +0100)]
typefinding: detect MSS manifests without using g_convert()

Embedded systems often have limited charset conversion
functionality, so don't rely on g_convert() (i.e. iconv)
for UTF-16 to UTF-8 conversions, we can easily enough do
that ourselves by converting to native endianness and
then using GLib's helper functions.

9 years agoaudiovisualizer: fix the license from GPL to LGPL
Stefan Sauer [Sat, 25 Apr 2015 16:45:50 +0000 (18:45 +0200)]
audiovisualizer: fix the license from GPL to LGPL

This was a copy'n'paste buf in the initial commit done by myself.

9 years agoxmptag: fix invalid reads in GST_DEBUG statement
Guillaume Desmottes [Fri, 24 Apr 2015 12:59:21 +0000 (14:59 +0200)]
xmptag: fix invalid reads in GST_DEBUG statement

Don't try to print a string that is not NUL-terminated. This
log line does not really seem useful so let's just drop it.

https://bugzilla.gnome.org/show_bug.cgi?id=748413

9 years agoremove unused enum items PROP_LAST
Luis de Bethencourt [Fri, 24 Apr 2015 16:10:59 +0000 (17:10 +0100)]
remove unused enum items PROP_LAST

This were probably added to the enums due to cargo cult programming and are
unused. Removing them.

9 years agoaudiodecoder: Add sink and src query virtual method
Wonchul Lee [Thu, 2 Apr 2015 15:44:12 +0000 (00:44 +0900)]
audiodecoder: Add sink and src query virtual method

API: GstAudioDecoderClass::src_query()
API: GstAudioDecoderClass::sink_query()

https://bugzilla.gnome.org/show_bug.cgi?id=747293

9 years agotests: define GST_CHECK_TEST_ENVIRONMENT_BEACON
Tim-Philipp Müller [Thu, 23 Apr 2015 14:57:37 +0000 (15:57 +0100)]
tests: define GST_CHECK_TEST_ENVIRONMENT_BEACON

Make sure the test environment is set up.

https://bugzilla.gnome.org//show_bug.cgi?id=747624

9 years agoconfigure: bump automake requirement to 1.14 and autoconf to 2.69
Tim-Philipp Müller [Thu, 23 Apr 2015 14:42:41 +0000 (15:42 +0100)]
configure: bump automake requirement to 1.14 and autoconf to 2.69

This is only required for builds from git, people can still
build tarballs if they only have older autotools.

https://bugzilla.gnome.org//show_bug.cgi?id=747624

9 years agoUpdate .gitignore
Tim-Philipp Müller [Thu, 23 Apr 2015 14:14:07 +0000 (15:14 +0100)]
Update .gitignore

9 years agovideo-converter: n_lines member should be a guint not a boolean
Ravi Kiran K N [Thu, 23 Apr 2015 04:20:12 +0000 (09:50 +0530)]
video-converter: n_lines member should be a guint not a boolean

https://bugzilla.gnome.org/show_bug.cgi?id=748348

9 years agooggdemux: fix event leaks
Guillaume Desmottes [Tue, 21 Apr 2015 13:27:57 +0000 (15:27 +0200)]
oggdemux: fix event leaks

gst_event_replace() takes its own reference on the event so we should drop
ours after creating and storing an event using it.

This fix leaks which can be reproduced using the
validate.http.media_check.vorbis_theora_1_ogg scenario.

https://bugzilla.gnome.org/show_bug.cgi?id=748247

9 years agoRemove INSTALL file
Sebastian Dröge [Wed, 22 Apr 2015 08:34:09 +0000 (10:34 +0200)]
Remove INSTALL file

autotools automatically generate this, and when using different versions
for autogen.sh there will always be changes to a file tracked by git.

9 years agoRemove LICENSE_readme
Sebastian Dröge [Wed, 22 Apr 2015 08:33:58 +0000 (10:33 +0200)]
Remove LICENSE_readme

It's completely outdated and just confusing, better if people are
forced to look at the actual code in question than trusting this file.

9 years agovideo-scaler: fix YUY2 scaling some more
Wim Taymans [Tue, 21 Apr 2015 11:31:44 +0000 (13:31 +0200)]
video-scaler: fix YUY2 scaling some more

Take into account the different steps between Y and UV when calculating
the line size for vertical resampling or else we might not resample
enough pixels and leave bad lines.

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=747790

9 years agovideo-scaler: scale enough pixels in YUY2 (and friends) mode
Wim Taymans [Tue, 21 Apr 2015 11:16:29 +0000 (13:16 +0200)]
video-scaler: scale enough pixels in YUY2 (and friends) mode

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=747790

9 years agotests: rtpbasedepayload: fix crash in test when passing varargs
Hyunjun Ko [Fri, 17 Apr 2015 07:21:05 +0000 (16:21 +0900)]
tests: rtpbasedepayload: fix crash in test when passing varargs

Need to pass 64 bits where 64 bits are expected.

https://bugzilla.gnome.org/show_bug.cgi?id=748027

9 years agovideo-converter: Remove unused variables
Ravi Kiran K N [Fri, 17 Apr 2015 05:48:22 +0000 (11:18 +0530)]
video-converter: Remove unused variables

Remove unused variables n_taps, max_taps in setup_scale()

https://bugzilla.gnome.org/show_bug.cgi?id=748021

9 years agovideo: add missing part of documentation text
Thiago Santos [Thu, 16 Apr 2015 13:03:05 +0000 (10:03 -0300)]
video: add missing part of documentation text

9 years agodiscoverer: fix GstToc leak when parsing toc messages
Guillaume Desmottes [Tue, 31 Mar 2015 11:26:21 +0000 (13:26 +0200)]
discoverer: fix GstToc leak when parsing toc messages

gst_message_parse_toc() returns a reffed GstToc which is owned by the
GstDiscovererInfo. But we have to make sure we unref its previous value before
setting the new one.

https://bugzilla.gnome.org/show_bug.cgi?id=747103

9 years agowin32: Update defs for new API
Edward Hervey [Fri, 17 Apr 2015 09:45:34 +0000 (11:45 +0200)]
win32: Update defs for new API

9 years agoallocators: make GstFdAllocator non-abstract
Wim Taymans [Fri, 17 Apr 2015 07:31:40 +0000 (09:31 +0200)]
allocators: make GstFdAllocator non-abstract

Make the GstFdAllocator non-abstract because it is perfectly possible
to make memory from a generic fd. Mark the memory as simply "fd".

9 years agoaudioconvert: fix mixed usage of gint and gint32 in int matrix
Bernhard Miller [Wed, 15 Apr 2015 09:24:17 +0000 (11:24 +0200)]
audioconvert: fix mixed usage of gint and gint32 in int matrix

This is a fixup for b2db18cda2e4e7951655cb2a34108a8523b6eca9
audioconvert: avoid float calculations when mixing integer-formatted channels

The int matrix was using gint and gint32 synonymously, which can theoretically
cause problems if gint and gint32 are actually different types.

https://bugzilla.gnome.org/show_bug.cgi?id=747005

9 years agogio: fix gvfs plugin dependencies
Tim-Philipp Müller [Tue, 14 Apr 2015 11:47:07 +0000 (12:47 +0100)]
gio: fix gvfs plugin dependencies

Try harder to look for gvfs backend changes in the right
place, to make sure the plugin gets reloaded when backends
are removed or installed. We watch the gvfs mounts directory
because the files there contain absolute paths to the
backend executables, and those may not be in the usual gio
path.

https://bugzilla.gnome.org/show_bug.cgi?id=747841

9 years agoexamples: disconnect scale callback in scrubby
Luis de Bethencourt [Tue, 14 Apr 2015 14:08:09 +0000 (15:08 +0100)]
examples: disconnect scale callback in scrubby

When the position slider's button is released, disconnect the "value_changed"
callback to avoid triggering false seek callbacks.

9 years agoexamples: keep scrubby command consistent
Luis de Bethencourt [Mon, 13 Apr 2015 16:35:36 +0000 (17:35 +0100)]
examples: keep scrubby command consistent

scrubby has two options, wav and playbin. Wav takes a file location so make
the playbin option take a file location as well instead of an uri. This also
means the usage help string will be correct for the playbin option.

9 years agoexamples: no need to set intermediate states
Luis de Bethencourt [Mon, 13 Apr 2015 16:28:45 +0000 (17:28 +0100)]
examples: no need to set intermediate states

9 years agoexamples: wavparse doesn't need dynamic linking
Luis de Bethencourt [Mon, 13 Apr 2015 15:09:26 +0000 (16:09 +0100)]
examples: wavparse doesn't need dynamic linking

In scrubby, there is no need to link wavparse with the sink dynamically.
The pad is available when the element is generated.

Change video and audio sinks to the automatically detected sinks.

9 years agovideodecoder: Break instead of return if default negotiation on GAP events fails
Sebastian Dröge [Sat, 11 Apr 2015 17:51:54 +0000 (19:51 +0200)]
videodecoder: Break instead of return if default negotiation on GAP events fails

Otherwise we're going to leak the event.

9 years agoapp, videorate: fix CFLAGS and LIBADD order
Tim-Philipp Müller [Fri, 10 Apr 2015 23:03:29 +0000 (00:03 +0100)]
app, videorate: fix CFLAGS and LIBADD order

Make sure local headers are included before installed -base.

9 years agoexamples: remove reference to 0.10 in playrec
Luis de Bethencourt [Fri, 10 Apr 2015 13:30:36 +0000 (14:30 +0100)]
examples: remove reference to 0.10 in playrec

9 years agoexamples: remove deprecated function in gtk-videooverlay
Luis de Bethencourt [Fri, 10 Apr 2015 12:41:39 +0000 (13:41 +0100)]
examples: remove deprecated function in gtk-videooverlay

gtk_widget_set_double_buffered () has been deprecated since GTK 3.14.
Also, widgets are realized automatically and gtk_wiget_realize () is only
meant to be used in widget implementations.

9 years agobasetextoverlay: fix buffer leak in chain function
Guillaume Desmottes [Thu, 9 Apr 2015 15:03:11 +0000 (17:03 +0200)]
basetextoverlay: fix buffer leak in chain function

If we don't consume the buffer by passing its reference to
overlay->text_buffer then we need to unref it.

Fix a leak with validate.file.playback.fast_forward.test5_mkv
when running inside Valgrind.

https://bugzilla.gnome.org/show_bug.cgi?id=747602

9 years agoappsrc: docs grammar fixes
Ilya Konstantinov [Wed, 8 Apr 2015 15:32:29 +0000 (18:32 +0300)]
appsrc: docs grammar fixes

https://bugzilla.gnome.org/show_bug.cgi?id=747516

9 years agoexamples: add example description to giosrc-mounting
Luis de Bethencourt [Thu, 9 Apr 2015 15:49:44 +0000 (16:49 +0100)]
examples: add example description to giosrc-mounting

Also, use GST_MESSAGE_TYPE instead of accessing the GstMessage structure

9 years agoaudiobasesink: fix ring buffer leak on open failure
Vincent Penquerc'h [Thu, 9 Apr 2015 12:00:02 +0000 (13:00 +0100)]
audiobasesink: fix ring buffer leak on open failure

9 years agoaudiobasesrc: fix ring buffer leak on open failure
Vincent Penquerc'h [Thu, 9 Apr 2015 11:59:38 +0000 (12:59 +0100)]
audiobasesrc: fix ring buffer leak on open failure

9 years agoexamples: reuse variables in encoding example
Luis de Bethencourt [Thu, 9 Apr 2015 10:23:25 +0000 (11:23 +0100)]
examples: reuse variables in encoding example

9 years agoaudiodecoder: Don't post error messages while holding the stream lock
Sebastian Dröge [Thu, 9 Apr 2015 03:49:24 +0000 (20:49 -0700)]
audiodecoder: Don't post error messages while holding the stream lock

9 years agoaudiodecoder: Don't get and parse the current srcpad caps
Sebastian Dröge [Thu, 9 Apr 2015 03:48:39 +0000 (20:48 -0700)]
audiodecoder: Don't get and parse the current srcpad caps

We only get here if we don't have any srcpad caps, and we're going
to override the GstAudioInfo a few lines below anyway without ever
using it if for whatever reason we get caps here.

9 years agovideodecoder: Try to invent default caps instead of setting none at all when getting...
Sebastian Dröge [Thu, 9 Apr 2015 03:45:58 +0000 (20:45 -0700)]
videodecoder: Try to invent default caps instead of setting none at all when getting a GAP event before CAPS

Otherwise we would forward the GAP event without ever providing any caps,
which then would make decodebin expose a srcpad without any caps set. That's
confusing for applications and can lead to all kinds of interesting bugs.

Instead do the same as already is done in GstAudioDecoder, and try to invent
caps based on the sinkpad caps and the caps allowed by downstream and the
srcpad template caps.

https://bugzilla.gnome.org/show_bug.cgi?id=747190

9 years agodecodebin: Also log the pointer value of sticky events in debug output
Sebastian Dröge [Thu, 9 Apr 2015 03:44:15 +0000 (20:44 -0700)]
decodebin: Also log the pointer value of sticky events in debug output

Makes it easier to follow them in the debug logs.

9 years agoexamples: remove unused return value in addstream
Luis de Bethencourt [Wed, 8 Apr 2015 16:12:22 +0000 (17:12 +0100)]
examples: remove unused return value in addstream

Removing unused return value of pause_play_stream ().
Fixing code style to satisfy the git hook.

9 years agoexamples: avoid sprinkle running endlessly
Luis de Bethencourt [Wed, 8 Apr 2015 14:31:39 +0000 (15:31 +0100)]
examples: avoid sprinkle running endlessly

Quit sprinkle when there are no more frequencies to remove.
Also rename for readability the check for linking elements.

9 years agotests: Use AM_TESTS_ENVIRONMENT
Edward Hervey [Wed, 8 Apr 2015 14:15:43 +0000 (16:15 +0200)]
tests: Use AM_TESTS_ENVIRONMENT

Needed by the new automake test runner

9 years agortp: rtcpbuffer: fix typo in enum
Tim-Philipp Müller [Tue, 7 Apr 2015 15:43:59 +0000 (16:43 +0100)]
rtp: rtcpbuffer: fix typo in enum

and in docs. Spotted by Rob Swain.

9 years agotests: remove unused filename string from appsink-src2
Luis de Bethencourt [Tue, 7 Apr 2015 14:32:35 +0000 (15:32 +0100)]
tests: remove unused filename string from appsink-src2

9 years agotests: check file exists before running appsink-src
Luis de Bethencourt [Tue, 7 Apr 2015 14:30:30 +0000 (15:30 +0100)]
tests: check file exists before running appsink-src

9 years agotests: add missing license headers for example apps
Luis de Bethencourt [Tue, 7 Apr 2015 14:16:41 +0000 (15:16 +0100)]
tests: add missing license headers for example apps

9 years ago{audio,video}decoder: Forward SEGMENT_DONE events immediately and drain decoders
Sebastian Dröge [Tue, 7 Apr 2015 02:20:00 +0000 (19:20 -0700)]
{audio,video}decoder: Forward SEGMENT_DONE events immediately and drain decoders

Otherwise we're going to wait with draining until the next data comes, which
is a bit suboptimal and might take a long time... or maybe never happens.

9 years agotests: appsrc: clean up block_deadlock test and make it work in valgrind
Tim-Philipp Müller [Sun, 5 Apr 2015 12:53:38 +0000 (13:53 +0100)]
tests: appsrc: clean up block_deadlock test and make it work in valgrind

Remove all the bus watch and main loop code from the block_deadlock
test, it's not needed: neither pipeline will ever post an EOS or ERROR
message on the bus, and we're the only ones posting an error, from a
timeout. Might just as well just sleep for a bit and then do whatever
we want to do.

Don't gratuitiously set tcase timeout, just use whatever is the
default (or set via the environment).

Make individual pipeline runs shorter.

Check for valgrind and only do a handful iterations when running
in valgrind, not 100 (each iteration takes about 4s on a core i7).

Make videotestsrc output smaller buffers than the default resolution,
we don't care about the buffer contents here anyway.

Fixes test timeouts when run in valgrind.

9 years agotests: multisocketsink: fix flaky unit test
Tim-Philipp Müller [Sun, 5 Apr 2015 11:30:39 +0000 (12:30 +0100)]
tests: multisocketsink: fix flaky unit test

On slower systems, or under high system load (e.g. check-valgrind),
the sending_buffers_with_9_gstmemories test would sometimes fail,
because the read call only returns 32 bytes instead of the full
36 bytes expected. This is because multisocketsink might end up
doing a partial write of 32 bytes first, and then write the
missing 4 bytes later, but since we don't wait for all of data
to be written, there's a short window where our read call in the
unit test might then only receive the 32 bytes written so far,
which makes it deeply unhappy.

Instead, make sure we loop to read all bytes.

9 years agotcpserversink: don't error out if clients send us something, just ignore it
Tim-Philipp Müller [Sat, 4 Apr 2015 20:38:40 +0000 (21:38 +0100)]
tcpserversink: don't error out if clients send us something, just ignore it

We don't expect clients to send us any data, but if they do, just
ignore it. Web browsers might send us an HTTP request for example,
but some will still be happy if we just send them data without
a proper HTTP response.

There was a bug in the reading code path. We only have a small
read buffer and would provoke an EWOULDBLOCK trying to read
because we don't bail out of the loop early enough.

https://bugzilla.gnome.org/show_bug.cgi?id=743834

9 years agotests: basetime: fix timeouts when running under valgrind
Tim-Philipp Müller [Sat, 4 Apr 2015 00:23:48 +0000 (01:23 +0100)]
tests: basetime: fix timeouts when running under valgrind

This test sets a rather short timeout, increase this when
we run under valgrind. Also add a short sleep to the
fakesrc ! fakesink pipeline to avoid thrashing the CPU,
which would often not stop the main loop when it should.

Also fix wrong (0.10) return value from pad probe callback.

9 years agovideorate: downgrade left-over ERROR debug message
Tim-Philipp Müller [Fri, 3 Apr 2015 23:46:46 +0000 (00:46 +0100)]
videorate: downgrade left-over ERROR debug message

9 years agovideorate: fix a couple of memory leaks
Tim-Philipp Müller [Fri, 3 Apr 2015 23:42:52 +0000 (00:42 +0100)]
videorate: fix a couple of memory leaks

tests: videorate: fix leak in unit test

9 years agodoc: Add gst_video_encoder_get_allocator() to doc
Nicolas Dufresne [Fri, 3 Apr 2015 22:18:32 +0000 (18:18 -0400)]
doc: Add gst_video_encoder_get_allocator() to doc

9 years agotag: exiftag: don't try to convert utf-8 to latin1 if string is ASCII already
Tim-Philipp Müller [Fri, 3 Apr 2015 20:00:53 +0000 (21:00 +0100)]
tag: exiftag: don't try to convert utf-8 to latin1 if string is ASCII already

Bypass g_convert/iconv if there's nothing to convert. That way,
conversion won't fail on systems where iconv doesn't support
converting utf-8 to latin1 and there's nothing to convert.

https://bugzilla.gnome.org/show_bug.cgi?id=723252

9 years agoAutomatic update of common submodule
Tim-Philipp Müller [Fri, 3 Apr 2015 17:57:43 +0000 (18:57 +0100)]
Automatic update of common submodule

From bc76a8b to c8fb372

9 years agooggdemux: fix wrong duration on partial streams with a skeleton index
Vincent Penquerc'h [Thu, 12 Mar 2015 16:01:48 +0000 (16:01 +0000)]
oggdemux: fix wrong duration on partial streams with a skeleton index

When a stream has a skeleton index, the stream time is taken from that
index. However, when part of the stream is captured, the index is
invalid as its offsets are now wrong. To avoid this, we ignore the index
when the last offset points beyond the end of the stream (when its
byte length is known).

https://bugzilla.gnome.org/show_bug.cgi?id=744070

9 years agotextoverlay: fix disappearing text with high deltax
Vincent Penquerc'h [Wed, 18 Mar 2015 16:32:53 +0000 (16:32 +0000)]
textoverlay: fix disappearing text with high deltax

When deltax is large enough to cause the text to push past the
width of the frame, it would disappear due to a bug in setting
the layout width.

While there, fix a log printing an incorrect width to set.

https://bugzilla.gnome.org/show_bug.cgi?id=739689

9 years agooggmux: fix deadlock when not pulling a buffer from collectpads
Vincent Penquerc'h [Wed, 17 Dec 2014 12:17:09 +0000 (12:17 +0000)]
oggmux: fix deadlock when not pulling a buffer from collectpads

oggmux keeps a cached buffer per pad, and pulls buffers from
collectpads to this cached buffer for all pads before processing
the best pad. In some cases, the move from collectpads buffer
to cached buffer is delayed till next call. However, when there
is only one pad, this can't be delayed till next call as there
will be a deadlock since collectpads has no other pad to push to.

https://bugzilla.gnome.org/show_bug.cgi?id=740565

9 years agodecodebin2: fix deadlock on chain shutdown
Vincent Penquerc'h [Wed, 25 Mar 2015 15:36:38 +0000 (15:36 +0000)]
decodebin2: fix deadlock on chain shutdown

When shutting down the chain, we can get a deadlock when removing
a pad, if that chain was being busy streaming but blocked (eg, while
waiting for a queue to have free space).

https://bugzilla.gnome.org/show_bug.cgi?id=746480

9 years agoexamples: add license header to scrubby
Luis de Bethencourt [Fri, 3 Apr 2015 12:20:58 +0000 (13:20 +0100)]
examples: add license header to scrubby