platform/upstream/gst-plugins-good.git
6 years agoMerge the tizen patch and fix build err based on 1.12.2 30/158530/1
Eunhae Choi [Wed, 1 Nov 2017 12:44:43 +0000 (21:44 +0900)]
Merge the tizen patch and fix build err based on 1.12.2

Change-Id: Ica5cd6b1a865b8367584aeb40e33ea6f3e4017e8

6 years agoExtract spherical video and spatial audio metadata and send it to the bus 78/152978/7 accepted/tizen/unified/20171031.055723 submit/tizen/20171030.055657
Mykola Alieksieiev [Wed, 27 Sep 2017 11:51:28 +0000 (14:51 +0300)]
Extract spherical video and spatial audio metadata and send it to the bus

Change-Id: I0126c6afa12e5843587030fd5ec236f6ba2c507b
Signed-off-by: Mykola Alieksieiev <m.alieksieie@samsung.com>
6 years agoqtdemux: fix memory leak 48/143448/1 accepted/tizen/unified/20170920.065406 submit/tizen/20170918.052251
Gilbok Lee [Thu, 10 Aug 2017 04:06:39 +0000 (13:06 +0900)]
qtdemux: fix memory leak

Change-Id: Ic27af3fcd40885695ea041160af7b56258461ef2
Signed-off-by: Gilbok Lee <gilbok.lee@samsung.com>
6 years agoRelease 1.12.2 upstream/1.12 1.12.2
Sebastian Dröge [Fri, 14 Jul 2017 11:03:05 +0000 (14:03 +0300)]
Release 1.12.2

6 years agoUpdate .po files
Sebastian Dröge [Fri, 14 Jul 2017 10:31:58 +0000 (13:31 +0300)]
Update .po files

6 years agopo: Update translations
Sebastian Dröge [Fri, 14 Jul 2017 10:22:45 +0000 (13:22 +0300)]
po: Update translations

6 years agoqtdemux: Fix parsing of RLE depth
Sebastian Dröge [Thu, 13 Jul 2017 09:47:02 +0000 (12:47 +0300)]
qtdemux: Fix parsing of RLE depth

Regression introduced by 86b427dc70562f891a551ffc9f96cefe1cafcddd

https://bugzilla.gnome.org/show_bug.cgi?id=784812

6 years agoosxaudio: fixes playback of mono streams with no channel-mask field in caps
Josep Torra [Sat, 20 May 2017 15:09:52 +0000 (17:09 +0200)]
osxaudio: fixes playback of mono streams with no channel-mask field in caps

Fixes a negotiation error seen when trying to playback of a .MOV file with
a mono AAC audio stream decoded by avcdec_aac that doesn't set channel-mask
field but sink was requiring channel-mask=0x3.

6 years agortpgsmpay: fix accidental garbage data before actual payload
Yasushi SHOJI [Fri, 7 Jul 2017 12:15:57 +0000 (21:15 +0900)]
rtpgsmpay: fix accidental garbage data before actual payload

Do not allocate payload size outbuf if appending payload buffer.

The commit 137672ff1824948bda4b1b1967de8c24a0055b67 attached payload
to the output buffer but forgot to remove payload allocation.  That
effectively doubled payload size and add zero'ed or random bytes.

Makes the following pipeline work again:

gst-launch-1.0 -v audiotestsrc wave=2 ! gsmenc ! rtpgsmpay ! rtpgsmdepay ! gsmdec ! autoaudiosink

https://bugzilla.gnome.org/show_bug.cgi?id=784616

6 years agortprtxreceive: Add memory and boundary checks
Nicolas Dufresne [Mon, 3 Jul 2017 15:47:13 +0000 (11:47 -0400)]
rtprtxreceive: Add memory and boundary checks

This element was not checking if mapping the RTP buffer and the payload
worked, and was not checking if the RTX payload was large enough.

https://bugzilla.gnome.org/show_bug.cgi?id=784484

6 years agoimagefreeze: fix use-after-free on seek event
Tim-Philipp Müller [Mon, 3 Jul 2017 19:27:29 +0000 (20:27 +0100)]
imagefreeze: fix use-after-free on seek event

Get seqnum before unreffing the seek event.

https://bugzilla.gnome.org/show_bug.cgi?id=784486

6 years agortspsrc: Create send/recv mutexes once, not on every connect()
Sebastian Dröge [Thu, 29 Jun 2017 15:59:58 +0000 (18:59 +0300)]
rtspsrc: Create send/recv mutexes once, not on every connect()

Also fixes a crash caused by freeing an uninitialized mutex in an error
case.

https://bugzilla.gnome.org//show_bug.cgi?id=784282

6 years agortspsrc: Actually use the receive lock when receiving, not the send lock
Sebastian Dröge [Thu, 22 Jun 2017 08:38:56 +0000 (11:38 +0300)]
rtspsrc: Actually use the receive lock when receiving, not the send lock

6 years agomatroska,videofilter: fix caps leak 67/136367/1 accepted/tizen/4.0/unified/20170816.012648 accepted/tizen/4.0/unified/20170816.015421 accepted/tizen/4.0/unified/20170828.222442 accepted/tizen/unified/20170710.154311 submit/tizen/20170707.065345 submit/tizen_4.0/20170811.094300 submit/tizen_4.0/20170814.115522 submit/tizen_4.0/20170828.100005 submit/tizen_4.0_unified/20170814.115522
Eunhae Choi [Thu, 29 Jun 2017 08:13:09 +0000 (17:13 +0900)]
matroska,videofilter: fix caps leak

Change-Id: I447b10735eed17e52769b939cc88be9f6a9781c2

6 years agoFix build error with TV profile 50/135850/1
Sangjin, Sim [Tue, 27 Jun 2017 05:02:05 +0000 (14:02 +0900)]
Fix build error with TV profile

Signed-off-by: Sangjin, Sim <sangjin0924.sim@samsung.com>
Change-Id: I6e28b1909192538e9e446a247da0fefaa6038c5c

6 years agoRelease 1.12.1
Sebastian Dröge [Tue, 20 Jun 2017 09:06:22 +0000 (12:06 +0300)]
Release 1.12.1

6 years agoUpdate .po files
Sebastian Dröge [Tue, 20 Jun 2017 08:20:12 +0000 (11:20 +0300)]
Update .po files

6 years agopo: Update translations
Sebastian Dröge [Tue, 20 Jun 2017 08:08:32 +0000 (11:08 +0300)]
po: Update translations

6 years agosplitmux: Drop allocation queries
Vivia Nikolaidou [Tue, 13 Jun 2017 14:40:19 +0000 (17:40 +0300)]
splitmux: Drop allocation queries

They can cause us to deadlock, while we're waiting for a new frame and
upstream is waiting for the allocation query to be answered before
sending a frame

https://bugzilla.gnome.org/show_bug.cgi?id=783753

6 years agortspsrc: Use a mutex for protecting against concurrent send/receives
Sebastian Dröge [Thu, 15 Jun 2017 07:40:51 +0000 (10:40 +0300)]
rtspsrc: Use a mutex for protecting against concurrent send/receives

We currently send data to the RTSP connection from multiple threads:
whenever a command is to be handled and whenever RTCP is generated. This
can cause data corruption or worse if both happen at the same time.

As such, protect gst_rtsp_connection_send() and gst_rtsp_connection_receive()
calls with a mutex. While this means that we hold a mutex during the IO
operation, this is not actually a problem as the IO operation can be
interrupted (gst_rtsp_connection_flush()) at any time and is blocking by
itself anyway.

6 years agortph264depay : fix mem leak 24/134424/1
Eunhae Choi [Fri, 16 Jun 2017 09:45:07 +0000 (18:45 +0900)]
rtph264depay : fix mem leak

Change-Id: Ic6255805254c15b9c01cdadcbecb2111e2a9bf7a

6 years agortp: fix mem leak 49/134349/1
Eunhae Choi [Fri, 16 Jun 2017 05:50:40 +0000 (14:50 +0900)]
rtp: fix mem leak

Change-Id: I26e98cdfac70d0adde686bf399620d82d9106269

6 years agoqtmux: Un-merge the last two stsc entries after serializing
Sebastian Dröge [Thu, 15 Jun 2017 08:50:44 +0000 (11:50 +0300)]
qtmux: Un-merge the last two stsc entries after serializing

The last entry will most likely get new samples added to it in "robust"
muxing mode, changing the samples_per_chunk and thus making it wrong to
keep the last two entries merged. It will run into an assertion later
when adding a new sample to the chunk.

Thanks to gdiener@cardinalpeak.com for the analysis of the bug and
proposal for a solution.

6 years agowavparse: Actually clip to upstream size instead of size of the data chunk
Sebastian Dröge [Tue, 13 Jun 2017 21:09:25 +0000 (00:09 +0300)]
wavparse: Actually clip to upstream size instead of size of the data chunk

There might be other chunks after the data chunk, so clipping the chunk
size with the data size can lead to a negative number and all following
calculations go wrong and cause crashes or worse.

This was introduced in 3ac119bbe2c360e28c087cf3852ea769d611b120.

https://bugzilla.gnome.org/show_bug.cgi?id=783760

6 years agortpsession: print value of unknown RTCP Payload Type
Juan Navarro [Tue, 30 May 2017 20:23:10 +0000 (22:23 +0200)]
rtpsession: print value of unknown RTCP Payload Type

This adds printing the actual value of any unknown RTCP PT
to the already existing WARNING log message.

https://bugzilla.gnome.org/show_bug.cgi?id=783248

6 years agortph265depay: fix caps leak
Tim-Philipp Müller [Fri, 2 Jun 2017 10:30:15 +0000 (11:30 +0100)]
rtph265depay: fix caps leak

6 years agoaacparse : Fix, Caps were not set while reusing aacparse
vijay [Wed, 24 May 2017 06:03:05 +0000 (11:33 +0530)]
aacparse : Fix, Caps were not set while reusing aacparse

While reusing aacparse caps were not set.This fix enables aacparse to reuse in same pipeline.

https://bugzilla.gnome.org/show_bug.cgi?id=783027

6 years agoapply CVE patch for security weakness 24/132124/2 accepted/tizen/unified/20170608.072420 submit/tizen/20170608.033315
Sejun Park [Thu, 1 Jun 2017 06:55:13 +0000 (15:55 +0900)]
apply CVE patch for security weakness

https://cgit.freedesktop.org/gstreamer/gst-plugins-good/commit/?id=153a8ae752c90d07190ef45803422a4f71ea8bff

Change-Id: I1472b6d6dcac2371c9d32d4ca0d9f5e98d4b9a1e

6 years agoqtmux: Do not check timecode data for mp4 container
Vivia Nikolaidou [Tue, 16 May 2017 09:56:15 +0000 (12:56 +0300)]
qtmux: Do not check timecode data for mp4 container

Timecode trak is only supported for mov right now, not for mp4. That
code would otherwise create an invalid trak if the muxed video contained
timecode metadata.

https://bugzilla.gnome.org/show_bug.cgi?id=782684

7 years agortpsrc, aacparse, mpegaudioparse: Fix build warning, check indent 30/130730/7
Gilbok Lee [Tue, 23 May 2017 12:18:17 +0000 (21:18 +0900)]
rtpsrc, aacparse, mpegaudioparse: Fix build warning, check indent

Change-Id: I2e4cff09dead99ed0ca5f365af6b1139032cb0de

7 years agoqtmux: Lateness is in QT timescale, diff in GstClockTime
Sebastian Dröge [Wed, 10 May 2017 13:58:41 +0000 (15:58 +0200)]
qtmux: Lateness is in QT timescale, diff in GstClockTime

Print the right one in debug output to get meaningful numbers.

7 years agovpxdec: Set fb->priv to NULL after freeing just in case
Sebastian Dröge [Tue, 9 May 2017 09:41:25 +0000 (11:41 +0200)]
vpxdec: Set fb->priv to NULL after freeing just in case

https://bugzilla.gnome.org/show_bug.cgi?id=782359

7 years agodirectsoundsink: Use GstClock API instead of Sleep() for waiting
Dustin Spicuzza [Mon, 8 May 2017 15:22:00 +0000 (15:22 +0000)]
directsoundsink: Use GstClock API instead of Sleep() for waiting

It's more accurate and allows cancellation.

https://bugzilla.gnome.org/show_bug.cgi?id=773681

7 years agovpx: fix build against older libvpx versions
Tim-Philipp Müller [Mon, 8 May 2017 15:05:45 +0000 (15:05 +0000)]
vpx: fix build against older libvpx versions

Such as 1.3.0 as on raspbian.

7 years agodirectsoundsink: Fix corner case causing large CPU usage
Nirbheek Chauhan [Wed, 3 May 2017 17:53:10 +0000 (23:23 +0530)]
directsoundsink: Fix corner case causing large CPU usage

We were unnecessarily looping/goto-ing repeatedly when we had exactly
the amount of data as the free space, and also when the free space was
too small. This, as it turns out, is a very common scenario with
Directsound on Windows.

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=773681

We have to do polling here because the event notification API that
Directsound exposes cannot be used with live playback since all events
must be registered in advance with the capture buffer, you cannot
add/remove them once playback has begun. Directsoundsrc had the same
problem.

See also: https://bugzilla.gnome.org/show_bug.cgi?id=781249

7 years agoRelease 1.12.0
Sebastian Dröge [Thu, 4 May 2017 12:38:34 +0000 (15:38 +0300)]
Release 1.12.0

7 years agoUpdate .po files
Sebastian Dröge [Thu, 4 May 2017 12:07:27 +0000 (15:07 +0300)]
Update .po files

7 years agopo: Update translations
Sebastian Dröge [Thu, 4 May 2017 10:47:20 +0000 (13:47 +0300)]
po: Update translations

7 years agoqtdemux: Fix crash on mss stream caused by invalid stsd entry access
Seungha Yang [Tue, 2 May 2017 01:32:30 +0000 (10:32 +0900)]
qtdemux: Fix crash on mss stream caused by invalid stsd entry access

Since mss has no moov, default stsd entry should be created with media-caps.

https://bugzilla.gnome.org/show_bug.cgi?id=782042

7 years agoRelease 1.11.91
Sebastian Dröge [Thu, 27 Apr 2017 14:29:58 +0000 (17:29 +0300)]
Release 1.11.91

7 years agoUpdate .po files
Sebastian Dröge [Thu, 27 Apr 2017 12:58:47 +0000 (15:58 +0300)]
Update .po files

7 years agopo: Update translations
Sebastian Dröge [Thu, 27 Apr 2017 12:28:02 +0000 (15:28 +0300)]
po: Update translations

7 years agoqtdemux: Don't crash in debug output if stream==NULL
Sebastian Dröge [Thu, 27 Apr 2017 09:56:27 +0000 (12:56 +0300)]
qtdemux: Don't crash in debug output if stream==NULL

That case is correctly handled below but not in the debug output.

https://bugzilla.gnome.org/show_bug.cgi?id=781270

7 years agoqtdemux: Don't perform seeks with inconsistent seek values
Sebastian Dröge [Tue, 25 Apr 2017 14:11:27 +0000 (17:11 +0300)]
qtdemux: Don't perform seeks with inconsistent seek values

If gst_segment_do_seek() fails, we shouldn't try seeking on that
resulting segment but just error out. Crashes further down the line
otherwise.

7 years agoAutomatic update of common submodule
Tim-Philipp Müller [Mon, 24 Apr 2017 19:27:49 +0000 (20:27 +0100)]
Automatic update of common submodule

From 60aeef6 to 48a5d85

7 years agotests: rtp-payloading: add test for rtph264depay avc/byte-stream output
Tim-Philipp Müller [Mon, 24 Apr 2017 16:31:04 +0000 (17:31 +0100)]
tests: rtp-payloading: add test for rtph264depay avc/byte-stream output

Make sure avc output doesn't contain SPS/PPS inline, but
byte-stream output does.

7 years agortph264depay: don't insert SPS/PPS inline for AVC output
Tim-Philipp Müller [Mon, 24 Apr 2017 16:29:37 +0000 (17:29 +0100)]
rtph264depay: don't insert SPS/PPS inline for AVC output

SPS/PPS are in the caps in this case and shouldn't be in
the stream data.

7 years agortspsrc: Chain up to the parent class' provide_clock() implementation
Sebastian Dröge [Fri, 21 Apr 2017 18:09:14 +0000 (19:09 +0100)]
rtspsrc: Chain up to the parent class' provide_clock() implementation

If no clock was provided directly by rtspsrc. This behaviour was removed
by f8013487c91a6ffc552a4b25aa1a70f0bd5377f8 and results in rtspsrc not
providing the system clock via the rtpjitterbuffer.

As a result, if another element like an audio sink, provides a clock,
the pipeline would select that (when going to PAUSED/PLAYING again later).
Audio clocks usually don't progress in PAUSED, and thus our live source
won't be able to use the clock to produce data, making the sink never
preroll and everything is stuck.

7 years agoqtdemux: reset sample_description_id to default
Jürgen Sachs [Thu, 20 Apr 2017 09:22:15 +0000 (11:22 +0200)]
qtdemux: reset sample_description_id to default

Fixes stream where sample_description_id is specified in the tfhd

https://bugzilla.gnome.org/show_bug.cgi?id=778337

7 years agosplitmuxsink: Don't use an explicit name for requesting audio pads
Sebastian Dröge [Thu, 20 Apr 2017 12:16:24 +0000 (13:16 +0100)]
splitmuxsink: Don't use an explicit name for requesting audio pads

... unless the muxer uses the same audio pad template name as
splitmuxsink. We can't request a pad called "audio_0" on a muxer that
wants pads to be "sink_%d".

7 years agoflvdemux: remove duplicated segment initialization
ChangBok Chae [Thu, 23 Feb 2017 00:31:36 +0000 (09:31 +0900)]
flvdemux: remove duplicated segment initialization

It's also done in gst_flv_demux_cleanup().

https://bugzilla.gnome.org/show_bug.cgi?id=779106

7 years agosplitmuxsink: Correctly catch FLUSH events in probes
Xavier Claessens [Thu, 20 Apr 2017 10:17:35 +0000 (20:17 +1000)]
splitmuxsink: Correctly catch FLUSH events in probes

https://bugzilla.gnome.org/show_bug.cgi?id=767498

7 years agoRevert "rtpbin: pipeline gets an EOS when any rtpsources byes"
Tim-Philipp Müller [Wed, 19 Apr 2017 11:28:12 +0000 (12:28 +0100)]
Revert "rtpbin: pipeline gets an EOS when any rtpsources byes"

This reverts commit eeea2a7fe88a17b15318d5b6ae6e190b2f777030.

It breaks EOS in some sender pipelines, see
https://bugzilla.gnome.org/show_bug.cgi?id=773218#c20

7 years agoqtdemux: Reset adapter in more discontinuity cases
Edward Hervey [Fri, 14 Apr 2017 15:01:49 +0000 (17:01 +0200)]
qtdemux: Reset adapter in more discontinuity cases

In push mode we process as much as possible in the adapter. When we receive
a DISCONT buffer which we can't match to an actual sample (based on the existing
sample table) and there is still data remaining in the incoming adapter,there is
one of two cases happening:
1) We are doing reverse playback, in which case we should flush out all pending
  data
2) We have leftover data from the previous incoming buffer... which we can't do
  anything about.

For the second case, make sure we flush out the remaining data so that we can start
parsing again from scratch.

https://bugzilla.gnome.org/show_bug.cgi?id=781319

7 years agortspsrc: Use GST_ELEMENT_ERROR_WITH_DETAILS
Edward Hervey [Fri, 14 Apr 2017 08:56:41 +0000 (10:56 +0200)]
rtspsrc: Use GST_ELEMENT_ERROR_WITH_DETAILS

Allows the application to know the exact status code that was returned
by the server in a programmatic fashion.

https://bugzilla.gnome.org/show_bug.cgi?id=781304

7 years agoqtdemux: Fix leak on QtDemuxStreamStsdEntry
Seungha Yang [Sun, 16 Apr 2017 09:47:56 +0000 (18:47 +0900)]
qtdemux: Fix leak on QtDemuxStreamStsdEntry

Fix unit test failure

https://bugzilla.gnome.org/show_bug.cgi?id=781362

7 years agoqtmux: Fix timescale of timecode tracks
Sebastian Dröge [Fri, 14 Apr 2017 10:38:53 +0000 (13:38 +0300)]
qtmux: Fix timescale of timecode tracks

They should have ideally the same timescale of the video track, which we
can't guarantee here as in theory timecode configuration and video
framerate could be different. However we should set a correct timescale
based on the framerate given in the timecode configuration, and not just
use the framerate numerator.

7 years agoqtdemux: Properly reset demuxer when all streams are EOS
Edward Hervey [Thu, 13 Apr 2017 11:25:06 +0000 (13:25 +0200)]
qtdemux: Properly reset demuxer when all streams are EOS

Make sure offset and neededbytes are properly resetted when all
streams are EOS in push-mode.

Avoids cases when some data might still be pushed by upstream (because
it didn't yet see the resulting GST_FLOW_EOS yet) and qtdemux gets
completely lost.

https://bugzilla.gnome.org/show_bug.cgi?id=781266

7 years agosouphttpsrc: Make more usage of error macro
Edward Hervey [Thu, 13 Apr 2017 06:00:30 +0000 (08:00 +0200)]
souphttpsrc: Make more usage of error macro

And make sure we actually use the provided soup_msg argument in the macro

7 years agomeson: Print message when disabling taglib on MSVC
Nirbheek Chauhan [Wed, 12 Apr 2017 13:16:53 +0000 (18:46 +0530)]
meson: Print message when disabling taglib on MSVC

7 years agoqtmux: Don't forget to update pad->last_buf
Edward Hervey [Wed, 12 Apr 2017 11:26:59 +0000 (13:26 +0200)]
qtmux: Don't forget to update pad->last_buf

buf is the current pad->last_buf value. If ever it gets copied/unreffed,
we need to make sure to write back the new  pointer to the last_buf
variable.

Fixes using wrong pointer values in the case of decrasing DTS value

7 years agotests: Add vp9enc to gitignore
Edward Hervey [Wed, 12 Apr 2017 09:33:05 +0000 (11:33 +0200)]
tests: Add vp9enc to gitignore

7 years agoqtdemux: fix: sample description index override in tfhd not evaluated
Jürgen Sachs [Tue, 11 Apr 2017 11:41:48 +0000 (13:41 +0200)]
qtdemux: fix: sample description index override in tfhd not evaluated

https://bugzilla.gnome.org/show_bug.cgi?id=778337

7 years agoqtdemux: Add out-of-bound check
Edward Hervey [Wed, 12 Apr 2017 09:03:24 +0000 (11:03 +0200)]
qtdemux: Add out-of-bound check

Make sure we don't read invalid memory

7 years agoqtdemux: move parsing of tkhd out of stsd entry loop
Thiago Santos [Wed, 27 Apr 2016 15:17:37 +0000 (12:17 -0300)]
qtdemux: move parsing of tkhd out of stsd entry loop

It needs only to be read once.

7 years agoqtdemux: check for a different stsd entry before pushing a sample
Thiago Santos [Thu, 7 Apr 2016 15:23:35 +0000 (12:23 -0300)]
qtdemux: check for a different stsd entry before pushing a sample

Before pushing a sample, check if there was a change in the current
stsd entry. This patch also assumes that the first stsd entry is
used as default for the first sample. It might cause an uneeded
caps renegotiation when this isn't the case.

7 years agoqtdemux: parse all stsd entries
Thiago Santos [Wed, 6 Apr 2016 15:55:18 +0000 (12:55 -0300)]
qtdemux: parse all stsd entries

stsd can have multiple format entries, parse them all.

This is required to play DVB DASH profile that uses multiple entries
to identify the different available bitrates/options on dash streams

The stream format-specific data is not stored into QtDemuxStreamStsdEntry

7 years agoqtdemux: rework stsd sample entries access
Thiago Santos [Tue, 5 Apr 2016 17:34:00 +0000 (14:34 -0300)]
qtdemux: rework stsd sample entries access

Instead of using the stsd as a base pointer, use the actual stsd
entry as the stsd can have multiple entries. This is rarely used
for file playback but is a possible profile with in DVB DASH specs.

This still doesn't support stsd with multiple entries but makes it
easier to do so.

7 years agoqtdemux: get stsd child by index instead of type
Thiago Santos [Tue, 5 Apr 2016 21:00:10 +0000 (18:00 -0300)]
qtdemux: get stsd child by index instead of type

There might be multiple children with the same type

7 years agotests/check/rtprtx: add checks for rtprtxqueue's max-size-{time,packets} properties
George Kiagiadakis [Fri, 7 Apr 2017 13:33:18 +0000 (16:33 +0300)]
tests/check/rtprtx: add checks for rtprtxqueue's max-size-{time,packets} properties

https://bugzilla.gnome.org/show_bug.cgi?id=780867

7 years agortprtxqueue: implement handling of the max-size-time property
George Kiagiadakis [Tue, 4 Apr 2017 14:33:31 +0000 (17:33 +0300)]
rtprtxqueue: implement handling of the max-size-time property

https://bugzilla.gnome.org/show_bug.cgi?id=780867

7 years agoAutomatic update of common submodule
Tim-Philipp Müller [Mon, 10 Apr 2017 22:49:06 +0000 (23:49 +0100)]
Automatic update of common submodule

From 39ac2f5 to 60aeef6

7 years agov4l2object: Copy timestamp when importing buffers
Todor Tomov [Mon, 10 Apr 2017 08:56:00 +0000 (08:56 +0000)]
v4l2object: Copy timestamp when importing buffers

This is needed for V4L2_OUTPUT interface, and is harmless of
V4L2_CAPTURE interfaces. This will fix timestamp in cases like:

  v4l2src io-mode=dmabuf ! v4l2videoNenc output-io-mode=dmabuf-import !  ...

Same apply for userptr.

https://bugzilla.gnome.org/show_bug.cgi?id=781119

7 years agoqtmux: Fix last_dts tracking for raw audio and similar formats
Sebastian Dröge [Mon, 10 Apr 2017 12:55:30 +0000 (15:55 +0300)]
qtmux: Fix last_dts tracking for raw audio and similar formats

Accumulate the durations directly and don't scale yet another time by
the number of samples.

7 years agotests: fix leak in splitmux test
Vincent Penquerc'h [Fri, 7 Apr 2017 09:48:50 +0000 (10:48 +0100)]
tests: fix leak in splitmux test

https://bugzilla.gnome.org/show_bug.cgi?id=781025

7 years agoscaletempo: Scale GAP event timestamp and duration like for buffers
Lyon Wang [Fri, 7 Apr 2017 07:29:43 +0000 (15:29 +0800)]
scaletempo: Scale GAP event timestamp and duration like for buffers

https://bugzilla.gnome.org/show_bug.cgi?id=781008

7 years agov4l2dec: Fix race when going from PAUSED to READY
Thibault Saunier [Fri, 17 Feb 2017 13:01:08 +0000 (10:01 -0300)]
v4l2dec: Fix race when going from PAUSED to READY

Running `gst-validate-launcher -t validate.file.playback.change_state_intensive.vorbis_vp8_1_webm`
on odroid XU4 (s5p-mfc v4l2 driver) often leads to:

  ERROR:../subprojects/gst-plugins-good/sys/v4l2/gstv4l2videodec.c:215:gst_v4l2_video_dec_stop: assertion failed: (g_atomic_int_get (&self->processing) == FALSE)

This happens when the following race happens:

- T0: Main thread
- T1: Upstream streaming thread
- T2. v4l2dec processing thread)

[The decoder is in PAUSED state]

T0. The validate scenario runs `Executing (36/40) set-state: state=null repeat=40`
T1- The decoder handles a frame
T2- A decoded frame is push downstream
T2- Downstream returns FLUSHING as it is already flushing changing state
T2- The decoder stops its processing thread and sets `->processing = FALSE`
T1- The decoder handles another frame
T1- `->process` is FALSE so the decoder restarts its streaming thread
T0- In v4l2dec-> stop the processing thread is stopped
NOTE: At this point the processing thread loop never started.
T0- assertion failed: (g_atomic_int_get (&self->processing) == FALSE)

Here I am removing the whole ->processing logic to base it all on the
GstTask state to avoid duplicating the knowledge.

https://bugzilla.gnome.org/show_bug.cgi?id=778830

7 years agoRelease 1.11.90
Sebastian Dröge [Fri, 7 Apr 2017 13:31:56 +0000 (16:31 +0300)]
Release 1.11.90

7 years agoUpdate .po files
Sebastian Dröge [Fri, 7 Apr 2017 12:18:11 +0000 (15:18 +0300)]
Update .po files

7 years agopo: Update translations
Sebastian Dröge [Fri, 7 Apr 2017 12:06:30 +0000 (15:06 +0300)]
po: Update translations

7 years agoaacparse: streamline and improve AudioSpecificConfig parsing
Edward Hervey [Thu, 6 Apr 2017 10:01:00 +0000 (12:01 +0200)]
aacparse: streamline and improve AudioSpecificConfig parsing

AudioSpecifigConfig is used in a variety of AAC streams but was
being parsed differently. Instead, make everyone use the same parsing.

* Remove unused 'bits' field (it was always set to 0 if present)
* Add proper GAConfig parsing (to know the  number of samples per frame
  if present).

Fixes wrong rate/channels configuration in streams coming from qtdemux

https://bugzilla.gnome.org/show_bug.cgi?id=780966

7 years agov4l2videodec: Fix 32bit only printf format
Nicolas Dufresne [Wed, 5 Apr 2017 13:46:31 +0000 (09:46 -0400)]
v4l2videodec: Fix 32bit only printf format

The previous patch was using %llu for 64bits printf, which is 32bit
specific. We also trace the latency in time human readable form now.

7 years agov4l2object: set streamparm for outputs that support it
Philipp Zabel [Wed, 16 Mar 2016 15:22:48 +0000 (16:22 +0100)]
v4l2object: set streamparm for outputs that support it

Without a specified framerate from the sink, the decoder frame interval
should be set using the framerate of the encoded video stream.
Therefore, the v4l2object should be able to change the framerate on the
output if the V4L2 device accepts it.

This is also necessary for mem2mem encoders so that their bitrate
calculation code may work correctly and they may report the correct
frame duration on the capture queue.

https://bugzilla.gnome.org/show_bug.cgi?id=779466

7 years agov4l2videodec: only set latency if the frame duration is valid
Philipp Zabel [Wed, 16 Mar 2016 15:24:55 +0000 (16:24 +0100)]
v4l2videodec: only set latency if the frame duration is valid

If the duration of the v4l2object is GST_CLOCK_TIME_NONE, because the
sink did not specify a framerate in the caps and the driver accepts the
framerate, the decoder element uses GST_CLOCK_TIME_NONE to calculate and
set the element latency.

While this is a bug of the capture driver, the decoder element should
not use the invalid duration to calculate a latency, but print a warning
instead.

https://bugzilla.gnome.org/show_bug.cgi?id=779466

7 years agov4l2sink: Block in preroll_wait on unlock
Olivier Crête [Wed, 23 Nov 2016 17:17:55 +0000 (12:17 -0500)]
v4l2sink: Block in preroll_wait on unlock

The correct behaviour of anything stuck in the ->render() function
between ->unlock() and ->unlock_stop() is to call
gst_base_sink_wait_preroll() and only return an error if this returns an
error, otherwise, it must continue where it left off!

https://bugzilla.gnome.org/show_bug.cgi?id=774945

7 years agovp9dec: Add warnings for unsupported frame formats
Jan Schmidt [Wed, 5 Apr 2017 05:55:20 +0000 (15:55 +1000)]
vp9dec: Add warnings for unsupported frame formats

At least output an element warning on the bus when we
encounter a frame format GStreamer doesn't currently support.

7 years agoaacparse: Handle Parametric Stereo with HE-AAC(v2)
Edward Hervey [Tue, 4 Apr 2017 15:55:13 +0000 (17:55 +0200)]
aacparse: Handle Parametric Stereo with HE-AAC(v2)

According to ISO/IEC:14496-2:2009 , in the case of HE-AACv2 (audioObjecType
29) parametric stereo is used (a single mono track is used and then
transformations are applied to it to provide a stereo output).

We therefore report two channels in the case where there is one reported
in the audioChannelConfiguration.

Fixes the various issues where a demuxer would report two channels, but
then the parser would say there's only one channel, and then the decoder
would output two channels.

7 years agoqtmux: Simplify buffer refcounting in add_buffer() and remove unneeded NULL checks
Sebastian Dröge [Tue, 4 Apr 2017 12:22:25 +0000 (15:22 +0300)]
qtmux: Simplify buffer refcounting in add_buffer() and remove unneeded NULL checks

7 years agoqtmux: Select the best pad based on the cached last_buf if any
Sebastian Dröge [Tue, 4 Apr 2017 12:08:33 +0000 (15:08 +0300)]
qtmux: Select the best pad based on the cached last_buf if any

last_buf is the one we're going to write next, not buf. As such we
should check timestamps against that one if there is one to select the
earliest pad.

Also remember the currently selected pad in the very beginning when
storing the first last_buf.

This both solves some edge cases where not the correct next pad was
selected corresponding to the target interleave.

7 years agoqtmux: Error out immediately if a timecode is to be written but downstream return...
Sebastian Dröge [Tue, 4 Apr 2017 12:07:40 +0000 (15:07 +0300)]
qtmux: Error out immediately if a timecode is to be written but downstream return not-OK

7 years agoqtdemux: Update variables before early exit
Edward Hervey [Mon, 3 Apr 2017 09:34:49 +0000 (11:34 +0200)]
qtdemux: Update variables before early exit

This is an update of d78d5896272d78df41e696fac929e7dfb3bb3dfa

We still exit as early as possible in case of non-ok/non-unlinked combined
flow, but we first make sure that we update the internal position variables.

This ensures that if upstreams "ignores" the flow return (and carries on pushing),
we don't end up processing data with completely bogus variables/positions.

7 years agointerleave: avoid using uninitialised ordering_map
Douglas Bagnall [Thu, 23 Mar 2017 11:11:13 +0000 (00:11 +1300)]
interleave: avoid using uninitialised ordering_map

If self->channel_positions == NULL (which seems unlikely),
self->default_channels_ordering_map will be used unintialised.

We avoid that by keeping track of the channel_mask, which is set when
the ordering map is initialised.

https://bugzilla.gnome.org/show_bug.cgi?id=780331

7 years agointerleave: don't overflow channel map with >64 channels
Douglas Bagnall [Thu, 23 Mar 2017 10:56:31 +0000 (23:56 +1300)]
interleave: don't overflow channel map with >64 channels

When there are more than 64 channels, we don't want to exceed the
bounds of the ordering_map buffer, and in these cases we don't want to
rempa at all. Here we avoid doing that.

https://bugzilla.gnome.org/show_bug.cgi?id=780331

7 years agov4l2videodec: Set rank to GST_RANK_PRIMARY 42/122042/1 accepted/tizen_common accepted/tizen_ivi accepted/tizen_mobile accepted/tizen_tv accepted/tizen_wearable accepted/tizen/common/20170404.142251 accepted/tizen/ivi/20170403.130012 accepted/tizen/mobile/20170403.125534 accepted/tizen/tv/20170403.125818 accepted/tizen/unified/20170403.130113 accepted/tizen/wearable/20170403.125904 submit/tizen/20170403.050513 tizen_4.0.m1_release
Gilbok Lee [Thu, 30 Mar 2017 04:01:05 +0000 (13:01 +0900)]
v4l2videodec: Set rank to GST_RANK_PRIMARY

Change-Id: If9c79874dc4af246f485b63c151ff732bc31abc9

7 years agomeson: Use get_pkgconfig_variable instead of calling pkg-config ourself
Thibault Saunier [Tue, 28 Mar 2017 17:23:16 +0000 (14:23 -0300)]
meson: Use get_pkgconfig_variable instead of calling pkg-config ourself

It is avalaible in meson 0.36 which is now are requirement

7 years agopkgconfig: Do not ever build an installed .pc file
Thibault Saunier [Tue, 28 Mar 2017 17:22:41 +0000 (14:22 -0300)]
pkgconfig: Do not ever build an installed .pc file

7 years agomeson: test: Fix environment object usage
Thibault Saunier [Tue, 28 Mar 2017 14:15:53 +0000 (11:15 -0300)]
meson: test: Fix environment object usage

7 years agopkgconfig: Generate the pkg-config with meson too
Thibault Saunier [Tue, 28 Mar 2017 14:14:47 +0000 (11:14 -0300)]
pkgconfig: Generate the pkg-config with meson too

7 years agoqtdemux: In gap mode, consider the mdat offset when calculating the remaining mdat...
Sebastian Dröge [Mon, 27 Mar 2017 18:52:00 +0000 (21:52 +0300)]
qtdemux: In gap mode, consider the mdat offset when calculating the remaining mdat size

The mdat generally does not start at offset 0, we have to include the
size of the moof and whatever else was in front of the mdat.

7 years agoatomsrecovery: Error out when fseek() fails instead of silently ignoring
Sebastian Dröge [Mon, 27 Mar 2017 08:43:31 +0000 (11:43 +0300)]
atomsrecovery: Error out when fseek() fails instead of silently ignoring

CID 1403262