platform/upstream/gst-plugins-good.git
9 years agov4l2allocator: Use qdata instead of parenting to DmabufMemory
Nicolas Dufresne [Thu, 10 Jul 2014 22:11:20 +0000 (18:11 -0400)]
v4l2allocator: Use qdata instead of parenting to DmabufMemory

Parenting V4l2Memory to DmabufMemory was in conflict with recent
optimization in DmabufMemory to avoid dup(), and didn't work with
memory sharing. Instead, use a qdata and it's destroy notify.

https://bugzilla.gnome.org/show_bug.cgi?id=730441

9 years agopo: Update translations
Sebastian Dröge [Fri, 11 Jul 2014 06:52:39 +0000 (08:52 +0200)]
po: Update translations

9 years agov4l2bufferpool: Workaround elements not requesting any buffers
Nicolas Dufresne [Tue, 8 Jul 2014 21:50:47 +0000 (17:50 -0400)]
v4l2bufferpool: Workaround elements not requesting any buffers

This is a workaround for element that don't request buffers when
they should.

https://bugzilla.gnome.org/show_bug.cgi?id=732288

9 years agotests: fix pipeline leak in videocrop test
Sebastian Rasmussen [Sun, 6 Jul 2014 09:27:36 +0000 (11:27 +0200)]
tests: fix pipeline leak in videocrop test

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=732976

9 years agoexamples: client-rtpaux: Release reference to parent when done
Sebastian Rasmussen [Sun, 6 Jul 2014 09:26:46 +0000 (11:26 +0200)]
examples: client-rtpaux: Release reference to parent when done

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=732976

9 years agortspsrc: fix query leak
Tim-Philipp Müller [Thu, 10 Jul 2014 16:19:42 +0000 (17:19 +0100)]
rtspsrc: fix query leak

https://bugzilla.gnome.org/show_bug.cgi?id=733003

9 years agowavenc: Return not-negotiated if we got no caps or caps negotiation failed
Sebastian Dröge [Thu, 10 Jul 2014 10:10:45 +0000 (12:10 +0200)]
wavenc: Return not-negotiated if we got no caps or caps negotiation failed

And do it always, not inside a g_return_val_if_fail().

See https://bugzilla.gnome.org/show_bug.cgi?id=732939

9 years agov4l2src: Ensure internal pool activation
Nicolas Dufresne [Tue, 8 Jul 2014 17:34:28 +0000 (13:34 -0400)]
v4l2src: Ensure internal pool activation

Before we would hit an assertion "'gst_buffer_pool_is_active (bpool)' failed"
if the internal pool was not used to push buffer downstrea, hence not
given to the baseclass.

https://bugzilla.gnome.org/show_bug.cgi?id=732912

9 years agovideomixer: fix double unlock in segment seek segment code path
Tim-Philipp Müller [Fri, 4 Jul 2014 19:22:10 +0000 (20:22 +0100)]
videomixer: fix double unlock in segment seek segment code path

We only want to unlock if we push an event downstream and
jump to done_unlock label afterwards. We would also unlock
in case of a segment seek and then unlock again later, and
nothing good can come of that.

(This code looks a bit dodgy anyway though, shouldn't it
also bail out with FLOW_EOS here in case of a segment seek
scenario, just without the event?)

9 years agotests: qtmux: suppress glib criticals caused by testing deprecated dts methods
Tim-Philipp Müller [Fri, 4 Jul 2014 18:45:55 +0000 (19:45 +0100)]
tests: qtmux: suppress glib criticals caused by testing deprecated dts methods

9 years agoavidemux, wavparse: Print invalid fourcc in hex
Sebastian Rasmussen [Fri, 4 Jul 2014 01:21:30 +0000 (03:21 +0200)]
avidemux, wavparse: Print invalid fourcc in hex

Previously this was printed as characters which caused later processing
of the error message to sometimes warn about non-UTF-8 characters.

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=732714

9 years agov4l2object: Pool might be NULL in decide allocation
Nicolas Dufresne [Thu, 3 Jul 2014 19:21:18 +0000 (15:21 -0400)]
v4l2object: Pool might be NULL in decide allocation

If special stride is needed and downstream don't support VideoMeta,
pool might be NULL in order to let the baseclass create a generic
pool­. This would lead to assertion with on Exynos with:

gst-launch-1.0 -v filesrc location=mov ! qtdemux ! h264parse ! \
                    v4l2video8dec ! fakesink

https://bugzilla.gnome.org/show_bug.cgi?id=732707

9 years agov4l2bufferpool: Handle FD error during poll
Nicolas Dufresne [Thu, 3 Jul 2014 19:29:54 +0000 (15:29 -0400)]
v4l2bufferpool: Handle FD error during poll

This will ensure we fail earlier if something unrecoverable
happens.

9 years agov4l2bufferpool: Wait before polling if queue is empty
Nicolas Dufresne [Thu, 3 Jul 2014 19:28:45 +0000 (15:28 -0400)]
v4l2bufferpool: Wait before polling if queue is empty

In kernel before 3.17, polling during queue underrun would unblock right
away and trigger POLLERR. As we are not handling POLLERR, we would endup
blocking in DQBUF call, which won't be unblocked correctly when going
to NULL state. A deadlock at start caused by locking error in libv4l2 was
also seen before this patch. Instead, we wait until the queue is no longer
empty before polling.

https://bugzilla.gnome.org/show_bug.cgi?id=731015

9 years agortspsrc: fix for mikey api change
Wim Taymans [Wed, 2 Jul 2014 14:01:47 +0000 (16:01 +0200)]
rtspsrc: fix for mikey api change

9 years agov4l2: fix probing and enumeration of stepwise frame sizes
Tim-Philipp Müller [Mon, 30 Jun 2014 09:29:54 +0000 (10:29 +0100)]
v4l2: fix probing and enumeration of stepwise frame sizes

The code enumerating STEPWISE framesizes would start from
(min_w, min_h) and then add (step_w, step_h) to get the
next framesize. However, it should really allow any width
from min_w to max_w with step_w and same for heights.
Secondly, we would add and probe each individual stepped
frame size to the caps as separate structure, which would
lead to hundreds if not thousands of structs ending up in
the probed caps. Use integer ranges with steps instead.

This was particularly noticable with the Raspberry Pi Cam.

https://bugzilla.gnome.org/show_bug.cgi?id=724521
https://bugzilla.gnome.org/show_bug.cgi?id=732458
https://bugzilla.gnome.org/show_bug.cgi?id=726521

9 years agov4l2object: drop workaround for misbehaving TRY_FMT
Daniel Drake [Fri, 27 Jun 2014 10:33:06 +0000 (11:33 +0100)]
v4l2object: drop workaround for misbehaving TRY_FMT

This workaround from 2011 was causing 25 S_FMT ioctls to be sent
to my UVC webcam from under gst_v4l2_object_get_caps as it probes
all the formats. In total, this adds up to about 5 seconds of
execution time, or a 10 second delay while starting up cheese.

These ioctls come from a workaround from 2011 where TRY_FMT might
make changes to hardware settings, so S_FMT was used to restore
the original config:
https://bugzilla.gnome.org/show_bug.cgi?id=649067

The driver bug is now assumed fixed. Remove the workaround to fix the
long startup delay.

https://bugzilla.gnome.org/show_bug.cgi?id=732326

9 years agovideomixer: reset QoS on segment event
Vincent Penquerc'h [Tue, 1 Jul 2014 11:50:31 +0000 (12:50 +0100)]
videomixer: reset QoS on segment event

https://bugzilla.gnome.org/show_bug.cgi?id=732540

9 years agomatroskademux: send gap events instead of segment tricks
Vincent Penquerc'h [Tue, 1 Jul 2014 14:14:34 +0000 (15:14 +0100)]
matroskademux: send gap events instead of segment tricks

This fixes missing frames from being time skipped.

https://bugzilla.gnome.org/show_bug.cgi?id=732372

9 years agortpsession: Fix memory leaks in unit test
Sebastian Dröge [Sun, 29 Jun 2014 22:00:32 +0000 (00:00 +0200)]
rtpsession: Fix memory leaks in unit test

9 years agortpbin: Don't leak caps
Sebastian Dröge [Sun, 29 Jun 2014 21:55:19 +0000 (23:55 +0200)]
rtpbin: Don't leak caps

9 years agopulsesrc: Fix compiler warning when compiling with G_DISABLE_ASSERT
Sebastian Dröge [Sun, 29 Jun 2014 18:02:14 +0000 (20:02 +0200)]
pulsesrc: Fix compiler warning when compiling with G_DISABLE_ASSERT

9 years agortpssrcdemux: Fix compiler warning when compiling with G_DISABLE_ASSERT
Sebastian Dröge [Sun, 29 Jun 2014 17:59:53 +0000 (19:59 +0200)]
rtpssrcdemux: Fix compiler warning when compiling with G_DISABLE_ASSERT

9 years agomatroskamux: Fix compiler warnings when compiling with G_DISABLE_ASSERT
Sebastian Dröge [Sun, 29 Jun 2014 17:57:57 +0000 (19:57 +0200)]
matroskamux: Fix compiler warnings when compiling with G_DISABLE_ASSERT

9 years agodeinterlace: Fix compiler warnings when compiling with G_DISABLE_ASSERT
Sebastian Dröge [Sun, 29 Jun 2014 17:54:44 +0000 (19:54 +0200)]
deinterlace: Fix compiler warnings when compiling with G_DISABLE_ASSERT

9 years agopulse: fix compiler warnings when compiling with -DG_DISABLE_ASSERT
Tim-Philipp Müller [Sun, 29 Jun 2014 16:05:13 +0000 (17:05 +0100)]
pulse: fix compiler warnings when compiling with -DG_DISABLE_ASSERT

Compiler complains about uninitialised variables in the impossible
'default' code path in device provider source/sink switch-case.

9 years agov4l2: fix compiler warnings when compiling with -DG_DISABLE_ASSERT
Tim-Philipp Müller [Sun, 29 Jun 2014 16:03:17 +0000 (17:03 +0100)]
v4l2: fix compiler warnings when compiling with -DG_DISABLE_ASSERT

Compiler complains about uninitialised variables in the impossible
'default' code path in device provider source/sink switch-case.

9 years agotests: matroskaparse: fail on errors and disable pull mode test
Tim-Philipp Müller [Sat, 28 Jun 2014 16:40:45 +0000 (17:40 +0100)]
tests: matroskaparse: fail on errors and disable pull mode test

Actually look for error messages on the bus and fail if there
is one before the EOS message. Disable pull mode test which is
pointless as long as matroskaparse only supports push mode
(pull mode support has not been ported over to 1.0).

9 years agomatroskaparse: don't error out if there's not enough data in the adapter
Tim-Philipp Müller [Sat, 28 Jun 2014 16:37:23 +0000 (17:37 +0100)]
matroskaparse: don't error out if there's not enough data in the adapter

gst_matroska_parse_take() would return FLOW_ERROR instead of
FLOW_EOS in case there's less data in the adapter than requested,
because buffer is NULL in that case which triggers the error
code path. This made the unit test fail (occasionally at least,
because of a bug in the unit test there's a race and it would
happen only sporadically).

9 years agovideomixer: Update dist generated ORC files
Sebastian Dröge [Sat, 28 Jun 2014 14:53:58 +0000 (16:53 +0200)]
videomixer: Update dist generated ORC files

9 years agovideomixer: Update videoconvert code from -base
Sebastian Dröge [Sat, 28 Jun 2014 14:48:13 +0000 (16:48 +0200)]
videomixer: Update videoconvert code from -base

And also rename the remaining symbols to prevent conflicts
during static linking.

https://bugzilla.gnome.org/show_bug.cgi?id=728443

9 years agoautovideosrc: use videotestsrc as fallback element instead of fakesrc
Tim-Philipp Müller [Sat, 28 Jun 2014 12:01:46 +0000 (13:01 +0100)]
autovideosrc: use videotestsrc as fallback element instead of fakesrc

fakesrc doesn't announce video caps, so most video pipelines will
just error out with not-negotiated if a fallback element is created.

9 years agoautoaudiosrc: use audiotestsrc as fallback element instead of fakesrc
Tim-Philipp Müller [Sat, 28 Jun 2014 11:44:31 +0000 (12:44 +0100)]
autoaudiosrc: use audiotestsrc as fallback element instead of fakesrc

fakesrc doesn't announce audio caps, so most audio pipelines will
just error out with not-negotiated if a fallback element is created.

9 years agoRelease 1.3.90 1.3.90
Sebastian Dröge [Sat, 28 Jun 2014 09:21:15 +0000 (11:21 +0200)]
Release 1.3.90

9 years agoUpdate .po files
Sebastian Dröge [Sat, 28 Jun 2014 09:08:33 +0000 (11:08 +0200)]
Update .po files

9 years agoRename GstDeviceMonitor to GstDeviceProvider
Olivier Crête [Thu, 26 Jun 2014 18:52:57 +0000 (14:52 -0400)]
Rename GstDeviceMonitor to GstDeviceProvider

9 years agovideobox: Add unit test
Ravi Kiran K N [Tue, 24 Jun 2014 03:44:40 +0000 (09:14 +0530)]
videobox: Add unit test

https://bugzilla.gnome.org/show_bug.cgi?id=732144

9 years agovideomixer: Declare as Compositor in 'klass'
Thibault Saunier [Mon, 16 Jun 2014 09:35:39 +0000 (11:35 +0200)]
videomixer: Declare as Compositor in 'klass'

9 years agoflvdemux: fix speex caps
Tim-Philipp Müller [Thu, 26 Jun 2014 12:50:19 +0000 (13:50 +0100)]
flvdemux: fix speex caps

Decoder complains about "notification: Invalid mode encountered.
The stream is corrupted" though, even if it works, so there's
probably something wrong with the generated codec headers.

9 years agoflvmux: fix speex in FLV
Tim-Philipp Müller [Thu, 26 Jun 2014 12:43:33 +0000 (13:43 +0100)]
flvmux: fix speex in FLV

Speex in FLV is always mono @ 16kHz, see
http://download.macromedia.com/f4v/video_file_format_spec_v10_1.pdf
section E.4.2.1: "If the SoundFormat indicates Speex, the audio is
compressed mono sampled at 16 kHz, the SoundRate shall be 0, the
SoundSize shall be 1, and the SoundType shall be 0"

Also see https://bugzilla.gnome.org/show_bug.cgi?id=683622

9 years agoisomp4: Add object type id and fourcc for DTS/DTS-HD
Jan Schmidt [Wed, 25 Jun 2014 19:19:57 +0000 (05:19 +1000)]
isomp4: Add object type id and fourcc for DTS/DTS-HD

Enables playback for files with DTS audio tracks.
Also add an extra AC-3 variant fourcc from Nero

9 years agovideomixer2: Solve segmentation fault when src caps are configured
David Fernandez [Thu, 13 Mar 2014 09:35:30 +0000 (10:35 +0100)]
videomixer2: Solve segmentation fault when src caps are configured

Change function pointers to NULL while holding the lock to avoid
race conditions

https://bugzilla.gnome.org/show_bug.cgi?id=701110

9 years agojitterbuffer: improve SR packet handling
Wim Taymans [Wed, 25 Jun 2014 12:34:21 +0000 (14:34 +0200)]
jitterbuffer: improve SR packet handling

Implement 3 different cases for handling the SR:

 1) we don't have enough timing information to handle the SR packet and
    we need to wait a little for more RTP packets. In that case we keep
    the SR packet around and retry when we get an RTP packet in the
    chain function.

 2) the SR packet has a too old timestamp and should be discarded. It is
    labeled invalid and the last_sr is cleared.

 3) the SR packet is ok and there is enough timing information, proceed
    with processing the SR packet.

Before this patch, case 2) and 1) were handled in the same way,
resulting that SR packets with too old timestamps were checked over and
over again for each RTP packet.

10 years agotests: add udpsink test to check client add/remove
Tim-Philipp Müller [Tue, 24 Jun 2014 09:47:33 +0000 (10:47 +0100)]
tests: add udpsink test to check client add/remove

10 years agotests: port udpsink tests to 1.0
Tim-Philipp Müller [Mon, 23 Jun 2014 15:13:27 +0000 (16:13 +0100)]
tests: port udpsink tests to 1.0

They all seem a bit pointless though.

10 years agoavimux: Add UYVY format
Olivier Crête [Mon, 23 Jun 2014 23:55:29 +0000 (19:55 -0400)]
avimux: Add UYVY format

10 years agogstrtpssrcdemux: manage ssrc of RTCP RR packets
Miguel París Díaz [Fri, 6 Jun 2014 09:20:21 +0000 (11:20 +0200)]
gstrtpssrcdemux: manage ssrc of RTCP RR packets

https://bugzilla.gnome.org/show_bug.cgi?id=731324

10 years agowavparse: Update offset after parsing adtl chunk
Sebastian Dröge [Mon, 23 Jun 2014 18:53:50 +0000 (20:53 +0200)]
wavparse: Update offset after parsing adtl chunk

Otherwise we will parse it over and over again without ever
getting past it.

https://bugzilla.gnome.org/show_bug.cgi?id=731533

10 years agoosxvideosink: remove legacy code for passing a window handle
Andoni Morales Alastruey [Sun, 7 Jul 2013 18:18:27 +0000 (20:18 +0200)]
osxvideosink: remove legacy code for passing a window handle

"have-ns-view" and the "embed" property was kept in 0.10 for
backwards compatibility but it's no longer used in favor of
the GstVideoOverlay interface

https://bugzilla.gnome.org/show_bug.cgi?id=703753

10 years agoBack to development
Sebastian Dröge [Sun, 22 Jun 2014 17:36:14 +0000 (19:36 +0200)]
Back to development

10 years agomatroskademux: Don't call GST_DEBUG_OBJECT() and other macros with non-GObject objects
Sebastian Dröge [Sun, 22 Jun 2014 17:26:03 +0000 (19:26 +0200)]
matroskademux: Don't call GST_DEBUG_OBJECT() and other macros with non-GObject objects

It will crash with latest GLib GIT and was never supposed to work before
either.

10 years agoRelease 1.3.3 1.3.3
Sebastian Dröge [Sun, 22 Jun 2014 16:08:03 +0000 (18:08 +0200)]
Release 1.3.3

10 years agoUpdate .po files
Sebastian Dröge [Sun, 22 Jun 2014 15:36:28 +0000 (17:36 +0200)]
Update .po files

10 years agopo: Update translations
Sebastian Dröge [Sun, 22 Jun 2014 12:24:24 +0000 (14:24 +0200)]
po: Update translations

10 years agopulse, v4l2: update for device "klass" -> "device-class" rename
Tim-Philipp Müller [Sat, 21 Jun 2014 00:32:03 +0000 (01:32 +0100)]
pulse, v4l2: update for device "klass" -> "device-class" rename

10 years agomultiudpsink: optimisation: avoid unnecessary memory ref/unrefs
Tim-Philipp Müller [Fri, 20 Jun 2014 11:21:05 +0000 (12:21 +0100)]
multiudpsink: optimisation: avoid unnecessary memory ref/unrefs

We know the buffer will stay valid and we will also not
modify the buffer, we just want to send out the data.

10 years agomultiudpsink: avoid some unnecessary run-time type checks
Tim-Philipp Müller [Thu, 19 Jun 2014 13:59:48 +0000 (14:59 +0100)]
multiudpsink: avoid some unnecessary run-time type checks

10 years agortspsrc: pass the stream id when asking for crypto params
Wim Taymans [Thu, 19 Jun 2014 14:17:23 +0000 (16:17 +0200)]
rtspsrc: pass the stream id when asking for crypto params

This way the app can choose different parameters for each stream.

10 years agortspsrc: add support for key length parameters
Aleix Conchillo Flaqué [Tue, 20 May 2014 21:58:07 +0000 (14:58 -0700)]
rtspsrc: add support for key length parameters

This patch adds supports for the incoming key management parameters for
encryption and authentication key lengths.

It also adds a new signal request-rtcp-key that allows the user to
provide the crypto parameters and key for the RTCP stream.

https://bugzilla.gnome.org/show_bug.cgi?id=730473

10 years agovp8depay: fix header size checking
Wim Taymans [Thu, 19 Jun 2014 13:25:01 +0000 (15:25 +0200)]
vp8depay: fix header size checking

Use a different variable name to make it clear that we are calculating
the header size.
Correctly check that we have enough bytes to read the header bits. We
were checking if there were 5 bytes available in the header while we
only needed 3, causing the packet to be discarded as too small.

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=723595

10 years agortph264pay: propagate the GST_BUFFER_FLAG_DISCONT flag
Guillaume Desmottes [Tue, 20 May 2014 10:39:31 +0000 (12:39 +0200)]
rtph264pay: propagate the GST_BUFFER_FLAG_DISCONT flag

Similarly to what we did with the DELTA_UNIT flag, this patch
propagates the DISCONT flag to the first RTP packet being used to transfer a
DISCONT buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=730563

10 years agortph264pay: propagate the GST_BUFFER_FLAG_DELTA_UNIT flag
Guillaume Desmottes [Tue, 6 May 2014 15:42:14 +0000 (17:42 +0200)]
rtph264pay: propagate the GST_BUFFER_FLAG_DELTA_UNIT flag

Downstream elements may be interested knowing if a RTP packet is the start
of a key frame (to implement a RTP extension as defined in the
ONVIF Streaming Spec for example).

We do this by checking the GST_BUFFER_FLAG_DELTA_UNIT flag we receive from
upstream and propagate it to the *first* RTP packet outputted to transfer this
buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=730563

10 years agogstrtpmp4gpay: propagate the GST_BUFFER_FLAG_DISCONT flag
Guillaume Desmottes [Tue, 20 May 2014 11:58:20 +0000 (13:58 +0200)]
gstrtpmp4gpay: propagate the GST_BUFFER_FLAG_DISCONT flag

Propagate the DISCONT flag to the first RTP packet being used to transfer
a DISCONT buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=730563

10 years agortpjpegpay: propagate the GST_BUFFER_FLAG_DISCONT flag
Guillaume Desmottes [Tue, 20 May 2014 11:58:20 +0000 (13:58 +0200)]
rtpjpegpay: propagate the GST_BUFFER_FLAG_DISCONT flag

Propagate the DISCONT flag to the first RTP packet being used to transfer
a DISCONT buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=730563

10 years agoavidemux: don't leak flow combiner
Tim-Philipp Müller [Wed, 18 Jun 2014 14:03:25 +0000 (15:03 +0100)]
avidemux: don't leak flow combiner

10 years agortpjp2kpay: pre-allocate buffer-list of the right size
Tim-Philipp Müller [Wed, 18 Jun 2014 13:38:55 +0000 (14:38 +0100)]
rtpjp2kpay: pre-allocate buffer-list of the right size

10 years agortpjpegpay: pre-allocate buffer list of the right size
Tim-Philipp Müller [Wed, 18 Jun 2014 13:34:09 +0000 (14:34 +0100)]
rtpjpegpay: pre-allocate buffer list of the right size

10 years agortpmp4vpay: pre-allocate buffer list of the right size
Tim-Philipp Müller [Wed, 18 Jun 2014 13:19:28 +0000 (14:19 +0100)]
rtpmp4vpay: pre-allocate buffer list of the right size

10 years agortpvp8pay: allocate bitreader on the stack
Tim-Philipp Müller [Wed, 18 Jun 2014 12:44:31 +0000 (13:44 +0100)]
rtpvp8pay: allocate bitreader on the stack

10 years agortpvp8pay: post error message on bus on error and don't use g_message()
Tim-Philipp Müller [Wed, 18 Jun 2014 12:29:47 +0000 (13:29 +0100)]
rtpvp8pay: post error message on bus on error and don't use g_message()

10 years agortpvp8pay: couple of minor optimisations
Tim-Philipp Müller [Wed, 18 Jun 2014 12:20:44 +0000 (13:20 +0100)]
rtpvp8pay: couple of minor optimisations

Pre-allocate buffer list of the right size to avoid re-allocs.
Avoid plenty of double runtime cast checks and re-doing the
same calculation over and over again in rtp_vp8_calc_payload_len().
Only call gst_buffer_get_size() once.

10 years agortpgstpay: pre-allocate buffer list of the right size
Tim-Philipp Müller [Wed, 18 Jun 2014 07:10:03 +0000 (08:10 +0100)]
rtpgstpay: pre-allocate buffer list of the right size

To avoid re-allocs.

10 years agortph264pay: pre-allocate bufferlist of the right size
Tim-Philipp Müller [Wed, 18 Jun 2014 06:52:05 +0000 (07:52 +0100)]
rtph264pay: pre-allocate bufferlist of the right size

To avoid unnecessary re-allocs.

10 years agortph264pay: push single buffer directly, no need to wrap it in a bufferlist
Tim-Philipp Müller [Mon, 16 Jun 2014 19:15:43 +0000 (20:15 +0100)]
rtph264pay: push single buffer directly, no need to wrap it in a bufferlist

No point in a buffer list if we just have one single
buffer to push. Fix up unit test to handle that case
as well.

10 years agortpvrawpay: make chunks per frame configurable
Tim-Philipp Müller [Mon, 16 Jun 2014 14:35:12 +0000 (15:35 +0100)]
rtpvrawpay: make chunks per frame configurable

Bit of a misnomer because it's really chunks per field
and not per frame, but we're going to ignore that for
the time being.

10 years agortpvrawpay: remove unused variables
Tim-Philipp Müller [Mon, 16 Jun 2014 13:52:16 +0000 (14:52 +0100)]
rtpvrawpay: remove unused variables

10 years agortpvrawpay: pre-allocate buffer lists of sufficient size
Tim-Philipp Müller [Mon, 16 Jun 2014 13:44:27 +0000 (14:44 +0100)]
rtpvrawpay: pre-allocate buffer lists of sufficient size

Avoids unnecessary reallocs when appending buffers
to the bufferlist.

10 years agortpvrawpay: micro-optimise variable access in inner loop
Tim-Philipp Müller [Mon, 16 Jun 2014 12:51:03 +0000 (13:51 +0100)]
rtpvrawpay: micro-optimise variable access in inner loop

Store some values that don't change during the execution
of the inner loops locally, so the compiler knows that too.

10 years agortpvrawpay: use buffer lists
Tim-Philipp Müller [Mon, 16 Jun 2014 12:38:47 +0000 (13:38 +0100)]
rtpvrawpay: use buffer lists

Collect buffers to send out in buffer lists instead of
pushing out single buffers one at a time. For HD video
each frame might easily add up to a couple of thousand
packets, multiply that by the frame rate and that's a
lot of push() and sendmsg() calls per second.

A good reason to push out buffers as early as possible is
latency, so we don't accumulate the whole frame in a single
buffer list, but instead push it out in a few chunks, which
is hopefully a reasonable compromise.

10 years agoudp: improve element descriptions for dynudpsink and multiudpsink
Tim-Philipp Müller [Mon, 16 Jun 2014 15:40:07 +0000 (16:40 +0100)]
udp: improve element descriptions for dynudpsink and multiudpsink

10 years agoudp: remove suppression of compiler warnings for deprecated GLib API
Tim-Philipp Müller [Mon, 16 Jun 2014 15:17:16 +0000 (16:17 +0100)]
udp: remove suppression of compiler warnings for deprecated GLib API

Not needed any more.

10 years agovideobox: Fix caps negotiation issue
Ravi Kiran K N [Tue, 17 Jun 2014 07:46:27 +0000 (13:16 +0530)]
videobox: Fix caps negotiation issue

Make sure that if AYUV is received it will detect that it can produce
both RGB and YUV formats

Signed-off-by: Ravi Kiran K N <ravi.kiran@samsung.com>
https://bugzilla.gnome.org/show_bug.cgi?id=725248

10 years agortptheoradepay: fix double frees
Tim-Philipp Müller [Mon, 16 Jun 2014 11:02:41 +0000 (12:02 +0100)]
rtptheoradepay: fix double frees

Fix double-frees introduced to fix another coverity report.

CID 1223053

10 years agodynudpsink: return FLUSHING when sendto got canceled, not an error
Tim-Philipp Müller [Fri, 13 Jun 2014 09:12:07 +0000 (10:12 +0100)]
dynudpsink: return FLUSHING when sendto got canceled, not an error

10 years agooss: simplify probed caps before returning them
Tim-Philipp Müller [Fri, 13 Jun 2014 08:52:03 +0000 (09:52 +0100)]
oss: simplify probed caps before returning them

Exposes all formats in the first structure if the
rest is the same for all of them.

10 years agooss: make sure 16-bit formats are before 8-bit formats in probed caps
Tim-Philipp Müller [Fri, 13 Jun 2014 08:45:28 +0000 (09:45 +0100)]
oss: make sure 16-bit formats are before 8-bit formats in probed caps

Probe supported formats in order of desirability rather than in
what order they may happen to be in the formats bitmask. Fixes
accidentally exposure of 8-bit formats in caps before 16-bit formats
(in case where U16 was not supported S8 might be listed before S16).

https://bugzilla.gnome.org/show_bug.cgi?id=706884

10 years agov4l2bufferpool: Cleanly handle v4l2_allocator_new failure
Nicolas Dufresne [Thu, 12 Jun 2014 20:36:24 +0000 (16:36 -0400)]
v4l2bufferpool: Cleanly handle v4l2_allocator_new failure

10 years agortptheordepay: fix leaks
Vincent Penquerc'h [Thu, 12 Jun 2014 10:24:15 +0000 (11:24 +0100)]
rtptheordepay: fix leaks

Coverity 1212163

10 years agortpg729pay: leak fixes
Vincent Penquerc'h [Thu, 12 Jun 2014 10:16:08 +0000 (11:16 +0100)]
rtpg729pay: leak fixes

Coverity 1212159

10 years agortph263pay: fix leak
Vincent Penquerc'h [Thu, 12 Jun 2014 10:11:38 +0000 (11:11 +0100)]
rtph263pay: fix leak

Coverity 1212157

10 years agortph263pay: fix leaks
Vincent Penquerc'h [Thu, 12 Jun 2014 09:43:53 +0000 (10:43 +0100)]
rtph263pay: fix leaks

Coverity 1212149

10 years agortpdvpay: catch failures to map buffer
Vincent Penquerc'h [Thu, 12 Jun 2014 09:31:47 +0000 (10:31 +0100)]
rtpdvpay: catch failures to map buffer

Coverity 1139741

10 years agomultipartdemux: guard against having no MIME type
Vincent Penquerc'h [Wed, 11 Jun 2014 16:43:42 +0000 (17:43 +0100)]
multipartdemux: guard against having no MIME type

The code would previously crash trying to insert a NULL string
into a hash table.
It does seem a little broken that indexing is done by MIME type
and not by index though, unless the spec says there cannot be
two parts with the same MIME type.

https://bugzilla.gnome.org/show_bug.cgi?id=659573

10 years agomultipartdemux: Send stream-start event
Nicolas Dufresne [Tue, 10 Jun 2014 19:42:14 +0000 (15:42 -0400)]
multipartdemux: Send stream-start event

This event was not sent. Send it before caps, this requires the pad to
be parented. This removes warning like: "Got data flow before
stream-start event".

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=731475

10 years agoqtdemux: avoid looping indefinitely in broken svq3 files
Thiago Santos [Tue, 10 Jun 2014 18:33:33 +0000 (15:33 -0300)]
qtdemux: avoid looping indefinitely in broken svq3 files

Abort if an atom with size 0 is read from within the svq3 stsd
atoms

https://bugzilla.gnome.org/show_bug.cgi?id=726512

10 years agoflacdec: add const where appropriate
Vincent Penquerc'h [Tue, 10 Jun 2014 09:52:23 +0000 (10:52 +0100)]
flacdec: add const where appropriate

10 years agospeexenc: add missing va_end in variadic function
Edward Hervey [Mon, 9 Jun 2014 08:39:20 +0000 (10:39 +0200)]
speexenc: add missing va_end in variadic function

Coverity 1139944

10 years agoflvdemux: Attempt upstream seek first
Edward Hervey [Mon, 9 Jun 2014 08:04:38 +0000 (10:04 +0200)]
flvdemux: Attempt upstream seek first

If we have an upstream element that can handle the seek (such as
rtmpsrc), try to do that first before attempting it ourself.

10 years agowavparse: do not include codec_data on raw audio caps
Vincent Penquerc'h [Wed, 4 Jun 2014 10:34:27 +0000 (11:34 +0100)]
wavparse: do not include codec_data on raw audio caps

If the wav header contains an extended chunk, we want to keep
the codec_data field, but not for raw audio.

This fixes some elements (such as adder) from failing to intersect
raw audio caps which would otherwise be intersectable.

10 years agoflvdemux: Query duration upstream first
Edward Hervey [Thu, 5 Jun 2014 07:38:29 +0000 (09:38 +0200)]
flvdemux: Query duration upstream first

Upstream elements (like rtmpsrc) might be able to provide the duration
more accurately than flvdemux. Especially with index-less vod files