platform/upstream/gstreamer.git
9 years agoosxaudiosrc: no resampling on OS X
Ilya Konstantinov [Wed, 15 Apr 2015 17:13:27 +0000 (20:13 +0300)]
osxaudiosrc: no resampling on OS X

Unlike Remote IO, AUHAL doesn't have built-in resampling
for sources -- confirmed by Core Audio engineer Doug Wyatt:
http://lists.apple.com/archives/coreaudio-api/2006/Sep/msg00088.html

https://bugzilla.gnome.org/show_bug.cgi?id=743758

9 years agoosxaudiosrc: avoid get_channel_layout
Ilya Konstantinov [Wed, 15 Apr 2015 15:29:14 +0000 (18:29 +0300)]
osxaudiosrc: avoid get_channel_layout

This only produces a warning and serves no purpose.

https://bugzilla.gnome.org/show_bug.cgi?id=743758

9 years agoosxaudio: Avoid making a duplicate structure in caps for mono/stereo case
Arun Raghavan [Tue, 7 Apr 2015 10:10:14 +0000 (15:40 +0530)]
osxaudio: Avoid making a duplicate structure in caps for mono/stereo case

For 1ch or 2ch devices, we just need to set the caps to allow both
options since CoreAudio will up/downmix appropriately.

Also fixes the condition for the 2ch case to be exact, rather than at
least 2 channels since the downmix will not take place in the >stereo
case.

9 years agoosxaudio: Don't set the format on an initialized AudioUnit
Arun Raghavan [Mon, 6 Apr 2015 10:52:34 +0000 (16:22 +0530)]
osxaudio: Don't set the format on an initialized AudioUnit

We need to initialize the AudioUnit early to be able to probe the
underlying device, but according to the AudioUnitInitialize() and
AudioUnitUninitialize() documentation, format changes should be done
while the AudioUnit is uninitialized. So we explicitly uninitialize the
AudioUnit during a format change and reinitialize it when we're done.

9 years agoosxaudio: Minor spelling fix (unitialize -> uninitialize)
Arun Raghavan [Mon, 6 Apr 2015 10:25:59 +0000 (15:55 +0530)]
osxaudio: Minor spelling fix (unitialize -> uninitialize)

9 years agoosxaudio: Fix lockup in _audio_unit_property_listener
Ilya Konstantinov [Sat, 21 Mar 2015 18:34:25 +0000 (20:34 +0200)]
osxaudio: Fix lockup in _audio_unit_property_listener

_audio_unit_property_listener is called either from a Core Audio thread
or as a result of a Core Audio API (e.g. AudioUnitInitialize)
from our own thread. In the latter case, osxbuf can be already locked
(GStreamer's mutex is not recursive).

We introduce the flag cached_caps_valid and use it instead of nullifying
cached_caps when we cannot lock on osxbuf.

https://bugzilla.gnome.org/show_bug.cgi?id=743758

9 years agoosxaudio: Invalidate cached caps on format change
Ilya Konstantinov [Thu, 12 Mar 2015 10:15:12 +0000 (12:15 +0200)]
osxaudio: Invalidate cached caps on format change

Listen for changes in hardware stream format and channel layout, and
invalidate cached caps (since they contain the preferred caps).

https://bugzilla.gnome.org/show_bug.cgi?id=743758

9 years agoosxaudio: Overhaul of probing caps
Ilya Konstantinov [Mon, 9 Mar 2015 21:34:06 +0000 (23:34 +0200)]
osxaudio: Overhaul of probing caps

- Probing caps is unified between source and sink
- Hardware stream format is now reported as preferred capabilities
  (dynamically updated when hardware configuration changes)
- Get hardware channel layout from Remote IO just like from HAL
- More comprehensive mapping between AudioChannelLabel and
  GstAudioChannelPosition
- Support for unpositioned channel layouts
- Announce stereo-mono upmixing/downmixing in caps

https://bugzilla.gnome.org/show_bug.cgi?id=743758

9 years agoosxaudio: AudioUnitInitialize on open
Ilya Konstantinov [Mon, 9 Mar 2015 21:15:56 +0000 (23:15 +0200)]
osxaudio: AudioUnitInitialize on open

Call AudioUnitInitialize upon open. Otherwise, we cannot get
(hardware) stream format nor channel layout from the outer scope.

9 years agortp: depayloaders: implement process_rtp_packet() vfunc
Tim-Philipp Müller [Sun, 12 Jul 2015 13:27:15 +0000 (14:27 +0100)]
rtp: depayloaders: implement process_rtp_packet() vfunc

For more optimised RTP packet handling: means we don't
need to map the input buffer again but can just re-use
the mapping the base class has already done.

https://bugzilla.gnome.org/show_bug.cgi?id=750235

9 years agortpvrawdepay: implement process_rtp_packet() vfunc
Tim-Philipp Müller [Wed, 27 May 2015 18:19:27 +0000 (19:19 +0100)]
rtpvrawdepay: implement process_rtp_packet() vfunc

For more optimised RTP packet handling: means we don't
need to map the input buffer again but can just re-use
the map the base class has already done.

https://bugzilla.gnome.org/show_bug.cgi?id=750235

9 years agortpjitterbuffer: Fix indention
Sebastian Dröge [Thu, 9 Jul 2015 21:13:32 +0000 (00:13 +0300)]
rtpjitterbuffer: Fix indention

9 years agortpjitterbuffer: Always estimate DTS from the current clock time
Sebastian Dröge [Thu, 9 Jul 2015 20:59:10 +0000 (23:59 +0300)]
rtpjitterbuffer: Always estimate DTS from the current clock time

Estimating it from the RTP time will give us the PTS, so in cases of PTS!=DTS
we would produce wrong DTS. As now the estimated DTS is based on the clock,
don't store it in the jitterbuffer items as it would otherwise be used in the
skew calculations and would influence the results. We only really need the DTS
for timer calculations.

https://bugzilla.gnome.org/show_bug.cgi?id=749536

9 years agogitignore: ignore rtph263 test
Thiago Santos [Thu, 9 Jul 2015 12:26:09 +0000 (09:26 -0300)]
gitignore: ignore rtph263 test

9 years agortpjitterbuffer: fix build error with gcc (Debian 4.9.2-21) 4.9.2
Thiago Santos [Thu, 9 Jul 2015 02:47:44 +0000 (23:47 -0300)]
rtpjitterbuffer: fix build error with gcc (Debian 4.9.2-21) 4.9.2

Replace static constants with macros to make gcc happy

  CC       elements/elements_rtpjitterbuffer-rtpjitterbuffer.o
elements/rtpjitterbuffer.c:387:1: error: initializer element is not constant
 static const GstClockTime PCMU_BUF_DURATION = PCMU_BUF_MS * GST_MSECOND;
 ^
elements/rtpjitterbuffer.c:388:1: error: initializer element is not constant
 static const guint PCMU_BUF_SIZE = 64000 * PCMU_BUF_MS / 1000;
 ^
elements/rtpjitterbuffer.c:390:5: error: initializer element is not constant
     PCMU_BUF_CLOCK_RATE * PCMU_BUF_MS / 1000;

9 years agortpjitterbuffer: run indent and fix some comments
Thiago Santos [Thu, 9 Jul 2015 02:40:45 +0000 (23:40 -0300)]
rtpjitterbuffer: run indent and fix some comments

Fix indent on this file and break some comment lines into two to make
it fit 80 chars per line

9 years agoqtdemux: rework segment event handling for adaptive streaming
Thiago Santos [Wed, 8 Jul 2015 18:02:24 +0000 (15:02 -0300)]
qtdemux: rework segment event handling for adaptive streaming

When a new time segment is received upstream is going to restart
with a new atom. Make the neededbytes and todrop variables
reflect that to avoid waiting too much or dropping the
initial bytes that contain the header.

9 years agoqtdemux: push data from adapter before starting new segment
Thiago Santos [Wed, 8 Jul 2015 15:35:55 +0000 (12:35 -0300)]
qtdemux: push data from adapter before starting new segment

The adapter might have data remaining from the previous segment,
push it all before clearing the adapter and starting a new segment.

It can accumulate data if it had pushed and got not-linked, returning
immediately without processing all the data. Before starting a new
segment this data should be handled.

9 years agortpjitterbuffer: Calculate DTS from the clock if we had none for the first packet...
Sebastian Dröge [Wed, 8 Jul 2015 16:59:13 +0000 (19:59 +0300)]
rtpjitterbuffer: Calculate DTS from the clock if we had none for the first packet after a reset

https://bugzilla.gnome.org/show_bug.cgi?id=749536

9 years agortpjitterbuffer: fix gap-time calculation and remove "late"
Havard Graff [Wed, 8 Jul 2015 19:08:36 +0000 (21:08 +0200)]
rtpjitterbuffer: fix gap-time calculation and remove "late"

The amount of time that is completely expired and not worth waiting for,
is the duration of the packets in the gap (gap * duration) - the
latency (size) of the jitterbuffer (priv->latency_ns). This is the duration
that we make a "multi-lost" packet for.

The "late" concept made some sense in 0.10 as it reflected that a buffer
coming in had not been waited for at all, but had a timestamp that was
outside the jitterbuffer to wait for. With the rewrite of the waiting
(timeout) mechanism in 1.0, this no longer makes any sense, and the
variable no longer reflects anything meaningful (num > 0 is useless,
the duration is what matters)

Fixed up the tests that had been slightly modified in 1.0 to allow faulty
behavior to sneak in, and port some of them to use GstHarness.

https://bugzilla.gnome.org/show_bug.cgi?id=738363

9 years agoRevert "rtpjitterbuffer: Fix expected_dts calc in calculate_expected"
Stian Selnes [Tue, 30 Jun 2015 09:21:31 +0000 (11:21 +0200)]
Revert "rtpjitterbuffer: Fix expected_dts calc in calculate_expected"

This reverts commit 05bd708fc5e881390fe839803b53144393d95ab0.

The reverted patch is wrong and introduces a regression because there
may still be time to receive some of the packets included in the gap
if they are reordered.

9 years agoqtdemux: flush samples before adding more from moof
Thiago Santos [Wed, 8 Jul 2015 02:53:02 +0000 (23:53 -0300)]
qtdemux: flush samples before adding more from moof

Avoids accumulating all samples from a fragmented stream that could
lead to a 'index-too-big' error once it goes over 50MB of data. It
could reach that before 2h of playback so it doesn't take that long.

As upstream elements are providing data in time format they should
be the ones that have more information about the full media index
and should be able to seek if possible.

9 years agoqtdemux: rename upstream_newsegment to upstream_format_is_time
Thiago Santos [Wed, 8 Jul 2015 02:56:12 +0000 (23:56 -0300)]
qtdemux: rename upstream_newsegment to upstream_format_is_time

upstream_newsegment isn't really clear on what it means, it is set
to TRUE when the upstream element sends a segment in TIME format, so
rename it to be more clear about it.

It is important to know this because it means that upstream has
a notion of time and qtdemux is likely being driven by an upstream
element that is reading from a higher level abstraction than a file,
such as a DASH, MSS or DLNA element.

9 years agoqtdemux: fix leak by flushing previous sample info from trak
Thiago Santos [Wed, 8 Jul 2015 00:31:08 +0000 (21:31 -0300)]
qtdemux: fix leak by flushing previous sample info from trak

In fragmented streaming, multiple moov/moof will be parsed and their
previously stored samples array might leak when new values are parsed.
The parse_trak and callees won't free the previously stored values
before parsing the new ones.

In step-by-step, this is what happens:

1) initial moov is parsed, traks as well, streams are created. The
   trak doesn't contain samples because they are in the moof's trun
   boxes. n_samples is set to 0 while parsing the trak and the samples
   array is still NULL.
2) moofs are parsed, and their trun boxes will increase n_samples and
   create/extend the samples array
3) At some point a new moov might be sent (bitrate switching, for example)
   and parsing the trak will overwrite n_samples with the values from
   this trak. If the n_samples is set to 0 qtdemux will assume that
   the samples array is NULL and will leak it when a new one is
   created for the subsequent moofs.

This patch makes qtdemux properly free previous sample data before
creating new ones and adds an assert to catch future occurrences of
this issue when the code changes.

9 years agoqtdemux: fix index size check and debug message
Thiago Santos [Tue, 7 Jul 2015 19:46:33 +0000 (16:46 -0300)]
qtdemux: fix index size check and debug message

It is allocating samples_count + n_samples, not only n_samples

9 years agortpjitterbuffer: Calculate receive time if we don't have any
Sebastian Dröge [Wed, 8 Jul 2015 14:02:05 +0000 (17:02 +0300)]
rtpjitterbuffer: Calculate receive time if we don't have any

This is required to properly schedule packet loss timers and make
sure all our calculations work properly.

https://bugzilla.gnome.org/show_bug.cgi?id=749536

9 years agortpjitterbuffer: Handle seqnum gaps in TCP streams without erroring out or overflowin...
Sebastian Dröge [Wed, 8 Jul 2015 12:13:17 +0000 (15:13 +0300)]
rtpjitterbuffer: Handle seqnum gaps in TCP streams without erroring out or overflowing calculations

That is, handle DTS==GST_CLOCK_TIME_NONE correctly.

https://bugzilla.gnome.org/show_bug.cgi?id=749536

9 years agoavidemux: fix event leak
Vineeth T M [Wed, 8 Jul 2015 11:31:42 +0000 (20:31 +0900)]
avidemux: fix event leak

when seek fails in avidemux, event is not being freed.

https://bugzilla.gnome.org/show_bug.cgi?id=752117

9 years agortph263depay: Make sure payload is large enough
Stian Selnes [Wed, 8 Jul 2015 10:02:22 +0000 (12:02 +0200)]
rtph263depay: Make sure payload is large enough

Plus new unit test.

https://bugzilla.gnome.org/show_bug.cgi?id=752112

9 years agortpklvdepay: fix printf format compiler warning
Vineeth TM [Tue, 7 Jul 2015 23:59:49 +0000 (08:59 +0900)]
rtpklvdepay: fix printf format compiler warning

v_len is of type guint64, but while print the value(16 + len_size + v_len)
G_GSIZE_FORMAT is being used instead of G_GUINT64_FORMAT

https://bugzilla.gnome.org/show_bug.cgi?id=752100

9 years agodocs: add new RTP elements to docs
Tim-Philipp Müller [Tue, 7 Jul 2015 19:25:47 +0000 (20:25 +0100)]
docs: add new RTP elements to docs

9 years agotests: rtp-payloading: add basic unit test for KLV payloading
Tim-Philipp Müller [Tue, 7 Jul 2015 19:07:31 +0000 (20:07 +0100)]
tests: rtp-payloading: add basic unit test for KLV payloading

Also make it so that the mtu is always set if specified, not
only in case of the rather weird bufferlist test code path.
This allows us to easily make the payloader fragment a payload
across multiple output packets by setting a small MTU on it.

9 years agortpklvdepay: improve start detection and handle fragmented KLV units
Tim-Philipp Müller [Tue, 7 Jul 2015 18:58:42 +0000 (19:58 +0100)]
rtpklvdepay: improve start detection and handle fragmented KLV units

9 years agortp: add SMPTE 336M KLV metadata depayloader
Tim-Philipp Müller [Sun, 5 Jul 2015 19:25:10 +0000 (20:25 +0100)]
rtp: add SMPTE 336M KLV metadata depayloader

http://tools.ietf.org/html/rfc6597

9 years agortp: add SMPTE 336M KLV metadata payloader
Tim-Philipp Müller [Sat, 9 Aug 2014 09:08:42 +0000 (10:08 +0100)]
rtp: add SMPTE 336M KLV metadata payloader

http://tools.ietf.org/html/rfc6597

9 years agodocs: fix "Symbol name not found at the start of the comment block"
Stefan Sauer [Tue, 7 Jul 2015 14:59:20 +0000 (16:59 +0200)]
docs: fix "Symbol name not found at the start of the comment block"

Add symbols or change comment into a regular comment.

9 years agodocs: remove outdated doc strings
Stefan Sauer [Tue, 7 Jul 2015 14:58:53 +0000 (16:58 +0200)]
docs: remove outdated doc strings

9 years agodocs: add missing plugins and ensure master doc is sorted
Stefan Sauer [Fri, 3 Jul 2015 21:10:40 +0000 (23:10 +0200)]
docs: add missing plugins and ensure master doc is sorted

9 years agoRevert "imagefreeze: Remove impossible error condition"
Luis de Bethencourt [Tue, 7 Jul 2015 14:54:41 +0000 (15:54 +0100)]
Revert "imagefreeze: Remove impossible error condition"

This reverts commit d46631c5c7312ad613397f8238c7a9714ae3ae94.

pad only handle EOS events but not EOS flow, and will push the buffer again
resulting in an assertion error. So we should not handle the buffer
and return EOS flow.

9 years agortpg729depay: unmap rtp buffer in error path
Tim-Philipp Müller [Tue, 7 Jul 2015 14:50:50 +0000 (15:50 +0100)]
rtpg729depay: unmap rtp buffer in error path

9 years agortpg729pay: fix buffer leak
Tim-Philipp Müller [Tue, 7 Jul 2015 14:48:40 +0000 (15:48 +0100)]
rtpg729pay: fix buffer leak

The handle_buffer vfunc takes ownership of the input buffer.
Fixes elements/rtp-payloading under valgrind.

9 years agogoom: Initialised variables to remove compiler warnings
Tobias Mueller [Thu, 2 Jul 2015 06:52:43 +0000 (08:52 +0200)]
goom: Initialised variables to remove compiler warnings

goom_core.c: In function 'goom_update':
goom_core.c:685:5: error: 'param2' may be used uninitialized in this function [-Werror=maybe-uninitialized]
     goom_lines_switch_to (goomInfo->gmline2, mode, param2, amplitude, couleur);
     ^
goom_core.c:684:5: error: 'param1' may be used uninitialized in this function [-Werror=maybe-uninitialized]
     goom_lines_switch_to (goomInfo->gmline1, mode, param1, amplitude, couleur);
     ^

https://bugzilla.gnome.org/show_bug.cgi?id=752053

9 years agortph261pay: fix indentation
Tim-Philipp Müller [Tue, 7 Jul 2015 08:18:39 +0000 (09:18 +0100)]
rtph261pay: fix indentation

9 years agortph261pay: Fix uninitialized variable compiler error
Jimmy Ohn [Mon, 6 Jul 2015 10:11:00 +0000 (19:11 +0900)]
rtph261pay: Fix uninitialized variable compiler error

endpos variable does not correctly understand in the
4.6.3 GCC version. So compile error appears when we do
compile rtph261pay using jhbuild.
This patch is fixed the compile error in 4.6.3 GCC version.

https://bugzilla.gnome.org/show_bug.cgi?id=751985

9 years agoflvdemux: Handle seek flags properly
Jan Alexander Steffens (heftig) [Wed, 12 Nov 2014 11:08:58 +0000 (12:08 +0100)]
flvdemux: Handle seek flags properly

Allows for non-keyframe seeks.

https://bugzilla.gnome.org/show_bug.cgi?id=738570

9 years agoqtdemux: avoid looping reading the 'moof' atom forever
Thiago Santos [Tue, 24 Feb 2015 13:50:52 +0000 (10:50 -0300)]
qtdemux: avoid looping reading the 'moof' atom forever

It gets stuck if it only finds a moof and no mfra/mfro or moov
atoms. Skip the moof to continue the parsing to have it either
play or error out.

https://bugzilla.gnome.org/show_bug.cgi?id=745089

9 years agoflacdec: improve error handling
Vineeth TM [Fri, 26 Jun 2015 04:24:17 +0000 (13:24 +0900)]
flacdec: improve error handling

for files which have corrupted header, libflac is not able to
process the metadata properly. We just try to ignore the error
and continue with the processing, since metadata parsing is not
making much of a difference to libflac

https://bugzilla.gnome.org/show_bug.cgi?id=751334

9 years agoximagesrc: add meta transform function
Hyunjun Ko [Mon, 6 Jul 2015 11:16:38 +0000 (20:16 +0900)]
ximagesrc: add meta transform function

ximage metadata can't be transformed or copied, but provide an empty
transformation function instead of NULL to allow unconditional calling
of metas' transform functions.

https://bugzilla.gnome.org/show_bug.cgi?id=751778

9 years agortph263pdepay: init debug category
Stian Selnes [Mon, 16 Jun 2014 14:14:28 +0000 (16:14 +0200)]
rtph263pdepay: init debug category

https://bugzilla.gnome.org/show_bug.cgi?id=752012

9 years agortpv8depay: ignore reserved bit in payload descriptor
Stian Selnes [Fri, 20 Jun 2014 08:59:14 +0000 (10:59 +0200)]
rtpv8depay: ignore reserved bit in payload descriptor

Draft 16 of "RTP Payload Format for VP8" states in section 4.2 that:

R: Bit reserved for future use.  MUST be set to zero and MUST be
   ignored by the receiver.

https://bugzilla.gnome.org/show_bug.cgi?id=751929

9 years agortph261pay: rtph261depay: Add documentation
Stian Selnes [Sat, 4 Jul 2015 18:56:42 +0000 (20:56 +0200)]
rtph261pay: rtph261depay: Add documentation

https://bugzilla.gnome.org/show_bug.cgi?id=751982

9 years agoAutomatic update of common submodule
Stefan Sauer [Fri, 3 Jul 2015 19:58:14 +0000 (21:58 +0200)]
Automatic update of common submodule

From f74b2df to 9aed1d7

9 years agortph261pay: Fix compiler warning
Sebastian Dröge [Fri, 3 Jul 2015 12:29:16 +0000 (14:29 +0200)]
rtph261pay: Fix compiler warning

gstrtph261pay.c: In function 'gst_rtp_h261_pay_class_init':
gstrtph261pay.c:1003:17: error: variable 'gobject_class' set but not used [-Werror=unused-but-set-variable]
   GObjectClass *gobject_class;

9 years agortph261depay: Let the base class push the buffer so it can deal with the flow return
Sebastian Dröge [Fri, 3 Jul 2015 12:03:05 +0000 (14:03 +0200)]
rtph261depay: Let the base class push the buffer so it can deal with the flow return

9 years agortph261pay: Remove unused adapter
Sebastian Dröge [Fri, 3 Jul 2015 12:11:35 +0000 (14:11 +0200)]
rtph261pay: Remove unused adapter

9 years agospeexpay: Directly attach payload to the output buffer instead of copying it
Sebastian Dröge [Fri, 3 Jul 2015 11:17:24 +0000 (13:17 +0200)]
speexpay: Directly attach payload to the output buffer instead of copying it

9 years agosbcpay: Attach payload directly to the output instead of copying
Sebastian Dröge [Fri, 3 Jul 2015 11:07:20 +0000 (13:07 +0200)]
sbcpay: Attach payload directly to the output instead of copying

9 years agortp: add H.261 RTP payloader and depayloader
Stian Selnes [Mon, 1 Dec 2014 13:18:40 +0000 (14:18 +0100)]
rtp: add H.261 RTP payloader and depayloader

Implementation according to RFC 4587.

Payloader create fragments on MB boundaries in order to match MTU size
the best it can. Some decoders/depayloaders in the wild are very strict
about receiving a continuous bit-stream (e.g. no no-op bits between
frames), so the payloader will shift the compressed bit-stream of a
frame to align with the last significant bit of the previous frame.

Depayloader does not try to be fancy in case of packet loss. It simply
drops all packets for a frame if there is a loss, keeping it simple.

https://bugzilla.gnome.org/show_bug.cgi?id=751886

9 years agortpmpvdepay: Don't forget to unmap the input buffer
Sebastian Dröge [Fri, 3 Jul 2015 10:18:52 +0000 (12:18 +0200)]
rtpmpvdepay: Don't forget to unmap the input buffer

9 years agortpmpvpay: Create buffer lists instead of pushing each buffer individually
Sebastian Dröge [Fri, 3 Jul 2015 10:14:47 +0000 (12:14 +0200)]
rtpmpvpay: Create buffer lists instead of pushing each buffer individually

9 years agortpmpapay: Use buffer lists instead of pushing each fragment individually
Sebastian Dröge [Fri, 3 Jul 2015 10:03:59 +0000 (12:03 +0200)]
rtpmpapay: Use buffer lists instead of pushing each fragment individually

9 years agortpmp4apay: Create buffer lists and don't copy payload memory
Sebastian Dröge [Fri, 3 Jul 2015 08:51:57 +0000 (10:51 +0200)]
rtpmp4apay: Create buffer lists and don't copy payload memory

9 years agortpjitterbuffer: Consider timers len to compare with RTP_MAX_DROPOUT
Miguel París Díaz [Mon, 29 Jun 2015 14:14:18 +0000 (16:14 +0200)]
rtpjitterbuffer: Consider timers len to compare with RTP_MAX_DROPOUT

When there are a lot of small gaps, we can consider that there is
a big gap (too losses) to reset the buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=751636

9 years agortpjitterbuffer: If possible, always update the current time before looping over...
Sebastian Dröge [Mon, 29 Jun 2015 13:53:52 +0000 (15:53 +0200)]
rtpjitterbuffer: If possible, always update the current time before looping over all timers

If we have a clock, update "now" now with the very latest running time we have.
If timers are unscheduled below we otherwise wouldn't update now (it's only updated
when timers expire), and also for the very first loop iteration now would otherwise
always be 0.

Also the time is used for the timeout functions, e.g. to calculate any times
for the next timeouts and we would otherwise pass too old times there.

https://bugzilla.gnome.org/show_bug.cgi?id=751636

9 years agov4l2transform: fix memory leak
Luis de Bethencourt [Thu, 2 Jul 2015 13:34:57 +0000 (14:34 +0100)]
v4l2transform: fix memory leak

tmp needs to be freed before going out of scope in 'done'.

CID #1308954

9 years agortph263ppay: Generate buffer lists and attach the payload directly instead of copying it
Sebastian Dröge [Thu, 2 Jul 2015 10:23:45 +0000 (12:23 +0200)]
rtph263ppay: Generate buffer lists and attach the payload directly instead of copying it

9 years agortph263pdepay: Simplify code a bit and do less direct memcpy and let GstBuffer do...
Sebastian Dröge [Thu, 2 Jul 2015 07:48:02 +0000 (09:48 +0200)]
rtph263pdepay: Simplify code a bit and do less direct memcpy and let GstBuffer do that for us

9 years agortph263pay: Stop using an adapter and directly use the buffer
Sebastian Dröge [Thu, 2 Jul 2015 07:17:59 +0000 (09:17 +0200)]
rtph263pay: Stop using an adapter and directly use the buffer

We always pushed one buffer into the adapter, then handled exactly that one
buffer and flushed it from the adapter. Now also don't memcpy() the actual
payload but just attach the input buffer's data to the output buffer.

This code still needs some serious refactoring/rewriting.

9 years agortpgsmpay: Remove non-existing includes for now
Sebastian Dröge [Wed, 1 Jul 2015 19:57:28 +0000 (21:57 +0200)]
rtpgsmpay: Remove non-existing includes for now

git add -p mistake.

9 years agortpgstpay: Use the return value of gst_buffer_append()
Sebastian Dröge [Wed, 1 Jul 2015 17:29:07 +0000 (19:29 +0200)]
rtpgstpay: Use the return value of gst_buffer_append()

9 years agortpgsmpay: Attach payload to the output buffer instead of copying it
Sebastian Dröge [Wed, 1 Jul 2015 17:19:13 +0000 (19:19 +0200)]
rtpgsmpay: Attach payload to the output buffer instead of copying it

9 years agortpg729pay: Attach payload directly to output buffers instead of copying
Sebastian Dröge [Wed, 1 Jul 2015 15:58:56 +0000 (17:58 +0200)]
rtpg729pay: Attach payload directly to output buffers instead of copying

9 years agortpg723pay: Attach payload buffer to the output instead of copying
Sebastian Dröge [Wed, 1 Jul 2015 15:43:51 +0000 (17:43 +0200)]
rtpg723pay: Attach payload buffer to the output instead of copying

9 years agortpdvdepay: Map the output buffer once instead of once every 80 bytes
Sebastian Dröge [Wed, 1 Jul 2015 15:30:39 +0000 (17:30 +0200)]
rtpdvdepay: Map the output buffer once instead of once every 80 bytes

9 years agoavidemux: fix return type of index_entry_offset_search()
Jimmy Ohn [Wed, 1 Jul 2015 12:46:46 +0000 (21:46 +0900)]
avidemux: fix return type of index_entry_offset_search()

It's a compare function and may return a negative value,
so should for correctness and consistency return a signed
integer.

https://bugzilla.gnome.org/show_bug.cgi?id=751780

9 years agortpjitterbuffer: refactor handle_next_buffer
Miguel París Díaz [Wed, 1 Jul 2015 12:12:57 +0000 (14:12 +0200)]
rtpjitterbuffer: refactor handle_next_buffer

The goal of this patch is making handle_next_buffer function
more readable avoiding unnecesary gotos and adding other
cosmetic changes.

9 years agortpac3pay: Attach the payload to the output buffer instead of copying it
Sebastian Dröge [Wed, 1 Jul 2015 13:40:25 +0000 (15:40 +0200)]
rtpac3pay: Attach the payload to the output buffer instead of copying it

Might also want to produce buffer lists here if needed.

9 years agortp: Fix indention
Sebastian Dröge [Wed, 1 Jul 2015 13:38:47 +0000 (15:38 +0200)]
rtp: Fix indention

9 years agortp: Add examples with VTS/ATS for VP8/OPUS
Sebastian Dröge [Wed, 1 Jul 2015 10:37:11 +0000 (12:37 +0200)]
rtp: Add examples with VTS/ATS for VP8/OPUS

Let's have an example with modern codecs.

9 years agortph264pay: Use GST_WARNING_OBJECT() instead of GST_WARNING()
Sebastian Dröge [Tue, 30 Jun 2015 16:11:33 +0000 (18:11 +0200)]
rtph264pay: Use GST_WARNING_OBJECT() instead of GST_WARNING()

9 years agovp8depay: Don't lock/map every non-keyframe buffer twice
Sebastian Dröge [Tue, 30 Jun 2015 12:06:20 +0000 (14:06 +0200)]
vp8depay: Don't lock/map every non-keyframe buffer twice

Just copy the complete header instead of first looking at the first byte
and then at the remaining 10 bytes.

9 years agov4l2: document fallthrough cases
Luis de Bethencourt [Mon, 29 Jun 2015 15:05:44 +0000 (16:05 +0100)]
v4l2: document fallthrough cases

Pacify coverity and document fallthrough cases in switch statements.

CID #1308948, #1308947, #1308946

9 years agoRevert "rtpjitterbuffer: If we have an immediate timeout, don't try to find an earlie...
Sebastian Dröge [Mon, 29 Jun 2015 08:36:58 +0000 (10:36 +0200)]
Revert "rtpjitterbuffer: If we have an immediate timeout, don't try to find an earlier timeout"

This reverts commit 0c21cd7177ea883c710999147ddcedb19004d182.

If we have multiple immediate timers, we want to first handle the one with the
lowest sequence number... which would be broken now.

Instead of this we should just use a GSequence for the timers, and have them
sorted first by timestamp, and for equal timestamps by sequence number. Then
we would always only have to take the very first timer from the list and never
have to look at any others.

9 years agortpjitterbuffer: If we have an immediate timeout, don't try to find an earlier timeout
Sebastian Dröge [Mon, 29 Jun 2015 08:14:05 +0000 (10:14 +0200)]
rtpjitterbuffer: If we have an immediate timeout, don't try to find an earlier timeout

If we have lots of such immediate timeouts, we would otherwise have quadratic
runtime in the number of timeouts.

9 years agosplitmuxsrc: sticky events are sent automatically from the pad
Thiago Santos [Fri, 19 Jun 2015 21:01:03 +0000 (18:01 -0300)]
splitmuxsrc: sticky events are sent automatically from the pad

No need to send them explicitly from the element

https://bugzilla.gnome.org/show_bug.cgi?id=751240

9 years agosplitmuxsrc: make sure to push sticky events before adding pad
Thiago Santos [Fri, 19 Jun 2015 21:00:40 +0000 (18:00 -0300)]
splitmuxsrc: make sure to push sticky events before adding pad

It allows the caps to be set on the pad before being added for
dynamic autoplugging to work.

https://bugzilla.gnome.org/show_bug.cgi?id=751240

9 years agortspsrc: Add new ntp-time-source property and deprecate use-pipeline-clock property
Hyunjun Ko [Thu, 25 Jun 2015 15:05:29 +0000 (00:05 +0900)]
rtspsrc: Add new ntp-time-source property and deprecate use-pipeline-clock property

Enable to use new ntp-time-source property of rtpbin

https://bugzilla.gnome.org/show_bug.cgi?id=751496

9 years agortpbin/session: fix description
Hyunjun Ko [Thu, 25 Jun 2015 14:19:58 +0000 (23:19 +0900)]
rtpbin/session: fix description

https://bugzilla.gnome.org/show_bug.cgi?id=751496

9 years agodocs: decodebin2 -> decodebin
Luis de Bethencourt [Thu, 25 Jun 2015 09:57:25 +0000 (10:57 +0100)]
docs: decodebin2 -> decodebin

9 years agodeinterlace: update example pipeline
Luis de Bethencourt [Thu, 25 Jun 2015 09:47:06 +0000 (10:47 +0100)]
deinterlace: update example pipeline

Update reference to decodebin2 to decodebin

9 years agodeinterlace: remove dead assignments
Luis de Bethencourt [Thu, 25 Jun 2015 09:45:35 +0000 (10:45 +0100)]
deinterlace: remove dead assignments

Values in fields_required and same_buffer are overwritten before used. Removing
assignment

9 years agomikmod: remove ancient unported plugin
Tim-Philipp Müller [Thu, 25 Jun 2015 09:06:07 +0000 (10:06 +0100)]
mikmod: remove ancient unported plugin

This hasn't been touched in 11 years, and
clearly no one's been missing it.

9 years agoqtdemux: does not detect orientation
Gilbok Lee [Tue, 23 Jun 2015 11:15:13 +0000 (20:15 +0900)]
qtdemux: does not detect orientation

Most files don't contain the values for transposing the coordinates
back to the positive quadrant so qtdemux was ignoring the rotation
tag. To be able to properly handle those files qtdemux will also ignore
the transposing values to only detect the rotation using the values
abde from the transformation matrix:

[a b c]
[d e f]
[g h i]

https://bugzilla.gnome.org/show_bug.cgi?id=738681

9 years agoBack to development
Sebastian Dröge [Wed, 24 Jun 2015 22:04:16 +0000 (00:04 +0200)]
Back to development

9 years agoRelease 1.5.2
Sebastian Dröge [Wed, 24 Jun 2015 21:30:41 +0000 (23:30 +0200)]
Release 1.5.2

9 years agoUpdate .po files
Sebastian Dröge [Wed, 24 Jun 2015 20:56:12 +0000 (22:56 +0200)]
Update .po files

9 years agopo: Update translations
Sebastian Dröge [Wed, 24 Jun 2015 09:15:00 +0000 (11:15 +0200)]
po: Update translations

9 years agoqtmux: Correctly test each segments
Nicolas Dufresne [Tue, 23 Jun 2015 22:42:59 +0000 (18:42 -0400)]
qtmux: Correctly test each segments

In presence of gaps, qtdemux will emit multiple segments. The
second segment start should match the CTTS.

https://bugzilla.gnome.org/show_bug.cgi?id=751361

9 years agoqtmux: Correctly calculate the elst media start
Nicolas Dufresne [Tue, 23 Jun 2015 21:54:31 +0000 (17:54 -0400)]
qtmux: Correctly calculate the elst media start

The media start has nothing to do with the shift we have applied
but with the value of the first PTS. This is defined as:

  Dt(0) = 0
  Ct(0) = Dt(0) + CTTS(0)

So the media start is always the first CTTS.

https://bugzilla.gnome.org/show_bug.cgi?id=751361

9 years agoqtdemux: accumulate previous edts entries into segment.base
Thiago Santos [Tue, 23 Jun 2015 14:49:32 +0000 (11:49 -0300)]
qtdemux: accumulate previous edts entries into segment.base

Allows playing edts editted files with proper synchronization of
streams. This patch fixes the regression introduced by
bf95f93c0189aa04f18e264b86b6527e431c5d53 that was added to fix
segment seeks handling.

Having the accumulated_base separated from the main segment.base
allows handling both segment seeks and edts editted files.

https://bugzilla.gnome.org/show_bug.cgi?id=751361