platform/upstream/gstreamer.git
13 years agobaseparse: Fix slightly broken buffer-in-segment check (aacparse, amrparse, flacparse)
René Stadler [Tue, 7 Apr 2009 01:53:02 +0000 (04:53 +0300)]
baseparse: Fix slightly broken buffer-in-segment check (aacparse, amrparse, flacparse)

13 years agobaseparse: Fix push mode seeking (aacparse, amrparse)
René Stadler [Sun, 5 Apr 2009 00:50:19 +0000 (03:50 +0300)]
baseparse: Fix push mode seeking (aacparse, amrparse)

Sending the flush-start event forward before taking the stream lock actually
works, in contrast to deadlocking in downstream preroll_wait (hunk 1).

After that we get the chain function being stuck in a busy loop. This is fixed
by updating the minimum frame size inside the synchronization loop because the
subclass asks for more data in this way (hunk 2).

Finally, this leads to a very probable crash because the subclass can find a
valid frame with a size greater than the currently available data in the
adapter. This makes the subsequent gst_adapter_take_buffer call return NULL,
which is not expected (hunk 3).

13 years agobaseparse: Delay newsegment as long as possible.
Mark Nauwelaerts [Tue, 31 Mar 2009 14:07:46 +0000 (16:07 +0200)]
baseparse: Delay newsegment as long as possible.

If newsegment is sent (too) early, caps may not yet be fixed/set,
and downstream may not have been linked.

13 years agoaacparse: Fix busyloop when seeking. Fixes #575388
René Stadler [Wed, 18 Mar 2009 23:17:25 +0000 (01:17 +0200)]
aacparse: Fix busyloop when seeking. Fixes #575388

The problem is that after a discont, set_min_frame_size(1024) is called when
detect_stream returns FALSE. However, detect_stream calls check_adts_frame
which sets the frame size on its own to something larger than 1024. This is the
same situation as in the beginning, so the base class ends up calling
check_valid_frame in an endless loop.

13 years agoaacparse: Refactor check_valid_frame to expose broken code
René Stadler [Wed, 18 Mar 2009 22:32:40 +0000 (00:32 +0200)]
aacparse: Refactor check_valid_frame to expose broken code

Just moving code around and removing an unhelpful/misleading comment.

13 years agobaseparse: revert last change and properly fix
Stefan Kost [Fri, 27 Feb 2009 09:24:37 +0000 (11:24 +0200)]
baseparse: revert last change and properly fix

Baseparse internaly breaks the semantics of a _chain function by calling it with
buffer==NULL. The reson I belived it was okay to remove it was that there is
also an unchecked access to buffer later in _chain. Actually that code is wrong,
as it most probably wants to set discont on the outgoing buffer.

13 years agobaseparse: remove checks for buffer==NULL
Stefan Kost [Thu, 26 Feb 2009 09:02:06 +0000 (11:02 +0200)]
baseparse: remove checks for buffer==NULL

Accordifn to docs for GstPadChainFunction buffer cannot be NULL. If we would
leave the check, we would also need more such check below.

13 years agoaacparse: Fix license specified in plugin details.
René Stadler [Tue, 10 Feb 2009 22:15:43 +0000 (00:15 +0200)]
aacparse: Fix license specified in plugin details.

13 years agoFix the return value of the default parse_frame function.
Jan Schmidt [Fri, 30 Jan 2009 18:18:10 +0000 (18:18 +0000)]
Fix the return value of the default parse_frame function.

Fix the return value of the default parse_frame function in both
copies of GstBaseParse

13 years agoLog aac details found in codec_data.
Stefan Kost [Fri, 23 Jan 2009 14:00:10 +0000 (16:00 +0200)]
Log aac details found in codec_data.

13 years agogst/aacparse/gstaacparse.c: Don't autoplug aacparse until it works.
Wim Taymans [Thu, 13 Nov 2008 17:24:58 +0000 (17:24 +0000)]
gst/aacparse/gstaacparse.c: Don't autoplug aacparse until it works.

Original commit message from CVS:
* gst/aacparse/gstaacparse.c: (plugin_init):
Don't autoplug aacparse until it works.

13 years agotests/check/: Add unit tests for new parsers.
Stefan Kost [Thu, 13 Nov 2008 15:20:15 +0000 (15:20 +0000)]
tests/check/: Add unit tests for new parsers.

Original commit message from CVS:
* tests/check/Makefile.am:
* tests/check/elements/aacparse.c:
* tests/check/elements/amrparse.c:
Add unit tests for new parsers.

13 years agogst/: Fix baseparse type name.
Stefan Kost [Thu, 13 Nov 2008 14:21:39 +0000 (14:21 +0000)]
gst/: Fix baseparse type name.

Original commit message from CVS:
* gst/aacparse/gstbaseparse.c:
* gst/amrparse/gstbaseparse.c:
Fix baseparse type name.

13 years agoAdd two new baseparse based parsers (aac and amr) from Bug #518857.
Stefan Kost [Thu, 13 Nov 2008 12:59:34 +0000 (12:59 +0000)]
Add two new baseparse based parsers (aac and amr) from Bug #518857.

Original commit message from CVS:
* configure.ac:
* gst/aacparse/Makefile.am:
* gst/aacparse/gstaacparse.c:
* gst/aacparse/gstaacparse.h:
* gst/aacparse/gstbaseparse.c:
* gst/aacparse/gstbaseparse.h:
* gst/amrparse/Makefile.am:
* gst/amrparse/gstamrparse.c:
* gst/amrparse/gstamrparse.h:
* gst/amrparse/gstbaseparse.c:
* gst/amrparse/gstbaseparse.h:
Add two new baseparse based parsers (aac and amr) from Bug #518857.

13 years agojitterbuffer: Make src_query MT-safe
Havard Graff [Sun, 20 Mar 2011 00:08:38 +0000 (01:08 +0100)]
jitterbuffer: Make src_query MT-safe

It is possible that the element might be going down while the event arrives

13 years agojpegdec: Unref event if the parent element disappeared
Sebastian Dröge [Fri, 8 Apr 2011 13:22:47 +0000 (15:22 +0200)]
jpegdec: Unref event if the parent element disappeared

13 years agojitterbuffer: Unref event if the parent element disappeared
Sebastian Dröge [Fri, 8 Apr 2011 13:22:19 +0000 (15:22 +0200)]
jitterbuffer: Unref event if the parent element disappeared

13 years agojpegdec: Make upstream events MT-safe
Havard Graff [Mon, 21 Mar 2011 15:04:34 +0000 (16:04 +0100)]
jpegdec: Make upstream events MT-safe

13 years agojitterbuffer: Make upstream events MT-safe
Havard Graff [Mon, 21 Mar 2011 15:04:34 +0000 (16:04 +0100)]
jitterbuffer: Make upstream events MT-safe

13 years agortp: Unref events if the parent element disappeared
Sebastian Dröge [Fri, 8 Apr 2011 13:20:51 +0000 (15:20 +0200)]
rtp: Unref events if the parent element disappeared

13 years agortpmanager: fix pad callbacks so they handle when parent goes away
Ole André Vadla Ravnås [Thu, 6 Jan 2011 17:24:36 +0000 (18:24 +0100)]
rtpmanager: fix pad callbacks so they handle when parent goes away

1) We need to lock and get a strong ref to the parent, if still there.
2) If it has gone away, we need to handle that gracefully.

This is necessary in order to safely modify a running pipeline. Has been
observed when a streaming thread is doing a buffer_alloc() while an
application thread sends an event on a pad further downstream, and from
within a pad probe (holding STREAM_LOCK) carries out the pipeline plumbing
while the streaming thread has its buffer_alloc() in progress.

13 years agortpsession: make iterate_internal_links MT-safe
Havard Graff [Fri, 26 Nov 2010 14:20:04 +0000 (15:20 +0100)]
rtpsession: make iterate_internal_links MT-safe

13 years agoRevert "Pulsesink: Allow chunks up to bufsize instead of segsize"
Sebastian Dröge [Fri, 8 Apr 2011 12:35:04 +0000 (14:35 +0200)]
Revert "Pulsesink: Allow chunks up to bufsize instead of segsize"

This reverts commit 1e2c1467ae042a3c6bb1a6bc0c07aeff13ec5edb.

The commit causes pulsesink to ignore the latency-time baseaudiosink property.

13 years agortpspeexpay: Do not transmitt samples with GAP flag
Alexey Fisher [Fri, 8 Apr 2011 09:13:07 +0000 (11:13 +0200)]
rtpspeexpay: Do not transmitt samples with GAP flag

If we get GAP samples, there is no need to transmitt it.
In some situations, microphone is muted, we can drop net traffick
usage to ~1 kbit/s. Without patch it will stay ~20 kbit/s

13 years agospeexenc: Use speex intern silence detection
Alexey Fisher [Fri, 8 Apr 2011 09:11:58 +0000 (11:11 +0200)]
speexenc: Use speex intern silence detection

Speex has build in silence detection. If speex_encode_int returns 0,
than there is silence and sample do not need to be transmitted.
This work only if vbr=1 and dtx=1 optionas are enabled.
So if we get 0, we add GAP flag to the sample.

13 years agortspsrc: handle * control correctly
Wim Taymans [Tue, 5 Apr 2011 15:12:28 +0000 (17:12 +0200)]
rtspsrc: handle * control correctly

Parse session control attributes when no media control attribute is
present. Threat * control attributes as an empty string, just like the
spec says.

Fixes #646800

13 years agomatroskamux: Add support for A-Law and µ-Law
Sebastian Dröge [Tue, 5 Apr 2011 12:28:54 +0000 (14:28 +0200)]
matroskamux: Add support for A-Law and µ-Law

Fixes bug #646567.

13 years agojack: Fix build with jack 0.120.1
Jon Nordby [Tue, 5 Apr 2011 07:44:01 +0000 (09:44 +0200)]
jack: Fix build with jack 0.120.1

9544622674c0d0a3147a9b51145159b02eec68e9 checked
for 0.120.2 and later, but the deprecation was introduced in
0.120.1

13 years agodocs: fix docuemntation warnings (and reindent)
Stefan Kost [Tue, 5 Apr 2011 09:05:19 +0000 (12:05 +0300)]
docs: fix docuemntation warnings (and reindent)

13 years agovideomixer: update orc dist files
Alessandro Decina [Mon, 4 Apr 2011 15:34:17 +0000 (17:34 +0200)]
videomixer: update orc dist files

13 years agoAutomatic update of common submodule
Stefan Kost [Mon, 4 Apr 2011 12:57:10 +0000 (15:57 +0300)]
Automatic update of common submodule

From 1ccbe09 to c3cafe1

13 years agopulsesink: Always call pa_stream_new_with_proplist()
Arun Raghavan [Tue, 1 Mar 2011 08:38:12 +0000 (14:08 +0530)]
pulsesink: Always call pa_stream_new_with_proplist()

pa_stream_new_with_proplist() can take a NULL proplist, so we don't need
to concern ourselves with whether it's NULL or not.

13 years agortspsrc: perform post-flush state tricks downstream to upstream
Mark Nauwelaerts [Mon, 4 Apr 2011 09:33:10 +0000 (11:33 +0200)]
rtspsrc: perform post-flush state tricks downstream to upstream

... so downstream is set when upstream resumes data flow.

13 years agortspsrc: distribute new base_time to manager children following flush seek
Mark Nauwelaerts [Mon, 4 Apr 2011 09:27:29 +0000 (11:27 +0200)]
rtspsrc: distribute new base_time to manager children following flush seek

... by forcing a state changed to PLAYING, which should otherwise be a
no-op as elements should already be in that state.

In particular, jitterbuffer needs new base_time as soon as possible to perform
proper timing (e.g. eos timeout handling) and can't wait for the new base_time
that will be distributed when the whole pipeline returns to PLAYING.

See bug #646397.

13 years agoRevert "jitterbuffer: reset element base_time upon flush"
Mark Nauwelaerts [Mon, 4 Apr 2011 09:35:59 +0000 (11:35 +0200)]
Revert "jitterbuffer: reset element base_time upon flush"

This reverts commit f84b8a69cba9c538f5546869cb4ef454ad5efb9d.

Fixes bug #646397.

13 years agoflv: Specify the only possible stream-format for h264 in the pad templates.
Zaheer Abbas Merali [Mon, 4 Apr 2011 09:31:44 +0000 (10:31 +0100)]
flv: Specify the only possible stream-format for h264 in the pad templates.

13 years agoqtdemux: Check for invalid (empty) classification info entity strings
Sebastian Dröge [Mon, 4 Apr 2011 08:07:42 +0000 (10:07 +0200)]
qtdemux: Check for invalid (empty) classification info entity strings

Otherwise the classification string can be empty and gst_tag_list_add() will
complain or have a \0 in the first four bytes, which is wrong too.

13 years agoqtdemux: Year 0 is not a valid year for GDate and the proleptic gregorian calendar
Sebastian Dröge [Mon, 4 Apr 2011 08:01:26 +0000 (10:01 +0200)]
qtdemux: Year 0 is not a valid year for GDate and the proleptic gregorian calendar

13 years agoflacenc: Add support for writing METADATA_BLOCK_PICTURE blocks for GST_TAG_IMAGE...
Sebastian Dröge [Fri, 1 Apr 2011 11:18:55 +0000 (13:18 +0200)]
flacenc: Add support for writing METADATA_BLOCK_PICTURE blocks for GST_TAG_IMAGE and GST_TAG_PREVIEW_IMAGE

13 years agovideomixer[2]: Use orc_memset() instead of memset()
Sebastian Dröge [Fri, 1 Apr 2011 09:33:54 +0000 (11:33 +0200)]
videomixer[2]: Use orc_memset() instead of memset()

13 years agovideomixer: Add transparent background option for alpha channel formats
Lane Brooks [Thu, 20 Jan 2011 01:06:45 +0000 (18:06 -0700)]
videomixer: Add transparent background option for alpha channel formats

13 years agovideomixer2: Add transparent background option for alpha channel formats
Lane Brooks [Wed, 19 Jan 2011 19:07:17 +0000 (12:07 -0700)]
videomixer2: Add transparent background option for alpha channel formats

This option allows the videomixer2 element to output a valid alpha
channel when the inputs contain a valid alpha channel. This allows
mixing to occur in multiple stages serially.

The following pipeline shows an example of such a pipeline:

gst-launch videotestsrc background-color=0x000000 pattern=ball ! video/x-raw-yuv,format=\(fourcc\)AYUV ! videomixer2 background=transparent name=mix1 ! videomixer2 name=mix2 ! ffmpegcolorspace ! autovideosink  videotestsrc ! video/x-raw-yuv,format=\(fourcc\)AYUV ! mix2.

The first videotestsrc in this pipeline creates a moving ball on a
transparent background. It is then passed to the first videomixer2.
Previously, this videomixer2 would have forced the alpha channel to
1.0 and given a background of checker, black, or white to the
stream. With this patch, however, you can now specify the background
as transparent, and the alpha channel of the input will be
preserved. This allows for further mixing downstream, as is shown in
the above pipeline where the a second videomixer2 is used to mix in a
background of an smpte videotestsrc. So the result is a ball hovering
over the smpte test source. This could, of course, have been
accomplished with a single mixer element, but staged mixing is useful
when it is not convenient to mix all video at once (e.g. a pipeline
where a foreground and background bin exist and are mixed at the final
output, but the foreground bin needs an internal mixer to create
transitions between clips).

Fixes bug #639994.

13 years agopulsesink: also uncork during EOS waiting (and after EOS is rendered)
Mark Nauwelaerts [Thu, 31 Mar 2011 11:25:00 +0000 (13:25 +0200)]
pulsesink: also uncork during EOS waiting (and after EOS is rendered)

Pulsesink was recently changed to defer uncorking until there is data
to write. This condition will however never occur when EOS in being
rendered (since that marks the end of data). Changing to PAUSED state
while EOS is being waited on results in a hang: pausing corks the
stream, which will never be undone since there is no more data when
going back to PLAYING. If pulsesink is the clock provider, deadlock
ensues since time doesn't continue in corked state and the clock id
for EOS wait never fires.

Fixes #645961.

13 years agortpbin: Don't try to request the same request pad twice
Sebastian Dröge [Tue, 29 Mar 2011 14:33:43 +0000 (16:33 +0200)]
rtpbin: Don't try to request the same request pad twice

13 years agoflacdec: fix issues with large metadata blocks when streaming unframed flac
Tim-Philipp Müller [Mon, 28 Mar 2011 22:46:47 +0000 (23:46 +0100)]
flacdec: fix issues with large metadata blocks when streaming unframed flac

Parse metadata blocks when handling unparsed flac in push mode. This
works around a bunch of issues with the flac decoder when handling
metadata blocks that are larger than the max. flac framesize, which
coverart blocks often are. We need to have all the data for these
blocks available when we pass data to libflac.

http://gstreamer-devel.966125.n4.nabble.com/Flac-files-that-will-playback-but-not-stream-td3338198.html#a3395276

https://bugzilla.gnome.org/show_bug.cgi?id=566769

13 years agoflvdemux: Do not build an index if upstream is not seekable
Jan Urbański [Sun, 27 Mar 2011 19:39:50 +0000 (21:39 +0200)]
flvdemux: Do not build an index if upstream is not seekable

An index is not useful if upstream cannot handle seeks and building it
for infinite files, for instance FLV streams, results in a memory leak.

13 years agov4l2: new v4l2radio element to control analog radio devices
Alexey Chernov [Sat, 26 Mar 2011 22:19:58 +0000 (01:19 +0300)]
v4l2: new v4l2radio element to control analog radio devices

https://bugzilla.gnome.org/show_bug.cgi?id=640118

13 years agoAutomatic update of common submodule
Sebastian Dröge [Fri, 25 Mar 2011 21:22:43 +0000 (22:22 +0100)]
Automatic update of common submodule

From 193b717 to 1ccbe09

13 years agoAutomatic update of common submodule
Stefan Kost [Fri, 25 Mar 2011 12:56:06 +0000 (14:56 +0200)]
Automatic update of common submodule

From b77e2bf to 193b717

13 years agocairo: fix the name of the *-marshall.list file to unbreak make distcheck
Stefan Kost [Fri, 25 Mar 2011 10:53:43 +0000 (12:53 +0200)]
cairo: fix the name of the *-marshall.list file to unbreak make distcheck

13 years agoAutomatic update of common submodule
Sebastian Dröge [Fri, 25 Mar 2011 08:31:03 +0000 (09:31 +0100)]
Automatic update of common submodule

From d8814b6 to b77e2bf

13 years agoAutomatic update of common submodule
Sebastian Dröge [Fri, 25 Mar 2011 08:06:16 +0000 (09:06 +0100)]
Automatic update of common submodule

From 6aaa286 to d8814b6

13 years agospectrum: refactor processing loop for block based operation
Stefan Kost [Thu, 24 Mar 2011 22:10:56 +0000 (00:10 +0200)]
spectrum: refactor processing loop for block based operation

Previously the chain function was working sample frame based. In each cycle it
was checking if it is time to run a fft or if it is time to send a message.
Now we changed the data transform functions to work on a block of data and
calculate the max length until either {end-of-data, do-fft, do-msg}. This allows
us also to avoid the duplicated code for the single and multi-channel case (as
the transformers have the same signature now).

13 years agojack: unbreak the build for jack2 users
Stefan Kost [Thu, 24 Mar 2011 21:47:33 +0000 (23:47 +0200)]
jack: unbreak the build for jack2 users

Jack2 (versions 1.X.X) does only have that API in svn. Limmit the use of the new
API for jack1 versions.

13 years agoAutomatic update of common submodule
Stefan Kost [Thu, 24 Mar 2011 16:49:19 +0000 (18:49 +0200)]
Automatic update of common submodule

From 6aec6b9 to 6aaa286

13 years agospectrum: fix the error accumulation and frames_todo handling
Stefan Kost [Thu, 24 Mar 2011 12:14:09 +0000 (14:14 +0200)]
spectrum: fix the error accumulation and frames_todo handling

Even though we wrap around the accumulated second, we still need to add the
error in the same cycle. Increase the todo in the same conditional as afterwards
the accumulated error will be below one second.

13 years agospectrum: fix broken code resulting for a wrong splitup of changes
Stefan Kost [Thu, 24 Mar 2011 11:53:12 +0000 (13:53 +0200)]
spectrum: fix broken code resulting for a wrong splitup of changes

13 years agospectrum: simplify the have_interval calculation
Stefan Kost [Tue, 22 Mar 2011 14:29:53 +0000 (16:29 +0200)]
spectrum: simplify the have_interval calculation

Move some of the conditions to the places where the dependent variables change.

13 years agospectrum: use local var for input_data function
Stefan Kost [Tue, 22 Mar 2011 14:26:45 +0000 (16:26 +0200)]
spectrum: use local var for input_data function

Avoid dereferencing the input_data from the instance from within an inner loop.

13 years agospeexdec: Get and use streamheader from the caps if possible
Sebastian Dröge [Wed, 23 Mar 2011 15:34:16 +0000 (16:34 +0100)]
speexdec: Get and use streamheader from the caps if possible

This allows playback of streams where the streamheader buffers
were dropped from the stream for some reason.

13 years agoflvmux: use running time for synchronization
Mark Nauwelaerts [Tue, 22 Mar 2011 18:36:31 +0000 (19:36 +0100)]
flvmux: use running time for synchronization

Fixes #432612.

13 years agomatroskamux: use running time for synchronization
Mark Nauwelaerts [Tue, 22 Mar 2011 18:36:21 +0000 (19:36 +0100)]
matroskamux: use running time for synchronization

Fixes #432612.

13 years agoavimux: use running time for synchronization
Mark Nauwelaerts [Tue, 22 Mar 2011 18:35:58 +0000 (19:35 +0100)]
avimux: use running time for synchronization

See bug #432612.

13 years agoconfigure.ac: redundant uses of AC_MSG_RESULT()
Luis de Bethencourt [Tue, 22 Mar 2011 11:53:22 +0000 (12:53 +0100)]
configure.ac: redundant uses of AC_MSG_RESULT()

cleaned the redundant uses of AC_MSG_RESULT() in configure.ac

13 years agoautogen: wingo signed comment
Luis de Bethencourt [Fri, 18 Mar 2011 18:34:57 +0000 (19:34 +0100)]
autogen: wingo signed comment

13 years agojackaudiosink: Fix typo from 9544622674c0d0a3147a9b51145159b02eec68e9
Robert Swain [Wed, 16 Mar 2011 09:43:47 +0000 (10:43 +0100)]
jackaudiosink: Fix typo from 9544622674c0d0a3147a9b51145159b02eec68e9

13 years agomatroska: Mark tag mapping tables as static const
Sebastian Dröge [Wed, 16 Mar 2011 08:38:43 +0000 (09:38 +0100)]
matroska: Mark tag mapping tables as static const

13 years agomatroskamux: Use ARTIST instead of AUTHOR for GST_TAG_ARTIST
Sebastian Dröge [Wed, 16 Mar 2011 08:37:58 +0000 (09:37 +0100)]
matroskamux: Use ARTIST instead of AUTHOR for GST_TAG_ARTIST

13 years agomatroskademux: Use ARTIST Matroska tag instead of AUTHOR for GST_TAG_ARTIST
Sebastian Dröge [Wed, 16 Mar 2011 08:35:50 +0000 (09:35 +0100)]
matroskademux: Use ARTIST Matroska tag instead of AUTHOR for GST_TAG_ARTIST

AUTHOR only existed in an old version of the spec and ARTIST is
the new replacement for this. We are still reading both to still
be compatible with old files.

Fixes bug #644875.

13 years agotests: enable more formats in videofilter unit test, check more resolutions
Tim-Philipp Müller [Tue, 15 Mar 2011 20:19:48 +0000 (20:19 +0000)]
tests: enable more formats in videofilter unit test, check more resolutions

13 years agovideoflip: Fix buffer overflow bug for odd resolutions and Y422 colorspaces
Youness Alaoui [Mon, 14 Mar 2011 23:14:07 +0000 (19:14 -0400)]
videoflip: Fix buffer overflow bug for odd resolutions and Y422 colorspaces

https://bugzilla.gnome.org/show_bug.cgi?id=644773

13 years agospeexdec: silence warning message when appropriate
Vincent Penquerc'h [Tue, 15 Mar 2011 17:36:01 +0000 (19:36 +0200)]
speexdec: silence warning message when appropriate

If we did not know how many frames to expect, then we get an unexpected
end of stream when trying to decode more frames that are there, if there
are leftover bits to pad to the next byte

13 years agovideoflip: Add support for YUY2, UVYV and YVYU colorspaces
Youness Alaoui [Mon, 14 Mar 2011 23:14:07 +0000 (19:14 -0400)]
videoflip: Add support for YUY2, UVYV and YVYU colorspaces

https://bugzilla.gnome.org/show_bug.cgi?id=644773

13 years agotests: in videofilter unit test also check with 'odd' widths and heights
Tim-Philipp Müller [Tue, 15 Mar 2011 09:43:35 +0000 (09:43 +0000)]
tests: in videofilter unit test also check with 'odd' widths and heights

And only use one test suite.

13 years agospeexdec: Always process the number of frames per packet as specified in the header
Sebastian Dröge [Mon, 14 Mar 2011 18:28:07 +0000 (19:28 +0100)]
speexdec: Always process the number of frames per packet as specified in the header

Looking at the remaining bits in the bitstream after decoding a
single frame can't be used as loop condition. The remaining
bits might not give a complete frame and the speex decoder will
then output nothing but access uninitialized memory, which leads
to valgrind warnings.

Fixes bug #644669.

13 years agomatroskamux: return TRUE from sink pad event function for tag events, which are handled
Andoni Morales Alastruey [Mon, 14 Mar 2011 14:46:50 +0000 (15:46 +0100)]
matroskamux: return TRUE from sink pad event function for tag events, which are handled

https://bugzilla.gnome.org/show_bug.cgi?id=644730

13 years agopulsesink: Better fix for deadlock on failed connect
Philip Jägenstedt [Fri, 11 Mar 2011 19:14:31 +0000 (00:44 +0530)]
pulsesink: Better fix for deadlock on failed connect

This reverts the previous fix that would cause a double-unlock when the
stream connect failed.

https://bugzilla.gnome.org/show_bug.cgi?id=644510

13 years agopulsesink: Fix deadlock if connecting to PA fails
Arun Raghavan [Fri, 11 Mar 2011 17:36:31 +0000 (23:06 +0530)]
pulsesink: Fix deadlock if connecting to PA fails

Commit dd4ec22e introduced a deadlock in the failure path while trying
to connect to PulseAudio. This makes sure we drop the lock on the
resource mutex to avoid this.

https://bugzilla.gnome.org/show_bug.cgi?id=644510

13 years agotests: order state-test blacklist and add jack elements
Stefan Kost [Fri, 11 Mar 2011 14:59:10 +0000 (16:59 +0200)]
tests: order state-test blacklist and add jack elements

Jack audio src/sink elements recently got moved from bad and should be excluded
from the test (like the other device specific source and sinks).

Fixes #644288

13 years agodvdemux: Chain up to the parent class' ::send_event for non-seek events
Sebastian Dröge [Fri, 11 Mar 2011 12:47:26 +0000 (13:47 +0100)]
dvdemux: Chain up to the parent class' ::send_event for non-seek events

13 years agodvdemux: Fix refcount issues with the seek event
Sebastian Dröge [Fri, 11 Mar 2011 12:46:05 +0000 (13:46 +0100)]
dvdemux: Fix refcount issues with the seek event

Fixes bug #642963.

13 years agodocs: fix pulsesink gtk-doc markup
Tim-Philipp Müller [Fri, 11 Mar 2011 09:54:02 +0000 (09:54 +0000)]
docs: fix pulsesink gtk-doc markup

13 years agojack: fix build against jack 0.120.2
Philippe Normand [Fri, 11 Mar 2011 09:29:08 +0000 (10:29 +0100)]
jack: fix build against jack 0.120.2

jack_port_get_total_latency() has been deprecated in favor of
jack_port_get_latency_range().

https://bugzilla.gnome.org/show_bug.cgi?id=644477

13 years agospectrum: more comments and tune and logging
Stefan Kost [Thu, 10 Mar 2011 12:29:25 +0000 (14:29 +0200)]
spectrum: more comments and tune and logging

13 years agospectrum: avoid unneccesary extra fft runs
Stefan Kost [Thu, 10 Mar 2011 12:15:42 +0000 (14:15 +0200)]
spectrum: avoid unneccesary extra fft runs

Before it was possible that we run an extra fft when the time for sending a new
message is due. Only do this if we have not run the fft for the interval at all.

13 years agospectrum: only scale the vectors that we are processing
Stefan Kost [Thu, 10 Mar 2011 12:12:01 +0000 (14:12 +0200)]
spectrum: only scale the vectors that we are processing

Phase is not produced by default, so lets not scale it unconditionally to save a
few cycles.

13 years agospectrum: put number of channels to instance variable
Stefan Kost [Thu, 10 Mar 2011 12:10:25 +0000 (14:10 +0200)]
spectrum: put number of channels to instance variable

When freeing data the format might have changed. Thus we need to remember for
which format we allocated memory.

13 years agospectrum: update doc review stamp
Stefan Kost [Thu, 10 Mar 2011 08:27:14 +0000 (10:27 +0200)]
spectrum: update doc review stamp

13 years agospectrum: use function pointers for data readers
Stefan Kost [Thu, 10 Mar 2011 08:22:29 +0000 (10:22 +0200)]
spectrum: use function pointers for data readers

Don't check the format for each sample frame to read. We can make that decission
in _setup already. This is still not ideal as we call the function per frame.
Ideally we determine how many samples we can copy and have a loop in the input
reader. As an alternative we might also consider to use the fft variants for the
various formats and not convert to float for all cases - we would still need to
mix or deinterleave though.

13 years agortspsrc: improve recovery from failed seek
Mark Nauwelaerts [Wed, 9 Mar 2011 16:07:47 +0000 (17:07 +0100)]
rtspsrc: improve recovery from failed seek

In case server-side fails to perform seek, i.e. PLAY at non-zero requested
position, recovery so far would arrange for streaming to continue, albeit
having lost position tracking in the process.  So, query position prior
to seek and use upon failed seek.

13 years agojitterbuffer: handle position query
Mark Nauwelaerts [Wed, 9 Mar 2011 15:51:00 +0000 (16:51 +0100)]
jitterbuffer: handle position query

13 years agospectrum: multi-channel support
Stefan Kost [Wed, 9 Mar 2011 14:57:28 +0000 (16:57 +0200)]
spectrum:  multi-channel support

Add a boolean multi-channel property with a default of FALSE. When set to TRUE
the element won't mix all input channels to mono, but instead run a FFT on each
channel. In that case the result message would contain a 2 dimensional array
of channel x data for magnitude and phase.

API: GstSpectrum:multi-channel
https://bugzilla.gnome.org/show_bug.cgi?id=593482

13 years agospectrum: more xrefs in the docs
Stefan Kost [Wed, 9 Mar 2011 14:55:56 +0000 (16:55 +0200)]
spectrum: more xrefs in the docs

13 years agospectrum: factor out the code that accumulated samples into the ring-buffer
Stefan Kost [Wed, 9 Mar 2011 10:41:15 +0000 (12:41 +0200)]
spectrum: factor out the code that accumulated samples into the ring-buffer

Use a separate function to read a sample frame into a ringbuffer slot. In the
future we can use format-specific function pointer to avoid the reoccuring
format checks.

13 years agospectrum: pull format to temp var to improve readability of lines using it
Stefan Kost [Wed, 9 Mar 2011 10:38:52 +0000 (12:38 +0200)]
spectrum: pull format to temp var to improve readability of lines using it

13 years agospectrum: code cleanup for copying data to ring-buffer
Stefan Kost [Wed, 9 Mar 2011 10:20:11 +0000 (12:20 +0200)]
spectrum: code cleanup for copying data to ring-buffer

Rename fp to is_float and restructure if-else part for handling the different formats.

13 years agospectrum: add a GstSpecrtumChannel context structure
Stefan Kost [Wed, 9 Mar 2011 09:40:48 +0000 (11:40 +0200)]
spectrum: add a GstSpecrtumChannel context structure

We now keep the fft data that is related to one channel in a separate structure
to prepare for multichannel support. We also refactor the code to operate more
often on the channel context.

13 years agospectrum: call the instance var spectrum instead of filter
Stefan Kost [Wed, 9 Mar 2011 09:18:19 +0000 (11:18 +0200)]
spectrum: call the instance var spectrum instead of filter

13 years agospectrum: don't value we already took from the gvalue
Stefan Kost [Wed, 9 Mar 2011 09:14:37 +0000 (11:14 +0200)]
spectrum: don't value we already took from the gvalue

13 years agoMerge ad-hoc release branch '0.10.28'
Tim-Philipp Müller [Tue, 8 Mar 2011 16:28:27 +0000 (16:28 +0000)]
Merge ad-hoc release branch '0.10.28'