platform/upstream/gstreamer.git
13 years agoghostpad: Only implement the iterate_internal_links function on proxypads
Sebastian Dröge [Fri, 6 May 2011 13:25:04 +0000 (15:25 +0200)]
ghostpad: Only implement the iterate_internal_links function on proxypads

ghostpads inherit it from their parent class. Also make it threadsafe.

13 years agoghostpad: API: Expose gst_proxy_pad_get_internal()
Sebastian Dröge [Fri, 6 May 2011 13:16:09 +0000 (15:16 +0200)]
ghostpad: API: Expose gst_proxy_pad_get_internal()

This allows to get the internal pad of ghostpads and
proxypads without using gst_pad_iterate_internal_links()
and is much more convenient.

The internal pad of a ghostpad is the pad of the opposite direction
that is used to link to the ghostpad target.

13 years agobasetransform: When trying to fixate the sink suggestion prefer its structure order
Sebastian Dröge [Thu, 5 May 2011 15:54:56 +0000 (17:54 +0200)]
basetransform: When trying to fixate the sink suggestion prefer its structure order

13 years agocaps: Merge structures when intersecting instead of appending them
Sebastian Dröge [Thu, 5 May 2011 09:28:38 +0000 (11:28 +0200)]
caps: Merge structures when intersecting instead of appending them

This prevents adding duplicates over and over again to the resulting
caps if they already describe the new intersection result.

While this changes intersection from O(n*m) to O(n^2*m), it results in
smaller caps, which in the end will decrease further processing times.

For example in an audioconvert ! audioconvert ! audioconvert pipeline,
when forwarding the downstream caps preference in basetransform
(see e26da72de25a91c3eaad9f7c8b2f53ba888a0394) this results in
16 instead of 191 caps structures.

13 years agobasetransform: In getcaps() prefer the caps order and caps of downstream if possible
Sebastian Dröge [Wed, 4 May 2011 09:29:15 +0000 (11:29 +0200)]
basetransform: In getcaps() prefer the caps order and caps of downstream if possible

13 years agobasetransform: Prefer caps order given by the subclass of the template caps order
Sebastian Dröge [Tue, 3 May 2011 15:26:53 +0000 (17:26 +0200)]
basetransform: Prefer caps order given by the subclass of the template caps order

13 years agoparse: don't unescape inside quotes
Miguel Angel Cabrera Moya [Wed, 20 Apr 2011 20:52:36 +0000 (22:52 +0200)]
parse: don't unescape inside quotes

Escaped characters inside quoted strings are supposed to be unescaped by
deserialization functions, not by parsing functions.

https://bugzilla.gnome.org/show_bug.cgi?id=648025

13 years agopad: Remove unnecessary FIXME
Sebastian Dröge [Mon, 18 Apr 2011 08:04:24 +0000 (10:04 +0200)]
pad: Remove unnecessary FIXME

Resetting the result is not necessary when resyncing because
pads that previously got the event will be skipped and we
need to consider the results of the previous pushes.

13 years agoelement: If activating one pad failed error out early instead of trying to activate...
Sebastian Dröge [Mon, 18 Apr 2011 07:53:55 +0000 (09:53 +0200)]
element: If activating one pad failed error out early instead of trying to activate the next pads

If one pad fails to activate the complete activation process will fail
anyway and trying to activate the other pads only wastes time.

13 years agobin: If activating one pad failed error out early instead of trying to activate the...
Sebastian Dröge [Mon, 18 Apr 2011 07:49:04 +0000 (09:49 +0200)]
bin: If activating one pad failed error out early instead of trying to activate the next pads

If one pad fails to activate the complete activation process will fail
anyway and trying to activate the other pads only wastes time.

13 years agoBack to development
Tim-Philipp Müller [Sat, 14 May 2011 08:31:33 +0000 (09:31 +0100)]
Back to development

13 years agoRelease 0.10.34 RELEASE-0.10.34
Tim-Philipp Müller [Sat, 14 May 2011 00:00:23 +0000 (01:00 +0100)]
Release 0.10.34

13 years agobasesrc: do not set first buffer timestamp to 0 for live sources
Vincent Penquerc'h [Wed, 4 May 2011 12:31:56 +0000 (15:31 +0300)]
basesrc: do not set first buffer timestamp to 0 for live sources

Doing so avoids a large timestamp gap between first and second buffer
for live sources which take time to start up.
The first buffer now has a "live" timestamp based on the running time,
as other buffers do.

https://bugzilla.gnome.org/show_bug.cgi?id=649369

13 years agomultiqueue: ensure thread safety when adding a pad
Mark Nauwelaerts [Tue, 10 May 2011 09:11:15 +0000 (11:11 +0200)]
multiqueue: ensure thread safety when adding a pad

This seems to be a regression, and was causing crashes.

https://bugzilla.gnome.org/show_bug.cgi?id=649878

13 years agoRelease 0.10.33 RELEASE-0.10.33
Tim-Philipp Müller [Tue, 10 May 2011 07:55:10 +0000 (08:55 +0100)]
Release 0.10.33

Highlights:

  - new parser base class: GstBaseParse
  - new core element: funnel
  - OSX multi-arch fixes
  - new QoS type for QoS events
  - new progress message API to notify applications of asynchronous operations
  - countless other fixes and improvements

13 years agobaseparse: don't post loads of empty taglists
Tim-Philipp Müller [Fri, 6 May 2011 22:40:58 +0000 (23:40 +0100)]
baseparse: don't post loads of empty taglists

Only post bitrate updates if there's something to post, don't
post empty taglists if nothing changed.

13 years agoinfo: avoid redefinition of symbols when debugging is off
Stefan Kost [Fri, 6 May 2011 08:35:36 +0000 (11:35 +0300)]
info: avoid redefinition of symbols when debugging is off

The refactoring of gst_debug_add_log_function() now causes build failure when
debug-logging is turned off. Just move it to the conditional part of the header.

13 years agotests: fix compiler warning in new miniobject test
Tim-Philipp Müller [Wed, 4 May 2011 14:29:42 +0000 (15:29 +0100)]
tests: fix compiler warning in new miniobject test

gst/gstminiobject.c: In function ‘test_dup_null_mini_object’:
gst/gstminiobject.c:459:7: warning: assignment from incompatible pointer type

13 years agominiobject: Fix dup_mini_object function to handle NULL gvalues
Thiago Santos [Mon, 2 May 2011 14:30:06 +0000 (11:30 -0300)]
miniobject: Fix dup_mini_object function to handle NULL gvalues

g_value_dup_object handles gvalues that contain NULL pointers,
gst_value_dup_mini_object should do the same.

https://bugzilla.gnome.org/show_bug.cgi?id=649195

13 years agodocs: it its %TRUE (constant)
Stefan Kost [Tue, 3 May 2011 10:55:43 +0000 (13:55 +0300)]
docs: it its %TRUE (constant)

As spotted by Tim.

13 years agodocs: fix copy'n'paste doc header mistake
Stefan Kost [Mon, 2 May 2011 13:22:56 +0000 (16:22 +0300)]
docs: fix copy'n'paste doc header mistake

13 years agodocs: add two trivial doc blobs
Stefan Kost [Mon, 2 May 2011 13:20:24 +0000 (16:20 +0300)]
docs: add two trivial doc blobs

13 years agodocs: add missing parameter docs
Stefan Kost [Mon, 2 May 2011 13:03:29 +0000 (16:03 +0300)]
docs: add missing parameter docs

13 years agodocs: add docs for GstDPPacketizer
Stefan Kost [Mon, 2 May 2011 13:00:52 +0000 (16:00 +0300)]
docs: add docs for GstDPPacketizer

13 years agodocs: improve the syntax for the capsintersectmode docs
Stefan Kost [Mon, 2 May 2011 12:52:58 +0000 (15:52 +0300)]
docs: improve the syntax for the capsintersectmode docs

13 years agodocs: fixup broken xref
Stefan Kost [Mon, 2 May 2011 12:48:01 +0000 (15:48 +0300)]
docs: fixup broken xref

13 years agodocs: add new api to -section.txt
Stefan Kost [Mon, 2 May 2011 12:46:59 +0000 (15:46 +0300)]
docs: add new api to -section.txt

13 years agodocs: fix gtk-doc syntax
Stefan Kost [Mon, 2 May 2011 12:35:52 +0000 (15:35 +0300)]
docs: fix gtk-doc syntax

13 years agodocs: don't duplicate info that we take from element-details
Stefan Kost [Mon, 2 May 2011 12:30:13 +0000 (15:30 +0300)]
docs: don't duplicate info that we take from element-details

13 years agodocs: remove non existing symbol
Stefan Kost [Thu, 28 Apr 2011 12:37:02 +0000 (15:37 +0300)]
docs: remove non existing symbol

13 years agodocs: we don't need to document private members in opaque structs
Stefan Kost [Thu, 28 Apr 2011 12:05:28 +0000 (15:05 +0300)]
docs: we don't need to document private members in opaque structs

13 years ago0.10.32.4 pre-release
Tim-Philipp Müller [Sat, 30 Apr 2011 15:55:36 +0000 (16:55 +0100)]
0.10.32.4 pre-release

13 years agopluginfeature: include plugin.h in header where we use a GstPlugin pointer
Tim-Philipp Müller [Fri, 29 Apr 2011 22:44:55 +0000 (23:44 +0100)]
pluginfeature: include plugin.h in header where we use a GstPlugin pointer

Should fix issue with gstreamermm build where <gst/gstindex.h> is included
directly instead of gst/gst.h.

13 years ago0.10.32.3 pre-release
Tim-Philipp Müller [Wed, 27 Apr 2011 10:49:11 +0000 (11:49 +0100)]
0.10.32.3 pre-release

13 years agoRevert lockfree GstBus for the release
Tim-Philipp Müller [Tue, 26 Apr 2011 14:42:46 +0000 (15:42 +0100)]
Revert lockfree GstBus for the release

Drop in old GstBus code for the release to play it safe, since
regressions that are apparently hard to track down and reproduce
have been reported (on windows/OSX mostly) against the lockfree
version, and more time is needed to fix them.

This reverts commit 03391a897001d35d1d290f27dd12e98a8b729fb4.
This reverts commit 43cdbc17e6f944cdf02aeed78d1d5f6bde5190c9.
This reverts commit 80eb160e0f62350271f061daa5f289d9d4277cf4.
This reverts commit c41b0ade28790ffdb0e484b41cd7929c4e145dec.
This reverts commit 874d60e5899dd5b89854679d1a4ad016a58ba4e0.
This reverts commit 79370d4b1781af9c9a65f2d1e3498124d8c4c413.
This reverts commit 2cb3e5235196eb71fb25e0a4a4b8749d6d0a8453.
This reverts commit bd1c40011434c1efaa696dc98ef855ef9cce9b28.
This reverts commit 4bf8f1524f6e3374b3f3bc57322337723d06b928.
This reverts commit 14d7db1b527b05f029819057aef5c123ac7e013d.

https://bugzilla.gnome.org/show_bug.cgi?id=647493

13 years agominor inline documentation fix
David Schleef [Sat, 16 Apr 2011 05:00:11 +0000 (22:00 -0700)]
minor inline documentation fix

13 years agoAutomatic update of common submodule
Tim-Philipp Müller [Sun, 24 Apr 2011 13:02:51 +0000 (14:02 +0100)]
Automatic update of common submodule

From c3cafe1 to 46dfcea

13 years agotests: clean up properly in the bin test_link_structure_change unit test
Tim-Philipp Müller [Sun, 24 Apr 2011 10:44:19 +0000 (11:44 +0100)]
tests: clean up properly in the bin test_link_structure_change unit test

Don't forget to set the pipeline back to NULL state, which makes
valgrind happy again.

13 years agoregistry: when removing a cached-but-no-longer-existing plugin, only remove features...
Tim-Philipp Müller [Sun, 24 Apr 2011 08:58:53 +0000 (09:58 +0100)]
registry: when removing a cached-but-no-longer-existing plugin, only remove features that belong to it

When a plugin file no longer exists, e.g. because it's been removed or
renamed, don't remove all features in the registry based on the *name*
of the plugin they belong to, but only remove those who actually belong
to that particular plugin (object/pointer).

This fixes issues of plugin features disappearing when a plugin .so file
is renamed.

https://bugzilla.gnome.org/show_bug.cgi?id=604094

13 years agopluginfeature: store pointer to plugin in addition to the plugin name
Tim-Philipp Müller [Sun, 24 Apr 2011 08:53:39 +0000 (09:53 +0100)]
pluginfeature: store pointer to plugin in addition to the plugin name

So we can reliably remove plugin features for a specific plugin later.

https://bugzilla.gnome.org/show_bug.cgi?id=604094

13 years agoregistry: use TRACE log level to log files that don't look like plugins
Tim-Philipp Müller [Sun, 24 Apr 2011 10:05:58 +0000 (11:05 +0100)]
registry: use TRACE log level to log files that don't look like plugins

Cuts down the noise in uninstalled setups.

13 years agobasetransform: fix negotiation regression
Thiago Santos [Tue, 19 Apr 2011 23:35:04 +0000 (20:35 -0300)]
basetransform: fix negotiation regression

Fixup patch for 83597767b169dd6c39a07b6144a650c1f098825a

Use a separate variable for knowing if a pad alloc has been made
instead of checking for the flow return that might not be the
result of the pad alloc

https://bugzilla.gnome.org/show_bug.cgi?id=648220

13 years agotests: add simple pipeline-in-pipeline unit test
Tim-Philipp Müller [Thu, 21 Apr 2011 11:33:10 +0000 (12:33 +0100)]
tests: add simple pipeline-in-pipeline unit test

https://bugzilla.gnome.org/show_bug.cgi?id=648297

13 years agobus: also allow popping a message without timeout if no poll available
Mark Nauwelaerts [Wed, 20 Apr 2011 13:39:16 +0000 (15:39 +0200)]
bus: also allow popping a message without timeout if no poll available

... which happens in particular flushing a bus, possibly as part
of a state change, e.g. when having a pipeline in a pipeline
and then changing state back to NULL. The interior pipeline
will/might then flush the bus, which is a child bus from the
parent which does not have a poll anymore these days.

https://bugzilla.gnome.org/show_bug.cgi?id=648297

13 years agopad: unlock before freeing the cache to avoid deadlock
Havard Graff [Tue, 19 Apr 2011 12:05:23 +0000 (14:05 +0200)]
pad: unlock before freeing the cache to avoid deadlock

https://bugzilla.gnome.org/show_bug.cgi?id=648199

13 years agobasetransform: don't unref trans until the function is done using it
Havard Graff [Thu, 14 Apr 2011 08:15:26 +0000 (10:15 +0200)]
basetransform: don't unref trans until the function is done using it

trans->priv->force_alloc = FALSE would crash if the ref held is the last

https://bugzilla.gnome.org/show_bug.cgi?id=648215

13 years agodocs: add note/warning to gst_index_get_writer_id() docs about the OBJECT_LOCK
Tim-Philipp Müller [Tue, 19 Apr 2011 12:23:19 +0000 (13:23 +0100)]
docs: add note/warning to gst_index_get_writer_id() docs about the OBJECT_LOCK

https://bugzilla.gnome.org/show_bug.cgi?id=646811

13 years agobaseparse: don't deadlock when setting external index
Tim-Philipp Müller [Tue, 19 Apr 2011 12:05:53 +0000 (13:05 +0100)]
baseparse: don't deadlock when setting external index

Protect index with its own lock. gst_index_get_writer_id() may take
the object lock internally (the default resolver, GST_INDEX_RESOLVER_PATH,
will anyway), so if we're using that to protect the index as well,
we'll deadlock.

https://bugzilla.gnome.org/show_bug.cgi?id=646811

13 years agobaseparse: make fmtlist constant
Tim-Philipp Müller [Tue, 19 Apr 2011 10:51:30 +0000 (11:51 +0100)]
baseparse: make fmtlist constant

13 years agoquery: const-ify formats arguments to gst_query_set_formatsv()
Tim-Philipp Müller [Tue, 19 Apr 2011 10:48:05 +0000 (11:48 +0100)]
query: const-ify formats arguments to gst_query_set_formatsv()

13 years agodocs: remove reference to baseparse API that didn't make it
Tim-Philipp Müller [Sat, 16 Apr 2011 14:20:08 +0000 (15:20 +0100)]
docs: remove reference to baseparse API that didn't make it

13 years ago0.10.32.2 pre-release
Tim-Philipp Müller [Sat, 16 Apr 2011 13:56:03 +0000 (14:56 +0100)]
0.10.32.2 pre-release

13 years agopo: update translations
Tim-Philipp Müller [Sat, 16 Apr 2011 13:54:00 +0000 (14:54 +0100)]
po: update translations

13 years agogst-launch: remove newline from translatable string
Tim-Philipp Müller [Sat, 16 Apr 2011 13:52:40 +0000 (14:52 +0100)]
gst-launch: remove newline from translatable string

13 years agogst: gobject-introspection scanner doesn't need to scan or update plugin info
Tim-Philipp Müller [Sat, 16 Apr 2011 12:49:45 +0000 (13:49 +0100)]
gst: gobject-introspection scanner doesn't need to scan or update plugin info

13 years agogst: make sure gobject-introspection scanner calls gst_init()
Sebastian Dröge [Sat, 16 Apr 2011 12:34:08 +0000 (14:34 +0200)]
gst: make sure gobject-introspection scanner calls gst_init()

https://bugzilla.gnome.org/show_bug.cgi?id=647922

13 years agolibs: gobject-introspection scanner doesn't need to scan or update plugin info
Tim-Philipp Müller [Sat, 16 Apr 2011 09:45:16 +0000 (10:45 +0100)]
libs: gobject-introspection scanner doesn't need to scan or update plugin info

Make sure the scanner doesn't load or introspect or check any plugins,
(especially not outside the build directory).

13 years agolibs: make sure gobject-introspection scanner calls gst_init()
Tim-Philipp Müller [Sat, 16 Apr 2011 09:33:21 +0000 (10:33 +0100)]
libs: make sure gobject-introspection scanner calls gst_init()

https://bugzilla.gnome.org/show_bug.cgi?id=647922

13 years agowin32: add new baseparse API to libgstbase.def
Tim-Philipp Müller [Sat, 16 Apr 2011 09:17:24 +0000 (10:17 +0100)]
win32: add new baseparse API to libgstbase.def

13 years agowin32: Add exports for the GstParseContext and GstBufferListIterator types
Sebastian Dröge [Sat, 16 Apr 2011 07:32:17 +0000 (09:32 +0200)]
win32: Add exports for the GstParseContext and GstBufferListIterator types

13 years agopluginloader: only run gst-plugin-scanner with /usr/bin/arch wrapper on OS X >= 10.5
Tim-Philipp Müller [Fri, 15 Apr 2011 19:58:51 +0000 (20:58 +0100)]
pluginloader: only run gst-plugin-scanner with /usr/bin/arch wrapper on OS X >= 10.5

Based on patch by: Daniel Macks <dmacks@netspace.org>

Earlier versions of OSX don't support proper multiarch and
trying to use /usr/bin/arch -foo with those versions would
just break things.

https://bugzilla.gnome.org/show_bug.cgi?id=615357

13 years agobaseparse: expose gst_base_parse_frame_free() for completeness
Tim-Philipp Müller [Fri, 15 Apr 2011 18:07:55 +0000 (19:07 +0100)]
baseparse: expose gst_base_parse_frame_free() for completeness

API: gst_base_parse_frame_free()

13 years agobaseparse: init frames on the stack with gst_base_parse_frame_init()
Tim-Philipp Müller [Fri, 15 Apr 2011 17:52:18 +0000 (18:52 +0100)]
baseparse: init frames on the stack with gst_base_parse_frame_init()

Frames must now be inited this way, can't just zero them
out and use them.

13 years agobaseparse: more debug logging, minor clean-up
Tim-Philipp Müller [Fri, 15 Apr 2011 17:38:46 +0000 (18:38 +0100)]
baseparse: more debug logging, minor clean-up

Trace frames, split out code to queue a frame for later.

13 years agobaseparse: change gst_base_parse_frame_init() to not take a GstBaseParse argument
Tim-Philipp Müller [Fri, 15 Apr 2011 17:00:21 +0000 (18:00 +0100)]
baseparse: change gst_base_parse_frame_init() to not take a GstBaseParse argument

13 years agobaseparse: make GstBaseParseFrame handling more bindings-friendly
Tim-Philipp Müller [Fri, 15 Apr 2011 16:41:02 +0000 (17:41 +0100)]
baseparse: make GstBaseParseFrame handling more bindings-friendly

Change semantics of gst_base_parse_push_frame() and make it take
ownership of the whole frame, not just the frame contents. This
is more in line with how gst_pad_push() etc. work. Just transfering
the content, but not the container of something that's not really
known to be a container is hard to annotate properly and probably
won't work. We mark frames allocated on the stack now with a private
flag in gst_base_parse_frame_init(), so gst_base_parse_frame_free()
only frees the contents in that case but not the frame struct itself.

https://bugzilla.gnome.org/show_bug.cgi?id=518857

API: gst_base_parse_frame_new()

13 years agobaseparse: register boxed type for GstBaseFrameParse
Tim-Philipp Müller [Fri, 15 Apr 2011 14:02:20 +0000 (15:02 +0100)]
baseparse: register boxed type for GstBaseFrameParse

To make this usable for bindings.

https://bugzilla.gnome.org/show_bug.cgi?id=518857

13 years agoqueue2: Add missing ) to the ring-buffer-max-size property description
Sebastian Dröge [Fri, 15 Apr 2011 11:57:47 +0000 (13:57 +0200)]
queue2: Add missing ) to the ring-buffer-max-size property description

13 years agobaseparse: Remove unused but set variable
Robert Swain [Fri, 15 Apr 2011 08:53:56 +0000 (10:53 +0200)]
baseparse: Remove unused but set variable

GCC 4.6.0 spits warnings about these.

13 years agobufferlist: Add boxed type for GstBufferListIterator for gobject-introspection
Sebastian Dröge [Thu, 14 Apr 2011 14:06:16 +0000 (16:06 +0200)]
bufferlist: Add boxed type for GstBufferListIterator for gobject-introspection

13 years agoparse: Add boxed type for GstParseContext for gobject-introspection
Sebastian Dröge [Thu, 14 Apr 2011 13:59:28 +0000 (15:59 +0200)]
parse: Add boxed type for GstParseContext for gobject-introspection

13 years agogst: Add some more gobject-introspection annotations
Sebastian Dröge [Thu, 14 Apr 2011 13:51:24 +0000 (15:51 +0200)]
gst: Add some more gobject-introspection annotations

13 years agomultiqueue: Don't leak the sinkpad name
Sebastian Dröge [Thu, 14 Apr 2011 07:07:48 +0000 (09:07 +0200)]
multiqueue: Don't leak the sinkpad name

13 years agomultiqueue: Don't leak pads in the named pads unit test
Sebastian Dröge [Thu, 14 Apr 2011 07:07:25 +0000 (09:07 +0200)]
multiqueue: Don't leak pads in the named pads unit test

13 years agoutils: Fix caps leaks in gst_element_factory_can_accept_{any,all}_caps_in_direction()
Sebastian Dröge [Thu, 14 Apr 2011 06:59:14 +0000 (08:59 +0200)]
utils: Fix caps leaks in gst_element_factory_can_accept_{any,all}_caps_in_direction()

13 years agoparser: Allow element names to begin with digits
David Schleef [Wed, 13 Apr 2011 16:20:13 +0000 (09:20 -0700)]
parser: Allow element names to begin with digits

13 years agotests: Add test for greatest common divisor
David Schleef [Wed, 13 Apr 2011 17:24:33 +0000 (10:24 -0700)]
tests: Add test for greatest common divisor

13 years agoelements: Fix pad callbacks so they handle when parent goes away
Ole André Vadla Ravnås [Thu, 6 Jan 2011 17:11:31 +0000 (18:11 +0100)]
elements: 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 agobase: Fix pad callbacks so they handle when parent goes away
Ole André Vadla Ravnås [Thu, 6 Jan 2011 17:11:31 +0000 (18:11 +0100)]
base: 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 agoghostpad: Fix pad callbacks so they handle when parent goes away
Ole André Vadla Ravnås [Thu, 6 Jan 2011 17:11:31 +0000 (18:11 +0100)]
ghostpad: 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 agoqueue2: prevent calculation with GST_CLOCK_TIME_NONE in update_time_level()
Janne Grunau [Wed, 13 Apr 2011 15:26:54 +0000 (17:26 +0200)]
queue2: prevent calculation with GST_CLOCK_TIME_NONE in update_time_level()

13 years agotests: fix unusued-but-assigned-variable warnings with gcc 4.6
Tim-Philipp Müller [Mon, 11 Apr 2011 14:08:30 +0000 (15:08 +0100)]
tests: fix unusued-but-assigned-variable warnings with gcc 4.6

13 years agotests: disable test_many_bins unit test for now
Tim-Philipp Müller [Mon, 11 Apr 2011 12:04:32 +0000 (13:04 +0100)]
tests: disable test_many_bins unit test for now

It fails on the OSX bot (both with git and the last release), and
it doesn't really test anything useful, so may just as well disable
it for now.

13 years agopluginloader: fix compiler warnings
Tim-Philipp Müller [Mon, 11 Apr 2011 11:51:36 +0000 (12:51 +0100)]
pluginloader: fix compiler warnings

Cast string constants to make compiler happy.

13 years agotests: allow more time for the test_many_bins pipeline to preroll
Tim-Philipp Müller [Mon, 11 Apr 2011 11:04:34 +0000 (12:04 +0100)]
tests: allow more time for the test_many_bins pipeline to preroll

Hopefully makes this test work on the OSX build bot and other
not-so-powerful machines.

https://bugzilla.gnome.org/show_bug.cgi?id=646624

13 years agopluginloader: make sure gst-plugin-scanner is called with the right arch on OSX
Jan Schmidt [Mon, 11 Apr 2011 10:29:00 +0000 (11:29 +0100)]
pluginloader: make sure gst-plugin-scanner is called with the right arch on OSX

On OSX, GStreamer might be built as a 'fat/universal' binary containing
both 32-bit and 64-bit code. We must take care that gst-plugin-scanner
is executed with the same architecture as the GStreamer core, otherwise
bad things may happen and core/scanner will not be able to communicate
properly.

Should fix issues with (32-bit) firefox using a 32-bit GStreamer core
which then spawns a 'universal' gst-plugin-scanner binary which gets
run in 64-bit mode, causing 100% cpu usage / busy loops or just hanging
firefox until killed.

https://bugzilla.gnome.org/show_bug.cgi?id=615357

13 years agopad: Allow tracking of buffers in GST_SCHEDULING debug output
Robert Swain [Mon, 11 Apr 2011 09:05:24 +0000 (11:05 +0200)]
pad: Allow tracking of buffers in GST_SCHEDULING debug output

As GST_SCHEDULING reports when buffers pass through pads due to
gst_pad_push calls, they are a good way of tracking the progress of
buffers through pipelines. As such, adding output of the buffer pointers
to these messages allows tracking of specific buffers, easing debugging.

13 years agoandroid: make it ready for androgenizer
Alessandro Decina [Thu, 27 Jan 2011 13:33:08 +0000 (14:33 +0100)]
android: make it ready for androgenizer

Remove the android/ top dir
Fixe the Makefile.am to be androgenized

To build gstreamer for android we are now using androgenizer which generates the needed Android.mk files.
Androgenizer can be found here: http://git.collabora.co.uk/?p=user/derek/androgenizer.git

13 years agotrace: don't put code with side effects into g_return_if_fail()
Tim-Philipp Müller [Sat, 9 Apr 2011 22:54:20 +0000 (23:54 +0100)]
trace: don't put code with side effects into g_return_if_fail()

13 years agodocs: minor fixes for baseparse docs
Tim-Philipp Müller [Sat, 9 Apr 2011 21:57:46 +0000 (22:57 +0100)]
docs: minor fixes for baseparse docs

Class vfunc references still aren't right, no idea what
the correct markup for those is.

13 years agoelement: unref event in default_send_event in case element has no pads
Tim-Philipp Müller [Sat, 9 Apr 2011 17:04:55 +0000 (18:04 +0100)]
element: unref event in default_send_event in case element has no pads

Spotted by  Haakon Sporsheim.

13 years agobaseparse: minor variable name clean-up
Tim-Philipp Müller [Fri, 8 Apr 2011 18:07:02 +0000 (19:07 +0100)]
baseparse: minor variable name clean-up

13 years agobaseparse: rename _set_frame_props() to _set_frame_rate()
Tim-Philipp Müller [Fri, 8 Apr 2011 14:31:14 +0000 (15:31 +0100)]
baseparse: rename _set_frame_props() to _set_frame_rate()

Seems like the best fit to what it does, and is shorter than
set_frame_properties() which might also have been confusing
because of GstBaseParseFrame.

https://bugzilla.gnome.org/show_bug.cgi?id=518857

13 years agobaseparse: replace format flags with gst_base_parse_set_{passthrough,syncable,has_tim...
Tim-Philipp Müller [Wed, 6 Apr 2011 16:43:27 +0000 (17:43 +0100)]
baseparse: replace format flags with gst_base_parse_set_{passthrough,syncable,has_timing_info}

This is more in line with e.g. GstBaseTransform's API, and makes for nicer
to read code. No getters for now since I don't see any use case for them,
the API is for subclasses, which usually know what format they're
dealing with already and hence know what they've set.

https://bugzilla.gnome.org/show_bug.cgi?id=518857

13 years agobaseparse: make DRAIN and SYNC flags on baseparse, not the frame, and change to DRAIN...
Tim-Philipp Müller [Mon, 4 Apr 2011 16:58:59 +0000 (17:58 +0100)]
baseparse: make DRAIN and SYNC flags on baseparse, not the frame, and change to DRAINING and LOST_SYNC

The first because it seems a better fit conceptually, the second
to express booleanness. Also change the accessor macros for subclasses
to GST_BASE_PARSE_DRAINING and GST_BASE_PARSE_LOST_SYNC.

https://bugzilla.gnome.org/show_bug.cgi?id=518857

13 years agobaseparse: add some padding to GstBaseParseFrame
Tim-Philipp Müller [Sat, 2 Apr 2011 13:18:57 +0000 (14:18 +0100)]
baseparse: add some padding to GstBaseParseFrame

Esp. since it's usually allocated on the stack.

https://bugzilla.gnome.org/show_bug.cgi?id=518857

13 years agobaseparse: fix typo in docs for GST_BASE_PARSE_FORMAT_FLAG_PASSTHROUGH
Tim-Philipp Müller [Sat, 2 Apr 2011 13:08:46 +0000 (14:08 +0100)]
baseparse: fix typo in docs for GST_BASE_PARSE_FORMAT_FLAG_PASSTHROUGH

https://bugzilla.gnome.org/show_bug.cgi?id=518857

13 years agobaseparse: use GQueue instead of GList for queued frames
Tim-Philipp Müller [Sat, 2 Apr 2011 13:04:42 +0000 (14:04 +0100)]
baseparse: use GQueue instead of GList for queued frames

and make buffer metadata writable before setting caps on queued
buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=646341

13 years agobaseparse: add GST_BASE_PARSE_FLOW_QUEUED to queue buffers until caps are known
Zaheer Abbas Merali [Sat, 2 Apr 2011 12:02:01 +0000 (13:02 +0100)]
baseparse: add GST_BASE_PARSE_FLOW_QUEUED to queue buffers until caps are known

This is useful for parser like flacparse or h264parse which may need to process
some buffers before they can construct the final caps, in which case they may
want to delay pushing the initial buffers until the full and proper caps are
known.

https://bugzilla.gnome.org/show_bug.cgi?id=646341

13 years agobaseparse: add to docs and fix up gtk-doc markup a little
Tim-Philipp Müller [Thu, 31 Mar 2011 14:50:22 +0000 (15:50 +0100)]
baseparse: add to docs and fix up gtk-doc markup a little

And add Since markers.