platform/upstream/gstreamer.git
3 years agoglcolorconver: Return empty caps in transform_caps if fails.
He Junyan [Fri, 6 Nov 2020 08:18:14 +0000 (16:18 +0800)]
glcolorconver: Return empty caps in transform_caps if fails.

We should not return a NULL in transform_caps() function. The NULL
will generate a assert of:
  "transform_caps returned caps which are not a real subset of the
   filter caps"
in transform base class.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/922>

3 years agoglbasefilter: Need to check the display before lock it.
He Junyan [Fri, 6 Nov 2020 07:36:01 +0000 (15:36 +0800)]
glbasefilter: Need to check the display before lock it.

In find_gl_context_unlocked(), the display of filter may be NULL
and can cause crash if we directly access and lock it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/922>

3 years agoglbasefilter: Delete the un-paired unlock in change_state().
He Junyan [Fri, 6 Nov 2020 07:36:21 +0000 (15:36 +0800)]
glbasefilter: Delete the un-paired unlock in change_state().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/922>

3 years agoaudioencoder: Fix incorrect GST_LOG_OBJECT usage
Arun Raghavan [Thu, 3 Dec 2020 11:50:07 +0000 (06:50 -0500)]
audioencoder: Fix incorrect GST_LOG_OBJECT usage

GstBuffer is not a GstObject, so this causes a warning to be emitted.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/956>

3 years agortp/basepayload: implement support for rtp header extensions
Matthew Waters [Fri, 10 Jul 2020 05:33:46 +0000 (15:33 +1000)]
rtp/basepayload: implement support for rtp header extensions

New signals are added for managing the internal list of rtp header
extension implementations read by a specific depayloader instance.

If the 'extmap-$NUM' field is present in the src caps, then an
extension implementation will be requested but is not required to be able
to negotiate correctly.  An extension will be requested using the
'request-extension' signal if none could be found internally.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/748>

3 years agortp/basedepayload: implement support for rtp header extensions
Matthew Waters [Fri, 10 Jul 2020 05:30:57 +0000 (15:30 +1000)]
rtp/basedepayload: implement support for rtp header extensions

New signals are added for managing the internal list of rtp header
extension implementations read by a specific depayloader instance.

If the 'extmap-$NUM' field is present in the sink caps, then an
extension implementation will be requested but is not requited to be
able to negotiate correctly.  An extension will be requested using the
'request-extension' signal if none could be found internally.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/748>

3 years agortp: add base object for reading/writing rtp header extensions (RFC5285)
Matthew Waters [Fri, 10 Jul 2020 05:14:09 +0000 (15:14 +1000)]
rtp: add base object for reading/writing rtp header extensions (RFC5285)

Facilitates the creation of rtp header extension implementations that
can be reused across applications.

Implementations are registered into the GStreamer registry as elements
(idea from GstRTSPExtension) and can be retrieved by URI or filtered
manually.  RTP header extensions must have the classification
"Network/Extension/RTPHeader" to be considered as a RTP Header
extension.

Fixes: https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/777
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/748>

3 years agogl/eagl: Fix automatic resize behaviour
Mart Raudsepp [Thu, 26 Nov 2020 12:52:26 +0000 (14:52 +0200)]
gl/eagl: Fix automatic resize behaviour

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/681
added a layoutSubViews, which never gets called, because it should have been
called layoutSubviews (non-capital "v"). However after fixing that, it still
doesn't work correctly, because window_width/height values are immediately
updated and then draw_cb will never trigger the resize path, because the
values are already up to date.
Update the values inside the resize path again instead, so the check for
entering the resize path is logically always correct.
This makes the layoutSubviews unnecessary, as it only updated the internal
size values prematurely, so it is deleted instead of method naming fixed.

These changes were originally done to avoid accessing UIKit objects on the
main thread, but no additional accesses are added here, only internal
private variable assignments under the same draw_lock, so there should be
no threading issues reintroduced.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/945>

3 years agogl/eagl: Fix resize condition check in draw_cb to not get called unnecessarily
Mart Raudsepp [Thu, 26 Nov 2020 12:45:05 +0000 (14:45 +0200)]
gl/eagl: Fix resize condition check in draw_cb to not get called unnecessarily

A CGSize contains CGFloat values (a typedef to double or float), which means
that the values aren't equal, despite it being equal after they are cast to
int by assigning them to window_height/width private members. This leads to
excessive gst_gl_window_resize calls on each frame, at least if the CGFloat
value has a .5 decimal value, e.g. 103.5.
Fix it by storing them as CGFloat instead of gint.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/945>

3 years agovideo/converter: increase the number of cache lines for resampling
Matthew Waters [Fri, 20 Nov 2020 05:51:09 +0000 (16:51 +1100)]
video/converter: increase the number of cache lines for resampling

The exising hardcoded max default does not account for the possible
-1 offset when retrieving lines for resampling.  As a result, when
another chain has the same number of cache lines (4), the resample
operation would be attempting to generate 5 lines with a cache size
of 4 and would overwrite the first cache line.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/821

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/938>

3 years agocompositor: Don't crash in prepare_frame() if the pad was just removed
Sebastian Dröge [Tue, 1 Dec 2020 11:13:40 +0000 (13:13 +0200)]
compositor: Don't crash in prepare_frame() if the pad was just removed

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/950>

3 years agovideometa: gir annotate the size of plane array in new API
Guillaume Desmottes [Tue, 1 Dec 2020 08:31:48 +0000 (09:31 +0100)]
videometa: gir annotate the size of plane array in new API

Fix #838

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/949>

3 years agoencodebin: Enhance algorithm to check if profiles can be plugged to muxer
Thibault Saunier [Thu, 19 Nov 2020 22:06:12 +0000 (19:06 -0300)]
encodebin: Enhance algorithm to check if profiles can be plugged to muxer

In particular muxing sinks have specific set of sinkpads and we need
to handle that.

3 years agoencodebin: Ignore element classification when a muxer name specified
Thibault Saunier [Thu, 19 Nov 2020 22:03:29 +0000 (19:03 -0300)]
encodebin: Ignore element classification when a muxer name specified

Instead of going through the list of known muxers go ahead and
instantiate the muxer specified as 'preset name' as this specifies
the exact element factory name to use.

3 years agopbutils: Add support for muxing sinks usage in encoding profiles
Thibault Saunier [Fri, 13 Nov 2020 19:32:45 +0000 (16:32 -0300)]
pbutils: Add support for muxing sinks usage in encoding profiles

3 years agoencoding: Implement encodebin2 with an ON_REQUEST src pad
Thibault Saunier [Fri, 13 Nov 2020 19:32:45 +0000 (16:32 -0300)]
encoding: Implement encodebin2 with an ON_REQUEST src pad

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/304

3 years agoencodebin: Split implementation into a base class
Thibault Saunier [Fri, 13 Nov 2020 16:06:32 +0000 (13:06 -0300)]
encodebin: Split implementation into a base class

Create EncodeBaseBin as a base class for the existing
encodebin to allow other implementations.

Co-authored with Jan Schmidt <jan@centricular.com>

3 years agoglcontext: wgl: Implement check_feature vfunc
Seungha Yang [Fri, 27 Nov 2020 19:14:14 +0000 (04:14 +0900)]
glcontext: wgl: Implement check_feature vfunc

There are several WGL specific extenstions such as WGL_NV_DX_interop.
Currently we have no WGL specific extension support and
this commit is also only for debugging purpose.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/947>

3 years agoaudiodecoder: Move max_errors out of GstAudioDecoderContext
Sanchayan Maity [Fri, 27 Nov 2020 06:11:36 +0000 (11:41 +0530)]
audiodecoder: Move max_errors out of GstAudioDecoderContext

Currently max-errors gets set during init to default or via property.
However, if a decoder element calls gst_audio_decoder_reset with 'full'
argument set to TRUE, it would result in all the fields of context being
zeroed with memset. This effectively results in max-errors getting a
value of 0 overriding the default or user requested value set during
init.

This would result in calls to GST_AUDIO_DECODER_ERROR which track error
counts and allow max-errors, to be ineffective.

To fix this move max-errors out of GstAudioDecoderContext, as changes to
context should not affect this. The error_count is anyways also in
GstAudioDecoderPrivate and not in context.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/946>

3 years agoopus: disable docs option in meson fallback
Stéphane Cerveau [Thu, 26 Nov 2020 16:37:01 +0000 (17:37 +0100)]
opus: disable docs option in meson fallback

As the docs are not necessary, no need to keep it
enabled.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/943>

3 years agovideo: Provide "deprecated in" version for gst_video_color_transfer fns
Marijn Suijten [Wed, 25 Nov 2020 19:16:53 +0000 (20:16 +0100)]
video: Provide "deprecated in" version for gst_video_color_transfer fns

As requested in [1].

[1]: https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/940#note_706437

Fixes: d0f36c7e1 ("video: Rename video_color_transfer to video_transfer_function")

3 years agoaudio: Move fill_silence into audio_format_info
Marijn Suijten [Tue, 24 Nov 2020 19:54:49 +0000 (20:54 +0100)]
audio: Move fill_silence into audio_format_info

With the function named gst_audio_format_fill_silence it would get
associated to the GstAudioFormat type in .gir which is incorrect and
confusing. See [1] for the discussion sparking this change.

https://gitlab.freedesktop.org/gstreamer/gstreamer-rs/-/merge_requests/630#note_694795

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/940>

3 years agoxvimagesink: Add support for the XV_COLORSPACE attribute.
Jan Schmidt [Wed, 25 Nov 2020 12:54:08 +0000 (23:54 +1100)]
xvimagesink: Add support for the XV_COLORSPACE attribute.

The XV_COLORSPACE attribute exists on some Xv adapters, with the same
semantics as the XV_ITURBT_709 attribute that was already supported.

A value of 0 is bt601, and 1 is for bt709 colorspace.

Fixes color shifting issues displaying bt709 content on some Xv
adapters.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/941>

3 years agodecodebin3: Take into account decoder latency for interleave size
Edward Hervey [Fri, 13 Nov 2020 16:45:12 +0000 (17:45 +0100)]
decodebin3: Take into account decoder latency for interleave size

Some decoders might introduce quite large latencies, which would result in
multiqueue draining out on some streams. In order to avoid that, check the
latency of decoders and adjust the minimum interleave time of multiqueue
accordingly.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/800

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/931>

3 years agodecodebin3: Don't leak filter caps
Edward Hervey [Thu, 19 Nov 2020 13:35:32 +0000 (14:35 +0100)]
decodebin3: Don't leak filter caps

Introduced by previous commit

See https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/933

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/934>

3 years agodecodebin3: Properly handle caps query with no filter
Edward Hervey [Sat, 14 Nov 2020 08:43:59 +0000 (09:43 +0100)]
decodebin3: Properly handle caps query with no filter

There's no guarantee that upstream elements will do queries with specified
filter caps. If that's the case, just return GST_CAPS_ANY

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/933>

3 years agoopus: add subproject fallback
Tim-Philipp Müller [Sat, 7 Nov 2020 17:30:19 +0000 (17:30 +0000)]
opus: add subproject fallback

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/924>

3 years agocompositor: expose zero-size-is-unscaled property
Mathieu Duponchelle [Sat, 7 Nov 2020 01:59:09 +0000 (02:59 +0100)]
compositor: expose zero-size-is-unscaled property

When that property is left to its default, the width and height
property considers frames from input pads with width or height <= 0
should be left unscaled in that dimension.

Setting this property to FALSE changes that behaviour to < 0, as when
animating these properties, 0 should be a valid end value (eg. shrinking
an input stream until it disappears).

The default value of the width and height properties is set to -1, so that
the default behaviour stays consistent whether that new property is set
or not.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/923>

3 years agovideo-converter: switch to using a task pool ..
Mathieu Duponchelle [Tue, 27 Oct 2020 23:01:16 +0000 (00:01 +0100)]
video-converter: switch to using a task pool ..

.. and make use of that API in videoaggregator.

When setting certain properties, such as cropping or the scaled
size of pads, a new converter is created by videoaggregator.

Before that patch, this implied spawning new threads, potentially
at each aggregate cycle when interpolating pad properties. This
is obviously wasteful, and re-using a task pool removes that
overhead.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/896>

3 years agoFix segfault when using invalid encoding profile
David Keijser [Tue, 10 Nov 2020 22:26:39 +0000 (23:26 +0100)]
Fix segfault when using invalid encoding profile

Trying to use gst_encoding_profile_get_file_extension on a
GstEncodingProfile with a cap containing a typo would result in strcmp
being called with NULL. Instead use g_strcmp0 that handles this case.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/929>

3 years agovideoaggregator: Fix locking around vagg->info
Jan Alexander Steffens (heftig) [Tue, 3 Nov 2020 16:00:53 +0000 (17:00 +0100)]
videoaggregator: Fix locking around vagg->info

Take `GST_OBJECT_LOCK` when writing `vagg->info`, so that reading in
subclasses is protected against races, as documented in the struct.

    /*< public >*/
    /* read-only, with OBJECT_LOCK */
    GstVideoInfo                  info;

`gst_video_aggregator_default_negotiated_src_caps` should take the
`GST_VIDEO_AGGREGATOR_LOCK` to avoid racing with
`gst_video_aggregator_reset` called by
`gst_video_aggregator_release_pad` of the last sinkpad. Otherwise it can
happen that `latency = gst_util_uint64_scale (...` gets called with a
zero framerate.

There doesn't seem to be any reason not to use the local `info` instead
of `vagg->info`, so do that.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/915>

3 years agovideo: fix doc warning
Guillaume Desmottes [Mon, 26 Oct 2020 07:59:34 +0000 (08:59 +0100)]
video: fix doc warning

@mode has been renamed to
gst_video_decoder_set_interlaced_output_state() but not in the header
file, raising a doc warning.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/894>

3 years agovideo-converter: Skip input lines where possible.
Jan Schmidt [Wed, 4 Nov 2020 17:43:47 +0000 (04:43 +1100)]
video-converter: Skip input lines where possible.

There is a case where there are no lines in the temp cache, and
it's possible to skip straight to the request line and not generate
intermediate ones. This is really only beneficial when doing
nearest-neighbour downscaling, as other methods generally require
all input lines sequentially to generate the output. In that case,
this change has no effect and all lines are generated and cached
as before.

As a side effect however, this fixes corruption when downscaling
using nearest-neighbour, as interactions with the pass_alloc flag
and reuse of temporary lines causes the unecessarily-generated
cache lines to overwrite the final output.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/919>

3 years agoexamples: Add a compositor mosaic example.
Jan Schmidt [Tue, 3 Nov 2020 15:11:30 +0000 (02:11 +1100)]
examples: Add a compositor mosaic example.

Add an example of configuring compositor to display 2 videos
side-by-side in a mosaic, and additionally configure some
video-converter properties to demonstrate how that works..

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/919>

3 years agogluploadelement: Avoid race condition in propose_allocation().
He Junyan [Wed, 4 Nov 2020 15:05:27 +0000 (23:05 +0800)]
gluploadelement: Avoid race condition in propose_allocation().

The inside upload and context may have race condition in the function
of propose_allocation(). They may be destroyed while this function is
stilling using it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/916>

3 years agogluploadelement: Avoid race condition of base class' context.
He Junyan [Wed, 4 Nov 2020 09:02:13 +0000 (17:02 +0800)]
gluploadelement: Avoid race condition of base class' context.

The base class' context may change, we should use the common API
with lock to access it, rather than the just directly access the
struct field.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/916>

3 years agomeson: Enable some MSVC warnings for parity with GCC/Clang
Nirbheek Chauhan [Wed, 4 Nov 2020 13:12:11 +0000 (18:42 +0530)]
meson: Enable some MSVC warnings for parity with GCC/Clang

This makes it easier to do development with MSVC by making it warn
on common issues that GCC/Clang error out for in our CI configuration.

Continuation from https://gitlab.freedesktop.org/gstreamer/gst-build/-/merge_requests/223

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/918>

3 years agogluploadelement: Avoid race condition of inside upload creation.
He Junyan [Tue, 3 Nov 2020 12:19:16 +0000 (20:19 +0800)]
gluploadelement: Avoid race condition of inside upload creation.

The operations for the inside GstGLUploadElement->upload have race
condition. The _transform_caps() will creates this object if it does
not exist, while the _stop() and change_state() can destroy this object.
The _transform_caps() is called by the gst_base_transform_query(),
so it does not hold the stream lock. It may use the upload while the
_stop() and change_state() has already destroy that object, and then
crash.

Fix: #645
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/913>

3 years agogl: Fix prototype of glGetSynciv()
Sebastian Dröge [Tue, 3 Nov 2020 14:40:38 +0000 (16:40 +0200)]
gl: Fix prototype of glGetSynciv()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/914>

3 years agovideoaggregator: Guarantee that the output format is supported
Thibault Saunier [Fri, 30 Oct 2020 16:56:16 +0000 (13:56 -0300)]
videoaggregator: Guarantee that the output format is supported

In the case `videoaggregator` is set as allowing format conversions,
and as we convert only on the sinkpads, we should ensure that the
chosen format is usable by the subclass. This in turns implies
that the format is usable on the srcpad.

When doing conversion *any* format can be used on the sinkpads, and this
is the only way that we can avoid race conditions during renegotiations
so we can not change that fact, we just need to ensure that the chosen
intermediary format is usable, which was not actually ensured before
that patch.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/834

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/909>

3 years agovideodecoder: Don't assume GstVideoChromaSite and GstVideoColorimetry
Seungha Yang [Fri, 30 Oct 2020 18:14:20 +0000 (03:14 +0900)]
videodecoder: Don't assume GstVideoChromaSite and GstVideoColorimetry

Even if given GstVideoChromaSite and/or GstVideoColorimetry has unknown
value(s), assumption for an unknown value should be done by subclass or
downstream element, not a role of video decoder. And subclass might
want to output unknown value as is.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/910>

3 years agovideodecoder: Remove trailing whitespace
Seungha Yang [Fri, 30 Oct 2020 18:16:00 +0000 (03:16 +0900)]
videodecoder: Remove trailing whitespace

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/910>

3 years agotcpserversink: Don't assume g_socket_get_remote_address succeeds
Jan Alexander Steffens (heftig) [Fri, 30 Oct 2020 12:52:29 +0000 (13:52 +0100)]
tcpserversink: Don't assume g_socket_get_remote_address succeeds

When the client disconnects immediately after connecting, the remote
address is no longer available.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/908>

3 years agovideo-aggregator: Fix renegotiation when using convert pads
Thibault Saunier [Thu, 29 Oct 2020 19:38:44 +0000 (16:38 -0300)]
video-aggregator: Fix renegotiation when using convert pads

Since 23189c60f4cff998c7880e1768cee2f6d1b719d0 we started using the
useless result of `modified_caps` which, was never used since it was
introduced 7 years ago (in videomixer2). The intersection is useless and
we should just avoid doing it at all (which was always the case before)
as it can end up failing renegotiation for bad reasons.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/905>

3 years agogst-play: Add support for audio mute control
Seungha Yang [Wed, 27 May 2020 14:20:25 +0000 (23:20 +0900)]
gst-play: Add support for audio mute control

Assign 'm' keyboard input to control audio mute state

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/674>

3 years agotests: allocator: Fix FDMemory portability issue
Nicolas Dufresne [Wed, 28 Oct 2020 19:51:27 +0000 (15:51 -0400)]
tests: allocator: Fix FDMemory portability issue

This fixes few issues in the test but mainly some portability issue reported
on Ubutun. The test now uses a randomly name tempory file located into system
default tempory location and uses glib wrappers when available.

Fixes !895

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/901>

3 years agogl: optimize glcolorbalance by precomputing shader math
Dinesh Manajipet [Sat, 24 Oct 2020 18:15:19 +0000 (23:45 +0530)]
gl: optimize glcolorbalance by precomputing shader math

Precompute the rgb -> yuv conversion and color balance adjustment math
so that the shader does minimal work per pixel.

Merging these 15+ steps into 3 steps let us jump from choppy
360p video to smooth 720p video on our underpowered embedded system.

If we can remove the clamp() step inside the shader, or apply it after
rgba
conversion, there are more performance benefits to reap. But I am not
sure
what the side effects will be in that case.

<https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/893>

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/893>

3 years agogl/build: use the brcm GL libraries on the rpi
Matthew Waters [Wed, 28 Oct 2020 12:26:27 +0000 (23:26 +1100)]
gl/build: use the brcm GL libraries on the rpi

Upstream RPi has moved to a completely separate GL library names now due
to conflicts.

See https://github.com/RPi-Distro/repo/issues/134

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/898>

3 years agofdmemory: Allow for change of protection mode
Tobias Ronge [Tue, 27 Oct 2020 13:12:42 +0000 (14:12 +0100)]
fdmemory: Allow for change of protection mode

After a memory has been unmapped, protection mode can now be changed
when mapping it again.

See https://bugzilla.gnome.org/show_bug.cgi?id=789952.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/895>

3 years agovideoaggregator: document and fix locking in convert pad
Mathieu Duponchelle [Wed, 28 Oct 2020 00:19:37 +0000 (01:19 +0100)]
videoaggregator: document and fix locking in convert pad

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/897>

3 years agogldeinterlace: fix doc blurb
Tim-Philipp Müller [Wed, 28 Oct 2020 13:22:09 +0000 (13:22 +0000)]
gldeinterlace: fix doc blurb

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/899>

3 years agoMeson: Use pkg-config generator
Xavier Claessens [Sun, 8 Apr 2018 00:52:12 +0000 (20:52 -0400)]
Meson: Use pkg-config generator

3 years agourisourcebin: Fix crash caused by use after free
Seungha Yang [Thu, 22 Oct 2020 10:16:42 +0000 (19:16 +0900)]
urisourcebin: Fix crash caused by use after free

OutputSlotInfo doesn't hold ref of queue, so gst_bin_remove()
will free the queue memory.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/888>

3 years agotheoradec: Set telemetry options only if they are nonzero
Zebediah Figura [Wed, 21 Oct 2020 19:55:40 +0000 (14:55 -0500)]
theoradec: Set telemetry options only if they are nonzero

Setting telemetry options, even to zero, causes libtheora to enable an expensive code path. For large enough videos (e.g. 1920x1080) this can increase the time to decode each frame by 30-40 ms, which can be enough to cause noticeable stutter.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/887>

3 years agoglslstage: delete shader on finalize of stage
Michael Tretter [Wed, 21 Oct 2020 09:42:54 +0000 (11:42 +0200)]
glslstage: delete shader on finalize of stage

GLSLstage creates the glShader using glCreateShader, but never calls
glDeleteShader if the glShader is not used anymore. This forces the GL
library to keep the compiled shader around, because it might be used in
the future. Therefore, the glShader is leaked whenever a GLSLStage is
destroyed.

Fix the leak by deleting the glShader when finishing the GLSLStage.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/886>

3 years agoaudio/videodecoder: Initialize max_errors in instance_init()
Sebastian Dröge [Tue, 20 Oct 2020 09:44:21 +0000 (12:44 +0300)]
audio/videodecoder: Initialize max_errors in instance_init()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/882>

3 years agoaudio/videodecoder: Don't reset max-errors in reset()
Sebastian Dröge [Tue, 20 Oct 2020 08:51:08 +0000 (11:51 +0300)]
audio/videodecoder: Don't reset max-errors in reset()

Otherwise setting the property on the elements has no effect at all
because it's immediately reset during startup.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/881>

3 years agoexamples: Add an uridecodebin3 example
Seungha Yang [Mon, 12 Oct 2020 18:34:07 +0000 (03:34 +0900)]
examples: Add an uridecodebin3 example

Demonstrate a way to select multiple streams

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/880>

3 years agodecodebin3: Store stream-start event on output pad before exposing it
Seungha Yang [Mon, 19 Oct 2020 11:03:40 +0000 (20:03 +0900)]
decodebin3: Store stream-start event on output pad before exposing it

It's required for users to be able to figure out associated GstStream
object with a pad on pad-added callback.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/880>

3 years agoAdd some missing nullable annotations
Sebastian Dröge [Sat, 17 Oct 2020 09:35:16 +0000 (12:35 +0300)]
Add some missing nullable annotations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/876>

3 years agotypefind/xdgmime: Validate mimetypes to be valid GstStructure names before using...
Sebastian Dröge [Sat, 17 Oct 2020 07:42:49 +0000 (10:42 +0300)]
typefind/xdgmime: Validate mimetypes to be valid GstStructure names before using them

On macOS, for example, "text/*" can be returned as mimetype for
plaintext files but we don't allow '*' in structure names and this would
cause critical warnings.

It's a valid mimetype but not a valid structure name.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/616

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/875>

3 years agomeson: update glib minimum version to 2.56
Stéphane Cerveau [Thu, 15 Oct 2020 16:20:37 +0000 (18:20 +0200)]
meson: update glib minimum version to 2.56

In order to support the symbol g_enum_to_string in various
project using GStreamer ( gst-validate etc.), the glib minimum
version should be 2.56.0.

Remove compat code as glib requirement
is now > 2.56

Version used by Ubuntu 18.04 LTS

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/874>

3 years agovideoconvert: Enhance debug message to be more human readable
Seungha Yang [Sun, 20 Sep 2020 10:47:08 +0000 (19:47 +0900)]
videoconvert: Enhance debug message to be more human readable

Numerical representation of GstVideoFormat is not debugging
friendly

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/822>

3 years agouridecodebin3: Forward upstream events to decodebin3 directly
Seungha Yang [Tue, 13 Oct 2020 08:57:53 +0000 (17:57 +0900)]
uridecodebin3: Forward upstream events to decodebin3 directly

Otherwise default handler will try to forward it to sink element
or srcpad which might not be available at that moment.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/867>

3 years agotests: appsrc: Fix unstable test case
Seungha Yang [Mon, 14 Sep 2020 11:01:18 +0000 (20:01 +0900)]
tests: appsrc: Fix unstable test case

Wait all buffers to be consumed before sending flush seek event,
so that checking timestamp and segment as expected.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/816>

3 years agovideoaggregator: Limit accepted caps by template caps
Jan Alexander Steffens (heftig) [Tue, 13 Oct 2020 13:31:41 +0000 (15:31 +0200)]
videoaggregator: Limit accepted caps by template caps

This seems like an obvious mistake, since `modified_caps` was created
but not used.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/868>

3 years agovideoaggregator: Log the format name instead of index
Jan Alexander Steffens (heftig) [Tue, 13 Oct 2020 13:28:54 +0000 (15:28 +0200)]
videoaggregator: Log the format name instead of index

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/868>

3 years agogstrtpbuffer: fix header extension length validation
Will Miller [Mon, 12 Oct 2020 11:21:17 +0000 (12:21 +0100)]
gstrtpbuffer: fix header extension length validation

We validate the header extensions length of an RTP buffer by comparing
it against the block size. Since we multiply the length in words by 4 to
get the length in bytes, a suitably large length could cause a wrapround
of the uint16, giving a lower length which erroneously passes the check
and allows the buffer to be mapped.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/864>

3 years agogl: glslstage: Sync new_default_{fragment,vertex} documentation name
Marijn Suijten [Fri, 11 Sep 2020 18:28:55 +0000 (20:28 +0200)]
gl: glslstage: Sync new_default_{fragment,vertex} documentation name

Documentation for the parameter and return value don't show up when
this header mismatches the function it tries to describe.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/814>

3 years agodecodebin3: only force streams-selected seqnum after a select-streams
François Laignel [Sat, 10 Oct 2020 15:26:42 +0000 (17:26 +0200)]
decodebin3: only force streams-selected seqnum after a select-streams

The initial streams-selected message might not match any select-streams,
in which case it is currently overriden with GST_SEQNUM_INVALID.

Related to:

- https://gitlab.freedesktop.org/gstreamer/gstreamer-rs/-/merge_requests/601

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/860>

3 years agoexamples/gl/gtk: Add missing dependency on gstgl
Jan Alexander Steffens (heftig) [Fri, 9 Oct 2020 15:15:37 +0000 (17:15 +0200)]
examples/gl/gtk: Add missing dependency on gstgl

Occasionally, a clean build would fail like this:

    In file included from ../subprojects/gst-plugins-base/tests/examples/gl/gtk/gstgtk.c:24:
    ../subprojects/gst-plugins-base/gst-libs/gst/gl/gl.h:25:10: fatal error: gst/gl/gl-enumtypes.h: No such file or directory
       25 | #include <gst/gl/gl-enumtypes.h>
          |          ^~~~~~~~~~~~~~~~~~~~~~~

Add the missing dependency so that the headers are generated beforehand.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/855>

3 years agortsptransport: support single port for client_port and server_port
Mathieu Duponchelle [Fri, 4 Sep 2020 19:09:17 +0000 (21:09 +0200)]
rtsptransport: support single port for client_port and server_port

This is explicitly specified as valid in the RFC, where the
syntax for both parameters is:

";" "client_port" "=" port [ "-" port ]
";" "server_port" "=" port [ "-" port ]

This is useful for applications where RTCP is either not necessary
or not possible.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/846>

3 years agovideo-info: clarify GST_VIDEO_INTERLACE_MODE_ALTERNATE doc
Guillaume Desmottes [Fri, 9 Oct 2020 10:02:14 +0000 (12:02 +0200)]
video-info: clarify GST_VIDEO_INTERLACE_MODE_ALTERNATE doc

Fields really should alternate in this mode.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/852>

3 years agovideodecoder: fix output state interlace-mode
Guillaume Desmottes [Mon, 5 Oct 2020 10:07:22 +0000 (12:07 +0200)]
videodecoder: fix output state interlace-mode

When user is passing the actual interlace-mode when calling
gst_video_decoder_set_interlaced_output_state() it should not be
overidden by the input interlace-mode.

Needed to fix #825 as we want to keep interlace-mode=interleaved from
parsers and have the OMX decoder producing interlace-mode=alternate.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/852>

3 years agovideodecoder: add interlace-mode to debug output when setting output state
Guillaume Desmottes [Mon, 5 Oct 2020 09:38:09 +0000 (11:38 +0200)]
videodecoder: add interlace-mode to debug output when setting output state

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/852>

3 years agotests: add gl structs to abi check
Matthew Waters [Thu, 16 Jul 2020 08:10:45 +0000 (18:10 +1000)]
tests: add gl structs to abi check

Tested on x86, x86_64, armv7l, aarch64.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/854>

3 years agobuild: use cpu_family for arch checks
Matthew Waters [Fri, 9 Oct 2020 01:25:21 +0000 (12:25 +1100)]
build: use cpu_family for arch checks

e.g. on 32-bit arm, we may have armv6, armv7l, armv7hf, etc which all
generally have the same layouts.  cpu_family() groups all of these into
just 'arm' that the ABI check table is expecting.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/854>

3 years agomultihandlesink: Don't pass NULL caps to gst_caps_is_equal
Jan Alexander Steffens (heftig) [Thu, 1 Oct 2020 09:22:51 +0000 (11:22 +0200)]
multihandlesink: Don't pass NULL caps to gst_caps_is_equal

Apparently the sinkpad caps can get cleared when shutting down the
pipeline while the sink is working, provoking a critical warning.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/842>

3 years agotimeoverlay: Add elapsed-running-time mode
Sebastian Dröge [Thu, 8 Oct 2020 14:10:30 +0000 (17:10 +0300)]
timeoverlay: Add elapsed-running-time mode

This takes the first running time and the first after each flush-stop as
an offset for the running time that is rendered.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/737>

3 years agoaudio: video: Fix in/outbuf confusion of transform_meta
Jan Alexander Steffens (heftig) [Thu, 8 Oct 2020 11:52:33 +0000 (13:52 +0200)]
audio: video: Fix in/outbuf confusion of transform_meta

There are three instances where in- and outbuf have been swapped. This
didn't affect the correctness of the libs *filter code, but the
videoscale implementation swapped the arguments of meta->transform_func.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/853>

3 years agopango: Add fallback dependency
Xavier Claessens [Thu, 1 Oct 2020 18:45:24 +0000 (14:45 -0400)]
pango: Add fallback dependency

When we'll depend on Meson >=0.55 this can be replaced by [provide]
section in gst-build's pango.wrap file. The dependency variable name can
also be dropped when pango is updated to use override_dependency:
https://gitlab.gnome.org/GNOME/pango/-/merge_requests/250

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/844>

3 years agomeson: Always wrap "prefix" option with join_paths() to make Windows happy
Seungha Yang [Tue, 15 Sep 2020 19:30:40 +0000 (04:30 +0900)]
meson: Always wrap "prefix" option with join_paths() to make Windows happy

Both "\" and "/" separators are allowed on Windows but the problem
is that "\" is not converted to "\\" in this case.
It will result to unknown/unrecognized character escape sequence.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/818>

3 years agomeson: Add GL summary
Xavier Claessens [Fri, 25 Sep 2020 18:05:09 +0000 (14:05 -0400)]
meson: Add GL summary

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/836>

3 years agocompositor: fix YUY2/UYVY/VYUY blending
Matthew Waters [Mon, 21 Sep 2020 13:50:03 +0000 (23:50 +1000)]
compositor: fix YUY2/UYVY/VYUY blending

Don't overead memory.

Fixup for https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/755

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/823>

3 years agomeson: Add a qt5 feature
Nicolas Dufresne [Tue, 15 Sep 2020 18:15:36 +0000 (14:15 -0400)]
meson: Add a qt5 feature

This allow opting out items of the builds the depends on QT5 library.
Auto-detection of QT5 in cross-build requires host tools to match with the
sysroot, and detection of mis-match is not fully reliable.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/817>

3 years agomeson: add abi configuration for meson build
sohwan.park [Mon, 14 Sep 2020 08:56:07 +0000 (17:56 +0900)]
meson: add abi configuration for meson build

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/815>

3 years agocompositor: blend with multiple threads
Matthew Waters [Wed, 15 Jul 2020 05:46:56 +0000 (15:46 +1000)]
compositor: blend with multiple threads

Increases the throughput of compositing by using more CPU cycles across
multiple threads.  Simple cases (the output contains one pixel from at
most one input) can have up to a 70% increase in throughput.  Not so
simple cases are limited by the region with the most number of
composite operations.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/755>

3 years agovideo: Rename video_color_transfer to video_transfer_function
Marijn Suijten [Fri, 11 Sep 2020 18:15:14 +0000 (20:15 +0200)]
video: Rename video_color_transfer to video_transfer_function

Rename remaining `gst_video_color_transfer_{encode,decode}` functions on
the `GstVideoTransferFunction` enumeration to
`gst_video_transfer_function_{encode,decode}` permitting
gobject-introspection to turn these into associated functions and place
them under the respective `<enumeration>` block in gir XML files.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/805>

3 years agovideoaggregator: ensure peek_next_sample() uses the correct caps
Matthew Waters [Fri, 11 Sep 2020 11:40:26 +0000 (21:40 +1000)]
videoaggregator: ensure peek_next_sample() uses the correct caps

gst_pad_get_current_caps() may be wrong when there is a renegotiation in
progress for the pad and we have not yet received or selected the buffer
with different caps yet.

Fix by storing the caps through in a similar way to the existing code
for buffer/video-info selection machinery.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/813>

3 years agovideodecoder: Bump log level of marking a sync point
Stian Selnes [Fri, 18 Aug 2017 15:01:02 +0000 (17:01 +0200)]
videodecoder: Bump log level of marking a sync point

Sync points are often important when debugging and deserves DEBUG
level.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/730>

3 years agovideodecoder: Add API for subclasses to request a new sync point
Sebastian Dröge [Wed, 1 Jul 2020 15:19:09 +0000 (18:19 +0300)]
videodecoder: Add API for subclasses to request a new sync point

This allows subclasses that notice missing reference frames to request a
new sync point to allow seamless decoding again. While doing so the
subclass can also signal whether it wants a) all following input frames
until the sync point to be discarded or b) all output frames until the
sync point to be marked as corrupt.

Sending of force-keyunit events for this can be throttled by the
application via the "min-force-keyunit-interval" property.

This replaces custom behaviour for the same in various decoders, for
example openh264dec.

Based on patches by Haakon Sporsheim <haakon@pexip.com> and
Stian Selnes <stian@pexip.com>.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/730>

3 years agovideodecoder: Add API for subclasses to signal that they need the stream to start...
Sebastian Dröge [Wed, 1 Jul 2020 15:14:40 +0000 (18:14 +0300)]
videodecoder: Add API for subclasses to signal that they need the stream to start with a sync point

If the first frame(s) at the very beginning or after a flush are not a
sync point then the base class would discard them before passing them to
the subclass.

This also fixes the previously broken distance_from_sync handling: it
was never reset at sync points.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/730>

3 years agovideodecoder: Add "discard-corrupted-frames" property
Sebastian Dröge [Wed, 1 Jul 2020 15:11:25 +0000 (18:11 +0300)]
videodecoder: Add "discard-corrupted-frames" property

This can be used by applications to configure decoders so that corrupted
frames are directly discarded instead of being forwarded inside the
pipeline. It is a replacement for the "output-corrupt" property of the
ffmpeg decoders.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/730>

3 years agovideodecoder: Add API for marking output frames as corrupted
Sebastian Dröge [Wed, 1 Jul 2020 15:09:22 +0000 (18:09 +0300)]
videodecoder: Add API for marking output frames as corrupted

This can be used by subclasses to mark output frames as known to be
corrupted, for example if reference frames were missing. ffmpeg's
decoders can signal this.

In addition this flag is propagated downstream if the input frame had it
set.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/730>

3 years agovideodecoder: Remove unused reorder_depth variable
Sebastian Dröge [Tue, 23 Jun 2020 15:40:07 +0000 (18:40 +0300)]
videodecoder: Remove unused reorder_depth variable

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/730>

3 years agortpbasedepayload: Mark GAP events sent because of packet loss as such
Mikhail Fludkov [Wed, 1 Jul 2020 16:18:36 +0000 (19:18 +0300)]
rtpbasedepayload: Mark GAP events sent because of packet loss as such

This allows downstream to distinguish packet loss from normal GAP events
that are sent simply because of gaps in the timeline.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/731>

3 years agomeson: Actually build gstgl without implicit include dirs
Nirbheek Chauhan [Tue, 8 Sep 2020 10:11:55 +0000 (15:41 +0530)]
meson: Actually build gstgl without implicit include dirs

Fixes a typo from https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/800

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/810>

3 years agoglmixer: Fix unit test to actually work reliably
Sebastian Dröge [Thu, 10 Sep 2020 11:19:04 +0000 (14:19 +0300)]
glmixer: Fix unit test to actually work reliably

Don't run the harness in live mode, or otherwise it would output frames
already in the very beginning before a buffer was provided to it due to
timeout.

Also send EOS/a second buffer before pulling a buffer as videoaggregator
has one frame of latency.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/812>

3 years agogl: add support for GBR and GBRA
Haihao Xiang [Thu, 13 Aug 2020 04:49:19 +0000 (12:49 +0800)]
gl: add support for GBR and GBRA

example pipeline:
gst-launch-1.0 videotestsrc ! video/x-raw,format=GBR ! glimagesink

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/789>

3 years agovideoaggregator: Don't require any pads to be configured for negotiating source pad...
Sebastian Dröge [Thu, 20 Aug 2020 06:06:04 +0000 (09:06 +0300)]
videoaggregator: Don't require any pads to be configured for negotiating source pad caps

This is not actually required (anymore?). Source pad caps can be
negotiated at any time regardless of any configured (or existing) sink
pads and videoaggregator comes up with some fixated caps based on the
downstream caps.

Subclasses can override this behaviour as needed by overriding
update_src_caps().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/793>