Release 1.19.90

Update ChangeLogs for 1.19.90

souphttpsink: Protect against spurious wakeups during startup

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1594>

souphttpsrc: Don't use the source element after setup from the session thread

The source element might be gone already if the session is shared with
other source elements.

As a consequence, do all logging via the session object instead of using
the source element.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1594>

souphttpsrc: Don't abort all pending operations on the session if shutting down a source with a shared session

Only do it for a non-shared session. Other sources would otherwise get
their requests cancelled unexpectedly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1594>

souphttpsrc: Don't set boolean to FALSE right after checking that it is FALSE

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1594>

souphttpsrc: soup_session_new_with_options() can't fail with NULL

So don't check for it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1594>

tests: Mark adaptivedemux tests as sometimes timing out

... in addition to all the other issues that were ignored for them
already.

At least the reverse playback test regularly times out, waiting for a
download to finish that has already finished successfully.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1593>

autodetect: fix debug init category

Since the split of elements, the debug category
was default for autodetect.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1590>

soup: move libsoup session into its own thread

Starting with libsoup3, there is no attempt to handle thread safety
inside the library, and it was never considered fully safe before
either. Therefore, move all session handling into its own thread.

The libsoup thread has its own context and main loop. When some
request is made or a response needs to be read, an idle source
is created to issue that; the gstreamer thread issuing that waits
for that to be complete. There is a per-src condition variable to
deal with that.

Since the thread/loop needs to be longer-lived than the soup
session itself, a wrapper object is provided to contain them. The
soup session only has a single reference, owned by the wrapper
object.

It is no longer possible to force an external session, since this
does not seem to be used anywhere within gstreamer and would be
tricky to implement; this is because one would not have to provide
just a session, but also the complete thread arrangement made in
the same way as the system currently does internally, in order to
be safe.

Messages are still built gstreamer-side. It is safe to do so until
the message is sent on the session. Headers are also processed on
the gstreamer side, which should likewise be safe.

All requests as well as reads on the libsoup thread are issued
asynchronously. That allows libsoup to schedule things with as
little blocking as possible, and means that concurrent access
to the session is possible, when sharing the session.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/947

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1555>

vapostproc: Make cb max values symmetrical to their min values.

Intel drivers expose some colorbalance's maximum values much more
bigger than their minimum values, given their middle values (default
value). This means, in practice, that the real middle point between
the maximum and minimum values implies a major change in the color
balance, which is not expected by the GStreamer color balance logic.

This patch makes the given maximum value symmetrical to the minimum
value, given the middle one (default value).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1580>

ci: Add a Windows native cross-arm64 job with a new image

Needed an update to the windows Dockerfile to:

1. Install the 'UniversalBuildTools' workload for Cerbero
2. Install ARM and ARM64 workloads for cross-uwp-universal in Cerbero
3. Install VS 2019 since we need that for ARM64 NEON support in Opus

We can't test UWP in gstreamer.git because glib needs custom patches
for that. It will be tested in Cerbero.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1570>

ci: Remove windows-sdk-8.1 from docker image

We don't need this, we use the Windows 10 SDK provided by Visual
Studio Build Tools.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1570>

subprojects/pcre: Add the wrap so it's cached in the image

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1570>

gst-devtools: Add a cairo option for the cairo dep

So that it can be explicitly disabled to avoid pulling in pixman which
doesn't build on Windows ARM64 yet.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1570>

subprojects/lame: Update to latest wrap

Fixes an issue with xmmintrin on non-x86 machines:

https://github.com/mesonbuild/wrapdb/pull/295

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1570>

subprojects/libpng: Update to latest wrap file

It was getting pulled in automatically via cairo, but the version
there is too old for us. We need the latest to fix Windows ARM64 NEON
support:

> ERROR: No specified compiler can handle file subprojects\libpng-1.6.37\arm/filter_neon.S

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1570>

v4l2codecs: Fix debug assertion in register functions

As now, we warn if the decoder have no support src pixel format, but that
warning is called before the type (hence the debug category) is initialized.
Fix this by moving the debug category init out of the type initialization,
into the register funcitons.

This will fix an assertion that occures in the register function and allow
relevant log to be seen by the users.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1588>

webrtcbin: Chain up to parent constructed method

Failing to do so makes GstWebRTCBin invisible to the leaks tracer.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1587>

aggregator: don't forward reconfigure events

Those will cause us to renegotiate at the next aggregate cycle,
and while at that point we may decide to reconfigure upstream
branches (in practice we don't as this is inherently racy,
and that's the reason why mixer subclasses perform conversion
internally), we certainly don't want to just forward the event
willy-nilly to all our sinkpads.

An actual issue this is fixing is when caps downstream of a
compositor are changed at every samples-selected signal emission,
for the purpose of interpolating the output geometry, and the
compositor has a non-zero latency, the reconfigure events were
forwarded to basesrc, which triggered an allocation query, which
in turn caused aggregator to have to drain (thus not being able
to queue <latency> frames), leading to disastrous effects
(choppy output as compositor couldn't consume frames fast enough,
the higher the latency the choppier the output)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1464>

rtphdrext: Return non-floating references from `gst_rtp_header_extension_create_from_uri()`

The header extension objects are never getting a parent object and using
floating references only complicates usage, especially via dynamic API
like signals.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1582>

rtphdrext: Use `set_metadata()` instead of `set_static_metadata()`

The latter needs a static string.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1583>

VideoInfo, AudioInfo: fix usage with python bindings

* Expose an actual constructor from caps

* Error out in overrides for code that was using the "manual
  allocation" pattern which only worked by chance. Direct
  the script writer to the new_from_caps constructor instead.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-python/-/issues/47

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1571>

docs: Updates to contributing guide

Mostly some modifications for handling security issues.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1370>

docs: Explain how to open a merge request with screenshots

The merge request workflow can be confusing to people unfamiliar with
it, so add screenshots.

Also add a new section on how to revise merge requests, since a lot of
people tend to open new merge requests to make any changes.

Eliminate the separate "How to Prepare a Merge Request for Submission"
section -- merge it with the main section.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1367>

gst-plugins-bad: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1579>

gst-plugins-ugly: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1579>

gst-plugins-good: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1579>

gst-plugins-base: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1579>

gstreamer: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1579>

d3d11av1dec: Fix typo in debug message

Fixing copy and paste mistake, It's AV1 decoder not VP8

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1576>

rtspsrc: Fix critical while serializing timeout element message

The "cause" field wasn't registered as a GEnumValue, so do that.

Fixes this critical in gst_structure_to_string():

`gst_value_serialize: assertion 'G_IS_VALUE (value)' failed`

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1573>

audioaggregator: Return NOT_NEGOTIATED when the configuration is invalid

Otherwise we just end up outputting garbage.

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/957

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1558>

videoaggregator: Remove extra semicolon in macro usage

This is usually necessary to allow gst-indent to treat it as
a statement, but we do not run gst-indent on headers and we do not
have extra semicolons in other places that this macro is used in the
header. Fixes warnings when using the header:

```
In file included from gstreamer/subprojects/gst-plugins-base/gst-libs/gst/video/video.h:185,
                 from XYZ:9001:
gstreamer/subprojects/gst-plugins-base/gst-libs/gst/video/gstvideoaggregator.h:206:78: warning: ISO C does not allow extra ‘;’ outside of a function [-Wpedantic]
  206 | G_DEFINE_AUTOPTR_CLEANUP_FUNC(GstVideoAggregatorConvertPad, gst_object_unref);
      |                                                                              ^
gstreamer/subprojects/gst-plugins-base/gst-libs/gst/video/gstvideoaggregator.h:214:181: warning: ISO C does not allow extra ‘;’ outside of a function [-Wpedantic]
  214 | G_DECLARE_DERIVABLE_TYPE (GstVideoAggregatorParallelConvertPad, gst_video_aggregator_parallel_convert_pad, GST, VIDEO_AGGREGATOR_PARALLEL_CONVERT_PAD, GstVideoAggregatorConvertPad);
      |                                                                                                                                                                                     ^
```

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1572>

gstplugin: Fix for UWP build

SetThreadErrorMode() API is available on UWP but flag values
are desktop API only. Since error dialogs don't exist on UWP,
we don't need to suppress it

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1569>

va: basetransform: Pass component index not plane index.

This is an issue detected and fixed in commit 3897b24f for other
libraries and elements.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1566>

splitmuxsink: Warn when calculating the next fragment time in timecode mode fails

But only if timecode mode is enabled as it will fail all the time
otherwise.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1557>

qt: Retain compatibility with Qt 5.9

QSharedPointer.get() was added in Qt 5.11, and it does the same thing
as .data()

https://doc.qt.io/qt-5/qsharedpointer.html#get

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/867

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1565>

meson: Update subprojects to fix warnings

glib-2-70 commit 359a837e
meson: fix warnings for extract_all_objects

libdrm-2.4.109 commit 05b0a955
man: convert to reStructuredText
Fixes warning about run_command() check: kwarg

pixman commit adc07d46
meson: Fix warning about extract_all_objects usage

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1564>

ges: Include ges-image-source.h in the installed headers

It's referenced by `ges.h` and was previously available, so removing it
is an API change.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/963

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1563>

va: filter & postproc: Match color with caps features.

When fixating color, there might be "other caps" with color spaces not
supported by the caps features exposed in the vapostproc's source pad
caps template (perhaps it's a bug somewhere else in GStreamer).

This solution checks if the proposed format exists in the filter
within the caps feature associated with the proposed format.

The check is done with the new filter's function
gst_va_filter_has_video_format().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1559>

tsmux: Allow specifying PMT order via the prog-map

Look for an entry `PMT_<PID>` in the `prog-map`, which specifies the
relative index of the stream in the PMT.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1510>

tsmux: Deterministically order program streams by PID

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1510>

tsmux: Deterministically order PAT programs by number

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1510>

tsmux: Remove program_array_index

It's only used for removal. Let's just scan the array.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1510>

tsmux: Replace streams GArray with GPtrArray

This is more appropriate.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1510>

audiotestsrc: Produce a default channel mask if possible

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/957

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1560>

video-converter: Fix for broken gamma remap with high bitdepth YUV output

Scale down the matrix before calculating RGB -> YUV matrix
otherwise offset values will be wrong

Fixing pipeline
videotestsrc ! video/x-raw,format=ARGB ! videoconvert gamma-mode=remap ! \
  video/x-raw,format=P010_10LE,colorimetry="bt2020"

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1432>

gst-rtsp-server: Fix leak in appsrc2 example

In the need-data appsrc callback, a buffer is pulled from the
appsink. This buffer is then copied so that metadata is writable.
The copy is pushed to the appsrc but it doesn't take ownership
of the buffer so we need to manually unref it. The original buffer
is finally unreffed when the sample is freed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1548>

videodecoder: Fix property description for new properties

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1556>

webrtc: Fix memory leaks

Redundant condition and unreachable codes are also removed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1544>

v4l2codecs: Unify the src template caps format

Notably NV12_4L4 ended up being applied to only VP9 decoder. This fixes the
situation by using a central define for all static src pad templated formats.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/965>

v4l2codecs: Sync kernel headers against 5.16.0

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/965>

video4linux2: Sync kernel headers against 5.16.0

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/965>

v4l2videdec: Fix race condition between drain and state changes

This is due to an unsafe usage of the pad task. We didn't ensure proper
ownership of the task. That race involved the task being released too early,
and was detected, luckily, by the glib mutex implementationt that
reported the mutex being disposed while being locked.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1478>

msdk: Avoid noisy registry when no MSDK device.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1550>

va: libs: Avoid noisy registry when no VA device.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1550>

soup: Reintroduce compile-time libsoup dependency for static builds

On Android (especially) and for static builds in general it is safer to link
against libsoup and have the dynamic custom loading disabled. For those cases we
can safely assume the application will use either libsoup2 or libsoup3 and not
both.

Fixes #939

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1536>

soup: Initialize debug category before use

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1536>

soup: Fix return types for a couple VTable functions

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1536>

meson: Build cairo subproject when unavailable on the system

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/952

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1523>

decodebin2: raise multiqueue max-size-bytes

The earlier size of 2 MB was set back in 2009, it doesn't
seem unreasonable to raise it to 8 MB these days. The use
case at hand is matroskademux containing both a video stream
with a very low amount of compression but no decoding latency,
and a H265 stream.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1538>

imagefreeze: Fix example launch line format issue

The currently shown example launch line will not run, because it
cannot handle png images.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1547>

Revert "videodecoder: Forward hdr-format info downstream"

This reverts commit 9b852181d8a1fa2fb1bf3ba898c6ffa04eea3fd4.

It's leftover commit which should've been reverted as a part of
https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/1148

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1475>

wasapi2: Fix for device open failure on old OS

To open automatic stream routing aware device,
at least Windows10 Anniversary Update is required.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1545>

gstreamer: capsfeatures: Fix docs of `gst_caps_features_new_single()`

They were just a copy of the `new_any()` docs before.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1542>

tools: gst-play: Enable Windows high-resolution clock

Apply https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/817
to gst-play as well, especially for better high-framerate
(60fps or higher) video support, because
15ms default clock precision (actual value is system dependent)
is not sufficient for such scenario.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1541>

gstplugin: Better warnings on plugin load failure on Windows

It is an extremely common mistake on Windows to have incorrect PATH
values when loading a plugin, and the error from g_module_error()
(which just calls FormatMessageW()) is very confusing in this case:

 The specified module could not be found.

 https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--0-499-#ERROR_MOD_NOT_FOUND

It implies the plugin itself could not be found. The actual issue is
that a DLL dependency could not be found. We need to detect this case
and print a more useful error message.

We should still print the error fetched from FormatMessage() so that
people are able to google for it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1540>

Add FEC SDP message test

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1527>

sdpmessage: fix mapping single char fmtp params

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1527>

codec-utils: Mark various string return values as `nullable`

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1539>

rtphdrext: increase GstRTPHeaderExtensionClass padding to LARGE

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/748#note_1223253

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1537>

pbutils: Define one debug category per module

Follow-up of https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1505

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1526>

wasapi2ringbuffer: Fix for desynced buffer-size and segsize

GstAudioRingBufferSpec::segsize has been configured by using
device period but GstWasapi2RingBuffer was referencing the
buffer size returned by IAudioClient::GetBufferSize()
which is most likely larger than device period.
Fixing to sync them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1533>

waylandsink: Ensure correct mapping of area_surface

If the `area_surface` got unmapped when changing to the `READY` or
`NULL` state, we currently don't remap it when playback resumes and
`wp_viewporter` is supported. Without `wp_viewporter` we do remap
it, but rather unintentionally and also when not wanted.

On Weston this has not been a big problem as it so far wrongly maps
subsurfaces of unmapped surfaces anyway - i.e. only the black
background was missing on resume. On other compositors and future
Weston this prevents the `video_surface` to get remapped.

Shuffle things around to ensure `area_surface` is mapped in the
right situations and do some minor cleanup.

See also https://gitlab.freedesktop.org/wayland/weston/-/issues/426

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1483>

gst-libav: fix build with ffmpeg-5.0.0

Latest ffmpeg has removed avcodec_get_context_defaults(), and its
documentation says a new AVCodecContext should be allocated for this
purpose.  The pointer returned by avcodec_find_decoder() is now
const-qualified so we also need to adjust for it.  And, AVCOL_RANGE_MPEG
is now rejected with strict_std_compliance > FF_COMPLIANCE_UNOFFICIAL.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1531>

d3d11deinterlace: Do not restrict minimum resolution to 64x64

The value 64 was completely arbitrary one, and this element
will be able to support smaller resolutions

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1528>

pluginfeature: Fix object leak

Need to release GstPluginFeature object after use

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1521>

compositor: Always draw if all input buffers are GAP

Otherwise output buffers will contain random scene
(previously rendered or arbitrary initial values) since we don't
draw onto output buffer at all.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1498>

videoaggregator: Don't pass GAP buffer to prepare_frame_start()

Likewise we do check the same condition in prepare_frames_finish(),
ignore GAP buffer (zero size with GST_BUFFER_FLAG_GAP flag) without
any further processing.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1498>

d3d11compositor: Don't try to read empty buffer

The queued buffer may not be readable buffer in case that
upstream sends GAP event or so.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1497>

waylandsink: Fix alpha value for the test pattern in example

The background-color property is in big-endian ARGB, resulting in
a alpha value of `0`. This accidentally used to work on all common
compositors, but on Weston this now correctly results in a black
background.

See also https://gitlab.freedesktop.org/wayland/weston/-/issues/577

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>

waylandsink: Use wl_surface_damage_buffer() instead of wl_surface_damage()

The later, doing damage in surface coordinates instead of buffer
coordinates, has been deprecated. The reason for that is that it
is more prone to bugs, both on the client and the compositor side,
especially when paired with buffer scale, `wp_viewporter` or
buffer transforms.

Unfortunately, on Weston this risks running into
https://gitlab.freedesktop.org/wayland/weston/-/issues/446
(which causes trouble for several other projects as well). However,
that bug only affects cases where we run in sync mode, i.e. only
during resizes. In practise I haven't been able to observe the
issue.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>

waylandsink: Use G_MAXINT32 for surface damage

Each time we call `wl_surface_damage()` we want to do full surface
damage. Like Mesa, just use `G_MAXINT32` to ensure we always do
full damage, reducing the need to track the right dimensions.

`window->video_rectangle` is now unused, but we keep it around for
now as we may need it again in the future.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>

waylandsink: Only call wl_surface_damage() when buffer content changed

From the spec:
> This request is used to describe the regions where the pending
> buffer is different from the current surface contents

We currently also call `wl_surface_damage()` on surfaces without
new or still compositor-hold buffers, e.g. when resizing the window.
In that case we call it on `area_surface_wrapper`, even though it
gets resized via `wp_viewport_set_destination()`, in which case
the compositor is in charge of repainting the area on screen.

Doing so is currently not forbidden by the spec, however it might
be in the future, see
https://gitlab.freedesktop.org/wayland/wayland/-/issues/267

Thus lets stay close to the spec and only call `wl_surface_damage()`
when we just attached a buffer.

Right now this prevents runtime assertions in Mutter.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>

waylandsink: Simplify input region handling

We only need to unset the input region for the area surface when
we don't have our own toplevel surface. By default, the input region
covers the whole surface, thus no need to change it on resize.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>

waylandsink: Use G_MAXINT32 for opaque regions

`gst_wl_window_set_opaque` does not get called on window resizes,
potentially leaving opaque regions too small.
According to the spec opaque regions can be bigger than the surface
size - parts that fall outside of the surface will get ignored.
Thus we can can simply use `G_MAXINT32` and be sure that the whole
surfaces will always be covered.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>

uridecodebin: Fix critical warnings

Don't pass non-GstObject object to there.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1519>

pbutils: Simplify h264_caps_structure_get_profile_flags_level a bit

Refactoring, removing one level of indentation from the function.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1505>

pbutils: H.265 support for gst_codec_utils_caps_get_mime_codec()

The codec_data caps payload is parsed and a MIME codec string is generated
according to the ISO/IEC 14496-15 specification.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1505>

pbutils: Add a pbutils debug category

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1505>

webrtc: fix log error message in function gst_webrtc_bin_set_local_description

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1511>

mxfdemux: don't error out if VANC track only contains packets we don't handle

If the VANC track does contain packets, but we skip over all packets, just
treat it the same as if there hadn't been any packets at all and send a
GAP event instead of erroring out with "Failed to handle essence element".

We would error out because when we reach the end of the loop without having
found a closed caption packet the flow return variable is still FLOW_ERROR
which is what it has been initialised to.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1518>

codecparsers: h265parser: return invalid profile if len is 0.

Though the profiles[0] is inited as GST_H265_PROFILE_INVALID in the
gst_h265_profile_tier_level_get_profile(), the profile detecting may
change its content later. So the return of profiles[0] may not be an
invalid profile even the len is 0.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1517>

codecparsers: h265parser: Fix the index incrementation error in append_profile().

The current "*idx++" operation just refers the pointer and increment the pointer
itself, not the content of the pointer. This causes that the count of the profiles
is always 0.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1517>

kmssink: Support auto-detection of NVIDIA Tegra driver

NVIDIA Tegra SoCs have a separate (from GPU) display controller. It's
the primary display device on all Tegra SoCs. Add Tegra to the list
of primary DRM drivers.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1514>

gst-env: extend PATH to include plugin scanner and generator

cccombiner: fix s334-1a scheduling

The previous code was mistakenly trying to compute a cc_type out
of the first byte in the byte triplet, whereas it is to be interpreted
as:

> Bit b7 of the LINE value is the field number (0 for field 2; 1 for field 1).
> Bits b6 and b5 are 0. Bits b4-b0 form a 5-bit unsigned integer which
> represents the offset

The same mistake was made when creating padding packets.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1496>

cccombiner: merge buffers for both fields with caption type s334-1a

Other elements such as line21encoder expect both fields to be present
in the same meta, not one meta per field.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1496>

gtk: Add "video-aspect-ratio-override" property

Allow front-ends to override the pixel aspect ratio found inside the
video file itself, or most likely, missing from the video file.

This is a long-standing feature of totem.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1137>