vtenc/vtdec: Fix typo in RGBA64_LE availability define

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3962>

vtenc: Disable ARGB/RGBA64 caps on M1 Pro/Max with macOS <13

Fixes #1358.
Passing ARGB64/RGBA64 to vtenc caused the encoding to fail
when running on M1 Pro/Max variants with macOS 12.x, so let's
remove these formats from caps when such scenario is detected.
This issue appears to have been fixed OS-side in macOS 13.0.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3962>

audio: channel-mix: Fix channel count limit to be able to equal 64

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3958>

qtdemux: Handle moov atom length=0 case by reading until the end

Previously it would fail to demux the file by trying to read G_MAXUINT64
bytes.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3938>

qtdemux: Fix guint vs gsize type confusion

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3938>

vtdec: Fix not waiting for async frames when flushing

This was causing incorrect output when seeking, especially
when used with a multithreaded source like `videotestsrc n-threads=2`.
It should now correctly wait for frames still being processed by VT
while vtdec is flushing.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3937>

win32ipcutils: Add missing include

We are using std::isspace() with one parameter. That function is defined
in the cctype header.

```
win32ipcutils.cpp(34): error C2672: 'std::isspace': no matching overloaded function found
win32ipcutils.cpp(34): error C2780: 'bool std::isspace(_Elem,const std::locale &)': expects 2 arguments - 1 provided
```

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3936>

pad: Don't leak user_data in gst_pad_start_task

When the task already exists, we forgot to free the passed `user_data`.
This wasn't an issue for most C code, which doesn't pass a
`GDestroyNotify`, but bindings such as gstreamer-rs do!

That said, allocating a trampoline in gstreamer-rs just for it to get
thrown away again is awkward. Maybe we need a `gst_pad_resume_task`?

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3925>

av1parser: Don't consider unknown metadata OBUs a bitstream error

Just don't parse them.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1774

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3924>

subparse: Properly forward segment seqnum

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3917>

closedcaption: Don't leak caps event

All events that we handle should be unreffed

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3906>

meson: Allow sysdeps to be forced as fallback subprojects

The original code was too complicated; likely created before the
provide section existed for wraps:

https://mesonbuild.com/Wrap-dependency-system-manual.html#provide-section

Now you can do --force-fallback-for=pygobject and it'll actually work.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3905>

validate-scenario: fix g-i warning in annotation

gst-validate-scenario.c:2485: Warning: GstValidate: missing ":" at column 28:

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3842>

gstreamer: bin: Don't unlock unlocked mutex in gst_bin_remove_func()

Calling `g_mutex_unlock(mutex)` leads to an undefined behavior if the
mutex is not locked by the current thread.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3910>

webrtc examples: Use webrtc.gstreamer.net

Actually just a CNAME to webrtc.nirbheek.in for now, but it allows
replacement / hosting without my involvement, so reduces the bus
factor.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3892>

nvencoder: Fix min buffers parameter of allocation query in auto GPU mode

At the time when propose_allocation() get called, encoder session
would not be initialized yet.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3891>

gl/cocoa: Return a strong ref to the parent GstGLContext

If the GstGLCAOpenGLLayer was initialized via a parent
context, make sure to ref the context before returning it
from getGLContext as all callers will unref it.

Follow up to !3729

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3885>

va: Avoid the array index overflow when filling 8x8 scaling list.

The VA API has not defined the scaling list entries for U/V planes
for the 4:4:4 stream. In fact, we do not meet the 4:4:4 format output
for H264 so far, and scaling list is not used frequently, so we just
print out some warning and ignore these scaling list values.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3877>

Fix gstreamer-validate-1.0 dependency name

The gst-devtools project generates gstreamer-validate-1.0.pc, this
must match the dependency in gst-editing-services for detection
to work properly.

Fixes:
Run-time dependency gst-validate-1.0 found: NO (tried pkgconfig and cmake)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3872>

vtenc: Disable HW acceleration for interlaced ProRes

Due to a bug in the VT API, attempting to encode interlaced content
with ProRes results in an error, halting the pipeline instead of
gracefully falling back to software encoding.
Should be removed in the future if Apple ever fixes this issue.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3878>

aom: Include stream-format and alignment in the AV1 caps

The decoder does not work with arbitrary alignment and annexb stream
format and the encoder can give the information that it outputs
obu-stream/tu to downstream.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3873>

nvvp9dec: Fix return value

It should return GstFlowReturn value, not boolean

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3874>

rtspsrc: Also consider "Method Not Valid In This State" error in broken control URL handling workaround

Some servers send a 455 error instead of any reasonable error when using
a correctly constructed control URL.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3864>

ci: rebuild windows image

Needed for the backport of https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3805
to take effect, once cerbero is made to use the new image.

Should fix errors like

cp: will not create hard link `./cerbero-sources/sbc-1.3' to directory `./cerbero-sources/bzip2-1.0.8'

and similar.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3869>

rtpptdemux: set different stream-id on each src pad

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3866>

rtpssrcdemux: set different stream-id on each src pad

All the RTP src pads were sharing the same stream-id while each actually
carry a different stream.

This was causing problem for example when funneling the streams together
and then trying to split them using 'streamiddemux'.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3866>

gst-env: Handle installing python modules to dist-packages

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3848>

BaseSrc: fix transfer annotation for fixate virtual method

The fixate virtual method has the same semantics as gst_caps_fixate(),
so the caps parameter must be marked as (transfer full).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3839>

validate: Fix gst_validate_execute_action annotation

The action passed in is transfer full

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3838>

gldisplay: Mark `gst_gl_display_create_context()` `other_context` parameter as nullable

See https://gitlab.freedesktop.org/gstreamer/gstreamer-rs/-/issues/438

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3835>

uritranscodebin: Fix unref of NULL

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3837>

mpegpsdemux: Ignore DTS if PTS < DTS

It's possibly timestamp rollover case. But PTS < DTS is already
invalid case anyway.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3836>

BaseSrc, PushSrc: add nullable annotations to virtual methods

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3833>

webrtc_sendrecv.py: Fix PEP8 warnings in CI lint

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3816>

webrtc_sendrecv.py: Handle LATENCY messages

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3816>

webrtc_sendrecv.py: Add bus message handling

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3816>

webrtc_sendrecv.py: Add support for using H264 encoding

Currently only works when we are creating the offer or the offer only
contains H264.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3816>

webrtc_sendrecv.py: Use sine wave for audio instead of red-noise

Makes it easier to notice when there's packet loss or other audio
distortion.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3816>

ci/windows: Remove cargo-vendor sources from the cerbero source cache

Copying the source cache to a running Cerbero CI job on Windows causes
strange / corrupted errors that talk about hard links like:

++ cp -a C:/cerbero/cerbero-sources .
cp: will not create hard link `./cerbero-sources/gstreamer-1.0/subprojects/gst-plugins-good/sys/osxaudio' to directory `./cerbero-sources/cargo-c-0.9.12/cargo-vendor/adler/benches'
[...]

So, delete it. The vendored sources are cached by cargo in its own
cache inside CARGO_HOME, which is build/rust/cargo/ and is preserved
in the image. We won't need to redownload any of this.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3821>

wasapi2src: Fix loopback capture on Windows 10 Anniversary Update

... or older. Work around an OS bug that loopback capture
device doesn't notify event.

Fixes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1738
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3820>

aggregator: Warn about the right segment's format

We were checking the head segment here, not the current segment.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3819>

aggregator: Always lock aggpad around update_time_level

`aggpad->segment` is protected by the `aggpad`'s object lock. We need to
take the lock before calling `update_time_level`.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3819>

rtsp: gstrtspurl: gst_rtsp_url_get_request_uri: fix incorrect scheme for tls transport methods

gst_rtsp_url_get_request_uri returns rtsp://... url when requested url is rtsps://, this is not
in accordance with https://www.rfc-editor.org/rfc/rfc7826.html#section-19.2.
This also impedes setting up a rtsps session with a live555 rtsp server.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3818>

sdp: gstmikey: gst_mikey_message_to_caps: extract salt and append to srtp-key

A KEMAC can hold both key and salt, so if salt is present, we should extract it
and append it to the srtp-key cap because libsrtp expects the srtp-key to hold
the concantenated srtp master key and salt.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3818>

sdp: gstmikey: gst_mikey_message_to_caps: extract ROC from first crypto session

We need the ROC to decrypt a SRTP stream as the ROC is part of the AES IV.
So look for first crypto session, from which we can get the ROC and find corresponding
crypto policy.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3818>

gst: Fix gst_type_find_peek() return value annotation

It's not possible to annotate a in-parameter for a return value array as
the array length. Both are assumed to have the same direction and the
current annotation causes the size parameter to be considered an out
parameter.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3814>

sdp: Fix gst_mikey_payload_key_data_set_interval() array length annotation

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1744

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3814>

v4l2h264dec: Fix Raspberry Pi4 will not play video in application

Ensure object v4l2object->pool will be released by
correctly releasing the temporary thread-safety lock

Fixes issue #1729

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3815>

vulkan: memory: Flush non coherent memory after write.

Spec 7.1.3:

If a memory object does not have the VK_MEMORY_PROPERTY_HOST_COHERENT_BIT
property, then vkFlushMappedMemoryRanges must be called in order to guarantee
that writes to the memory object from the host are made available to the host
domain, where they can be further made available to the device domain via a
domain operation. Similarly, vkInvalidateMappedMemoryRanges must be called to
guarantee that writes which are available to the host domain are made visible to
host operations.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3817>

nvcodec: Log readable errors when initializing CUDA

It is really difficult for people to figure out why nvcodec has
0 features. Even the debug log is cryptic. Also make sure the errors
go to the ERROR log level, which is more likely to be enabled by
default.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3813>

nvcodec: Fix reporting of CuDeviceGetCount error

cuda_ret is was always going to be CUDA_SUCCESS in the error log.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3813>

redenc: fix setting of extension ID for twcc

1 was previously hardcoded in, and the bug went under the radar because
webrtcsink hardcodes the number too.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3812>

multiqueue: Handle use-interleave latency live pipelines

Due to the dynamic nature of multiqueue, when `use-interleave` is used we can't
report a maximum tolerated latency (when queried) since it is calculated
dynamically.

When in such live pipelines, we need to make sure multiqueue can handle the
lowest global latency (provided by this event). Failure to do that would
result in not providing enough buffering for a realtime pipeline.

Fixes #1732

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3810>

avvidenc: Don't take ffmpeg timestamps verbatim but only use them to calculate DTS

The ffmpeg timestamps are inaccurate and only in framerate granularity.
To avoid generating inaccurate output timestamps, especially with
variable framerate streams, only use the ffmpeg timestamps for
calculating the DTS.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1544
again.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3808>

Revert "avvidenc: Set timebase in the ffmpeg context to nanoseconds and set framerate"

This reverts commit 50db59449e9cdb637c5f0c1b7c5dccd582fac4ee.

This broke the MPEG-1 video encoder as it requires the framerate to be
used for the timebase.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1734

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3808>

vaapi: Skip plugin pc file for shared plugins

Following similar logic in the other subprojects.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3778>

udpsrc: GstSocketTimestampMessage only for SCM_TIMESTAMPNS

Deserialize socket control messages as GstSocketTimestampMessage only
if (level, type) is (SOL_SOCKET, SCM_TIMESTAMPNS).

Without this patch, messages with types SCM_RIGHTS or SCM_CREDENTIALS
could be deserialized as GstSocketTimestampMessage instead of
GUnixFDMessage or GUnixCredentialsMessage from gio.

Fixes #1736

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3777>

ci: update for 1.22 branch

Don't have validate do --check-bugs in the 1.22 branch, as
any issues fixed may only have been fixed in the main branch.

Back to development

Release 1.22.0

gst-plugins-ugly: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3773>

gst-plugins-bad: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3773>

gst-plugins-good: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3773>

gst-plugins-base: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3773>

gstreamer: update translations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3773>

d3d11compositor: Workaround blending artifacts on Intel platform

Disable vendor implemented converter in d3d11compositor to workaround
artifacts

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3770>

glvideomixer: Fixing error with mouse navigation events when no pixel-aspect-ratio

Handling mouse navigation events in glvideomixer element, if no
pixel-aspect-ratio info in the caps, an assertion error is produced
inside gst_util_fraction_multiply because default denominator is zero.

Error fixed:
```
(gst-launch-1.0:102654): GStreamer-CRITICAL **: 00:47:51.598: gst_util_fraction_multiply: assertion 'b_d != 0' failed
```

Simple pipeline to reproduce the issue:
```
gst-launch-1.0 -v glvideomixer name=mix ! glimagesinkelement  gltestsrc  ! mix.sink_0
```

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3766>

playback: New playback elements are no longer experimental

The API is not expected to change and it has been used extensively since 1.18.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3761>

examples: webrtc: rust: Update dependencies

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3758>

examples: webrtc: multiparty-sendrecv: rust: Remove unnecessary macro recursion limit annotation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3758>

examples: webrtc: sendrecv: rust: Implement `OFFER_REQUEST` handling

Allow requesting an offer from the peer if we're joining a call with a
peer, and allow the peer to request an offer from us if waiting for an
incoming call.

This implements all 4 variants the protocol allows for.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3758>

examples: webrtc: sendrecv: rust: Allow providing our ID via the commandline

Otherwise it continues to use a random ID as before.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3758>

examples: webrtc: sendrecv: rust: Implement TWCC support in both directions

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3758>

examples: webrtc: rust: Set keyframe-max-dist=2000 and picture-id-mode=15-bit for VP8 and perfect-timestamps=true for audio

This makes it in sync with the C sendrecv and generally behaves better.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3758>

examples: webrtc: sendrecv: rust: Use the correct payload types if the remote is the offerer

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3758>

wpe: Add 'run-javascript' action signal

Introduce way of running a script in the context of the internal
webView.

Fixes #1722

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3753>

decodebin3: Remove failing stream from selection

If no decoder is present for a given stream, remove it from the requested
selection.

Fixes #1581 and #1662

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3754>

decodebin3: Don't intercept queries if no parsebin present

If we don't use a parsebin, we forward the queries as-is from upstream. There is
no reconfiguration possible within identity.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3754>

decodebin3: Don't insert parsebin if input is already parsed

This is a temporary workaround until we find a generic solution to indicate that
a stream has already been "parsed".

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3754>

win32ipcvideosrc: Fix leak in fallback copy path

Clear shared memory after use

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3751>

va: Fix path leak

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3755>

gst: Mark caps parameters in ALLOCATION query API as nullable

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3752>

gst-play: add --no-position

It can be useful to reduce or remove the clutter in logs due to those
queries when debugging.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3698>

matroska: Add `stream-format = (string) obu-stream` to AV1 caps

Anything else is not allowed in Matroska/WebM.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3740>

isomp4: Add `stream-format = (string) obu-stream` to AV1 caps

Anything else is not allowed in MP4.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3740>

gst-examples: Update Rust dependencies

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3750>

gleglimage: Fix GL_OES_EGL_image_external name in debug trace

There is a debug log when this extension is missing, but it was
reporting the wrong extension name, which was confusing.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3748>

input-selector: Take the object lock while iterating sinkpads

Otherwise we can race with pad removal and crash from use-after-free.

Fixes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1717
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3745>

glvideomixer: don't clear pad->geometry_change when setting an identical value

Closes #1715

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3733>

gst-docs: include dmabuf and gapless playback design docs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3744>

uridecodebin3: clear subitem if suburi is set to NULL

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1649

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3730>

qtdemux: Add basic support for AVC-Intra video

AVC-Intra is a range of H.264-compliant intra-only codecs from
Panasonic. The codes and descriptions have been taken from VLC.

The (encumbered) sample I have here produces byte-stream H.264,
including SPS and PPS and no `avcC` box.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3739>

gl: Fix crash in cocoa due to race in gst_gl_context_activate()

We create a new context in `gst_gl_context_create_thread()` and then
activate it on the current thread. Thereafter we assume that the
current thread continues to be the active thread for that context and
call `gst_gl_context_fill_info()` which asserts that the current
thread is the active thread.

However, if at the same time a different thread calls
`send_message_async()`, it will call into
`gst_gl_window_cocoa_send_message_async()` which will schedule the
message to be invoked using GCD. That anonymous function will also
call `gst_gl_context_activate()`, which creates a race, which can lead
to:

```
gst_gl_context_fill_info: assertion 'context->priv->active_thread == g_thread_self ()' failed
```

Fix it by using `gst_gl_context_thread_add()` to invoke `fill_info()`
on the context.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3732>

ges: smart-mixer: Avoid video freeze on NLE stack changes

Since we support "nested compositors"[0] we were not finalizing the pads_infos
and thus not releasing the mixer pad on stack change, meaning that
the following stacks were never prerolling.

[0] https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2094

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3738>

ges: Fix issues avoiding to use `operator` when unavailable on mixer

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3738>

gl/cocoa: Store a weak reference to the `GstGLWindow` instead of the `GstGLContext`

We can't rely on the `GstGLContext` to stay alive and need to keep track
of it. For that we keep track of the `GstGLWindow` in a weak reference
to avoid a reference cycle, and get the corresponding `GstGLContext`
whenever needed.

With contributions from Nirbheek Chauhan.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1697

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3729>

Release 1.21.90

avvidenc: Offset PTS to zero to fix bitrate control

Otherwise ffmpeg's rate control algorithm will not work correctly as
it is based on the absolute PTS values.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-libav/-/issues/91

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3718>

urisourcebin: Cleanup on failure

This ensures there is a source present/valid *ONLY* in PAUSED and above.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3720>

urisourcebin: Unset locked state

Some elements still had it in failure cases

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3720>