hlsdemux2: m3u8: Use PDT to offset stream time when aligning playlist

When matching segments across playlists with Program-Date-Times,
use the difference in segment PDTs to adjust the stream time
that's being transferred. This can fix cases where the
segment boundaries don't align across different streams
and the first download gets thrown away once the PTS
is seen and found not to match.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3309>

hlsdemux2: Download new header when it changes

Check whether the init file / MAP data for a segment
is different to the current data and trigger an
update if so. Previously, the header would only
be checked in HLS after switching bitrate or
after a seek / first download.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3307>

m3u8: Expose GstM3U8InitFile methods

Exposure ref/unref methods for the GstM3U8InitFile type,
and add a gst_m3u8_init_file_equal() comparison method.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3307>

subprojects: orc: bump to 0.4.33

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3304>

d3d11decoder: Fix for VP9 decoding with odd resolution

Fixing off by one mismatch when width and/or height of stream
is odd number

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3302>

qsv: Add VP9 decoder

Recent Intel GPU supports 12bits VP9 decoding but only VP9
profile2 with 10bits is defined by DXVA spec.
Thus, we need this vendor specific decoder element

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3290>

adaptivedemux2: Improve minimum buffering threshold

Previously the minimum buffering threshold was hardcoded to a specific
value (10s). This is suboptimal this an actual value will depend on the actual
stream being played.

This commit sets the low watermark threshold in time to 0, which is an automatic
mode. Subclasses can provide a stream `recommended_buffering_threshold` when
update_stream_info() is called.

Currently implemented for HLS, where we recommended 1.5 average segment
duration. This will result in buffering being at 100% when the 2nd segment has
been downloaded (minus a bit already being consumed downstream)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3240>

d3d11vp9dec: Disallow Profile2 12bits stream

Since DXVA does not support the format, specify bit-depth field
to sinkpad template caps so that d3d11vp9dec can be skipped
during autoplugging

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3297>

vp9parse: Set subsampling to src caps even if GBR stream

Some muxer elements will need the subsampling information

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3297>

vp9parse: Delaying src caps until frame is parsed if profile > 0

subsampling and/or bitdepth information will be required for
negotiation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3297>

fdkaacenc: Update documentation to clarify bitrate and peak-bitrate

bitrate property is only applicable for constant bitrate and
peak-bitrate is only applicable for variable bitrate. Clarify
the same.

fdkaacenc: Add support for setting bitrate mode

fdkaacenc: Add support for setting peak bitrate

fdkaacenc: Add support for enabling afterburner

This is an additional quality parameter. In the default configuration this
quality switch is deactivated because it would cause a workload increase
which might be significant. If workload is not an issue in the application
it can be recommended to activate this feature.

fdkaacdec: Do not report decoding error for flush request

A flush request is done when set_format is called to empty internal bit
buffer maintained by fdk-aac. When this happens, during the explicit
call to handle_buffer, decodeFrame does not return a AAC_DEC_OK. This
gets reported as a decoding error while no decoding error in fact took
place. Since this can be confusing, just return a GST_FLOW_OK and log
that an explicit flush was requested.

wavparse: Speed up type finding for DTS

In order to figure out if the "raw" audio contained within the wav
container is actually DTS, right now we call the typefinder helper
which runs all typefinders.

Speed up this type finding process by specifying the extension.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3294>

playbin3: Fix missing pad unref

GST_TRACERS="leaks" GST_DEBUG="GST_TRACER:7,leaks:6" gst-play-1.0 --use-playbin3 test.mkv

When running a pipeline like above, leaks are observed.

0:00:56.882419132 240637 0x5562c528ccc0 TRACE             GST_TRACER :0:: object-alive, type-name=(string)GstConcatPad, address=(gpointer)0x7efd7c0d20a0, description=(string)<'':sink_0>, ref-count=(uint)1, trace=(string);
0:00:56.882429131 240637 0x5562c528ccc0 TRACE             GST_TRACER :0:: object-alive, type-name=(string)GstConcatPad, address=(gpointer)0x7efd7c0d2be0, description=(string)<'':sink_0>, ref-count=(uint)1, trace=(string);
0:00:56.882437056 240637 0x5562c528ccc0 TRACE             GST_TRACER :0:: object-alive, type-name=(string)GstConcatPad, address=(gpointer)0x7efd7c0d3720, description=(string)<'':sink_0>, ref-count=(uint)1, trace=(string);

gst_element_release_request_pad does not unref the pad. It needs to
be followed by gst_object_unref. Doing that fixes the above leaks.

Use g_ptr_array_new_with_free_func with gst_object_unref as the free
function to unref the pad after release.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3177>

pbutils: add correct mime mapping for vp9

Based on https://www.webmproject.org/vp9/mp4/#codecs-parameter-string

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3260>

mp4mux: enable muxing VP9 streams

As specified in https://www.webmproject.org/vp9/mp4/

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3260>

qtmux: add support for writing vpcC box for VP9

Increases compatibility for VP9 in .mov in at least VLC.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3260>

vp9parser: initialize subsampling to -1

The default value of 0 is a valid subsampling value and could be
confused with an 'unset' value.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3260>

gst-plugins-rs: Track 0.9 branch until 1.22 release

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3286>

vaencoder: Reset attribute index to one.

Further fix to 4ffb3663 where I forgot to reset the attribute index.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3285>

dashdemux2: Fix the way we determine current_position after seeks

Without that the current_position was off after seeks, potentially
leading to not properly push a last fragment when a `.stop` time was
set.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3159>

adaptivedemux: Minor typo fix

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3159>

dash: Fix computing `repeat_index` when seeking in stream with a start !=0 on the first fragment

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3159>

matroskademux: Let upstream handle seeking/duration query in time if possible

So proper response are given for dash streams

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3159>

matroskademux: Start support for upstream segments in TIME format

So we can use matroskademux for dash webm dash streams.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3159>

ximagesrc: grab the server while capturing screen image

Makes sure screen resolution doesn't change in the middle of the
process.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1562>

ximagesrc: change video resolution when X11 screen gets resized

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1562>

ci: Allow the documentation job to always run post-merge

We want the docs to be deployed on every post-merge pipeline
on the main branch. In order to do that however we have to
also have the fedora build job always running.

This won't affect the documentation job run by MRs.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3258>

ci: Make jobs explicitly depend on the trigger job to run

Instead of relying on the implicit dep from the image build
jobs.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3258>

ci: consolidate pre-build CI stages

Back in the day we kept separate stages around to define
build order, however with DAG/needs now jobs run asynchronously.
Additionally on recent version its possibel to same jobs depend on
other jobs from the same stage.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3258>

ci: Replace obselete variable

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3258>

core/base: Only post latency messages if the latency values have actually changed

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1525

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3282>

videodecoder: Only post latency message if it changed

Posting latency messages causes a full and potentially expensive latency
recalculation of the pipeline. While subclasses should check whether the latency
really changed or not before calling this function, we ensure that we do not
post such messages if it didn't change.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3282>

avdec_h265: Fix endless renegoation with alternate interlacing

The picture parameter picture->top_field_first is reused in this mode
to signal the TOP fields. As a side effect, it will change every frame
and current code assumed that if this changes then a renegotiation is
needed. Fixed this by ignoring that change whenever we are decoding one field
only.

Fixes #1523

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3276>

h265bitwriter: Correct the all API to byte aligned.

In fact, all the h265 bit writer have byte aligned output. So we
change the API from bit size in unit to byte size, which is easy
to use.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3193>

h264bitwriter: Correct the all API to byte aligned.

In fact, all the h264 bit writer have byte aligned output except
the slice header. So we change the API from bit size in unit to
byte size, which is easy to use. For slice header, we add a extra
"trail_bits_num" to return the unaligned bits number.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3193>

bitwriter: Fix a nal conversion bug when input is not byte aligned.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3193>

msdkvpp: Set va mem caps as higher priority

We use va pool as msdkvpp's bufferpool, which means both va memory
and dma memory will be allocated by va pool. Considering drm modifier
stuff is not ready, we use va memory with higher priortiry than
dma memory when deciding vpp caps.

Besides, this patch removes the specified "interlace-mode" in vpp caps.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3253>

concat: Properly propagate EOS seqnum

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3284>

msdkav1enc: Remove reorder TU workaround

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3225>

msdkav1enc: Remove pts workaround

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3225>

codecparsers: {h264,h265}bitwriter: Don't install headers yet

Those APIs are quite early stage of development, and we might want to
change them in the near future

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3275>

msdkvpp: Fix upper frc

Refact vpp commit 4c6b7194451ff6fbfe88834219d982302114d74b broke the
upper FRC, here to fix this by using outbuf_new.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3223>

bad: Add msdkav1enc docs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2971>

meson: Fix path for webrtc validate tests

This fixes a crash when trying to run gst-validate-launcher from inside
the meson devenv. The error was:

  ModuleNotFoundError: No module named 'observer'

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3273>

avviddec: Avoid flushing on framerate changes

A framerate change does not require flushing the decoder and causes
issues with some specific fragmented files if the two fragments have
different framerate.

Fixes #1522

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3270>

ximagesink, xvimagesink: set the _NET_WM_PID atom on the window

This makes it easier for X11 window management to work with windows created by these sinks

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3267>

alphacombine: Add missing query handler for gaps

The gap handling was in place, but there was no event handler to trigger it.
Implement the alpha sink event handler for the gaps. This fixes handling of
valid streams which may not refresh the alpha frames for every video frames.
It will also allow a clean error if the stream was missing the initial
alpha frame, at least until we find a better way to handle these
invalid frames.

Related to #1518

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3264>

avfdeviceprovider: do not leak the properties

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3268>

ci: Fix traceback when user doesn't have a cerbero fork

```
Traceback (most recent call last):
  File "ci/gitlab/trigger_cerbero_pipeline.py", line 59, in <module>
    if cerbero is None:
NameError: name 'cerbero' is not defined
```

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3063>

vaencoder: Fix caps semantics.

When using gst_va_caps_from_profiles() the semantics of sink/src caps
depends if the element is an encoder or a decoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3063>

va: Fix typos.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3063>

vah264enc: Fail if unsupported rate control.

Handle when encoder doesn't support rate control, which is set as
VA_RC_NONE, and if the set rate control mode is not supported by the
GStreamer element, the element configuration fails.

Also it logs out max and target bitrate.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3063>

vah264enc: Add todo item.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3063>

vaencoder: Honor entrypoint at constructor.

The entrypoint is set when the encoder helper is constructed,
nonetheless it was also passed as parameter when opening. That's
buggy.

In order to simplify the code, the entrypoint at construction is
honored.

But gst_va_encoder_has_profile_and_entrypoint() now doesn't rely in
the internal list of profiles since it only contains those that
belongs to codec and entrypoint, thus it queries directly the VA
driver.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3063>

vabaseenc: Scope error bail out.

Though this is not enforced by the GStreamer code style, it's clearer
to add a nested scope for error handling using label/goto.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3063>

vabaseenc: Use class entrypoint.

Add a macro to access to class entrypoint and use it instead of move
it to a variable.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3063>

d3d11: Move format defines to private header

Those defines will be likely changed/deleted in the future, and we
don't need to expose them to public at the moment

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3265>

ci: Fetch all user cerbero branches when matching branch names

Fixes CI on coordinated merges when the user has more than 20 branches
in their fork, which will happen very easily since new forks will
always have all the branches of the original remote.

```
ci/gitlab/trigger_cerbero_pipeline.py:48: UserWarning: Calling a `list()` method without specifying `get_all=True` or `iterator=True` will return a maximum of 20 items. Your query returned 20 of 37 items. See https://python-gitlab.readthedocs.io/en/v3.9.0/api-usage.html#pagination for more details. If this was done intentionally, then this warning can be supressed by adding the argument `get_all=False` to the `list()` call. (python-gitlab: /usr/local/lib/python3.7/site-packages/gitlab/client.py:979)
  if os.environ["CI_COMMIT_REF_NAME"] in [b.name for b in cerbero.branches.list()]:
```

xingmux: move from gst-plugins-ugly to gst-plugins-good

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/415

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3251>

build: Fix building ges with tools disabled

If you configure with `tools=disabled` then ges_launch is undefined.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3259>

Fix various warnings from gobject-introspection

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3261>

fdkaac: add minimal unit test

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1785>

fdkaacenc: fix output caps in case of implicit signaling and HE-AAC

Need to put the actual profile in the output caps otherwise any
capsfilter after the encoder that was used to force the output
profile will fail, such as

  fdkaacenc ! audio/mpeg,stream-format=adts,profile=he-aac-v1 ! ..

because we put profile=lc in there to match the profile signaled
in the ADTS header. This is expressed through the base-profile=lc
in the GStreamer caps though, the profile needs to carry the
'real' profile.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1785>

fdkaacenc: don't set base-profile=lc for non-backwards compatible output

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1785>

fdkaacenc: rename profile=sbr|ps to profile=he-aac-v1|he-aac-v2

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1785>

fdkaacenc: add support for AAC-LD

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1785>

fdkaacenc: add support for HE-AACv1 and HE-AACv2

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1785>

webrtc: Fix double free in webrtc-recvonly-h264 demo

The "message" signal does not transfer ownership of the GBytes passed
to it so calling g_bytes_unref() on it is incorrect.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3257>

gst-inspect: Hide GST_PARAM_DOC_SHOW_DEFAULT flag

It's known flag but only for documentation purpose. Don't show
its (and user cannot understand) value 0x2000

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3241>

ges: transition: Notify vtype change on ges_extractable_set_asset call

Fixes #1454

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3215>

gst-plugins-bad: Update Windows plugins cache

Add/update d3d11, qsv, nvcodec, wic, directshow, and amfcodec plugins cache

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

nvcodec: Update for documentation

* Use GST_PARAM_DOC_SHOW_DEFAULT flags for GPU ID related
  properties
* Fix some typos
* Add since markers

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

mediafoundation: Update for documentation

* Use GST_PARAM_DOC_SHOW_DEFAULT flags for GPU ID related
  properties
* Add doc caps
* Add since markers

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

mfvideoencoder: Document only hardware encoders

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

mfvideoencoder: Fix device enumeration

Although a GPU does not support the codec, the other GPU might
be able to support the codec

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

amfcodec: Update for documentation

* Use GST_PARAM_DOC_SHOW_DEFAULT flags for GPU ID related
  properties
* Add doc caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

qsv: Use GST_PARAM_DOC_SHOW_DEFAULT flag

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

d3d11: Update for documentation

* Use GST_PARAM_DOC_SHOW_DEFAULT flags for GPU ID related
  properties
* Add doc caps
* Increase deinterlace max resolution

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

d3d11: Run indent

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

wic: Add since marker

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

directshow: Add since marker

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3250>

examples: webrtc: Update to gstreamer-rs 0.19 release

Also update the macOS workaround for gstreamer-gl requiring a
`NSRunLoop` / `NSApp` on the main thread, and update from strucopt to
clap 4.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3255>

qsv: Enable MinGW toolchain support

Use PCRE regex method to work around (likely) GCC std::regex bug,
and enable building for non-MSVC

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3239>

rtspsrc: Only EOS on timeout if all streams are timed out/EOS

Otherwise a stream that is just temporarily inactive might time out and
then can never become active again because the EOS event was sent
already.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3238>

gl/vulkan: Fix static linking on macOS

duplicate symbol '__invoke_on_main' in:
    /Library/Frameworks/GStreamer.framework/Versions/1.0/lib/libgstvulkan-1.0.a(cocoa_gstvkwindow_cocoa.m.o)
    /Library/Frameworks/GStreamer.framework/Versions/1.0/lib/libgstgl-1.0.a(cocoa_gstglwindow_cocoa.m.o)
ld: 1 duplicate symbol for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)

Also make the same change in iOS for consistency.

Continuation of https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1132

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3242>

webrtcbin: configure rtpulpfecdec passthrough property

This allows downstream (payloaders mostly) to be able to correctly
detect actual packet loss from rtp sequence numbers.

See
https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/issues/581
for background.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1407

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3212>

rtpulpfecdec: add property for passthrough

Support for enabling and disabling decoding of FEC data decoding on
packet loss events and unconditional seqnum rewriting of packets.

See
https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/issues/581
for background.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3212>

webrtc: Fix critical in webrtc-recvonly-h264 example

This signal only takes 2 properties yet a third was passed.
This would cause a critical in GLib.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3252>

net: Add missing nullable annotation on the name parameter in the net/NTP clock constructors

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3249>

padtemplate: Fix annotations

gst_caps_replace() does not take ownership of the new caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3248>

gst-libav: Fix synchronization issues and errors created by the forwarding of
segment events by demuxer.

In order to play nicely with `ffmpeg`, demuxers in `gst-libav` have to make
buffers available to `ffmpeg` while taking the blocking I/O model in `ffmpeg`
into account, which results in buffers not being sent downstream until `ffmpeg`
has processed them in its separate thread.

In constrast, many `gstreamer` events are simply forwarded downstream.
Currently `GST_EVENT_SEGMENT` events are forwarded downstream without any
processing, which can potentially result in:
 * `GST_EVENT_SEGMENT` events being out of sync with buffers
 * `GST_EVENT_SEGMENT` events going out that are incorrect because they apply
   to data seen by the demuxer, but not necessarily seen by downstream elements

I came across this bug when I was attempting to enable G723.1 demuxing/decoding
using the G723.1 demuxer and decoder provided by `ffmpeg`.  I wrote tests to
verify support for the functionality, and found that, in push mode,
`GST_EVENT_SEGMENT` events pushed to the demuxer by the upstream `filesrc`
element would be forwarded to the decoder without modification, resulting in
an internal data streaming error.  With this patch, tests work in both push and
pull mode.

This patch solves the problem by disabling the forwarding of
`GST_EVENT_SEGMENT` events downstream (an initial `GST_EVENT_SEGMENT` event is
still pushed downstream by the demuxer).  It's possible there's a better way to
do this, but, having looked at how a few different `gstreamer` demuxers deal
with `GST_EVENT_SEGMENT` events, it seems like the processing is somewhat
specific to the demuxer implementation, whereas `gst-libav` has one general way
of handling the situation for any `ffmpeg` demuxer.  Perhaps there's a better
way to solve this using the `ffmpeg` API to take advantage of specific demuxer
details.  IDK.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3218>

gst-inspect: Don't print link to doc if it's known to be unavailable

"gst_element_factory_get_skip_documentation() == true" means
documentation was intentionally skipped for the element feature

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3243>

gst: uri query: fix inconsistent `uri` nullability  assertion

Functions `gst_query_set_uri` and `gst_query_set_uri_redirection`
can both set a `NULL` `uri`, as annotated in the documentation.
However the functions bodies reject `NULL` `uri`s.

See [1] for a discussion on that matter.

[1]: https://gitlab.freedesktop.org/gstreamer/gstreamer-rs/-/merge_requests/1134#note_1600988

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3237>

d3d11screencapture: Add WinRT API based capture mode

Add Windows Graphics Capture (WGC) API based screen capture mode.
The conditions where this mode is used:
* Explicitly requested by user (capture-api property)
* To capture specific window
* When DXGI desktop duplication API does not work on hybrid graphics systems
  (e.g., multi-gpu laptop)

Full features of this implementation require Windows 11. And Windows 11
SDK is required to build this feature.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3144>

d3d11screencapture: Subclassing capture implementation

Preparation to use WinRT capture API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3144>

qsv: Add plugin doc

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2738>