rtpg726depay: fix endian conversion

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1469>

webrtcstats: Fall back to last packet ssrc if caps dont provide it

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1448>

webrtcstats: Use our own caps instead of the sticky event

The sticky event seems to get cleared sometimes.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1448>

webrtcbin: Store the ssrc of the last received packet

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1448>

webrtc stats: Remove duplicate structure get

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1448>

webrtc stats: Add more details about codecs into the stats

This makes the output a little closer to what the upstream stats are.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1448>

rtpjitterbuffer: fix typo in tests

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1466>

doc: typo fix for streaming

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1463>

docs: interlaced video: small additions for alternate interlacing

Clarify that width/height in caps is still the frame height/width,
not field height width, just like framerate is frame rate not
field rate.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1465>

va: av1dec: Use named profiles to replace the numeric ones.

Use named AV1 profiles (i.e., main, high, and professional) to replace
the old 0, 1, 2 profiles.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1456>

av1enc: Update for newly designed AV1 profile signalling

Accept named AV1 profiles (i.e., main, high, and professional)
as well

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1456>

d3d11av1dec: Update sinkpad template for profile

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1456>

av1parse: Use descriptive profile name instead of numeric

As per AV1 specification Annex A, AV1 profiles have explicit and
descriptive names for each seq_profile.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1456>

av1parse: Remove trailing white space

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1456>

rtsp-server: add gst_dep to gst_rtsp_server_deps

Add gst_dep to gst_rtsp_server_deps, in the context of buildroot, this
will avoid the following build failure, because the correct girdir
location will be retrieved from gstreamer-1.0.pc:

/home/giuliobenetti/autobuild/run/instance-3/output-1/host/riscv32-buildroot-linux-gnu/sysroot/usr/bin/g-ir-compiler gst/rtsp-server/GstRtspServer-1.0.gir --output gst/rtsp-server/GstRtspServer-1.0.typelib --includedir=/usr/share/gir-1.0
Could not find GIR file 'Gst-1.0.gir'; check XDG_DATA_DIRS or use --includedir
error parsing file gst/rtsp-server/GstRtspServer-1.0.gir: Failed to parse included gir Gst-1.0
If the above error message is about missing .so libraries, then setting up GIR_EXTRA_LIBS_PATH in the .mk file should help.
Typically like this: PKG_MAKE_ENV += GIR_EXTRA_LIBS_PATH="$(@D)/.libs"

Fixes:
 - http://autobuild.buildroot.org/results/04af6b22cfa0cffb6a3109a3b32b27137ad2e0b0

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1460>

aatv: Fixes for rain-mode

Some rain-mode properties tried to read float from int value.
Initialize rain after setting rain-mode. Rain was non-functional if
width/height were left at default values.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1459>

uridecodebin: use non-floating object as signal argument

... as was the case with source-setup signal until change of order
in commit 52bca104e447309898ca8904b3914211ec7d4114

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1457>

d3d11av1dec: Fix for Cdef param

av1parser will increase the sec_strength values by 1 if parsed
values were equal to 3 as defined in spec. But DXVA wants unmodified
ones.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1455>

d3d11av1dec: Sync DXVA AV1 data structure with released header

Update AV1 data structure based on Windows 11 SDK header

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1455>

rtsp-stream: fix get_rates raciness

Prior to this patch, we considered that a stream was blocking
whenever a pad probe was triggered for either the RTP pad or
the RTCP pad.

This led to situations where we subsequently unblocked and expected
to find a segment on the RTP pad, which was racy.

Instead, we now only consider that the stream is blocking when
the pad probe for the RTP pad has triggered with a blockable object
(buffer, buffer list, gap event).

The RTCP pad is simply blocked without affecting the state of the
stream otherwise.

Fixes #929

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1452>

tests: h265parser: Add test for multiple compatibility profiles.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1440>

h265parser: Compare upstream profile with in SPS.

Compare if upstream profile in caps is the same as the one parsed in
the SPS. If they are different use the bigger for simplicity and
more chances to decode it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1440>

codecparsers: h265parser: Use a table map to get profile.

Instead of a sequence of if statements, declare a table to map profile
idc with profiles and traverse it.

Also, first add the profile from the parsed profile idc and later add,
into the profile array, the profile from the compatibility flags.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1440>

codecparsers: h265parser: Verify all possible profiles.

It's possible a HEVC stream to have multiple profiles given the
compatibility bits. Instead of returning a single profile, internal
gst_h265_profile_tier_level_get_profiles() returns an array with all
it possible profiles.

Profiles are appended into the array only if the generated profile
is not invalid.

gst_h265_profile_tier_level_get_profile() is rewritten in terms of
gst_h265_profile_tier_level_get_profiles(), returning the first
profile found the array.

And  gst_h265_get_profile_from_sps() is also rewritten in terms of
gst_h265_profile_tier_level_get_profiles(), but traversing the array
verifying if the proposed profile is actually valid by Annex A.3.x of
the specification.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1440>

video-info: Don't assume colorimetry of UHD resolution as BT.2020

BT.2020 color primaries are designed to cover much wider range of
CIE chromaticity than BT.709, and also it's used for both SDR and HDR
contents. So, the incorrect assumption (i.e., BT.709 as a BT.2020)
is risky and resulting image color tends to be visually very wrong.
Unless there's obvious clue, don't consider color space of high resolution
video stream as BT.2020

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1445>

webrtcbin: bind transceiver's fec-percentage to encoder percentage

Allows for dynamic control of the applied FEC overhead

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1429>

webrtcbin: fix ulpfec / red for the BUNDLE case

* Add fec / red encoders as direct children of webrtcbin, instead
  of providing them to rtpbin through the request-fec-encoder signal.

  That is because they need to be placed before the rtpfunnel, which
  is placed upstream of rtpbin.

* Update configuration of red decoders to set a list of RED payloads
  on them, instead of setting the pt property.

  That is because there may be one RED pt per media in the same session.

* Connect to request-fec-decoder-full instead of request-fec-decoder,
  in order to instantiate FEC decoders according to the payload type
  of the stream.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1429>

reddec: implement support for the BUNDLE case

When multiple streams are bundled together, there may be more
than one red payload type to handle.

In addition, as the red decoder works by filling in gaps in
the seqnums, there needs to be one rtp_history queue per sequence
domain.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1429>

rtpbin: add new request-fec-decoder-full signal for BUNDLE

When multiple streams are bundled together, the application needs
to know about the payload type in order to instantiate the appropriate
FEC decoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1429>

subprojects: Lower minimal pygobject version

It works just fine with PyGObject3.8, this min requirement was added in
547570cd790f2b2e390edc1dfb5df4c7a33de45c without any justification for
such a high requirement while 3.8 is the min requirement for
`gst-python`.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1419>

rtp/redenc|ulpfecenc: add support for TWCC

In redenc, when input buffers have a header for the TWCC extension,
we now add one to our wrapper buffers.

In ulpfecenc we add one in that case to our protection buffers.

This makes TWCC functional when UlpRed is used in webrtcbin.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1414>

rtph264pay: Handle 'profile' field

In order to allow "level-asymmetry-allowed" we now handle a new
"profile" field, which as the same semantics as the "profile" field in
H.264 stream so that we can force payloaded stream to have the right
format when using the `gst_sdp_media_get_caps_from_media` to set caps
filter after the payloader. This allows a simple negotiation in standard
RTP negotiation based on SDPs (like webrtc) for that particular case,
closely respecting the specs.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1410>

sdp: Handle level-asymmetry-allowed for H264 streams

The ["level-asymmetry-allowed"] field states that the peer wants the
profile specified in the "profile-level-id" fields but doesn't care
about the level. To express this in GStreamer caps term, we add a
"profile" field in the caps, which reuses the usual "profile" semantics
for H.264 streams and, and remove "profile-level-id" and
"level-asymmetry-allowed" fields.

["level-asymmetry-allowed"]: https://www.iana.org/assignments/media-types/video/H264

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1410>

d3d11videosink: Use only tested color space for swapchain

We are querying supported swapchain colorspace via
CheckColorSpaceSupport() but it doesn't seem to be reliable.
Use only tested full-range RGB formats which are:
- sRGB
- BT709 primaries with linear RGB
- BT2020 primaries with PQ gamma

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1433>

pitch: Specify layout as required for negotiation

There are cases where it might negotiate 'non-interleaved' while it
is wrong.

```
gst-launch-1.0 audiotestsrc !  "audio/x-raw, format=(string)F32LE, layout=(string)non-interleaved" ! audioconvert ! audioresample ! pitch tempo=1.2 ! audioconvert ! "audio/x-raw,format=S16LE" ! fakesink

Setting pipeline to PAUSED ...
Pipeline is PREROLLING ...
(gst-launch-1.0:3029628): GStreamer-Audio-CRITICAL **: 11:42:22.477: gst_audio_buffer_map: assertion '(!meta && info->layout == GST_AUDIO_LAYOUT_INTERLEAVED) || (meta && info->layout == meta->info.layout)' failed
ERROR: from element /GstPipeline:pipeline0/GstAudioConvert:audioconvert1: The stream is in the wrong format.
Additional debug info:
../subprojects/gst-plugins-base/gst/audioconvert/gstaudioconvert.c(876): gst_audio_convert_transform (): /GstPipeline:pipeline0/GstAudioConvert:audioconvert1:
failed to map input buffer
ERROR: pipeline doesn't want to preroll.
ERROR: from element /GstPipeline:pipeline0/GstAudioTestSrc:audiotestsrc0: Internal data stream error.
Setting pipeline to NULL ...
Additional debug info:
../subprojects/gstreamer/libs/gst/base/gstbasesrc.c(3127): gst_base_src_loop (): /GstPipeline:pipeline0/GstAudioTestSrc:audiotestsrc0:
streaming stopped, reason error (-5)
ERROR: pipeline doesn't want to preroll.
Freeing pipeline ...
```

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1441>

doc: Update vp9alphadecodebin doc cache

A new field was added to the template cpas.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>

alphacombine: Fix for early allocation queries

When using playbin3, it seems that the alpha decode is always first to
push caps and run an allocation query. As the format change from sink
and alpha were not synchronized, the allocation query could endup
being run before the caps are pushed. That may lead to failing query,
which makes the decoder thinks there is no GstVideoMeta downstream and
most likely CPU copy the frame.

This patch implements a format cookie to track and synchronize the
format changes on both pads fixing the racy performance issue.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>

v4l2codecs: vp9: Add alpha decodebin wrapper

This will allow HW accelerated decoding of WebM alpha videos.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>

v4l2codecs: plugin: Minor style fix

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>

v4l2codecs: decoder: Improve logging of timed out request

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>

v4l2codecs: mpeg2: Check that the decoder output formats

This is to avoid exposing a decoder for which we don't support any
output format. This happens on platform using vendor formats or
not yet supported tiles formats.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>

vp9alphadecodebin: Fix auto-pluging v4l2slvp9dec

This adds the alignment field to the template caps. Without this field
set, the auto-plugger will see fixed caps and will use
gst_caps_is_subset() against the caps produced by the parser. This is a
challenge for all cases where a parser can do conversion. This is fixed
by adding alignment field, which makes the auto-pluggers do an
intersection of the caps as it gets unfixed caps after intersection now.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>

ximagesrc: avoid blocking wait for X events

XNextEvent() blocks indefinitely in absence of X11 events, which can
prevent the pipeline from stopping.

This can cause problems when ximagesrc is used in "remote desktop"
scenarios and the GStreamer application itself, through which the user
is viewing and controlling the machine, is the only source of input
events.

Replace the call with non-blocking XCheckTypedEvent().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1438>

multifilesrc: fix caps leak

since `gst_caps_replace()` and `gst_pad_set_caps()` both ref the caps and neither of them takes the ownership of the caps -> it must be unreffed in `gst_multi_file_src_set_property()`

to test the leak (on Unix): `echo coucou > /tmp/file.txt && GST_TRACERS=leaks GST_DEBUG="GST_TRACER:7" gst-launch-1.0 multifilesrc location=/tmp/file.txt caps='txt' ! fakesink`

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1436>

d3d11window: Remove hack related to color space selection

Use input GstVideoColorPrimaries without any special case handling
otherwise rendered image color would be very wrong.

The hack was added to work around an issue that some Intel driver
couldn't handle wide color gamut image without HDR10 metadata, specifically PQ image.
But device capability can be checked via a method added in
https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1723
so there's no issue now.

Fixes: https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1175
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1430>

d3d11window: Fix typo in debug message

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1430>

elementfactory: Annotate create_full and make_full to take arrays

We need the array annotation for it to be usable from the introspection side.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1427>

codecs: h264decoder: Fix return value type

Return value should be GstFlowReturn not gboolean

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1426>

codecs: h265decoder: Fix return value if klass->new_picture isn't set

If klass->new_picture isn't set we need to initialize
ret with GST_FLOW_OK to avoid unwanted error case

Fixes: 5b405d15858b ("codecs: h265decoder: Use GstFlowReturn everywhere")
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1426>

gstnetclockclient: signal lost sync if remote time resets

When detecting the remote time has been reset which may occur if remote
device providing the clock server has been power reset, then clock is
no longer synced. Setting clock state will trigger a signal to client
informing on sync lost making it possibility to take appropriate action.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/975>

wpevideosrc: Use basesrc event vfunc

Allows for basic default handling from the base class.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1422>

ci: run fedora amd64 docker jobs on packet runners

Which have much better network connectivity to fdo gitlab.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1420>

jpegdec: only allow conversions from RGB

libjpeg-turbo only supports converting from RGB to other RGB formats.
Fix runtime error when trying to convert from a YUV format for example.

Fix #916

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1421>

teletextdec: fix minor string leak

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1416>

ci: fedora-docker: Cache media files

As of now all integration tests endup fetching 600MB of media every run. This
will increase the docker side, but this is to the traceof that the docker
image will quickly be cached on all runners.

Fixes #841

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1413>

ristsink: set properties on children early

The properties on the udpsink/udpsrc elements need to be set before
there is any state change. If not, in a network without default gateway,
udpsink tries to bind an a NULL interface and fails.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1411>

v4l2codecs: mpeg2: Fix selected sizeimage

Due to a copy paste bug, the bitdepth was never set and that was leading
to requesting sizeimage of 0. Previously that worked since the driver
would in that case pick a size for us. But now the we bumped the minimum
to 4KB, the driver happily allocate 4KB of bitstream which lead to
decoding error.

As MPEG2 have a fixed bitdeph of 8, use a define instead of the run-time
variable.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1415>

v4l2codecs: vp9: Drop frames on non-keyframe format change

V4L2 does not yet support this feature, this will skip over the
transition portion up to the next keyframe.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1404>

v4l2codecs: Fix renegotiation

If we hold the last reference to the allocator, leaving the device
streaming will cause an EBUSY error when trying to free the allocate
buffers.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1404>

d3d11vp9dec: Drop frames on non-keyframe format change

... in case of NVIDIA GPU

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1407>

nvvp9sldec: Drop frames on non-keyframe format change

NVDEC doesn't seem to be able to handle the case

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1407>

codecs: vp9: Drop frames on non-keyframe format change

... if subclass does not support the case

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1407>

codecs: vp9: Also consider render_width/height

Also emits new_sequence if on keyframe and the render_width/height have
change. The subclass can always optimize this if the frame resolution
didn't change, the output caps needs to reflect this though.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1407>

rtpsink: set properties on children early

The properties on the udpsink/udpsrc elements need to be set before
there is any state change. If not, in a network without default gateway,
udpsink tries to bind an a NULL interface and fails.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1398>

vaapidecodebin: Misc enhancements.

- Use GST_WARNING rather than g_critical

- Replace gst_ghost_pad_new_from_template() with
  gst_ghost_pad_new() to avoid using the template.

- Declare extern variable

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1405>

vaapidecode: Autogenerate caps template.

vaapidecode is used in vaapidecodebin and it exposes all the
theoretically supported caps, but that slows down autoplug. With this
autplug is negotiated faster, giving more option to decodebin to select
other decoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1405>

v4l2codecs: vp9: Remove uneeded picture data

The GstV4l2Request now holds a reference on the picture buffer and is
recounted already. This effectively removes usage of GRefCount which is only
available in GLib 2.58, while we support 2.56.

Fixes #910

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1402>

v4l2codecs: decoder: Add method to get the pic_buf

This helper will be needed for VP9 frame duplication.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1402>

v4l2codecs: vp9: Add missing error checks in decide_allocation

This could otherwise lead to crash.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1402>

ci: Avoid fetching tags for build jobs

Avoid fetching tags since we don't need them for builds.
Only caveat might be what happens when a pipeline from
a tag is triggered

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1400>

ci: Looks like we where not triggering pipelines for tags

Oops.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1400>

ci: Set GIT_DEPTH variable

This switches gitlab-runner to using shallow clones

related #841

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1400>

basesink: Add missing annotations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1392>

pushsrc: Add missing annotations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1392>

bitwriter: Add missing annotations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1392>

baseparse: Add missing annotations

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1392>

v4l2codecs: h264: Fix typo in comment

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1387>

v4l2codecs: h264: Properly set pic_num/frame_num

The V4L2 uAPI uses pic_num for both PicNum and ShortTermPicNum. It also
doe the same for both FrameNum and LongTermFrameIdx. This change does
not change the fluster score, but fixed a visual corruption noticed
with some third party streams.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1387>

va: Use a lock to protect the surface copy by using vpp.

 If we use vpp to do the surface copy, its operation is not atomic.
 We need to maintain the filter's context unchanged during the whole
 copy progress.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1373>

va: Use the GstVaSurfaceCopy of the allocator atomically.

The mem_copy() of the allocator can be called simultaneously from
different threads. We should use atomic pointer operations to create
and use the GstVaSurfaceCopy of the allocator.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1373>

va: DMA allocator: Set the copied memory properly when popped from pool.

The current code does not set the copied memory correctly when it is popped
from the surface cache pool.
1. We forget to ref the allocator, which causes the allocator to be freed
   unexpected, and we get a crash later because of the memory violation.
2. We forget to add ref_mems_count, which causes the surface leak because
   the surface can not be pushed back to the cache pool again.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1373>

v4l2codecs: gstv4l2decoder: set minimum sizeimage

Set minimum sizeimage such that there is enough space for any overhead
introduced by the codec.

Notably fix a vp9 issue in which a small image would not have a
bitstream buffer large enough to accomodate it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1012>

v4l2codecs: gstv4l2codecsvp9dec: implement a render delay

The v4l2 backend support delayed output for performance reasons.
It is then possible to use render delays to queue multiple requests
simultaneously, thus increasing performance.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1012>

v4l2codecs: vp9dec: Implement VP9 v4l2 decoder

Implement a v4l2 based vp9 decoder element based on the preexisting vp8
v4l2 decoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1012>

v4l2codecs: update to the new uAPI

The new VP9 stateless API is in its way to be destaged. Update our
headers to match.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1012>

codecs: gstvp9decoder: copy frame->system_frame_number into picture

A comment in gstvp9picture.h states that picture->system_frame_number
should get its value from frame->system_frame_number, but in fact
it was never copied at all.

Fix it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1012>

codecs: gstvp9{decoder|statefulparser}: optionally parse compressed headers

Rework gstvp9{decoder|statefulparser} to optionally parse compressed headers.
The information in these headers might be needed for accelerators
downstream, so optionally parse them if downstream requests it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1012>

v4l2: handle 0x0 sizes gracefully

Also ignore 0x0 sizes in the fallback case and assume the size can be
anything between 1x1 and MAXxMAX.

This fixes the case where a width=0, height=0 caps are created. Whith
this patch the caps will contain width=[1,MAX], height=[1,MAX].

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1396>

webrtcbin: deduplicate extmaps

When an extmap is defined twice for the same ID, firefox complains and
errors out (chrome is smart enough to accept strict duplicates).

To work around this, we deduplicate extmap attributes, and also error
out when a different extmap is defined for the same ID.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1383>

codecs: mpeg2decoder: drain() only when significant sequence changes.

There are a lot of info in the mpeg2's sequence(also including ext
display_ext and scalable_ext). We need to notify the subclass about
its change, but not all the changes should trigger a drain(), which
may change the output picture order. For example, the matrix changes
in sequence header does not change the decoder context and so no need
to trigger a drain().

Fixes: #899
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1375>

openjpegenc: Fix build warning

Compiling C object subprojects/gst-plugins-bad/ext/openjpeg/gstopenjpeg.dll.p/gstopenjpegenc.c.obj
../subprojects/gst-plugins-bad/ext/openjpeg/gstopenjpegenc.c(416):
  warning C4133: '=': incompatible types - from 'GstFlowReturn (__cdecl *)(GstVideoEncoder *,GstVideoCodecFrame *)' to
  'gboolean (__cdecl *)(GstVideoEncoder *,GstVideoCodecFrame *)'

../subprojects/gst-plugins-bad/ext/openjpeg/gstopenjpegenc.c(418):
  warning C4133: '=': incompatible types - from 'GstFlowReturn (__cdecl *)(GstVideoEncoder *,GstVideoCodecFrame *)' to
  'gboolean (__cdecl *)(GstVideoEncoder *,GstVideoCodecFrame *)'

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1378>

rtpfunnel: fix extmap handling on accept-caps

Follow-up on 97d83056b315c56834eaa6776ae4c6a0848b5ef9, only check
for intersection with the current srccaps when checking if a sinkpad
can accept caps.

I must have been lucky in my firefox testing then, and always entered
the code path with audio getting negotiated first, thus not failing
the is_subset check when srccaps had been negotiated as
application/x-rtp, and an accept-caps query was made for the video
caps with a defined extmap.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1384>

tutorials: Fix typos

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1382>

gssink: add metadata property

This property can be used to set metadata on the storage object.

Similar API has been added to the S3 sink already, see
https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/613

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1377>

basesrc: Add annotation to virtual methods with (out) parameters

This allows to actually use these virtual methods from the GObject introspection.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1376>

d3d11: Update comments

Remove copy & paste mistake (this is not GstGL) and add more
description.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1374>

vapostproc, vadeinterlace: don't transform caps if no intersection.

If caps to transform don't intersect with those supported by the VA
filter (VAEntrypointVideoProc) then return them as is, because only
pass-through mode is the only possibility.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1369>

vapostproc: Append featured caps rather than merge.

So it would be possible to honor upstream preference.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1369>

vapostproc: Remove dead code.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1369>

va: filter: Validate input parameter in internal function.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1369>