trigger_cerbero_pipeline.py: Do not hardcode gitlab instance

When GStreamer is forked into a private GitLab instance we should
trigger cerbero into that private instance too. Otherwise the token
won't be accepted.

Also do not hardcode the cerbero project ID because it is instance
specific.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1056>

va: Use macro rather than VAMemory feature string.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1061>

ci: remove duplicate stage key in documentation job

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

ci: Remove "check allow-edit" job because gitlab does it by default

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

ci: Update fedora image for cerbero trigger

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

ci: Fix gst-indent path

It used to be downloaded into PATH, but we can now instead run it from
git. Also move it to top source dir instead of gstreamer subproject.

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

ci: Fix subprojects caching in Windows image

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

ci/docker/fedora: update rustc and meson versions

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

ci: port to fd.o citemplates

https://gitlab.freedesktop.org/freedesktop/ci-templates

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

ci: remove duplicate keys

They are redefined/copied just below

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

ci: Add a dummy trigger job/stage

This is to avoid triggering docker images for no reason, and bring back
running gst-indent always.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1054>

ci: Document the difference between the two doc jobs

Naming is hard, so I opted for a small comment to explain this one.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1054>

rtsp-media: Also mark receive-only (RECORD) medias as prepared when unsuspending

Previously the status was only changed for other medias.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1058>

docs: update "Building from Source using Meson" for mono repo

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1052>

splitmuxsink: Keep track of the pending input GOPs in a queue

This cleans up input GOP handling and makes it possible to handle more
than 2 pending GOPs, which could happen before if keyframes are arriving
with too short of a distance between them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1005>

splitmuxsink: Handle frame reordering due to B frames better

Instead of assuming that the PTS of a keyframe is the lowest PTS of a
GOP, wait until the DTS has passed this PTS and take the minimum PTS up
to that point. That way the minimum PTS of a GOP can be determined, at
least for closed GOP streams. Open GOP streams still can't be handled
properly.

By knowing the minimum PTS of each GOP, keyframes can be requested at
the correct time relative to the GOP (and thus fragment) start and
fragment overflow calculations can calculate the correct durations of
the GOPs.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1005>

splitmuxsink: Reset timecodes when resetting the sink

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1005>

splitmuxsink: Use GST_CLOCK_STIME_NONE instead of 0 to initialize max out running time

Otherwise streams with only DTS would misbehave while it is negative.

For outputting everything immediately at EOS, use G_MAXINT64 instead
which is bigger-or-equal to any other running time.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1005>

splitmuxsink: Only update max in/out running time if it's actually bigger

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1005>

splitmuxsink: Only count keyframes for the reference context, consistently

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1005>

ci: Rework workflow rules to prevent duplicate pipelines

Currently gitlab spawns 2 pipelines, one for the branch and
another for the merge request.

Update our rules so that only one of them is spawned depending
on the situation.

https://docs.gitlab.com/ee/ci/yaml/index.html#switch-between-branch-pipelines-and-merge-request-pipelines

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/958>

docs/running: Describe GST_REGISTRY_MODE

Change-Id: I45e9fe90137d4a8306c3a5e4f636fa43425b978f
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/825>

registrybinary: registry file mode via GST_REGISTRY_MODE

In an embedded system where all services run as seperate users it is
useful to have the gstreamer registry readable by all so it can be
re-used, in similar manner as a host system where one user have seperate
applications running but all share same registry.

To make this possible introducing GST_REGISTRY_MODE for adjusting the
changing mode of the registry binary when finishing up with the
temporary file (which has restricted access).

Fixes: #692
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/825>

mpegtsparse: Don't assert the packet_size when filling for EOS

If the packetizer got reset for any reason (failure to find PCR?) then
the packet_size can be zero here even though we already enqueued some
packets.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1038>

pbutils: Remove sample entry code "raw"

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1045>

multiqueue: Fix query unref race on flush

If the query has already been destroyed at this point, GST_IS_QUERY will
read garbage, can return false and we will try to unref it again.
Instead, make note of whether the item is a query when we dequeue it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1029>

handle-subprojects-cache.py: Fix shebang

codecs: mpeg2decoder: Use GstFlowReturn everywhere

boolean return value is not sufficient for representing the reason
of error in most cases. For instance, any errors around new_sequence()
would mean negotiation error, not just *ERROR*.
And some subclasses will allocate buffer/memory/surface on new_picture()
but it could be failed because of expected error, likely flushing

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1019>

codecs: h264decoder: Use GstFlowReturn everywhere

boolean return value is not sufficient for representing the reason
of error in most cases. For instance, any errors around new_sequence()
would mean negotiation error, not just *ERROR*.
And some subclasses will allocate buffer/memory/surface on new_picture()
but it could be failed because of expected error, likely flushing

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1019>

codecs: h265decoder: Use GstFlowReturn everywhere

boolean return value is not sufficient for representing the reason
of error in most cases. For instance, any errors around new_sequence()
would mean negotiation error, not just *ERROR*.
And some subclasses will allocate buffer/memory/surface on new_picture()
but it could be failed because of expected error, likely flushing

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1019>

closedcaption: Fix broken debug function macros with MSVC build

warning C4003: not enough arguments for function-like macro invocation 'warning'

G_STMT_END macro is extended to the below form with MSVC
__pragma(warning(push)) \
__pragma(warning(disable:4127)) \
while(0) \
__pragma(warning(pop))

So MSVC preprocessor will extend it further to
__pragma(VBI_CAT_LEVEL_LOG(push)) ...

Should rename warning() debug macro function therefore.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1018>

mpegtsmux: basetsmux: Don't try to return value from void function

gstbasetsmux.c(1508): warning C4098: 'free_splice': 'void' function returning a value

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1018>

move_mrs: Disable our commit hook during the move

As GNU indent version can be different on the user system, we see some
differences during migration thas causes conflicts. Making cherry-pick
difficults to recover without breaking the style temporily. Note that
cherry-pick continuation does not allow passing the -n option to skip
the hooks.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1017>

move_mrs: Fix coding style using autopep8

And some small edits for:
- E741 ambiguous variable name 'l'
- E722 do not use bare 'except'
- W504 line break after binary operator

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1017>

va: Make libgudev dependency optional.

libgudev is a problematic dependency, particularly in sandboxed
environments, such as flatpak.

This patch implements a way to get the available VA devices using
brute-forced traverse of /dev/drm/renderD* directory. Thus usable in
those sandboxed environments.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1027>

va: meson: Move back libgudev dependency to plugin.

When move the libgstva, libgudev dependency was moved as part of the
library, though it's not use by the library but the plugin. This patch
moves back libgudev dependency to the plugin.

Also HAVE_LIBDRM is move to the library which is the one who use it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1027>

nvcodec: nvh264sldec: Add support for interlaced stream

Implement missing interlaced stream support

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1026>

nvcodec: nvh264sldec: Consider additional render delay DPB pictures

At least additional 4 pictures are required

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1026>

nvcodec: nvvp9sldec: Fix for VP9 profile2 decoding

Fix for output video format to be selected correctly

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1026>

msdk: meson: Fix build on Windows

subprojects\gst-plugins-bad\sys\msdk\meson.build:160:2: ERROR: Unknown variable "libva_dep".

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1022>

d3d11: d3d11vp9dec: Fix use_prev_in_find_mv_refs value setting

"last_show_frame" should be updated based on
GstVp9FrameHeader::show_frame, not show_existing_frame

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1021>

codecs: {vp8,vp9}decoder: Drain on new_sequence()

Decoder should drain queued frame (if any) and empty DPB before
starting new sequence.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/987>

nvcodec: nv{vp8,vp9}sldec: Implement get_preferred_output_delay()

Equivalent to that of nvh264sldec. Use render delay in case of non-live
pipeline for the better throughput performance.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/987>

codecs: {vp8,vp9}decoder: Cleanup drain code

Make them consistent with h26x decoder baseclass

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/987>

codecs: vp8decoder: Fix typo

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/987>

codecs: vp9decoder: add support for render delay

Some decoding APIs support delayed output for performance reasons.
One example would be to request decoding for multiple frames and
then query for the oldest frame in the output queue.

This also increases throughput for transcoding and improves seek
performance when supported by the underlying backend.

Introduce support in the vp9 base class, so that backends that
support render delays can actually implement it.

Co-authored by Seungha Yang <seungha@centricular.com>

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/987>

codecs: vp9decoder: Fix class struct documentation

s/GstVp9Decoder/GstVp9DecoderClass

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/987>

ges:uri-asset: Ignore streams with no ID

Since 0d95d9258b97 we respect the asset stream-id in `GESUriSource` so
we can not work with unknown or broken stream ID in the assets.

We just ignore them, warning about it and we should fix that in
demuxer so they don't expose pad without providing a stream id for them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1001>

typefindfunctions: differentiate h265 from h264

in some cases, the algo gives the same probability
to h264 and h265 for h26x stream resulting in a h265
stream detected as a h264.
if sps/pps/vps detected, increase the probabilty.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/957>

va: meson: Update and enhance meson syntax usage.

This patch contains two updates:

1. Instead of checking for dependency already checked just to verify a
   version, we use the dependency version API.
2. Update the deprecated function get_pkgconfig_variable.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/997>

msdk: meson: Don't get dependency variable before it's valid.

It's possible to have installed MediaSDK environment
package (libmfx-dev in Debian) without libva environment package. This
setup will lead to a breakage of meson configuration.

The fix is to get the libva's driver directory variable after the
dependency is validated as found.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/998>

ristsink: set sync to FALSE on RTCP sink

See commit 921e9a54: rtpsink: set sync off on rtcp_sink

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/993>

rtpsink: set sync off on rtcp_sink

When using the following setup (the error can be reproduced using
simpler sender pipelines), the receiver resynchronises the clock on RTCP
packets. The effect was that a couple seconds were cut out of the
playback because an initial RTCP packet was dropped.

When sending out all RTCP packets (setting sync=FALSE on the RTCP
updsink), the playback is fine.

This syncs rtpsink with rtpsrc (where this property was already set).

gst-launch-1.0 filesrc location=899-en.mp3 \
    ! mpegaudioparse \
    ! mpg123audiodec \
    ! audioconvert \
    ! audioresample \
    ! avenc_g722 \
    ! rtpg722pay
    ! rtpsink uri=rtp://239.1.2.3:1234

gst-launch-1.0 uridecodebin rtp://239.1.2.3:1234?encoding-name=G722 \
    ! autoaudiosink

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/993>

rtsp-session: Don't unref medias twice if it is removed inside gst_rtsp_session_filter() while the mutex is shortly released

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/757

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1004>

rtpmanagerbad: do not set iface on sink

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/994>

ci: Rebuild windows image

Get updates and newer ca-certificates in.

Also update the way we install chocolatey

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1002>

ci:doc: Rewrite script to import doc on fdo ensuring the right job is used

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/999>

move-script: Do not expect user to be members of the gstreamer project

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/990>

move-script: Handle users not having forked gstreamer core

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/990>

core: remove outdated mention to gst-build

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

mfdeviceprovider: Add support for device update

Similar to the wasapi2 plugin, GstWinRT library will be used for UWP,
and adding new GstWin32DeviceWatcher object implementation for
Win32 desktop application.

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

wasapi2deviceprovider: Add support for device update

... by using newly implemented GstWinRT library

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

libs: Introduce GstWinRT library

Adding a helper library for various WinRT specific implementations.
Currently this library supports only DeviceWatcher abstraction object
which can be used for dynamic device add/remove detection.
See also
https://docs.microsoft.com/en-us/uwp/api/windows.devices.enumeration.devicewatcher?view=winrt-20348

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

video: make GST_VIDEO_FORMAT_INFO_IS_*() macros booleans

Spotted by Stirling Westrup.

Fixes #726

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/926>

audio: make GST_AUDIO_INFO_IS_UNPOSITIONED return a boolean

Luckily this worked right since the flag we check is the
only flag and its vale is 1, but feels more correct to
actually check for non-zeroness.

Spotted by Stirling Westrup.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/926>

scripts:move_mrs: Revert cherry-picking when aborting

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/959>

Add .editorconfig used by some IDE like vscode

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/969>

Fix gitignore

- `*build*/` is not needed because Meson puts a .gitignore in builddir
already.

- `meson/` why?

- `/gst*` is too wide, it includes e.g. gst-env.py, list symlinks
explicitly instead.

- `subprojects/*` excludes all .wrap files, and `!subprojects/gst*` is
not enough because e.g. `macos-bison-binary` is still ignored. Instead
ignore only directories except those we include in our git repository.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/969>

Delete .arcconfig files

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/951>

indent: Ignore any changes in medias/

It is a submodule and we do not care about it

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/952>

docs: Always build documentation on gstreamer/gstreamer main branch

So artifacts can the be imported on gst.fd.o.

And update the importation script accordingly

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/952>

ci: Remove useless GIT_LFS_SKIP_SMUDGE and GIT_STRATEGY

GIT_LFS_SKIP_SMUDGE is not needed because we don't have git-lfs any
more.

GIT_STRATEGY is not needed because "fetch" is the default strategy.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/939>

docs: faq: monorepo: add another question and clarify bisecting section

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/953>

docs: faq: monorepo: import gitlab access token dialog screenshot

Best not to rely on external hosting.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/953>

rtphdrext: Give "inherited" direction in set_attributes

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Use NULL-safe case insensitive comparison

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Print warnings when trying to parse caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Pass just the attributes to the subclass

Since the base class now does the parsing, there is no need
to reproduce that code in all the subclasses, just pass the attributes
which are the only relevant bit anyway.

Also, only store the direction if the subclass accepted the caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Use fail_unless_equals_string() to make tests easier to debug

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext test: Use helper function for caps

Also, let the base class parse the direction for us.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext-rfc6464: Use helper function to set caps field

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Add helper function to set fields in caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Parse direction from the caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Set the direction in the caps from the property

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Store the direction in the base class

Store the direction associated wit the RTP header extension in the
base class so it can use it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Set caps without attributes as the default

Most subclasses just use the simple function, so just let the base class
do it. It makes less code in subclasses.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

rtphdrext: Put simple caps generation as the base class default

Instead of having a helper function that gets called by almost every
subclass, just let the base class set the caps fields automatically.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/906>

docs: add mono repo FAQ

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

move-mrs-script: Add an option to select which modules and MR to work on

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

move-mrs-script: Add options to use python gitlab config files

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

move-mrs-script: Get back to checked out branch

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

scripts: Add a script to rebase branches from old modules into monorepo

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

move-mrs-script: Add prefix to mr titles if no provided

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

move-mrs-script: Add a notion of when comments where added

And resolve already resolved discussions

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

Add a scripts to ease moving pending MRs to the monorepo

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

gitignore: Ignore symlinks

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

ci: Wait for cerbero pipeline to finish

So we are sure the pipeline is marked as failed if the cerbero sub pipeline fails

See https://gitlab.com/gitlab-org/gitlab/-/issues/341737 for details

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

doc: update IRC links to OFTC

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

ci: trigger cerbero

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

d3d11videosink: Add support for crop meta

... when upstream element is d3d11.

Note that, if upstream element is not d3d11, crop meta is almost
pointless since d3d11videosink will upload video frame to GPU memory
in any case.

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