ci: Run tests post merge

This will make tracking new races much easier

Mark other jobs as manual when pipeline is running on master

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1129>

ges:test: Document some GES failures

Namely races in:
 * check.gst-editing-services.edit_while_seeked_with_stop_it
 * check.gst-editing-services.check_layer_activness_gaps_it

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/803
and https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/804

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1130>

ges:tests: Create shorter assets to avoid timeouts

And use a simple GStreamer pipeline as testsrcbin with GstTranscoder
doesn't let us easily set the framerate of the source and we end up
having videorate dropping frames leading to the rendered file having
an unprecise duration.

This should fix races with `check.gst-editing-services.pythontests.pyunittest.python.test_assets.TestTimeline.test_reload_asset`

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1130>

validate:launcher: Avoid spamming terminal when inspecting unit tests

And redirect GStreamer logs when necessary

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1130>

validate:launcher: Run validate unit tests as GstValidate tests

So we have all the features and we can properly document known issue
for them

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1130>

codecparsers: {h264,h265}parser: Fix typo around SEI nalu generator

Fix to create correct SEI nalu when the size of payloadType and/or
payloadType is larger than 255 (0xff)

Fixes: https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1601
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1106>

rebase-branch-from-old: few improvments

- Enhance the documentation
- Allow to revert cherry-pick
- coding style

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1094>

validate: flow: Fix for line-ending mismatch issue on Windows

Unlike POSIX system, Windows distinguishes "w" and "wb" and when
a file is opened with text mode, OS will translate \n into \r\n.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1092>

vapostproc: Negotiate interlaced.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1109>

vapostproc: Copy missing fields at fixate.

When caps negotiation implies a caps feature change, some fields might
get lost. This patch brings them back from input caps.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1109>

vapostproc: Simplify size fixate.

gst_va_vpp_fixate_size() returned the fixated caps, but that is not
needed since `othercaps` are modified inline.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1109>

vapostproc: Simplify fixate.

The first approach to fixate was simply a copy&paste of both
videoconvert and videoscale, trying to keep their logic as isolated
as possible. But that brought duplicated and sparse logic.

This patch merge both approaches simplifying the fixate operation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1109>

va: filter, deinterlace, vpp: Add gst_va_buffer_get_surface_flags().

Add a helper function to get, from GstVideoInfo and GstBuffers flags,
the VA interlace surface flags. This is used currently by vainterlace
element, but it will be used in vapostproc too if it can process
interlaced frames.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1109>

build: Disable pygobject submodule if system version exists

If pygobject is available on the system, its version is new enough,
don't build the older pygobject and rely on that system one.

This fixes totem not being able to use libpeas' Python support.

** (totem:544972): WARNING **: 12:04:05.407: Error initializing Python Plugin Loader: PyGObject initialization failed
ImportError: could not import gobject (version mismatch, 3.40.1 is required, found 3.38.1)

Closes: #806
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1135>

tests: Blacklist some tests that recently failed

See:
    - https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/739
    - https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/740

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/927>

gtkglsink: Avoid assertion when applying "auto" rotation method

Guard against the orientation not coming from an inexistant tag, nor
from the application (rotation set to "auto") which caused an assertion.

When the application requests the auto rotation method, make sure it is
resolved to a rotation that's applicable.

ERROR:gstreamer/subprojects/gst-plugins-good/ext/gtk/gtkgstglwidget.c:745:gtk_gst_gl_widget_set_rotate_method: code should not be reached

Fixes: 103ceb853a5f7dade07a1ac4aa517e9df6ed5ded
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1134>

soup: Runtime compatibility support for libsoup2 and libsoup3

The src and sink elements no longer link against libsoup. It is now loaded at
runtime. If any version is resident already, it is used. Otherwise we first try
to load libsoup3 and if it's not found we fallback to libsoup2.

For the unit-tests, we now build one version of the test unit file per libsoup
version found. So if both libsoup2 and libsoup3 are available on the host, the
CI will cover them both.

Based on initial patch by Daniel Kolesa <dkolesa@igalia.com> and
Patrick Griffis <pgriffis@igalia.com>.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1044>

examples: playback-test: Fix for UI resize scenario on Windows

Application needs to notify videosink element of video widget resize
via gst_video_overlay_set_render_rectangle() since WM_SIZE event
wouldn't be notified.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1115>

audio: Merge simd libs into the main one

Actually extract the .o objects from the convience libraries and put
them into the main one. Without this, they will just be referenced by
the .pc file, but it will be unusable because they are not installed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1122>

ci: have the fedora docker build depend on the trigger job

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1120>

meson: Remove duplicate definition of 'examples' option

This is a hard error starting with Meson 0.60.0rc1, due to
https://github.com/mesonbuild/meson/pull/9184#issuecomment-941603358

The upstream change might get reverted before release, but we should
of course fix this regardless.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1127>

rtopuspay: Set marker bit inside RTP packet too

At the end of a talk spurt, not only set the marker flag on the
GstBuffer, but also set the bit inside the RTP header as recommended
by the RFC.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1124>

gtksink: Fetch the default "widget" value in the docs

There's really no interesting "widget" value that could be shown in the
docs, so use the GST_PARAM_DOC_SHOW_DEFAULT flag to avoid showing
another value.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

gtksink: Avoid errors fetching widget property

Avoid errors when fetching the "widget" property and GTK initialisation
fails, such as when running in a non-graphical environment.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

gtksink: Return reference to GtkWidget in the acquire function

This should ensure thread safety.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

gtkglsink: Add rotate-method property

This mostly just takes code out of glimagesink and applies it here.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

glutils: Export affine transformation functions for gtkglsink

Also remove duplicated copy of those functions from the gl plugin

With contributions from Bastien Nocera

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

glvideoflip: Replace GstVideoFlipMethod -> GstVideoOrientationMethod

It's the same enum, just drop the renamed copy. But keep the registered
GEnum as it is part of the API.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

glvideoflip: Use the API to parse the image orientation

This will reduce the code duplication a little.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

glimagesink: Use the API to parse the image orientation

This will reduce the code duplication a little.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

video: Add API to parse the image orientation from a GstTagList

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

glimagesink: Replace GstGLRotateMethod with GstVideoOrientationMethod

It's the same enum, just drop the renamed copy. But keep the registered
GEnum as it is part of the API.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

video: Put nicer documentation in GstVideoOrientationMethod

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1088>

splitmuxsink: Don't assert on the input side if no GOP is available when shutting down

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1119>

move-script: fix type in dateutil package name

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1080>

tests: Mark check.gstreamer.gst_gstbin.test_watch_for_state_change flaky

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/778

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

ges:tests: Mark check.gst-plugins-base.validate.giosrc.read-growing-file as flaky

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/776

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

ges:tests: Mark ges 'check_edit_in_frames_with_framerate_mistmatch as flaky

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/775

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

tests:webrtcbin: Make known flaky webrtcbin as such

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/773

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

validate:launcher: Do not dump on failure test we will retry

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

gst-env: Ignore SIGINT when running on the CI server

We get spurious sigint and this is just a small temporary workaround

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

tests: aggregator: Take `TIMEOUT_FACTOR` env var into account

This env var is set in the CI so we grow the timeout as required
(when running in valgrind for example).

Trying to avoid hitting wrong timeout in valgrind job like in:
https://gitlab.freedesktop.org/tpm/gstreamer/-/jobs/14009456

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

ci: Split base and core valgrind on different runners

Those are very slow compared to other jobs and it should mitigate the problem

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

tests: Blacklist tests that are currently producing many CI failures

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

validate:launcher: Allow more flexible handling of flaky tests

And retry twice flaky tests

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

validate:launcher: Fix progress reporting with --n-runs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

validate:tests: Mark rtsp tests as generally flaky

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

ges:tests: Mark nested timelines as generally flaky

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

validate:launcher: Support retry on failure when using --n-runs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

validate:launcher: Fix potential invalid index access

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

validate:launcher: Fix adding logs to xunit

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1113>

gst-play: pick up minus and plus also from navigation events

Makes it easier to test playback rate changes with the video
window being in focus.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/928>

libsoup: Bump to 2.74 to fix gssapi build failure on macOS

See: https://gitlab.gnome.org/GNOME/libsoup/-/merge_requests/255

We also need to disable sysprof by default because it only builds on
native Linux. If someone really wants it, they can enable the option
on the command-line by passing -Dlibsoup:sysprof=enabled

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1114>

msdkenc: fix vp9enc initialization fail

MediaSDK does not support to handle extbuff with id
MFX_EXTBUFF_VIDEO_SIGNAL_INFO for mjpegenc and vp9enc. Hence, need to
exclude mjpeg and vp9 when passing color properties to MediaSDK during
msdkenc initialization.

Fix issue: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/764

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1101>

examples: d3d11: Add a desktop capture example

... with d3d11 desktop capture device provider

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1103>

d3d11: Add device provider for d3d11desktopdupsrc

... and add support for multi-GPU/multi-monitor

By using newly added "monitor-handle" property, user can specify a
monitor to be captured via HMONITOR handle.

Fixes: https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1673
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1103>

ttml: fix log init

The log system should be init before calling a log

Fix regression after:

https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2038

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1112>

move-mrs-script: add url option

You can give as option:

--url
https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/1

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/960>

move-mrs-script: add a list only options

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/960>

tests: va: Add VA buffer copy tests.

It should only work for raw buffers, but fails on dmabuf since it
should have a drm modifier.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1023>

va: allocator: dmabuf: Use GstVaSurfaceCopy, if possible.

If dmabuf-based buffer to copy contains only one memory, and there are
memories available in the allocator's pool, it's possible a fast
memory copy using GstVaSurfaceCopy, regardless the drm modifier.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1023>

va: allocator: Use GstVaSurfaceCopy.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1023>

va: Add GstVaSurfaceCopy class.

This new class is a helper for fast/tricky copy of surfaces. First it
tries to copy using the function vaCopy in libva 1.12. If it fails, or
it's not available, a GstVaFilter is tried to be instantiated with the
allocator's parameters, and if succeed, it's used for copying the
source surface.

This is required for dmabuf surfaces with drm modifier.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1023>

va: filter: Enable to pass VASurfaceID in GstVaSample.

Initially GstVaSample processed its GstBuffer member to get the
VASurfaceID. But it might cases where we already have the VASurfaceID
to process by the filter.

This patch enables the possibility to pass the surfaces rather than
the buffers. In order to validate the surfaces a function to check
surfaces were added.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1023>

va: Split VA memory handling in different files.

Take out the VA memory wrappers from gstvallocator.c to an external
file exposing the functions.

This is going to be needed for the copy helper object.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1023>

va: allocator: Add missing header file.

Added stdint.h because uintptr_t is used.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1023>

va: display: Add gst_va_display_has_vpp()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1023>

v4l2codecs: Align v4l2-controls header with kernel 5.15-rc3

Update v4l2-controls to be aligned with kernel 5.15-rc3.
Fix VP8 decoder to use the correct field name.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1081>

ci: remove outdated rule for gstreamer-full-minimal

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1065>

ci: update rules that pointed to gst-ci subproject

The files are now moved into ci/ at the root of the repo.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1065>

uridecodebin3/urisourcebin: Reusability fixes

Improvements to uridecodebin3 and urisourcebin so that they are
reusable across a PAUSED->READY->PAUSED transition.

Disconnect and release decodebin3 request pads when urisourcebin
removes src pads.

In urisourcebin, make sure to remove src pads that are exposed
directly (raw pads and static typefind srcpads) when
cleaning up.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/768

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1100>

playbin3: Always register 'playbin3' element.

If the USE_PLAYBIN3=1 env var is set, we want to replace
playbin with playbin3, but separate to that, we always
want to register the 'playbin3' element so that applications
which explicitly use playbin3 work regardless of the env var.

This fixes `USE_PLAYBIN3=1 gst-validate-launcher`, for example.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1102>

mediafoundation: mfvideoenc: Use DXGI adapter LUID

Make use of new DXGI adapter LUID based device context sharing.
Note that we were using DXGI adapter LUID to open MFT already.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1098>

d3d11: d3d11decoder: Use DXGI adapter LUID

... instead of index of DXGI adapter.

The order of IDXGIAdapter1 enumerated via IDXGIFactory1::EnumAdapters1
can be varying even there's no rebooting in case that GPU preference order
is updated by user (for example, it can be done by using NVIDIA Control Panel
in case of multi-GPU laptop system) and eGPU is another possible case.

So, for an element which requires fixed target GPU requirement,
index based device enumeration is unreliable.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1098>

d3d11: d3d11utils: Add support for DXGI Adapter LUID based D3D11 device context sharing

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1098>

d3d11: d3d11device: Add gst_d3d11_device_new_{for_adapter_luid,wrapped}

* gst_d3d11_device_new_for_adapter_luid()
  Used for creating D3D11 device for a DXGI adapter (i.e., GPU)
  corresponding to a LUID (Locally Unique Identifier).
  This method can be useful for interop with other APIs such as
  Direct3D12, MediaFoundation, CUDA, etc.
* gst_d3d11_device_new_wrapped()
  Allows creating a new GstD3D11Device object by using already
  configured ID3D11Device. This is conceptually equivalent to
  gst_gl_context_new_wrapped()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1098>

d3d11: d3d11device: Remove "allow-tearing" property

Plugin can query DXGI_FEATURE_PRESENT_ALLOW_TEARING without d3d11device
help

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1098>

webrtcbin: Use the same promise reply structure name everywhere

This was an inconsistent mix of different names in the past. The name
has no meaning at all so let's set all to "application/x-gst-promise".

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1099>

playbin3: Avoid group deactivation deadlock.

Change locking around group deactivation to avoid deadlocks
when shutting down exactly as a buffering message arrives.

The PLAYBIN3_LOCK now protects the active field of the
source group. Everything else is still protected by the
source-group-lock.

Also properly protect group switching operations with
the PLAYBIN3_LOCK everywhere.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1049>

chopmydata: Fix FIXMEs in gst_element_class_set_static_metadata

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1097>

FAQ: add mention to downstream cherry-pick

Give an help to cherry-pick an upstream patch to
downstream repository (old stable version)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1095>

d3d11: d3d11desktopdupsrc: Add support for non-D3D11 downstream element

By this commit, application doesn't need to configure d3d11download
element for software pipeline which will make things simpler

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1082>

validate:scenario: Allow waiting signal searching for element factory

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1064>

jpegdec: Support libjpeg-turbo colorspace conversion

The libjpeg-turbo has a built-in support for performing colorspace
conversion. The performance of this conversion is much better than
doing the same separately using videoconvert. Implement support for
this conversion to RGBx/xRGB/BGRx/xBGR formats. Other formats can
be easily added later.

- The decoding of various pixel formats can be tested and compared to
  non-libjpeg-turbo decoding as follows:
  for gfmt in {RGB,BGR}{,x} x{RGB,BGR} ; do
      echo "$gfmt"
      gst-launch-1.0 -q \
          videotestsrc pattern=colors ! \
          video/x-raw,format=${gfmt} ! \
          fakesink dump=true | \
  head -n 200 | tail -n 1
      gst-launch-1.0 -q --gst-plugin-path=build/ext/jpeg/ \
          videotestsrc pattern=colors ! \
          video/x-raw,format=${gfmt} ! \
  jpegenc ! \
  jpegdec ! \
  video/x-raw,format=${gfmt} ! \
  fakesink dump=true | \
  head -n 200 | tail -n 1
      done
  Result looks as follows, i.e. comparable:
  RGB
  00000c70 (0x7f7736fbdd10): 05 33 19 05 33 26 05 33 33 05 33 40 05 33 4c 05  .3..3&.33.3@.3L.
  00000c70 (0x7f389e8f7d10): 05 32 17 04 32 22 04 32 31 04 32 3e 04 32 4a 04  .2..2".21.2>.2J.
  RGBx
  00000c70 (0x7f79efd0ad10): cc 07 22 ff d9 07 22 ff e6 07 22 ff f3 07 22 ff  .."..."..."...".
  00000c70 (0x7fb6989f3d10): cd 06 22 00 d9 06 22 00 e6 06 22 00 f4 06 22 00  .."..."..."...".
  BGR
  00000c70 (0x7fa0a6c42d10): 05 0c 33 05 19 33 05 26 33 05 33 33 05 40 33 05  ..3..3.&3.33.@3.
  00000c70 (0x7fc74165fd10): 05 08 32 04 17 32 04 22 32 04 31 32 04 3e 32 04  ..2..2."2.12.>2.
  BGRx
  00000c70 (0x7fbf399f1d10): 22 07 cc ff 22 07 d9 ff 22 07 e6 ff 22 07 f3 ff  "..."..."..."...
  00000c70 (0x7f50e3d1cd10): 22 06 cd 00 22 06 d9 00 22 06 e6 00 22 06 f4 00  "..."..."..."...
  xRGB
  00000c70 (0x7f0b950a2d10): ff cc 07 22 ff d9 07 22 ff e6 07 22 ff f3 07 22  ..."..."..."..."
  00000c70 (0x7f4416b8dd10): 00 cd 06 22 00 d9 06 22 00 e6 06 22 00 f4 06 22  ..."..."..."..."
  xBGR
  00000c70 (0x7f237d74dd10): ff 22 07 cc ff 22 07 d9 ff 22 07 e6 ff 22 07 f3  ."..."..."..."..
  00000c70 (0x7f095547dd10): 00 22 06 cd 00 22 06 d9 00 22 06 e6 00 22 06 f4  ."..."..."..."..
                             ^^          ^^          ^^          ^^
  Notice how the alpha channel is set to arbitrary value in case of the
  libjpeg-turbo decoding into RGBx/BGRx/xRGB/xBGR pixel formats. This is
  documented in libjpeg-turbo README.md as follows:
    "
    When using the RGBX, BGRX, XBGR, and XRGB colorspaces during decompression, the
    X byte is undefined, and in order to ensure the best performance, libjpeg-turbo
    can set that byte to whatever value it wishes.
    "

- The interlaced num_fields=2 mjpeg stream can be generated and
  tested as follows (this does require mjpegtools):
  $ gst-launch-1.0 videotestsrc num-buffers=10 ! jpegenc ! multifilesink location=in%04d.jpg
  $ jpeg2yuv -f 25 -I t -L 0 -j in%04d.jpg | yuv2lav -f avi -o result.avi
  ...
  $ gst-launch-1.0 --gst-plugin-path=build/ext/jpeg/ filesrc location=result.avi ! \
       avidemux ! jpegdec ! video/x-raw,format=RGBx ! videoconvert ! autovideosink

Signed-off-by: Marek Vasut <marex@denx.de>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1091>

rtsp-media: Unprepare suspended medias too

Previously suspended medias immediately reached the UNPREPARED state
without going through the media's unprepare() vfunc. This didn't allow
the media subclass to do any additional cleanup, and for example the
shutdown-eos property of GstRTSPMedia was ignored.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1090>

tests: gstmeta: Fix failure on Windows

Windows doesn't support fork so every test will be performed in
one process. So the test_meta_custom_transform() is being
failed because "test-custom" custom meta is being used/defined in
another test test_meta_custom() as well.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1086>

codecs: Use GST_VIDEO_DECODER_ERROR() only for decoding error case

The GST_VIDEO_DECODER_ERROR() should be used only for robust/error-resilient
decoding purpose. Any other error codes such as not-negotiated or flushing
should be returned without modified for upstream to be able to handle
it immediately. (for example, application might want to try other
decoder element on not-negotiated)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1070>

codecs: h264decoder: Update for remaining gboolean to GstFlowReturn port

Fix for spurious/spammy warning and wrong function return type

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1085>

rtsp-media: Only unprepare a media if it was not already unpreparing anyway

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1083>

meson: Fix gst-launch build on UWP

The APIs we need from WinMM for higher timer resolution are only
available for non-UWP apps. The winmm library itself is still
available, so we need to disable the check.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/890>

meson: Fix build with -Dopenjpeg=disabled

tests/check/meson.build uses the openjpeg_dep variable
unconditionally, and the subdir_done() is useless anyway, since the
plugin is only built if openjpeg_dep.found() is true. Fixes:

..\tests\check\meson.build:23:0: ERROR: Unknown variable "openjpeg_dep".

In particular, this fixes the build on UWP since we disable openjpeg
explicitly in Cerbero when building for UWP.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1069>

bad: gs: update README

- add one missing dep
- change install path to match monorepo
- fix current dirs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1008>

rtsp-client: make sure sessmedia will not get freed while used

handle_*_request() functions were all retrieving the session media from
the session by calling gst_rtsp_session_get_media () which is a transfer-none
call. If a session timeout happens at that time, the session media may get freed
making the pointer invalid..

Fixes #757

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1053>

validate:launcher: Sync asset only when passing --sync

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1020>

ci: Fix typo in cerbero trigger

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

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>