h264parse: Add new H.264 levels

The spec now list 6, 6.1 and 6.2.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1509>

va: h264dec: remove spurious comment

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1499>

va: h264dec: check return value of gst_va_handle_set_context()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1499>

smoothstreaming: start closer to the edge in live streams

It is more appropriate to start closer to the live edge in
live streams. Some live streams maintain a large dvr window
(over few hours in some cases), so starting from the first
fragment will be too far away from the live edge.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1346>

cccombiner: Update for additional info parameter to the "samples-selected" signal

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/590

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1498>

build: update for gl pkg-config file split

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1462>

docs: include *.cc and *.hh in gst-c-sources

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

mpeg2enc: add disable-encode-retries property

MJPEG Tools may reencode pictures in a second pass to stick
closer to the target bitrate. This can result in slower than
real-time encoding for full HD content in certain situations,
as entire GOPs need reencoding when the reference picture is
reencoded.

See https://sourceforge.net/p/mjpeg/bugs/141/ for background

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

mpeg2enc: report a latency

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

mpeg2enc: finalize GstVideoEncoder port

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

mpeg2enc: store video encoder instance directly in stream writer class

Instead of storing the pad and then only using it to get the
element.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

mpeg2enc: remove unused streamwriter member 'buf'

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

mpeg2enc: remove some unused code

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

mpeg2enc: remove code paths for older mjpegtools versions

Gets rid of lots of code paths that no one has built,
used or tested for ages, and makes code more maintainable.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

mpeg2enc: initial port to GstVideoEncoder base class

https://bugzilla.gnome.org/show_bug.cgi?id=685414

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1491>

decklink: Re-order modes enum for backwards compatibility with 1.16

The PAL/NTSC widescreen modes were added after 1.16 but inserted before
the HD modes, which changed the integer values of the enums.

Move them to the very end instead to keep backwards compatibility.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1048

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1492>

srt: Add support for using hostnames instead of IP addresses

If an address can't be parsed as IP address, try resolving it via
GResolver instead. SRT URIs more often than not contain hostnames and
without trying to resolve them we won't be able to handle such URIs.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1493>

cccombiner: update to new samples selection API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1497>

opencv: compile with -Wno-format-nonliteral

opencv plugin is pulling a header which makses clang++ 10
complain a lot and blocks -werror.

```
/usr/include/opencv4/opencv2/flann/logger.h:83:36: error: format string is not a string literal [-Werror,-Wformat-nonliteral]
        int ret = vfprintf(stream, fmt, arglist);
                                   ^~~
```

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1494>

player: Add g_autoptr() support

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1495>

gstlv2utils.c: avoid implicit float to int conversion

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1487>

gstautoconvert.c: fix clang warnings

clang 10 is complaining about incompatible types due to the
glib typesystem.

```
gst-plugins-bad/gst/autoconvert/b5c3019@@gstautoconvert@sha/gstautoconvert.c.o' -c ../subprojects/gst-plugins-bad/gst/autoconvert/gstautoconvert.c
../subprojects/gst-plugins-bad/gst/autoconvert/gstautoconvert.c:898:8: error: incompatible pointer types passing 'typeof ((((void *)0))) *' (aka 'void **') to parameter of type 'GList **' (aka 'struct _GList **') [-Werror,-Wincompatible-pointer-types]
  if (!g_atomic_pointer_compare_and_exchange (&autoconvert->factories, NULL,
       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/include/glib-2.0/glib/gatomic.h:192:44: note: expanded from macro 'g_atomic_pointer_compare_and_exchange'
    __atomic_compare_exchange_n ((atomic), &gapcae_oldval, (newval), FALSE, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST) ? TRUE : FALSE; \
                                           ^~~~~~~~~~~~~~
1 error generated.
```

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1487>

gstladspautils.c: avoid implicit float to int conversion

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1487>

androidmedia: ignore additional camera effects if not present

Fixes https://gitlab.freedesktop.org/gstreamer/cerbero/-/issues/283

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1470>

meson: Only look for Obj-C/C++ compilers on macOS/iOS

On Windows, MinGW-GCC Objective-C/C++ compilers can be in PATH and
mess up the build since they may not match the CPU family of the C/C++
compilers we are using.

Also require them on macOS/iOS, because they should always be present.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-build/-/issues/88

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1488>

va: h264dec: log if upstream pool is kept

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1489>

va: utils: fix precondition check for handle_context_query()

display paramater can be NULL, but if it's not, it has to be a
GstVaDisplay.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1489>

va: tests: example: Fix memory leaks

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1489>

srtobject: Add support for IPv6

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1477>

srtobject: Reset parameters before setting URI

This makes `gst_srt_object_validate_parameters` work properly since
`localaddress` and `localport` will be missing if the URL did not
provide them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1477>

srtobject: Simplify gst_srt_object_set_*_value

This fixes `gst_srt_object_set_string_value` in particular because the
value might not be a static string.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1477>

srtobject: Store passphrase like other parameters

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1477>

webrtc, rtmp2: Warn if the user or password aren't escaped

If the user/pass aren't escaped, the userinfo will be ambiguous and we
won't know where to split. We will accidentally get it right if the :
belongs in the password.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1481>

webrtc, rtmp2: Fix parsing of userinfo in URI strings

While parsing the string, `gst_uri_from_string()` also unescapes the
userinfo. This is bad if your username contains a `:` character, since
we will then split the userinfo at the wrong location when parsing it.

To fix this, we can use the new `gst_uri_from_string_escaped()` API
that was added in
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/583

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/831

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1481>

tests: examples: add va-x11-render example

This a GTK+ example will share, through GstContext, a custom X11
VADisplay to a pipeline using vah264dec and appsink.

When the frames are processed for rendering, the VASurfaceID is
fetched from the buffer and it is rendered using vaPutSurface in a X11
widget.

tests: examples: Comply with compilation order

va: allocator: support for GST_MAP_VA map flag

This flag will return the VASurface value at mapping

va: context: instanciate VA display through GstContext

Add all the machinery to instanciate VA display through GstContext,
thus all va elements can share the same display and the user can set
a custom one.

va: h264dec: don't copy frames if VAMemory capsfeature is negotiated

Otherwise the VASurfaceID is lost.

va: h264dec: copy render_device_path in klass

It it's not copied both cdata and klass, the string is lost. Thus
also it's freed from cdata when freeing it.

va: display: wrapped: Fix property name

va: decoder: remove unused argument

And that changes function's namespace

va: caps: bail raw caps if driver doesn't report surface formats

This is a bug in Gallium RadeonSI driver for Polaris10, which doesn't
report sufrace formats for reported chroma.

If one chroma doesn't report surface formats, skip the generated caps.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1483>

va: allocator: get a surface format from a image format

For the allocator to create surfaces with the correct chroma an
fourcc, it should use a surface format, not necessarily the negotiated
format.

Instead of the previous arbitrary extra formats list, the decoder
extracts the valid pixel formats from the given VA config, and pass
that list to the allocator which stores it (full transfer).

Then, when the allocator allocates a new surface, it looks for a
surface color format chroma-compatible with the negotiated image color
format.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1483>

va: caps: add raw caps image formats with same chroma of surfaces

Instead of adding a list of ad-hoc formats for raw caps (I420 and
YV12), the display queries the available image formats and we assume
that driver can download frames in that available format with same
chroma of the config surfaces chroma.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1483>

va: display: add gst_va_display_get_image_formats()

For this it was also added gst_va_video_format_from_va_image_format()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1483>

va: decoder: initialize rt_formas to zero

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1483>

d3d11download: Allow linking with downstream d3d11 elements

It will make pipeline configuration easier since d3d11download
element can be placed unconditionally. This behavior is similar
to that of gldownload element.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1485>

hlssink2: Don't assert if we don't have a current location when receiving the fragment-closed message

This can happen if the application did not provide an output stream for
the fragment and didn't handle the error message before splitmuxsink
decided to consider the fragment closed.

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

opencv: allow compilation against 4.4.x

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1482>

mfvideosrc: Select common formats if both VideoPreview and VideoRecord are available

Some devices (e.g., Surface Book 2, Surface Pro X) will expose
both MediaStreamType_VideoPreview and MediaStreamType_VideoRecord types
for a logical device. And for some reason, MediaStreamType_VideoPreview
seems to be selected between them while initiailzing device.
But I cannot find any documentation for the decision rule.
To be safe, we will select common formats between them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1478>

mfvideosrc: Check framerate for target IMediaFrameFormat selection

Not only resolution and format, but framerate needs to be checked
for proper target IMediaFrameFormat selection.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1478>

mfvideosrc: Handle I420/IYUV subtypes for UWP cases

Microsoft defines two I420 formats, one is I420, and the other is
IYUV (but both are same, just names are different).
Since both will be converted to GST_VIDEO_FORMAT_I420,
we should check both I420 and IYUV subtypes during
GstVideoFormat to Microsoft's format conversion.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1478>

mfvideosrc: Add more debug log

It would be useful for finding the error reason.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1478>

docs: Update wasapi2 and mfvideosrc doc

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1480>

wasapi2, mfvideosrc: Update "dispatcher" property to be only writable

Disallow getting dispatcher pointer, since it doesn't seem to be useful
and might not be safe.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1480>

cccombiner: implement samples selection API

Call gst_aggregator_selected_samples() after identifying the
caption buffers that will be added as a meta on the next video
buffer.

Implement GstAggregator.peek_next_sample.

Add an example that demonstrates usage of the new API in
combination with the existing buffer-consumed signal.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1390>

wpesrc: timestamp buffers when working with SHM buffers

GLBaseSrc::fill() will take care of that when dealing with
images, but as we don't chain up when dealing with SHM buffers
this needs to be done in order for GLBaseSrc::get_times() to
work appropriately.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1476>

wpe: fix ready signalling

Receiving the WEBKIT_LOAD_COMMITTED event doesn't actually
mean we have committed an SHM buffer / image yet.

As this is the condition we are interested in, check it
instead.

Also wrap g_cond_wait in a loop for extra correctness points.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1476>

basecamerabinsrc: silence g-ir-scanner warnings

They're legit, but there's lots of other stuff that needs
fixing up in this API, so just silence for now and add a
FIXME and leave it for some other day.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1475>

ristsrc: drop stream-start & eos messages posted from the internal udp sink(s)

See #1368

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1472>

rtpsrc: drop stream-start & eos messages posted from the internal udp sink(s)

See #1368

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1472>

tsmux: Fix PCR calculation for CBR live streams

Take the first ever timestamp as an offset

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1431>

tsmux: Refactor get_current_pcr

No functional change.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1431>

camerabinpreview: fix potential crash on preview pipeline error

Post error message on actual element, not the allocated helper struct.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1474>

waylandsink: Update stale GstBuffer references in wayland buffer cache

"waylandsink: use GstMemory instead of GstBuffer for cache lookup"
changes the cache key to GstMemory, but the cached data still needs
a pointer to the GstBuffer to control the buffer lifecycle.
If the GstMemory used as the cache key moves from one GstBuffer to
another, the pointer in the cached data will be out-of-date.

Update the current GstBuffer pointer for each frame so that it always
represents the currently in use (from attach to release) GstBuffer
for each wl_buffer.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1473>

wasapi2device: Allow empty caps for UWP use case

If the device has not been activated yet, caps might not be available.

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

wasapi2: Activate device asynchronously if required

In case of UWP, documentation from MS is saying that
ActivateAudioInterfaceAsync() method should be called from UI thread.
And the resulting callback might not happen until user interaction
has been made.
So we cannot wait the activation result on constructed() method.
and therefore we should return gst_wasapi2_client_new()
immediately without waiting the result if wasapi2 elements are
running on UWP application.
In addition to async operation fix, this commit includes COM object
reference counting issue around ActivateAudioInterfaceAsync() call.

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

wasapi2: Add a new property for ICoreDispatcher setting

... so that ensure device activation on UI thread.

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

mfvideosrc: Only device activation needs to be running on UI thread

... and the other operations does not have the thread constraint.

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

mfvideosrc: Add a new property for ICoreDispatcher setting

Since the commit c29c71ae9d46cc70e29e0cca2313917f319ef6f2,
device activation method will be called from an internal thread.

A problem is that, CoreApplication::GetCurrentView()
method will return nullptr if it was called from non-UI thread,
and as a result, currently implemented method for accessing ICoreDispatcher
will not work in any case. There seems to be no robust way for
accessing ICoreDispatcher other then setting it by user.

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

mfvideosrc: Suppress more spammy debug messages

The failure on compressed format (e.g., MJPG, H264 subtypes) is expected.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1468>

directfb: suppress compiler warning from directfb headers

On debian sid, directfb 1.7.7

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1467>

d3d11window: Remove timeout and wait a task if possible

If the run_async() method is expected to be called from streaming
thread and not from application thread, use INFINITE as timeout value
so that d3d11window can wait UI dispatcher thread in any case.

There is no way to get a robust timeout value from library side.
So the fixed timeout value might not be optimal and therefore
we should avoid it as much as possible.

Rule whether a timeout value can be INFINITE or not is,
* If the waiting can be cancelled by GstBaseSink:unlock(), use INFINITE.
  GstD3D11Window:on_resize() is one case for example.
* Otherwise, use timeout value
  Some details are, GstBaseSink:start() and GstBaseSink:stop() will be called
  when NULL to READY or READY to NULL state change, so there will be no
  chance for GstBaseSink:unlock() and GstBaseSink:unlock_stop()
  to be called around them. So there is no other way then timeout way.
  GstD3D11Window:consturcted() and GstD3D11Window:unprepare() are the case.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1461>

d3d11window: Increase timeout to 10 sec for WinRT window implementations

No strong reason for this number, but we can follow other well known
UWP implementation ANGLE.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1461>

codecs: vp8decoder: Sync up with h264decoder implementation

Pass GstVideoCodecFrame with GstVp8Picture to new_picture() and
output_picture() methods for subclass to be able to reference
it directly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1458>

v4l2codecs: vp8dec: Remove white space

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1458>

meson: add a plugin summary

This summary displays a list of plugins which
have been enabled.

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

iqa: Add a 'mode' property

This property currently only supports a 'strict' that checks that
all the input streams have the exact same number of frames.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1424>

iqa: Implement child proxy

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1424>

codecs: vp9decoder: Sync up with h264decoder implementation

* Pass GstVideoCodecFrame to new_picture() and output_picture()
* Pass the last reference of GstVp9Picture to subclass if possible

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

wasapi2: Fix broken debug log

I hope that MSVC is smart enought to warn this at build time

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

v4l2codecs: h264dec: Remove outdated comment

... and unnecessary user data clear on GstVideoCodecFrame

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

v4l2codecs: h264dec: Fix leak on decoding failure

Given GstVideoCodecFrame and GstH264Picture should be cleared,
and the return must be GST_FLOW_ERROR, not boolean value.

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

d3d11decoder: Disable array-of-texture

Some 4K VP9 + super frame enabled streams would be broken with
array-of-texture configuration (i.e., allowing runtime texture registration
for DPB).

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

mfutils: Suppress spammy debug print

Remove FIXME debug print. It seems to be spammy.

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

mfvideosrc: Expose sorted caps

Sort the list of supported caps for downstream to be able to select
the best image in terms of quality (resolution and framerate) by default.

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

mfvideosrc: Don't expose unsupported formats

Some UVC cameras support H.264 stream but we don't support it yet.

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

docs: update plugins doc cache

Add va plugin

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

va: documentation since tag

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

vah264dec: add documentation caps for pad templates

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

d3d11window: Disable d3d11 video processor for HDR10 stream.

d3d11 video processor is showing buggy behaviors in various cases.
And this HDR10 is one case of them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1454>

codecs: h264decoder: Fix wrong annotations

What intended was GstH264Decoder actually.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1453>

codecs: h264decoder: Update document with Since marks

To make documentation CI happy with the newly added APIs.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1449>

codecs: h264decoder: Rename API arguments

Although it's not public ones, make them consistent with vfunc
for them to be nicer.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1449>

v4l2codecs: h264dec: Use the new picture lookup

This will avoid having to create a reference chain from CodecFrame
to GstH264Picture to GstV4l2Request.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1449>

h264decoder: Add ability to lookup by system_frame_number

This is to avoid having to create a reference chain in decoders from
GstVideoCodecFrame to GstH264PIcture to implementation wrapper.

So this change introduces:

 - gst_h264_dpb_remove_outputed (dpb)
 - gst_h264_dpb_get_picture(dpb, system_frame_num)
 - gst_h264_decoder_get_picture (dec, system_frame_num)

In order to ensure that frames can be looked up during the draining
process, we now first remove all (including reference) frames that
have been outputed but are still in the DPB. Then for each remaining
buffers, we remove it from the DPB to reach reference 1 and output it.

Previously we could take all not outputed outside of the DPB which would
prevent lookup by the base class.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1449>

codecs: h264decoder: Make GstVideoCodecFrame hold the last reference of the buffer

... when gst_video_decoder_finish_frame() is called.

Some subclasses attach GstBuffer to GstH264Picture as an user data
which will increase reference count of the buffer. It would result
to buffer copy per every frame.

Background here is, GstVideoDecoder wants writable output buffer for
GstMeta handling, and if the output buffer is not writable
(i.e., reference count is not one), the buffer will be copied.
Even if underlying GstMemory wouldn't be copied, buffer copy operation
will introduce extra memory allocation overhead which is not optimal.

By this modification, subclass might be able to receive the last
reference to GstH264Picture

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1449>

codecs: h264decoder: Pass GstVideoCodecFrame to output_picture()

All subclasses are retrieving list to get target output frame, which
can be done by baseclass. And pass the ownership of the GstH264Picture
to subclass so that subclass can clear implementation dependent resources
before finishing the frame.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1449>

codecs: h264decoder: Don't attach extra ref of GstH264Picture to GstVideoCodecFrame

The lifecycle of GstH264Picture is being managed by our DPB
implementation. If it's still required, subclass can do that
by itself in the new_picture() method.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1449>