rtspsrc: Don't blindly copy over sticky events from manager pad to external source pad

This would get around the code that modifies some events when they go
through the ghost pad's proxypad. Instead go via the event function.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3409>

rtspsrc: Don't make udpsrc segment events writable just to retrieve their seqnum

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3409>

rtspsrc: Reset EOS flag also on FLUSH_STOP and not only on ssrc-active

Also don't bother not sending EOS if EOS was sent already:
gst_pad_push_event() takes care of that for us already.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3409>

pbutils: Fix vp9 mime mapping when caps when no chroma-site.

It's possible to have 4:2:0 chroma format without chroma-site in caps, but it
isn't validated by the current code. This patch fixes it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3405>

gdkpixbufoverlay: fix docs - changing images at runtime is supported

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3415>

videodecoder: Consider having output data when the subclass drops a frame

The subclass might drop a frame for QoS reasons (e.g. vpxdec) and if all
frames are dropped because of that it wouldn't make sense to post an
error message on EOS.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3408>

codectimestamper: Remove duplicated GstSegment struct

Use the one in priv struct. And use gst_segment_is_equal()
to compare GstSegment

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3380>

h264/5timestamper: provide a workaround for h264/5parse producing pts=NONE buffers

A workaround for
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/649
and
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/287
which is hard to change baseparse behaviour for both video and audio
parsers.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3380>

rtpjitterbuffer: Unlock timer waits on flushing

If there is a pending EOS wait for example, we would never unblock on flushing

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3401>

inspect: Print default value of ValueArray properties

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3377>

nvcodec: Update plugin cache

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>

tests: cudaconvert: Update test code

Adding more formats, and rescale test with borders

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>

cudaconvertscale, cudascale: Add "add-borders" property and support 8bits RGB planar formats

Adding "add-borders" property which is identical to that of
videoscale and this will be enabled by default.
And adding RGBP/BGRP/GBR/GBRA format support.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>

cuda: Rewrite colorspace/rescale object

Rewriting GstCudaConverter object, since the old implementation was not
well organized and it's hard to add new features.
Moreover, the conversion operations were not very optimized.

Major change of this implementation:
* Remove redundant intermediate conversion operations such as
  any RGB -> ARGB(64) conversion or any YUV -> Y444 (or 16bits Y444).
  That's not required most of cases. The only required case is
  converting 24bits (such as RGB/BGR) packed format to 32bits format
  because CUDA texture object does not support sampling 24bits format
* Use normalized sample fetching (i.e., [0, 1] range float value)
  and also normalized coordinates system for CUDA texture.
  It's consistent with the other graphics APIs such as Direct3D
  and OpenGL, that makes sampling operations much easier.
* Support a kind of viewport and adopt math for colorspace conversion
  from GstD3D11 implementation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>

cudaupload,cudadownload: Add support for planar 8bits RGB formats

Defines RGBP, BGRP, and GBR formats, which have the same memory
layout as already supported Y444. And defines GBRA format as well
which needs just one additional alpha plane

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>

cudacontext: Store texture alignment

it was missed in the previous refactoring

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>

cudaconvert, cudascale: Port to GstCudaBaseCovert baseclass

Don't need to hold duplicated code in the source tree

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>

cuda: Add convertscale element

GstCudaConverter object can do colorspace conversion and scale at once.
Adding new element "cudaconvertscale" to do that, this can
save unnecessary GPU operation if colorspace conversion and
rescale is required for given input stream format.

Most of codes are taken from d3d11convert element

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>

typefindhelper: Use the new GstTypeFind * API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3296>

typefindhelper: Introduce a new API for working with a GstTypeFind *

Introduce a new API that can return a GstTypeFind * with helper functions
and data set around buffer data.

While at it, drop factory field from GstTypeFindBufHelper. While it was
useful for logging, it was not passed through function arguments and keeping
it for logging would require an additional API increasing the API surface
and making it harder to use.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3296>

typefindhelper: Add helpers to improve type finding given the caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3296>

typefindhelper: Update annotation for function parameters

Fix the nullable/optional situation. allow-none is deprecated and
replaced by either or both of the others.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3296>

validate-scenario: Avoid timeouts on EOS

If there is an EOS and the next action is waiting for a playback time, execute
it since we are guaranteed the position won't change anymore.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3403>

gstplay: add warning message details check before post it

For some warning message which has no detail data, need check to avoid printing critical log.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3382>

tests: Add DASH MPD baseURL with query test

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1147>

mpdparser: Fix missing baseURL query

When no initializationURL or mediaURL, return baseURL that also
contains original URI query if available. This fixes a problem
where URI query was being omitted in the HTTP requests.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1147>

mpdparser: Be consistent about returning duplicated URL

Instead of returning a "const gchar" or a "gchar" that should not be freed, always
return a duplicated string as those functions were used together with g_strdup anyway.

This is needed to prepare support for returning modified strings in next commit.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1147>

mpdparser: Return correct mediaURL value

This fixes a problem where get_mediaURL was returning NULL when segmentURL
was unavailable instead of baseURL as a fallback.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1147>

d3d11converter: Fix YUY2 conversion error

Always configure shader conversion path, then fallback to the
shader path if processor is not available

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3392>

d3d11: Add support for planar RGB formats

Adding RGBP, BGRP, GBR, GBR_10LE, GBR_12LE, GBRA, GBRA_10LE, and
GBRA_12LE format support

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3392>

d3d11screencapturesrc: Update plugin cache

Missed in
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3360

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3392>

oggdemux: Don't leak incoming EOS event

If we're going to drop it ... then do drop it :)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3400>

matroskademux: Handle element's duration query.

This is small regression from commit f7abd81a.

When calling `gst_element_query()` no pad is associated with that query, but the
current code always forwards the query to the associated pad, which is NULL in
previous case. This patch checks for the pad before forwarding the query.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3404>

aesdec: Fix padding removal for per-buffer-padding=FALSE

When per-buffer-padding is FALSE, the OpenSSL context needs
to be told to remove any padding at the end of the ciphertext

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1243

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3406>

adaptivedemux2: Fix parent object leak

gst_object_get_parent() method is transfer-full, thus unref is needed

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3396>

pbutils/encoding-profile: Streamline cleanup

Cleanup all local variables in one place, and ensure failure cases go through there.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

avvidenc: Don't leak AVPacket on EOF

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

avcfg: Don't leak the support GArray when registering flag/enums

* Only create it if we are attempting to put values in it
* If we sucessfully registered the values, only free the GArray
* IF we didn't, also free the backing memory

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

transcodebin: Use const arguments for clarity

And rename it when it classhes with existing local variable

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

transcodebin: Don't leak GstStream

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

transcode: Don't leak caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

transcode: Don't leak timer

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

validate-scenario: Don't leak GFile

And use specific variables for clarity in loops.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

validate: Don't leak discoverer information

The various streaminfo were being overwritten without being unreffed first

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

mpegtsbase: Don't leak GList

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

validate: Don't deinit too early

If we are going to use the validate functions ... do that *before*
deinit'ing.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

streamsynchronizer: Don't leak the syncstream object

It was leaked when breaking out early when handling GST_EVENT_STREAM_START

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

uritranscodebin: Don't leak uris, profile, sinkpad and filters

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

transcodebin: Don't leak profile

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

transcodebin: Actually free the TranscodingStream structure

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

transcoder: Don't leak API bus and GMainLoop

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

avvidenc: Don't leak filename

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

encodebasebin: Don't leak muxer pad

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

pbutils/encoding-profile: Don't leak string array

It was never freed

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

streams: Reset stream id field on finalize

If debugging is activated, the parent class finalize might attempt to read it

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3390>

meson: d3d11: Clarify the message for WGC support check

Even Windows10 SDK seems to have required symbols depending on installed
SDK version

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3376>

d3d11screencapturesrc: Fix build error with MSVC x86

Delete the code for now. Note that the code block is unnecessary already

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3375>

Updated the Pango broken link

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2723>

Updated the GObject tutorial link

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2723>

rawvideoparse: Fix support for sub-sampled tile

This element was missed as it does not use the new helpers to calculate the
plane size. The code is relatively simple though, so adding support for
subsample tiles was easy enough.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3385>

webrtc/datachannel: handle error messages from appsrc/sink

Fixes a possible race where closing a data channel may produce e.g.
not-linked errors.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3381>

sctpdec: fix stream reset (src pad removal) if no data is ever received

If we don't receive any data from usrsctp, then there will be no src pad
for the stream id and the stream reset will fail to remove the relevant
src pad.  Workaround by first attempting to add the relevant src pad, then
almost immediately removing it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3381>

test: Correct the test suite name of the h264 and h265 bitwriter.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3384>

videocrop, videobox: Simplify navigation event handling and support touch events

Signed-off-by: Colin Kinloch <colin.kinloch@collabora.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3053>

va: Fix vapostproc navigation event co-ordinate mapping

Signed-off-by: Colin Kinloch <colin.kinloch@collabora.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3053>

vaapi: Use gst_video_orientation_from_tag to parse orientation

Signed-off-by: Colin Kinloch <colin.kinloch@collabora.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3053>

videoflip: Use gst_video_orientation_from_tag to parse orientation

Signed-off-by: Colin Kinloch <colin.kinloch@collabora.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3053>

va: Use gst_video_orientation_from_tag to parse orientation

Signed-off-by: Colin Kinloch <colin.kinloch@collabora.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3053>

msdkvpp: delete the macros to make video memory work on Windows

Since gst_msdk_import_to_msdk_surface has been implemented for d3d11, we
delete the macros to make it work on Windows.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkvpp: modify the way of copying system memory to msdk surface

Replace video_copy with memcpy to fix the issue when the sizes of the
src frame and dst frame don't match. Moreover, for Windows, you have to
do the copy first and call gst_msdk_import_to_msdk_surface later.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkvpp: add d3d11 support during negotiation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkvpp: add d3d11 video memory to sink and src caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkvpp: delete use_sinkpad_va and use_srcpad_va

This two variables are no longer useful. Delete them for now.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkenc: modify the way of copying system memory to msdk surface

Replace video_copy with memcpy to fix the issue when the sizes of the
src frame and dst frame don't match. Moreover, for Windows, you have to
do the copy first and call gst_msdk_import_to_msdk_surface later.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkallocator_d3d: add d3d11 video memory and import it to msdk surface

This patch introduces d3d11 texture and index to GstMsdkMemoryID and
implements gst_msdk_import_to_msdk_surface in order to wrap the surface
to mfxMemId.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkallocator_d3d: implement GetHDL callback function

During encoding process, oneVPL runtime will call GetHDL callback function.
This patch implements this function for the use of msdk.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkallocator: add a parameter to function gst_msdk_import_to_msdk_surface

Add a flags parameter to function gst_msdk_import_to_msdk_surface and
move the function declaration to gstmsdkallcator.h to prepare for adding
Windows implementation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkenc: add d3d11 pool during negotiation

This patch adds Windows specific set_format and propose_allocation
functions to enable creating d3d11 pools during negotiation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

msdkenc: add D3D11 caps for msdkenc

This patch adds D3D11 feature to sink factory for msdkenc.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

MSDK: Introduce d3d11 device to MSDK context

Currently MSDK context does not support d3d11va. Now introduce d3d11va
device to MSDK context, making it able to create msdk session with d3d11
device and to easily share with upstream and donwstream.

Add environment variable to enable user to choose GPU device in multi-GPU
environment. This variable is only valid when there's no context
returned by upstream or downstream. Otherwise it will use the device
that created by upstream or downstream.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

MSDK: Add gst d3d11 lib into MSDK's dependency

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>

vafilter: Log out processed surfaces for debug.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3373>

vadeinterlace, vapostproc: Drop output buffer if process failed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3373>

debugutils: videocodectestsink: add support for more formats

Add support for more formats so as to run the libvpx high bit depth test suite.
This means the files under CONFIG_VP9_HIGHBITDEPTH

This also allows running the yuv444p 8bit file in the regular 8 bit vp9 suite.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3356>

gstplay: fix segmentation fault caused by subtitle info update

For subtitle stream played by subrui, gstplay will update subtitle
information before select the stream by using playbin3. Need check
the subtitle_sid value to avoid segmentation fault.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3383>

gst-rtsp-server: compile example for onvif backchannel testing

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3363>

rtspsrc: Introduce new action signal `push-backchannel-sample` with correct ownership semantics

Signals are not supposed to take ownership of their arguments but only
borrow them for the scope of the signal emission.

The old action signal `push-backchannel-buffer` is now marked deprecated.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3363>

gssink: add 'content-type' property

Useful when one wants to upload a video as `video/mp4` instead of
'video/quicktime` for example.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3371>

cccombiner: initial implementation of using CCBuffer helper

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

closedcaption: move CC buffering to helper object

Move most of the interesting code from ccconverter to this new helper
object.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

closedcaption: move cdp->cc_data into shared location

So it can be used by both ccconverter and cccombiner

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

closedcaption: move cc_data->cdp to shared file

Used by both ccconverter and cccombiner

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

ccconverter: avoid different indent versions indenting !! differently.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

closedcaption: move cdp framerate table to common file

shared by both cccombiner and ccconverter

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

cccombiner: don't assume a single cea608 data packet per buffer

e.g. 24fps can have up to 3 and would include either two field0 or
field1 cea608 data.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

ccconverter: drop data when overflow on extracting cea608 from cc_data

If the buffer overflows, then drop rather than causing a failure and
fropping the output buffer indefinitely.  This may have caused downstream to
be waiting for data the will never arrive.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

ccconverter: fix framerate passthrough with malformed input

If an input is malformed (only produces cea608 field 1 cc_data) then
when in passthrough we would effectively be dropping every second cea608
on output as we would not store any unused cea608 data.

Fix by having all code paths go through the framerate conversion code
which will store and retrieve any relevant data across buffers.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>

amfcodec: Fix typos

Fixing typo surfrace -> surface

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3378>

mfvideosrc: Return S_OK from ISampleGrabberCB callback

The Microsoft's ISampleGrabber implementation seems to be ignoring
the HRESULT return value from the callback though,
S_OK is the right HRESULT code.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3379>

gstgl: Fix handling of padded tile formats

When a tile format is padded and imported as DMABuf, the stride
contains the information about the actual width and height in
number of tiles. This information is needed by the detiling shader
in order accuratly calculate the location of pixels. To fix that,
we also copy the offset and strides into the otuput format and
the converter will ensure that the shader is recompiled whenever
the stride changes.

This fixes video corruptions observed when decoding on MT8195
with videos that aren't not aligned to 64bytes in width.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3365>