adaptivedemux2: Recheck for a pending track on drain

When a track is completely drained and EOS, but
there's a pending track on the slot loop again
to switch to that track.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2838>

adaptivedemux2: Check stream selected instead of state

When combining stream flows, ignore streams that
are not selected, instead of checking whether
the stream state has changed yet.

Fixes another issue with dashdemux2 where it fails to
change to the next period when playing content with
several video, audio and text streams, as with
Manifest_MultiPeriod_1080p.mpd when seeking to 730
just before the end of the first period.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2838>

tracers: leaks: delay type name lookup

Micro optimisation: Store the quark of the type name when tracking
objects and only do the quark to string conversion (hashtable lookup)
later when we actually need the string.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2832>

tracers: leaks: fix potentially invalid memory access when trying to detect object type

The is_gst_mini_object_check would sometimes detect a proper GObject
as a mini object, and then bad things happen.

We know whether a pointer is a proper GObject or a MiniObject here
though, so just pass that information to the right code paths and
avoid the heuristics altogether.

Eliminates all remaining uses of object_is_gst_mini_object().

Fixes #1334

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2832>

tracers: leaks: fix potentially invalid memory access when trying to detect object type

The is_gst_mini_object_check would sometimes detect a proper GObject
as a mini object, and then bad things happen.

We know whether a pointer is a proper GObject or a MiniObject here
though, so just pass that information to the right code paths and
avoid the heuristics altogether.

There are probably more cases where the check should be eliminated.

Fixes #1334, maybe

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2832>

gstalsaplugin: return the result of the element registration

Previously there were branches that would return FALSE, however
it looks like we forgot to return the new result variable.

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/900

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2695>

d3d11compositor: Add gamma-mode and primaries-mode properties

Allows controlling gamma remap and/or chromatic adaptation behavior.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2831>

d3d11videosink: Add gamma-mode and primaries-mode properties

Allows controlling gamma remap and/or chromatic adaptation behavior.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2831>

d3d11convert: Add gamma-mode and primaries-mode properties

Allows controlling gamma remap and/or chromatic adaptation behavior.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2831>

d3d11converter: Add options for gamma and primaries conversion

Gamma remap and/or primaries conversion requires additional
processing which might be something user want to avoid, performance
reason for example

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2831>

waylandsink: Logging code style updates

For better readability of debug messages and to keep similar code
in sync with `GstGtkWaylandsink`.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2694>

waylandsink: Rename occurrences of GstWaylandSink to 'self'

Rename all occurrences to `self`, making it consintent with `GstWl*`
and `GstGtkWaylandsink`.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2694>

ges: Update outdated comment

d3d11compositor is a videoaggregator subclass and no more wrapper bin
since the MR
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2631

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2825>

vapostproc: Check for colorimetry changes.

It uses what's merged in
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2765

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2777>

gtkwaylandsink test: Add navigationtest to example

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1515>

Add new gtkwaylandsink element

This is based on gtksink, but similar to waylandsink uses Wayland APIs
directly instead of rendering with Gtk/Cairo primitives.

Note that the long term plan is to move this into the existing extension
in `-good`, which requires the Wayland library to move the as well.

For this reason several files like `gstgtkutils.*` and `gtkgstbasewidget.*`
are straight copies and should be kept in sync.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1515>

smartencoder: fix detection of avc1

While avc1 is the FourCC, avc is the name used in caps

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1549>

decodebin2: don't reverse stream topology order

This can be important for instance when a container holds multiple
tracks with the same media type, with no indication (eg tags) of
which track is the default one.

In that case, players usually pick the first track by default.

This is especially useful when using smart editing with GES, as
it will result in the same ordering as the input file that was
used as a template.

For reference, this yields the same order as ffprobe.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1549>

ges: preserve discovery order

The previous code was storing container children in reverse
addition order, this was mitigated by the fact that track elements
were also stored in reverse order, thus restoring the original
order, but it seems more consistent to preserve order throughout,
the extra cost of append operations is negligible.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1549>

gstsmartencoder: don't make calculations for invalid DTS

Instead, as the current code relies on having a valid DTS (for lining
up passed through and re-encoded segments), simply compute a DTS
from the PTS if the DTS was invalid.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1549>

encoding-profile: don't order profiles by stream ID ..

when creating a profile from a discoverer info.

There is no justification for the existing code, and talking with
Thibault he cannot remember why the sort was in place.

On the other hand, this allows GES users to not have to implement
a callback for the select-tracks-for-object callback when using
it to trim a single clip, which the output profile was built from:
track elements will be placed in the appropriate track by default,
that is the one that will be connected to the matching profile.

For multi-clip timelines, the situation doesn't change, users will
still have to implement a callback and do the leg work of placing
track elements (if any) in a matching track (if any).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1549>

encoding-profile: ignore more output caps fields

chroma-format, bit-depth-chroma, bit-depth-luma are all informative
fields set by the H265 and H265 parser upon receiving an SPS.

They shouldn't be constrained downstream of the parser, instead
if a user wants those to ultimately match certain values they
should do so by constraining a profile.

In this case however, we also always remove the profile constraint
in order to let encoders pick a suitable one as a function of the
raw input video format and their own capabilities.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1549>

d3d11videosink: Translate mouse position

Converts mouse cursor position represented in display coordinates to
stream coordinates.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2824>

d3d11videosink: Early terminate mouse/keyboard event handling

... and add missing null check (plus coding style fix)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2824>

parsebin: Avoid crash with unknown streams

With the new addition of handling unknown sream types we *could* end up with a
chain which doesn't have a current_pad (it's an intermediary one)

Fixes #1287

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2822>

aggregator: fix reversed active/flushing arguments in debug log output

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2813>

webrtc: nice: WeakRef access fixes

The GstWebRTCNiceStream::ice property getter already hands-off a full reference,
so there is no need to call g_weak_ref_get() in call sites.

Fixes #1350

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2819>

webrtc: ice: Fix GstWebRTCICE parent class

It is a GstObject, not a GObject.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2819>

dtls: Properly name encoder/decoder logging categories

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2820>

dtls: Make agent and connection GstObjects

Facilitates debug logs interpretation of GST_DEBUG_OBJECT() calls.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2820>

meson: Update ogg.wrap from wrapdb

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2814>

build: Fix some compiler warnings by upgrading wraps

https://gitlab.freedesktop.org/gstreamer/orc/-/commit/03d9f144cb333057bb203cbecea3fc29bb336120

https://gitlab.gnome.org/GNOME/pygobject/-/commit/2913e72d85269d6b307dc054946a88cbba6707de

https://gitlab.gnome.org/GNOME/pygobject/-/commit/c945c99fbadb4496ff0fc01d42efd786abf5fba8

There's one more pygobject warning that will be fixed with this:

https://gitlab.gnome.org/GNOME/pygobject/-/merge_requests/210

Also bump the image versions to regen them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2814>

meson: Don't pass -Werror to vendored code

Do it the correct way with libusrsctp -- override the option so that
it's done in a compiler-agnostic and future-proof way.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2494>

rtsp+rtmp: Forward warning added to tls-validation-flags to our users

With the 2.72 release, glib-networking developers have decided that
TLS certificate validation cannot be implemented correctly by them, so
they've deprecated it.

In a nutshell: a cert can have several validation errors, but there
are no guarantees that the TLS backend will return all those errors,
and things are made even more complicated by the fact that the list of
errors might refer to certs that are added for backwards-compat and
won't actually be used by the TLS library.

Our best option is to ignore the deprecation and pass the warning onto
users so they can make an appropriate security decision regarding
this.

We can't deprecate the tls-validation-flags property because it is
very useful when connecting to RTSP cameras that will never get
updates to fix certificate errors.

Relevant upstream merge requests / issues:

https://gitlab.gnome.org/GNOME/glib/-/merge_requests/2214

https://gitlab.gnome.org/GNOME/glib-networking/-/issues/179

https://gitlab.gnome.org/GNOME/glib-networking/-/merge_requests/193

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2494>

dtls: Disable OpenSSL 3.0 deprecation warnings for now

Fedora 36 ships with OpenSSL 3.0, which deprecates all low-level APIs,
so this code needs to be rewritten. There is no easy fix in the
porting guide, and it recommends disabling the warnings if you can't
use the high-level API.

https://wiki.openssl.org/index.php/OpenSSL_3.0#Upgrading_to_OpenSSL_3.0_from_OpenSSL_1.1.1

Here's the replacement API:

https://www.openssl.org/docs/man3.0/man7/migration_guide.html#Deprecated-low-level-object-creation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2494>

harfbuzz.wrap: Fix wrap type, and remove unnecessary patch

Forgot to change the wrap type in e0014ef4fe which broke the
subproject. Wasn't noticed by CI because the subproject cache wasn't
regenerated.

The accompanied patch was included in 2.8.2, so it is not needed. It
was originally needed with 2.8.1

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2812>

windows/Dockerfile: replace ADD arguments with Invoke-WebRequest

Have the windows susbsystem handle the networking layer and avoid
layer invalidation until the strings/urls are changed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2812>

ci: Update the windows image

Mainly to get newer meson

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2812>

subprojects: fix pango wrap diff file location

Must match the version of the pango wrap.

Fixes ci windows docker image rebuild issue.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2812>

videobox: avoid dropping caps fields for passthrough caps transform

Fixes potential negotiation failure in case downstream element
is a bit picky regarding the fields in question.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2786>

compositor: Update plugins cache

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1486>

compositor: Warn when inputs are SDR/HDR mixed

Let user know that the result of mixed SDR/HDR is not guaranteed
to be a good visual quality.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1486>

compositor: Add support for all formats

For formats which we don't have fast-path implementation, compositor
will convert it to common unpack formats (AYUV, ARGB, AYUV64 and ARGB64)
then blending will happen using the intermediate formats.
Finally blended image will be converted back to the selected output format
if required.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1486>

compositor: Add support for Y444 high bitdepth formats

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1486>

compositor: Add support for I420/I422 high bitdepth formats

Implementation for {I420,I422}_{10,12}_{LE,BE} formats

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1486>

compositor: Calculate background color only once

... instead of do that per fill_color() call in case of RGB format.
Moreover, respect selected GstVideoColorRange

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1486>

wasapi2: Fix initial mute/volume setting

Fix up volume/mute change flag setting

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2809>

rtspconnection: protect cancellable by a mutex

It is entirely possible for the cancellable to be cancelled (and freed)
in gst_rtsp_connection_flush() while there may be an ongoing read/write
operation.

Nothing prevents gst_rtsp_connection_flush() from waiting for the
outstanding read/writes.

This could lead to a crash like (where cancellable has been freed
within gst_rtsp_connection_flush()):

 #0  0x00007ffff4351096 in g_output_stream_writev (stream=stream@entry=0x7fff30002950, vectors=vectors@entry=0x7ffe2c6afa80, n_vectors=n_vectors@entry=3, bytes_written=bytes_written@entry=0x7ffe2c6af950,  cancellable=cancellable@entry=0x7fff300288a0, error=error@entry=0x7ffe2c6af958) at ../subprojects/glib/gio/goutputstream.c:377
 #1  0x00007ffff44b2c38 in writev_bytes (stream=0x7fff30002950, vectors=vectors@entry=0x7ffe2c6afa80, n_vectors=n_vectors@entry=3, bytes_written=bytes_written@entry=0x7ffe2c6afb90, block=block@entry=1, cancellable=0x7fff300288a0) at ../subprojects/gst-plugins-base/gst-libs/gst/rtsp/gstrtspconnection.c:1320
 #2  0x00007ffff44b583e in gst_rtsp_connection_send_messages_usec (conn=0x7fff30001370, messages=messages@entry=0x7ffe2c6afcc0, n_messages=n_messages@entry=1, timeout=timeout@entry=3000000) at ../subprojects/gst-plugins-base/gst-libs/gst/rtsp/gstrtspconnection.c:2056
 #3  0x00007ffff44d2669 in gst_rtsp_client_sink_connection_send_messages (sink=0x7fffac0192c0, timeout=3000000, n_messages=1, messages=0x7ffe2c6afcc0, conninfo=0x7fffac019610) at ../subprojects/gst-rtsp-server/gst/rtsp-sink/gstrtspclientsink.c:1929
 #4  gst_rtsp_client_sink_try_send (sink=sink@entry=0x7fffac0192c0, conninfo=conninfo@entry=0x7fffac019610, requests=requests@entry=0x7ffe2c6afcc0, n_requests=n_requests@entry=1, response=response@entry=0x0, code=code@entry=0x0) at ../subprojects/gst-rtsp-server/gst/rtsp-sink/gstrtspclientsink.c:2845
 #5  0x00007ffff44d3077 in do_send_data (buffer=0x7fff38075c60, channel=<optimized out>, context=0x7fffac042640) at ../subprojects/gst-rtsp-server/gst/rtsp-sink/gstrtspclientsink.c:3896
 #6  0x00007ffff4281cc6 in gst_rtsp_stream_transport_send_rtp (trans=trans@entry=0x7fff20061f80, buffer=<optimized out>) at ../subprojects/gst-rtsp-server/gst/rtsp-server/rtsp-stream-transport.c:632
 #7  0x00007ffff4278e9b in push_data (stream=0x7fff40019bf0, is_rtp=<optimized out>, buffer_list=0x0, buffer=<optimized out>, trans=0x7fff20061f80) at ../subprojects/gst-rtsp-server/gst/rtsp-server/rtsp-stream.c:2586
 #8  check_transport_backlog (stream=0x7fff40019bf0, trans=0x7fff20061f80) at ../subprojects/gst-rtsp-server/gst/rtsp-server/rtsp-stream.c:2645
 #9  0x00007ffff42793b3 in send_tcp_message (idx=<optimized out>, stream=0x7fff40019bf0) at ../subprojects/gst-rtsp-server/gst/rtsp-server/rtsp-stream.c:2741
 #10 send_func (stream=0x7fff40019bf0) at ../subprojects/gst-rtsp-server/gst/rtsp-server/rtsp-stream.c:2776
 #11 0x00007ffff7d59fad in g_thread_proxy (data=0x7fffbc062920) at ../subprojects/glib/glib/gthread.c:827
 #12 0x00007ffff7a8ce2d in start_thread () from /lib64/libc.so.6
 #13 0x00007ffff7b12620 in clone3 () from /lib64/libc.so.6

Fix by adding a cancellable lock and returning an extra reference used
across all read/write operations.  gst_rtsp_connection_flush() can free
the in-use cancellable and it will no longer affect any in progress
read/write.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2799>

GST_CHECK_VERSION: Fix unexpected "git = next version" assumption

1.21.0.1 should not satisfy a check for 1.22.0.

If someone needs more control they should do a feature check for
the symbol in the headers or lib.

Based on a similar patch by Tim-Philipp Müller for libnice.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2501>

videoflip: Fix caps negotiation when method is selected

The caps negotiation should respect the selected method to the test pipeline below works properly.
gst-launch-1.0 videotestsrc ! video/x-raw,width=320,height=600 ! videoflip method=clockwise ! video/x-raw,width=600,height=320 ! fakesink

Signed-off-by: Adrian Fiergolski <adrian.fiergolski@fastree3d.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2803>

vaapi: h265dec: avoid possible NULL deref

Check "pi" before attempting to dereference it.

Captured by static analysis.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2801>

va: h264enc: Make AUD the first NAL unit

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2796>

meson: webrtc: ensure definition of libgstwebrtcnice_dep

... and skip if it's disabled.

Fixes #1344

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2797>

vaapi: h264dec: Remove dead code

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2795>

vaapi: h265dec: Refresh our sps/pps copy

The decoder uses a copy of the SPS/PPS. Since
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2575
the SPS/PPS is updated when the slice header is parsed. As a side
effect, the H265 decoder was using an partially filled copy.

Fixes: 5ae05bb42a ("codecparsers: Defering computation of pps and sps fields")
Fixes: #1343
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2795>

v4l2object.c: add support for Apple's full-range bt709 colorspace variant

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2793>

webrtc: Split WebRTCICE into base classes and implementation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2398>

v4l2: fix the critical log when unreference the buffer with no data

In the trick mode, driver may queue a valid buffer follow by an
empty buffer which has no valid data to indicate EOS.For the empty
buffer whose memory is multi-plane, need to resize it before
unreference it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2731>

qsvencoder: Add support for D3D11 shared texture

Although input texture belongs to the other d3d11 device,
we can do GPU copy when its physical device is identical.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2790>

d3d11: Port to GstD3D11Fence

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2790>

d3d11: Implement fence abstraction

Depending on device feature level, d3d11 runtime can support
ID3D11Fence which is equivalent to ID3D12Fence.
Waiting using fence has performance-wise benefit over pulling
ID3D11Query status. If ID3D11Fence is not supported by device,
then ID3D11Query will be used instead.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2790>

audiointerleave: fix property docs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2785>

video: Fix scaling in 4x horizontal co-sited chroma

4x downscaling of chroma with co-sited chroma has never worked
it seems.

Fixes incorrect videotestsrc output and videoconvert conversions
to Y41B, YUV9, YVU9 and IYU9 with co-sited chroma.

e.g.
gst-launch-1.0 videotestsrc ! video/x-raw,format=Y41B,width=1280,height=720 ! \
  videoconvert ! autovideosink

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2789>

codecs: h265: Make sure that sps is processed just before decoding

It may happens that bitstream doesn't provided SPS in decoding order
(like in VPSSPSPPS_A_MainConcept_1 conformance test file).
To be sure that the decoder got the correct SPS parameters process
SPS just before start decoding the frame.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2575>

codecs: h265: Fix typos

Fix multiple typos in warning and info messages.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2575>

videoparsers: h265: Be less restrictive between PPS and SPS parsing order

Allow to parse PPS to be parser even if SPS hasn't been parsered before.
This help to parse VPSSPSPPS_A_MainConcept_1 conformance test

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2575>

codecparsers: Defering computation of pps and sps fields

While possible defer computataion of pps and sps fields until
slice parsing since it may happens that bitstreams don't encoded
them in expected order.
A example weird ordered bitstreams is VPSSPSPPS_A_MainConcept_1
conformance test.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2575>

v4l2codecs: h264: Reset the number of slices on picture start

This way, if we aborted a frame in the middle, we don't endup thinking
there is multiple slices to send.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1079>

v4l2codecs: h264: Fix initialization of the slice array

The function g_array_sized_new() leaves the len to 0, but the slice
implementation assumes it would be set to 4. Sending multiple slices is
not yet support for H.264 as no driver needed it yet, but if that code
was to be used it would have overflowed as the array would never grow as
multiple 0 by 2 always results in 0.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1079>

v4l2codecs: h264: Rename count variable to num_controls

This is the counter for the used control slot. The term num_controls was
introduced in the HEVC decoder and felt like a better choice.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1079>

v4l2codecs: hevc: Add an API check

This adds an API check and bump recommended base kernel version to 5.20.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1079>

v4l2codecs: Add v4l2 stateless H265 decoder

Add v4l2 stateless H265 decoder.

Co-authored-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Co-authored-by: Sebastian Fricke <sebastian.fricke@collabora.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1079>

v4l2codecs: Increase minimum payload size

Fluster MAXBINS_{A-C}_TI_5 HEVC tests files require to store
a payload larger than usual given their resolutions.
Increase the minimum payload size to handle them correctly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1079>

v4l2codecs: Align v4l2 headers with HEVC final uAPI

HEVC uAPI have landed for kernel 5.10.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1079>

codecparsers: h265: Introduce long_term_ref_pic_set_size

This is the size of the `if( long_term_ref_pics_present_flag )` inside
slice_segment_header( ) syntax. This size in bit, and after de-emulation
is needed by some HW.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1079>

audiobuffersplit: Actually store number of samples to drop in gapless mode

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2780>

audiobuffersplit: Use input running time for comparison instead of the currently tracked running time

Otherwise gapless mode would do completely wrong calculations on
discontinuities and cause input/output to drift slowly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2780>

audiobuffersplit: Combine two if expressions to reduce indentation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2780>

videoaggregator: Don't force upstream alpha format

"video/x-raw,format=RGBA ! some-video-filter ! video/x-raw,format=NV12"
is a very common case and therefore such erroring at baseclass
does not seem to be a desired behavior.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2703>

d3d11compositor: Add support for crop meta

GstD3D11Converter supports cropping already. Cropping is just
a matter of setting source rectangle area to converter,
from d3d11 point of view

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2782>

examples: d3d11: Add GstD3D11Converter example

An example to demonstrate color space conversion, rescale and flip/rotation
using GstD3D11Converter

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2767>

d3d11: Move GstD3D11Converter to gst-libs

The object is equivalent to GstVideoConverter but for D3D11.
Application can use this object directly for various conversions,
such as color space conversion, rescale, crop and flip/rotation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2767>

d3d11: Define enum and flags types manually

gnome.mkenums_simple() doesn't work well for GstD3D11, seems to
be confused by numeric representation of D3D11, must be a bug
in GLib or so. Just don't rely on the incomplete automagic.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2767>

d3d11: Run gst-indent

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2767>

openmpt: update from now deprecated api

https://lib.openmpt.org/doc/classopenmpt_1_1module.html#ab2695af0baa274054f5687741fa7c05b

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2721>

msdkenc: Enable LowDelayBrc and MaxFrameSize for I/P frame

Enable these features for accurate bitrate control.
Feature introduction of LowDelayBRC, MaxFrameSizeI and MaxFrameSizeP could be found here:
https://github.com/Intel-Media-SDK/MediaSDK/blob/master/doc/mediasdk-man.md

Signed-off-by: Fan F He <fan.f.he@intel.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2742>

d3d11converter: Use GstStructure to set initialization options

Runtime updatable options are exposed via object properties
but there are a few options which should be known at initialization
time. Use GstStructure to signal it and to be easily extensible

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2776>

d3d11converter: Use primaries and transfer function compare methods

Different enum values could be functionally equal, then we don't need
to do full conversion.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2776>

d3d11format: Move DXGI color space mapping method to gst-libs

Expose color space mapping method to public so that application
can configure GstCaps or GstVideoInfo using DXGI color space

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2776>

gststructure: Add gst_structure_get_flags method

We don't prevent setting G_TYPE_FLAGS on GstStructure
but no helper method for getting the value.
Add a method similar to gst_structure_get_enum()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2770>

aggregator: Reset EOS flag after receiving a stream-start event

And also don't assert that there are no buffers queued up when handling
an EOS event. The pad's streaming thread might've already received a new
stream-start event and queued up a buffer in the meantime.

This still leaves a race condition where the srcpad task sees all pads
in EOS state and finishes the stream, while shortly afterwards a pad
might receive a stream-start event again, but this doesn't seem to be
solveable with the current aggregator design.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2769>

d3d11convert: Use new colorimetry compare function

Avoid conversion if both are functionally identical

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2765>

video-converter: Use primaries compare function

Avoid conversion if both color primaries are functionally equal

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2765>

video-color: Add primaries and colorimetry compare functions

SMPTE 170M and 240M use the same RGB and white point coordinates
and therefore both primaries can be considered functionally
equivalent.
Also, some transfer functions have different name but equal
gamma functions. Adding another colorimetry compare function
to deal with thoes cases at once

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2765>

qt: Fix another instance of Qt/GStreamer both defining `GLsync` differently

In file included from ../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/gstreamer-1.0/gst/gl/gstglfuncs.h:87,
                 from ../gst-plugins-good-1.20.3/ext/qt/qtglrenderer.cc:14:
../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/gstreamer-1.0/gst/gl/glprototypes/gstgl_compat.h:40:18: error: conflicting declaration 'typedef void* GLsync'
   40 | typedef gpointer GLsync;
      |                  ^~~~~~
In file included from ../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/QtGui/qopengl.h:127,
                 from ../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/QtQuick/qsggeometry.h:44,
                 from ../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/QtQuick/qsgnode.h:43,
                 from ../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/QtQuick/qsgrendererinterface.h:43,
                 from ../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/QtQuick/qquickwindow.h:44,
                 from ../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/QtQuick/QQuickWindow:1,
                 from ../gst-plugins-good-1.20.3/ext/qt/qtglrenderer.cc:6:
../gstreamer1.0-plugins-good/1.20.3-r0/recipe-sysroot/usr/include/QtGui/qopengles2ext.h:24:26: note: previous declaration as 'typedef struct __GLsync* GLsync'
   24 | typedef struct __GLsync *GLsync;
      |                          ^~~~~~

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2763>

meson: Improve certifi documentation on macOS

First, just installing certifi doesn't install the ca-cert in the
right location. The `Install Certificates.command` script also
symlinks the openssl cert.pem to the certifi ca cert file

Second, we can make it more likely that users will notice this if we
make it a warning. If we ever get a bug report about this despite
these measures, we can try to make this an error.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2707>

d3d11testsrc: Fix SMPTE pattern rendering

It was wrong vertex shader code

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2766>

ci: Fix project URL when triggering cerbero pipelines

When merge request pipelines are triggered on the gstreamer namespace,
CI_PROJECT_URL will be gitlab.[...]/gstreamer/gstreamer but we need to
use gitlab.[...]/$USER/gstreamer because that's where the source
branch is located.

This exhibits as cerbero pipelines failing because it can't find the
specified branch:

https://gitlab.freedesktop.org/gstreamer/cerbero/-/pipelines/639379

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2761>

osxaudio: Fix deprecation in macOS 12.0

kAudioObjectPropertyElementMaster has been renamed to
kAudioObjectPropertyElementMain

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2761>

meson: Fix compile failures on macOS in harfbuzz and pango

These patches are taken from upstream, and they fix compile failures
with latest clang. These can be dropped when upgrading these wraps.

This is currently causing a warning because we do not require the
version of meson that ships with this feature: 0.63.0. The version has
not been bumped because older Meson versions gracefully ignore the
wrap field, this fix is optional and only needed on macOS, and 0.63.0
is a very new release with a bug that partially breaks this feature:

https://github.com/mesonbuild/meson/pull/10602

We can consider bumping the requirement once 0.63.1 is released.

Also switch from git to tarballs, no reason to use git here anymore.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2761>