tests: Cast drop-messages-interval type properly

The rtpjitterbuffer test drop_messages_interval uses a GstClockTime for
the message drop interval.  This property is defined as a guint.  On
systems with 64-bit time_t but 32-bit uint, this can cause the
g_object_set function to fail to read the arguments properly.

Fixes: #1656
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3580>

d3d11videosink: Fix deadlock when parent window is busy

Deadlock sequence:
* From a streaming thread, d3d11videosink sends synchronous message
  to the parent window, so that internal (child) window can be
  constructed on the parent window's thread
* App thread (parent window thread) is waiting for pipeline's
  state change (to GST_STATE_NULL) but streaming thread is
  blocked and waiting for app thread

To avoid the deadlock, GstD3D11WindowWin32 should send message
to the parent window asynchronously.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3570>

base:navigation: Cleanup navigation key modifiers enum

We were exposing the 'ALT' modifier as if we were guaranteeing its
accuracy but truth is we were only exposing configuration dependent
values.

Make the API simpler for now, the same way as Gtk3 was exposing it, and
when we have time to guarantee more values by making them take backends
configuration into account, we will expose those values in a accurate
way.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1402

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3565>

glupload: add CAN_ACCEPT_RAW to all dma-buf uploaders

Fixes cases where a dma-buf would be uploaded using direct-dma-buf
into an external-oes texture (using video/x-raw caps) and then
attempting to reconfigure the same source.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3486>

gst: handle combinations in gst_stream_type_get_name()

This should handle the majority of the valid stream cases.
The element setting the stream type may set each type separately.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2804>

glvideomixer: override sink pad template

Allow us to pass the GType of its pad and so
improve the documentation when inspecting glvideomixer.

Fix #1253

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2527>

qtdemux: Clear protection events when we get new ones

If we keep the old events they can be end up being passed to the app, that could
discard the protection information because it has been seen before.

Drive by improvement: use g_queue_clear_full instead of foreach+clear for
protection events.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3547>

webrtc: implement support for msid values

Local msid values are taken from sink pad property, or fallback to the
previously used cname.

The remote msid values are exposed on the relevant src pads.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3106>

macos: Add wrapper API to run a NSApplication in the main thread

On macOS, a Cocoa event loop is needed in the main thread to ensure
things like opening a GL window work correctly. In the past, this was
patched into glib via Cerbero, but that prevented us from updating it.
This workaround simply runs an NSApplication and then calls the
main function on a secondary thread, allowing GStreamer to correctly
display windows and/or system permission prompts, for example.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3532>

systemclock: Use `futex_time64` syscall if available (32-bit systems) and use correct `struct timespec` definition

See also https://gitlab.gnome.org/GNOME/glib/-/issues/2634

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1648

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3561>

gst_init: Removed wrong warning in docstring

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3562>

h264parser: Update doc for GST_H264_FRMAE_PACKING_TOP_BOTTOM

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1318>

h264parser: typo fix in enum value

Old value is marked deprecated, new enum with same entry is added. Should be binary compatible.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1318>

video-format: Add macro checking for validity of GstVideoFormatInfo

Mostly to maintain consistency with the GST_AUDIO_FORMAT_INFO_IS_VALID_RAW
macro.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2800>

audio-format: Add macro checking for validity of GstAudioFormatInfo

`gst_audio_format_info_fill_silence()` not properly checking the validity of its
input may lead it into an infinite loop.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2800>

webrtc/signalling: Give a helpful error when starting a double-session

If the peer is already in a session and tries to start a new one, give
them a helpful error.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2460>

gst-examples: webrtc: signalling: simple-server Fix condition when calling a busy peer

When a session request is coming in, ERROR occurs when the callee is busy.
But peer_status is the status of the caller, which is of course None when
calling someone, while self.peers[callee_id][2] is that of the callee.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2460>

jpegdec: Enable packetized if sink caps contains parsed as true.

jpegdec is capable to parse input frames, but if jpegparse is before,
there's no need to reparse frames. This patch configure jpegdec as
packetized, skipping parsing, if negotiated sink caps has the boolean
field 'parsed' as true.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2464>

mpegts: Handle when iconv doesn't support ISO 6937

Systems like musl libc don't support ISO 6937 in iconv.  This ensures
that the MPEG-TS plugin can cope with that.  There is existing support
in the plugin for other methods, so it seems to have been the original
intent anyway.

Fixes: #1314
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3245>

pulsesink: Fix occasional period of silence on resume

According to comment in gst_pulsering_stream_latency_cb, latency updates
happen every 100 ms. The code in gst_pulsering_stream_latency_cb does
not care about the actual state of the ringbuffer, pbuf->acquired or
not.
Thus, every 100 ms the ringbuf->segdone may be set, even though the
object itself might be in 'destroyed' state. On next
gst_pulseringbuffer_acquire the segdone is not touched, so playback may
resume with invalid/wrong segdone value. This finally leads to a period
of silence playing after resuming the pipeline.

The problem was found on 'not-yet-released'-hardware and so far was not
reproducible on desktop computer.

Removing the callback as long as the ringbuffer is not in 'acquired'
state solves the problem reliably on the hardware device that the issue
was detected on.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3082>

timeoverlay: fix pad leak

Spotted by Jiri Uncovsky.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1336>

vaallocator: Fix gi annotations.

Remove spurious types and skip functions that returns VASurfaceID
which isn't exposed to introspection.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3557>

zxing: update to 1.4.0 tag

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3450>

vaapi: prefix USE_FOO defines to fix build with mesa 22.3.0

Apparently mesa 22.3.0 has updated the egl headers, and eglplatform.h now
contains commit
https://github.com/KhronosGroup/EGL-Registry/pull/130/commits/3670d645f4a26a0a9e87e7f3a8608e7cc1d53b5b
after which xlib headers don't get included by default anymore but are
dependent upon whether USE_X11 was defined.

This breaks headless builds of gstreamer-vaapi because we always define
an internal define USE_X11 as either 1 or 0.

Change these defines to GST_VAAPI_USE_XYZ instead to avoid this.

Fixes #1634

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3555>

docs: specify possibility of a NULL return

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3546>

ci: pin hotdoc to 0.13.7 for now, 0.14 has a regression

```
ERROR: [links]: (mandatory-link-not-found): Mandatory link Link GstGLSinkBin -> None (GstGLSinkBin) could not be resolved
ERROR: [links]: (mandatory-link-not-found): Mandatory link Link GstRTPMux -> None (GstRTPMux) could not be resolved
ERROR: [links]: (mandatory-link-not-found): Mandatory link Link GstSRTSink -> None (GstSRTSink) could not be resolved
ERROR: [links]: (mandatory-link-not-found): Mandatory link Link GstSRTSrc -> None (GstSRTSrc) could not be resolved
```

Same change was already made in the 1.20 branch:

https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1582#note_1669723

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3527>

ci: Bump image tags to rebuild new fedora / windows images

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3527>

meson: Update some more wraps

At least the libxml2 update fixes a failure on macOS, the rest are
happening for free.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3527>

mpegts: Check continuity counter on section streams

This wasn't really done, and is needed in order to detect potential section
changes for sections that have got identical information (such as when switching
between streams that have the same PAT/PMT pid and subtable information).

Other checks exist in tsbase to detect if the "new" PAT/PMT really is an update or not.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3530>

rtpvp9depay: expose keyframe-related properties

This simply brings in the wait-for-keyframe and request-keyframe
properties from rtpvp8depay.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/909>

multiqueue: update for renamed log id macros

See #1635.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3553>

info: rename new log macros from GST_*_OBJECT_ID -> GST_*_ID

Fixes #1635

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3553>

gstplugin: Handle static plugins in gst_plugin_load_by_name()

gst_plugin_load_by_name() assumed a plugin has a filename,
which isn't true for static plugins, leading to criticals.

If a plugin is already loaded, just return the loaded plugin,
which makes it work for static plugins as well as saving a
moment for already-loaded dynamic plugins.

Add locking in gst_plugin_is_loaded(), as a plugin may be
still being loaded in another thread.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3552>

examples: webrtc: fix unidirectional pipeline

'autoaudiosrc' does not have a 'is-live' property.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3550>

cuda: Fix runtime compiler library loading on Windows

The cuda is a part of GPU driver but runtime compiler is a part of
cuda toolkit, which means the version number can be different.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3545>

cuda: Hide memory copy util function

The method was intended to be used by only cudaupload/download elements
and not ready to be a part of public API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3545>

ksvideosrc: Warn for deprecated plugin use

This plugin should not be used any more

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3544>

winscreencap: Warn for deprecated plugin use

This plugin should not be used any more

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3544>

proxysink: Post EOS message on EOS event

proxysink is actual sink (GST_ELEMENT_FLAG_SINK flag has configured)
so it should post EOS message.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3472>

qtdemux: exit when protection caps are not defined during PIFF parsing

Reproduction testcase (uses PlayReady):
https://developers.canal-plus.com/rx-player/upc/?appTileLocation=[object%20Object]

In test streams we are using PIFF box, but caps did not had
present GST_PROTECTION_SYSTEM_ID_CAPS_FIELD. In consequence, invalid
system_id was returned which caused SIGSEGV crash.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3535>

h264parse: Avoid setting wrong colorimetry info if the stream doesn't contain it

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3543>

textrender: Negotiate caps on a GAP event if none were negotiated yet

Otherwise downstream wouldn't have received a segment event either and
wouldn't know what to do with the gap event

And also forward any pending segment event.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3541>

h265decoder: Do not abort when failed to prepare ref pic set

Currently the element calls abort when failed to prepare reference
picture set. This can happent when the input stream is somehow
corrupted, like a rtsp strem with lost packets. Now it will only
return with GST_FLOW_ERROR instead of terminating whole process.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3505>

meson: fix check for pthread_setname_np()

Need to define _GNU_SOURCE.

Fixes #1542

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3538>

dvbsubenc: Forward GAP events as-is if we wouldn't produce an end packet and are not in the middle of an existing subtitle

An end packet is only produced once for the last subtitle, so multiple
GAP events between subtitles would result only in a single end packet
and nothing else otherwise. This would potentially starve downstream
then, so instead forward the GAP events in that case.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3534>

ci: build less for integration testsuite

- skip gst-omx
- skip sharp bindings
- skip examples and gst-examples
- skip check unit tests (only needed for check job)
- skip microdns (not needed and pulls in subproject)
- skip avtp (ditto)
- skip webrtc (same)
- skip benchmarks
- skip qt5 and gtk

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3520>

examples: webrtc: fix plugins check

`videoconvert` and `videoscale` are now part of the `videoconvertscale`
plugin, see d11f13f476411b828387c3c26619bc77c255affb

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3529>

directshow: Fix build error with glib 2.75 and newer

Starting with glib 2.75, `NULL` is `nullptr`, which cannot be
implicitly coerced to `0`, unlike `NULL`. So explicitly pass `0`.

```
[3206/4524] Compiling C++ object subprojects/gst-plugins-bad/sys/directshow/gstdirectshow.dll.p/dshowvideosink.cpp.obj
FAILED: subprojects/gst-plugins-bad/sys/directshow/gstdirectshow.dll.p/dshowvideosink.cpp.obj
"cl" "-Isubprojects\gst-plugins-bad\sys\directshow\gstdirectshow.dll.p" "-Isubprojects\gst-plugins-bad\sys\directshow" "-I..\subprojects\gst-plugins-bad\sys\directshow" "-Isubprojects\gst-plugins-bad" "-I..\subprojects\gst-plugins-bad" "-Isubprojects\gst-plugins-base\gst-libs" "-I..\subprojects\gst-plugins-base\gst-libs" "-Isubprojects\gstreamer\libs" "-I..\subprojects\gstreamer\libs" "-Isubprojects\gstreamer" "-I..\subprojects\gstreamer" "-Isubprojects\orc" "-I..\subprojects\orc" "-I..\subprojects\gst-plugins-bad\sys\directshow\strmbase\baseclasses" "-Isubprojects\gst-plugins-base\gst-libs\gst\video" "-Isubprojects\gstreamer\gst" "-Isubprojects\gst-plugins-base\gst-libs\gst\audio" "-Isubprojects\gst-plugins-base\gst-libs\gst\tag" "-IC:/gst-install/include/glib-2.0" "-IC:/gst-install/lib/glib-2.0/include" "-IC:/gst-install/include" "/MD" "/nologo" "/showIncludes" "/utf-8" "/W2" "/EHsc" "/O2" "/Zi" "/wd4018" "/wd4146" "/wd4244" "/wd4305" "/utf-8" "/we4002" "/we4003" "/we4013" "/we4020" "/we4027" "/we4029" "/we4033" "/we4045" "/we4047" "/we4053" "/we4062" "/we4098" "/we4101" "/we4189" "/utf-8" "-D_MBCS" "/wd4189" "/wd4456" "/wd4701" "/wd4703" "/wd4706" "/wd4996" "-DHAVE_CONFIG_H" "/Fdsubprojects\gst-plugins-bad\sys\directshow\gstdirectshow.dll.p\dshowvideosink.cpp.pdb" /Fosubprojects/gst-plugins-bad/sys/directshow/gstdirectshow.dll.p/dshowvideosink.cpp.obj "/c" ../subprojects/gst-plugins-bad/sys/directshow/dshowvideosink.cpp
../subprojects/gst-plugins-bad/sys/directshow/dshowvideosink.cpp(62): warning C5051: attribute 'noinline' requires at least '/std:c++20'; ignored
../subprojects/gst-plugins-bad/sys/directshow/dshowvideosink.cpp(123): error C2664: 'LRESULT SendMessageA(HWND,UINT,WPARAM,LPARAM)': cannot convert argument 3 from 'nullptr' to 'WPARAM'
../subprojects/gst-plugins-bad/sys/directshow/dshowvideosink.cpp(123): note: A native nullptr can only be converted to bool or, using reinterpret_cast, to an integral type
C:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um\winuser.h(3690): note: see declaration of 'SendMessageA'
../subprojects/gst-plugins-bad/sys/directshow/dshowvideosink.cpp(635): error C2664: 'BOOL SystemParametersInfoA(UINT,UINT,PVOID,UINT)': cannot convert argument 2 from 'nullptr' to 'UINT'
../subprojects/gst-plugins-bad/sys/directshow/dshowvideosink.cpp(635): note: A native nullptr can only be converted to bool or, using reinterpret_cast, to an integral type
C:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um\winuser.h(13153): note: see declaration of 'SystemParametersInfoA'
../subprojects/gst-plugins-bad/sys/directshow/dshowvideosink.cpp(1593): error C2664: 'LRESULT SendMessageA(HWND,UINT,WPARAM,LPARAM)': cannot convert argument 3 from 'nullptr' to 'WPARAM'
../subprojects/gst-plugins-bad/sys/directshow/dshowvideosink.cpp(1593): note: A native nullptr can only be converted to bool or, using reinterpret_cast, to an integral type
C:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um\winuser.h(3690): note: see declaration of 'SendMessageA'
```

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3528>

gstreamer-full: use the basename of plugin_path to avoid the ':' detection

The absolute path on windows contains ':' which prevents
gstinitstaticplugins.py to work properly. Use the basename whic is good
enough for the script to make the list of plugins

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3526>

adaptivedemux2: Don't allow stream selection while switching periods

The stream selection is done on the currently outputting tracks, but in order to
(de)activate the backing streams we can only do it if the input and output
period are identical.

Fixes crash when doing stream selection during period migration

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3525>

gstinfo: Minor modification to avoid gst-indent pain

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3483>

multiqueue: Use new id-based debugging methods

Clarifies the debug logs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3483>

gstinfo: Allow passing a string identifier for debugging

This adds "id" variants to most debugging functions, and allows providing a
string identifier instead of a GObject.

This allows providing unified and clearer debug logs for all the
non-gobject-based items, and opens the way for more unified logging.

As an extension, copying the object name is avoided as much as possible, by
using it directly instead of going through another copy.

* API : gst_debug_message_get_object_id

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3483>

Back to development

Release 1.21.3

Remove ChangeLog files from git repository

This information is tracked fully in the git repository, so
no point having the ChangeLog duplicate it, and it interferes
with grepping the repository.

We are going to create the ChangeLogs on the fly when generating
tarballs going forward (with a limited history), since it's still
valuable for tarball consumers to be able to easily see a list of
recent changes.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer-project/-/issues/73

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3521>

meson: Generate ChangeLog files for release tarballs on dist

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3521>

flacparse: Fix handling of headers advertising 32bps

According to the flac bitstream format specification, the sample size in bits
corresponding to `111` is 32 bits per sample.

https://xiph.org/flac/format.html#frame_header

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3517>

v4l2src: Fix crash in renegotiation

This regression was introduce by fix for making buffer pool thread safe. When
we renegotiate, the pool will be setup after we set the format. But the code
has been simplified to only get the pool once before, which caused a null
pointer deref.

Fixes 94ba019 ("v4l2: Fix SIGSEGV on 'change state' during 'format change'")
Related to !3481
Fixes #1626

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3513>

Change name of property from Exists to Availability

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3510>

ges: Add API to disable timeline coherence checks

There are cases where user might want to be in full control of the
timeline and not be limited by the checks that are being done by GES
to go from one timeline layout to another, this should be doable as
it is a valid use case.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3501>

d3d11av1dec: Promote rank to primary + 1

... so that this element can have higher rank than the other software
AV1 decoders

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3495>

git: Show all suggestions on pre-commit hook

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3506>

uridecodebin3: Protect against NULL uri/suburi

Fixes #1625

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3511>

README: Add section how to build and visualize documentation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3442>

meson: Add toplevel "gst-doc" alias target

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3442>

meson: Add doc directory to XDG_DATA_DIRS

gst-env.py does it already, but it was missing from meson devenv.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3442>

multiqueue: Handle gapless with flushing

Don't reset the stream-start group-id when stop/pausing single queues. They are
only resetted when re-used (in READY->PAUSED).

Fixes #1586

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3487>

uridecodebin3: Always store pending buffering messages

Fixes #1586

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3487>

vajpegdec: Reenable element negotiation.

negotiation vmethod were overwritten by vabasedec, leading to errors.

This was a regression of commit b6538e0560.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3509>

gstinfo: Optimize color escape code creation

When coloring is in use, those escape codes are going to be created many times
for almost all debug lines.

Don't create plenty of temporary allocations, and instead build the escape code
ourselves statically

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3498>

webrtc: Fix possible use-after-free of GstWebRTCICETransport

Because of the asynchronous resolving of mDNS ICE candidates it is
possible that GstWebRTCICE outlives webrtcbin. This in turn prolongs
the lifetime of the GstWebRTCNiceStream objects via refs in
nice_stream_map. Thus the GstWebRTCICETransport objects held in
GstWebRTCNiceStream may be invalid at the time they are accessed by
the _on_candidate_gathering_done() callback since GstWebRTCNiceStream
doesn't take a reference to them. Doing so would create a circular
reference, so instead this commit introduces weak references to the
transport objects and then we can check if the objects are valid before
accessing them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3502>

rtspsrc: fix seek event leaks

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3500>

va: Handle input caps change.

Update output caps if it's notified by baseclass

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3328

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

vaav1dec: Use gst_va_base_dec_set_output_state().

And even that vaav1dec doesn't use vabasedec negotiate vmethod, it should align
with the new scheme of using base's width & height for surface size and
output_info structure for downstream display size negotiation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

vavp9dec: Use gst_va_base_dec_set_output_state().

As this element reopen the internal decoder differently, it only uses the helper
function to negotiate.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

vampeg2dec: Streams are progressive by default.

By initializating progressive, then interlaced streams are detected correctly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

va: Add and use common decode negotiate vmethod.

This vmethod can be used by decoders with the same VA decoder reopen logic:
same profile, chroma, width and height.

Also a new public method called gst_va_base_dec_set_output_state() with the
common GStreamer code for setting the output state, which is always called by
the negotiate vmethod.

In order to do this refactoring, new variables in vabasedec have to be populated
by the decoders:

* width and height define the resolution set in VA decoder. In the case of H264
  would be de coded_width and codec_height, or max_width and max_height in AV1.
* output_info is the downstream video info used for negotiation in
  gst_va_base_dec_set_output_state().
* input_state, from codec parent class shall be also held by vabasedec

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

vaav1dec: Use gst_va_base_dec_prepare_output_frame().

And simplify a bit the code flow.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

va: Add and use gst_va_base_dec_prepare_output_frame().

This helper will do downstream negotiation and later will
allocate the output frame.

H265 and AV1 decoders don't use this approach since their output
frame allocation is different.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

va: Remove last_ret error handling in decoders.

It was used in the early development of the base classes. Now it
shouldn't be needed.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

va: Add and use gst_va_base_dec_process_output().

This function will copy the frame, if it's needed, and will apply buffer flags.

The function is used by all the decoders.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3480>

va: Fix the caps memory leak by gst_va_pool_new_with_config().

The gst_va_pool_new_with_config() will ref the caps, and so we need
to unref after that.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3503>

va: Add render node name in non-first devices.

There could be multi-GPU setups where the non-first has more
entrypoints than the first one, and the elements names are not
homogeneous, leading to pipeline building error.

This patch add the render node in the elements names when they belong
to the non-first device.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3491>

vafilter: Increase the caps for HDR.

As they might be other medatadata types.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3473>

vafilter: Increase the number of 3DLUT caps to 16.

To fix the warning on Alderlake

vafilter gstvafilter.c:534:gst_va_filter_ensure_filters:<vafilter0>
vaQueryVideoProcFiltersCaps: list argument exceeds maximum number

Increase the number of caps to 16 as vadumpcaps does.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3473>

qt: deactivate context if fill_info fails

Now the OpenGL context is deactivated if call to gst_gl_context_fill_info()
fails in gst_qt_get_gl_wrapcontext(), preventing that the context is left
activated, which could lead to invalid memory reads.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3492>

av{aud,vid}dec: Report decoding errors to the base class

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1889>

v4l2: Fix SIGSEGV on 'change state' during 'format change'

Ensure all access to v4l2object->pool imply taking a lock and a hard ref on the pool

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3481>

gst-inspect: Don't leak list

Just iterate the list instead of trying to be smart...

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3497>

example: vaenc-dynamic-reconfigure: Support H265.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

va: Add H265 SCC profile support.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

va: Add extended formats support such as 10/12 bits, 4:2:2 and 4:4:4.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

va: enable multi tile support for H265 encoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

va: enable vah265lpenc for low power mode H265 encoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

va: Add the low-delay-b frame support for H265.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

va: Add prediction direction attribute support for H265 encoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

va: Register and enable and the vah265enc plugin.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

va: Implement the vah265enc plugin for va HEVC encoding.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2036>

opusdec: Use proper guint/guint8 type conversion

Do not cast, that might yield wrong results.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3494>