wpevideosrc: Give WebKit the keyboard, touch and pointer modifiers

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2156>

wpesrc: Convert from utf32 to support other keys

This makes all of the non-letter keys work.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2156>

wpesrc: Initialize key event to 0

Otherwise, WebKit sees random modifiers

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2156>

amfcodec: fix setting quality presets

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2159>

mxfdemux: Fix issue with re-syncing

In case of re-syncing (i.e. moving to another partition to avoid too much of an
interleave), there was previously no checks to figure out whether a given
partition was already fully handled (i.e. when coming across it again after a
previous resync).

In order to handle this at least for single-track partitions, check whether we
have reached the essence track duration, and if so skip the partition.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2150>

mxfdemux: Don't double-increase track position

The essence track position should only be overriden if we sucesfully switched to
another position. In case of EOS we do not want to override it else we would
increase the track position *again* at the end of this function

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2150>

amfcodec: Fix for MinGW build

timeapi.h header might not be missing depending on toolchain.
Also do hard meson error if winmm dep is not available but
amfcodec is explicitly enabled.
And fixing various GCC build warning errors.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2155>

gstreamer-sharp: fix typo in script name

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2152>

rtpbin: Fix division by zero when using ts-offset-smoothing-factor

avg_ts_offset may cause division by zero when calculating potential
overflow protection. This fix will avoid the division.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2151>

qsv: Add JPEG encoder

Note that YUY2 is not supported yet, since the format support is not
implemented now.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2142>

subparse: don't try to index string with -1

If the len of the string turns out to be 0, str[len - 1] resolved to
str[-1] which is not a good idea.

Fixes: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=46543
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2147>

Update openh264 wrap to v2.2.0 + fixes from master

Commit 0819f288dc5fc24fb34089113bc230aea65c8375 to be precise.

We need post-v2.2.0 build fixes from master, otherwise linking
fails because of missing symbols.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2145>

amfcodec: Update plugin cache

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2143>

nvencoder: Update property names and default values

Update properties of newly written D3D11/CUDA mode encoders
to be consistent with Intel QSV plugin. And rename "zerolatency"
to "zero-reorder-delay" because previous "zerolatency" may
misleading

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2140>

qsvencoder: Update property names and default values

Stop using MSDK plugins property naming, instead use unified names
and default values to be consistent with newly written NVIDIA encoders.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2140>

amfcodec: Add support for h264/h265 encoding

New encoder implementations for AMD GPU using AMF API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2109>

amfcodec: Import AMF SDK headers

SDK version v1.4.24 at the commit of
https://github.com/GPUOpen-LibrariesAndSDKs/AMF/commit/fbf12cd39fe1812ed902525a1c001307b94871b9

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2109>

msdk: Fix Windows build

Headers of gstva should not be included on Windows

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2141>

rtpopusdepay: assume 2 channels if sprop-stereo is missing

Fixes #1064

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2125>

d3d11window: Use ANSI version WIN32 API explicitly

We were using ANSI version APIs implicitly because UNICODE is not
defined by ourselves. But potentially it can be broken if user
defines UNICODE.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2131>

d3d11: Run indent

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2131>

ogg: fix possible buffer overrun

If an ogg stream does not match our expectations of how the end of a
buffer may be structured, it was possible to read memory past the end of
the buffer parsed by libogg.  Include a bounds check for this case and
stop parsing.

Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=3930

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2134>

rtpbasepayload: fix transfer annotation for push and push_list

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2133>

gst: remove custom logger for critical/warnings/fatals

It's not really needed anymore (only calls out to
g_log_default_handler() and interferes with applications using
g_log_set_default_handler().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2135>

riff-media: fix memory leak after usage for g_strjoin

This leak is observed with valgrind.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2129>

h265parser: Fix short_term_ref_pic_set() size calculation

This field is used by DXVA/NVDEC/VA, and each specification
describes (NVDEC is not well documented) that it's the number of
bits used in short_term_ref_pic_set().

DXVA doesn't explicitly mention that whether the size of
emulation preventation bytes (EPB) is inclusive or not, but
VA is clearly specifying that it's the size after removing
EPB. Excluding EPB size here makes more sense therefore.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1930>

build/fuzzing: integrate fuzz targets into the build system

Currently disabled but may be enabled later.

Updates the existing fuzzing to use shared libraries as that's easier
for meson to deal with if there is a mix of static and shared libraries
on the system.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2123>

meson: add the list of tools to summary

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2078>

gltransformation: let graphene alloc its structures memory aligned

With NEON instructions enabled, graphene expects the memory passed to it
16-byte-aligned. Otherwise unaligned memory access faults occur causing
SIGBUS signals.

graphene has alloc functions for its structures that take care of this,
so use them.

See also: https://github.com/ebassi/graphene/issues/215#issuecomment-794744829

Suggested-by: Sebastian Dröge <sebastian@centricular.com>
Signed-off-by: Bastian Krause <bst@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1321>

qmlglsrc: Fix deadlock when stopping

This fix makes sure that streaming thread stops waiting when the
qmlglsrc element transitions from playing to paused.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2115>

mxfvanc: Handle empty VANC packets

Some XDCAM recorders store empty packets for VANC, and don't even include the
2byte length.

Handle them in the same way as VANC packets with 0 packets.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2122>

qmlglsrc: Fix missing depth & stencil buffer

Qt Quick primitives which have some kind of alpha blending
(transparency, rounded corners) are z-sorted by Qt and rendered in the
correct order. For opaque primitives Qt relies on the OpenGL depth
buffer to correctly determine the visibility of stacked elements.

This change enables the depth buffer to make sure that opaque primitives
are correctly z-stacked.

https://doc.qt.io/qt-6/qtquick-visualcanvas-scenegraph-renderer.html#opaque-primitives

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2114>

docs: design: negotiation: Fix indentation in pseudo-code

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2120>

rtpstats: Remove non-existing twcc field docs from RTPPacketInfo and add missing field docs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2121>

rtpsession: Remove unused twcc fields from the struct

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2121>

gst-libav: fix build on systems without C++ compiler

Fix the following build failure on systems without C++ compiler:

The following exception(s) were encountered:
Running "/nvmedata/autobuild/instance-9/output-1/host/bin/or1k-buildroot-linux-musl-g++ --version" gave "[Errno 2] No such file or directory: '/nvmedata/autobuild/instance-9/output-1/host/bin/or1k-buildroot-linux-musl-g++'"

Fixes:
 - http://autobuild.buildroot.org/results/8ac0ba5eaaf7571857b4d8cfabf1488d640dc59a

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2112>

appsrc: Clarify buffer ref semantics in signals

The documentation could be read to mean that the caller continuous to
'own' the buffer, and that there is some other mechanism to find out
when to unref it.

Clarify that "not taking ownership" here means "taking a reference",
and specify that you can unref it at any time after calling the
function.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2110>

csharp: check format in the pre-commit hook

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/983>

gstreamer-sharp: apply new code style to alll sources

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/983>

gstreamer-sharp: ensure a consitent code format

Use EditorConfig and dotnet-format to ensure a consistent
code style.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/983>

webrtc/datachannel: Change initial data channel state to CONNECTING

It complies with http://w3c.github.io/webrtc-pc/#dom-rtcdatachannelstate.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2099>

qsvencoder: Don't use USAGE_HINT_ENCODER and enable derived image support

Peformance issue by disabled derived image support was reported

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2111>

Revert "vah264enc: Disable derived image support"

This reverts commit 0011feb0d660f331abde5b1e36f1c56e9cf3b014.

The commit caused performance-wise regression for some platforms

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2111>

msdk: use mfxU32 instead of uint32_t

Msdk should use mfx variables defined in mfxdefs.h. Replace uint32_t
with mfxU32.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2091>

gst-examples: add -xobjective-c for darwin

To avoid error when including #include <AppKit/AppKit.h>
the  build should include the c args xobjective-c

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2095>

bad: examples: fix unused res warning

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2095>

examples: fix build on macOS with gtk+-quartz-3.0

gdk_quartz_window_get_nsview is not declared in the header file now:
error: implicit declaration of function 'gdk_quartz_window_get_nsview'
is invalid in C99 [-Werror,-Wimplicit-function-declaration]

fixes #979

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2095>

ximagesink/xvimagesink: use GST_XINITTHREADS to ensure call to XInitThreads

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2098>

webrtc: Improve robustness of nice agent signal handlers

NiceAgent and it's associated thread is alive for as long as
GstWebRTCICE is alive so make sure any signal handlers connected to
NiceAgent do not access data that is deleted earlier.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2073>

nvcodec: Fix typos

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2107>

nvencoder: Fix operator overloading

GUID equality check was inverted

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2107>

libs: va: make sure va-enumtypes.h is generated before it's used

Need to add a dep for the generated header file to the dependency()
so that it gets generated before anyone tries to #include files
from it (which may happen before the actual lib is built because
that's only needed when stuff is linked together later).

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2057#note_1323236
and CI build failures in https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2096

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2103>

vah264enc: Disable derived image support

Behavior of derived image seems to be platform dependent and
it's buggy in some cases.

Fixes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1110
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2086>

va: h264enc: Do not use USAGE_HINT_ENCODER for input pool.

It seems that the VA_SURFACE_ATTRIB_USAGE_HINT_ENCODER is used only
for the recon surfaces, if we set this flags for input raw surfaces,
the alignment is wrong and we get broken result.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2100>

ci: Don't check indentation on post merge pipelines

We're probably safe to assume that it hasn't changed
since the MR pipeline last checked it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2063>

ci: stop automatically running tests post-merge

We did that in order to find flaky tests so we could add
them to our ignore lists, but that's all done now and I
doubt anyone still looks at the results, so this is just
wasting runner cycles at this point.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2063>

va: Fix printf format modifier for drm modifiers

DRM modifiers are 64bit, the printf modifier is platform dependant. Use
the G_GINT64_MODIFIER macro to make this printf portable.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2102>

aggregator: Improve debugging of arriving buffers

Log some details about the buffers that are arriving and
being enqueued on each sink pad.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2096>

gstreamer: Remove GST_DATAFLOW debug category

Nothing has logged anything to this category since event and
buffer dataflow was de-unified in 2005.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2097>

Use gmodule-no-export-2.0

We don't need `-Wl,--export-dynamic`, that's used only for executables
that needs to export an API to be used by plugins they load.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2031>

Remove glib and gobject dependencies everywhere

They are part of gst_dep already and we have to make sure to always have
gst_dep. The order in dependencies matters, because it is also the order
in which Meson will set -I args. We want gstreamer's config.h to take
precedence over glib's private config.h when it's a subproject.

While at it, remove useless fallback args for gmodule/gio dependencies,
only gstreamer core needs it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2031>

Fix cross build with mingw32

At least on Ubuntu 20.04 the x86_64-w64-mingw32-gcc toolchain defaults
to WinXP. We require at least Vista for FILE_STANDARD_INFO.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2022>

ges-smart-video-mixer: use the proper pad to get the positioner meta

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2094>

glcolorconvert: should copy metadatas from the incoming buffer

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2094>

Update android cross file

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/617>

Android: Implement JNI_OnLoad()

When building for Android, chances are that gstreamer is going to be
loaded from Java using System.loadLibrary(). In that case we can
initialize GStreamer (including static plugins), redirect log functions,
etc.

This code is copied from cerbero because it can be used with
gstreamer-full-1.0 too. Cerbero needs to be adapted to drop that code
and generate gst_init_static_plugins() function.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/617>

playbin/playbin3: Allow setting a NULL URI

The URI is already initialized to NULL at the beginning and GstPlayer
was assuming that it is possible to set to NULL at a later time too.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1124

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2090>

navigation: Rename parse_state to parse_modifier_state

`parse_state` sounds a bit weird and `parse_modifier_state` is clearer.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2087>

base:gl: add x11 deps to gstglx11_dep

On MacOS with homebrew the xlib-xcb.h is in
own cellar /opt/homebrew/Cellar/libx11/1.7.3.1/include
Need to add the windowing dependencies to gl tests

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2061>

qsvencoder: Add support for VA memory

Use VA allocator and buffer pool implementation for zero-copy
encoding with upstream VA elements

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2030>

qsv: Use pipeline's VA context

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2030>

Update openjp2 and libxml2 from wrapdb

This fix their static link for Windows.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2084>

libs: va: add VA allocator parameter for derived images usage.

Added GstVaFeature enum type, and new parameter for VA allocator's
set_format() and get_format(). Also added a new parameter in VA pool
gst_va_pool_new_with_config() and
gst_buffer_pool_config_set_va_allocation_params().

This new parameter will define if derived images will by used for
buffer mapping.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2057>

webrtcstats: Unify 'packets-lost' data type to int64

Previously, 'packets-lost' member of RTCReceivedRtpStreamStats had
a value of G_TYPE_INT from rtpsource or a value of G_TYPE_UINT64
from rtpjitterbuffer.
Because of the negative value of estimated amount of packets lost
in rtpsource as well as the description in
https://www.w3.org/TR/webrtc-stats/#dom-rtcreceivedrtpstreamstats
it is fixed to set this value with G_TYPE_INT64.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2049>

overlay: Fix qt support detection

On Ubuntu moc-qt5 command is called moc. This requires Meson 0.54.0 for
the new has_tools() method.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2075>

va: encoder: Remove unused allocation config.

Since it's also done inside of gst_va_pool_new_with_config().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2072>

vadisplay: Add description readonly property

Expose description of vendor for user information, similar to
the description property of d3d11device.
Also, set description and DRM device path on GstContext structure
so that user can read them and it will be printed on terminal
when gst-launch-1.0 is used

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2064>

gst-play: Improve Win32 keyboard input handling

The console HANDLE will be keep signalled state unless application
reads console input buffer immediately. So we should read and flush
console input buffer from the thread where the event is signalled,
instead of GMain context thread.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2058>

tools: Add support for building gstreamer tools against gst-full

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1581>

playsink: improve GL context sharing

Configure playsink tried element with the bus of the main pipeline.
That tried element can be a gl video sink, which would benefit from being
able to propagate context messages to the main pipeline and have other
internal pipeline elements configured with it. Having different elements
configured with the same GL context allows them to share buffers with
video/x-raw(memory:GLMemory) caps and achieving zero-copy.

Thanks to Alicia Boya García <aboya@igalia.com> for her work co-debugging
the issue and contributing to find a solution.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2056>

docs/design: Updates for upstream selection

Update existing documentation for various behaviours.

Add new information on how selection "upstream" of decodebin3 happens.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1905>

urisourcebin: When streams-aware, remove pads immediately

For the same reason we add them immediately

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1905>

urisourcebin: Don't wait for pads content when streams-aware

If the adaptive demux is streams-aware it can add/remove pads at any point in
time without the need for no-more-pads or data blocking

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1905>

urisourcebin: Don't do buffering if source already does

Sources that can internally handle buffering shouldn't have yet-another
buffering element after it. This can be simply detected by checking if it can
answer a TIME BUFFERING query just after creation.

If that is the case, we can expose the element source pads directly

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1905>

decodebin3: Handle upstream selection

Detect if upstream handles stream-selection, and if so bypass all stream
selection handling (streams are forwarded as-is).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1905>

query: Add a new stream selection query

This new API allows querying whether elements can handle stream selection
themselves or not.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1905>

ci: Don't tie up a full job slot for cerbero trigger

We don't have a huge amount of capacity on shared runners, so better if
we don't use that limited capacity on a job which sits there doing
~nothing.

There is a new runner class called 'placeholder-job' which accepts a
huge number of parallel jobs, separately to normal jobs, on the
assumption that they will consume ~no resources; start using that.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2071>

va: Don't expose internal classes.

VA allocators and pools classes don't need to be exposed as external
symbols.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2070>

gstplay: don't print error log in warning_cb

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2069>

vapostproc: Build classification dynamically.

By default, the classification is
"Converter/Filter/Colorspace/Scaler/Video/Hardware", but if VA
post-processor driver supports either color balance, skin tone
enhancement, sharpening or noise reduction, "Effect" is added.

Thus, if vapostproc ranking is raised, it can be chosen by
autovideosink.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2066>

webrtc: produce stats for all relevant streams

Instead of only using the last ssrc that was pushed into a sink pad.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

webrtc: avoid different versions of gnu-indent always wanting to change !!

Add some sneaky parenthesis to avoid always having to use git commit -n
or revert out hunk of the change.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

webrtc: implement initial simulcast fec/rtx usage

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

webrtc/datachannel: fix use-after-free in sctp state notification

g_signal_disconnect*() doesn't stop any existing callbacks from running
which means that if the notify::state callback is in progress in one
thread and the data channel object is finalize()ed in another thread,
then there could be a use-after-free trying lock the data channel
object.

We can't reasonably use a GWeakRef as we don't have a 'parent' object to
free the GWeakRef after the data channel is finalized.  This is also
complicated by the fact that the application can hold a reference to the
data channel object that would live beyond the lifetime of webrtcbin
itself.

We solve this by implementing a ghetto weak-ref solution internally with
a list of outstanding data channels.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

tests/webrtc: fix a use-after-free in test_data_channel_close

g_object_weak_ref() is not thread-safe and the data channel object's
refs/unrefs can happen on multiple threads.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

tests/webrtc: fix a race in the tests related to state tracking

If things progress fast enough, some state changes may not be seen be
the waiting code.

Fix by:
1. keeping a list of all the state changes
2. waiting checks each entry and if the relevant state is found, all
   states up to and including then are removed.

This ensures that any waits will see all the state sets.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

tests/webrtc: factor out src pad property checking to a separate function

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

webrtcbin: initial support for sending and receiving simulcast streams

Input (sink pads) is the already-ssrc-muxed stream with the relevant rtp
sdes header extensions already applied:
  - mid
  - stream-id
  - repaired-stream-id

Output (src pads) have the pads separated into individual ssrc's as
that's what rtpbin gives us.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

tests/webrtc: test for enabled bundled fec/rtx

Doesn't actually check that any fec/rtx happens, just that the pipeline
is vaguely sane and doesn't error.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>

webrtcbin: support multiple received streams for a single mline

Each rtpbin exposed recv_src pad is now exposed as webrtcbin src_%u pad
now with no meaining applied to the value of %u.  Previously this used
to mean the mline in the SDP.  If this is is still required, then the
transceiver can be retrieved from the pad and the "mlineindex" property
from the transciever.  The "mid" is also retrievable from the
transceiver.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1664>