wasapi2: Add support for process loopback capture

Adding loopback capture mode for specified PID.

Note that this feature requires Windows 10 build 20348
(Windows 11/Windows Server 2022 or later),
and any process loopback related properties will not be exposed
if OS does not support it.

Example launch lines:
* wasapi2src loopback-mode=include-process-tree loopback-target-pid=<PID>
 Captures audio generated by an application (specified by PID)
 and its child process
* wasapi2src loopback-mode=exclude-process-tree loopback-target-pid=<PID>
 Captures desktop audio excluding PID and its child process

Fixes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1278
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3195>

ci: Run windows jobs when win-* binary subprojects are updated

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3209>

meson: Update flex, bison, and nasm

Latest flex is 2.6.4, bison is 3.8.2, nasm is 2.15.04

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3209>

examples/webrtc/signalling: Fix compatibility with Python 3.10

Fix asyncio throwing a deprecation warning when using
asyncio.get_event_loop().

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3196>

rtmp2sink: Correctly return GST_FLOW_ERROR on error

If there is an error while connecting, the streaming task will be stopped, and
is_running() will be false, causing a GST_FLOW_FLUSHING to be returned. Instead,
we perform the error check (!self->connection) first, to return an error if
that's what occured.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3189>

gst-play: Don't leak the stream collection

We are given a reference to the collection when parsing it from the
message. Just store it (instead of referencing it again).

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3186>

webrtc/nice: fix small leak of split strings

g_strfreev previously stopped at our manual NULL-termination. Fix by
restoring the pointer after joining.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3188>

gst-docs: Fix typo in hotdoc kwarg

The hotdoc module passes unknown keyword arguments as arguments to
hotdoc, and the fatal warnings argument is --fatal-warnings.

https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2972#note_1586361

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3178>

wavparse: Avoid occasional crash due to referencing freed buffer.

We've seen occasional crashes in the `wavparse` module associated with
referencing a buffer in `gst_wavparse_chain` that's already been freed.  The
reference is stolen when the buffer is transferred to the adapter with
`gst_adapter_push` and, IIUC, assuming the source doesn't hold a reference to
the buffer, the buffer could be freed during interaction with the adapter in
`gst_wavparse_stream_headers`.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3179>

rtsp-client: Remove duplicate documentation

Confuses the documentation builder, since it's documented twice it complains
about a missing "Since:" marker whereas it's present in the documentation
comment further down

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3180>

riff: Mark jpeg as parsed

This is needed so that autoplugging works with avidemux and JPEG decoders that
need parsed sink caps (eg rockchip 'mppjpegdec' decoder). It also works fine
with 'jpegdec' decoder regardless.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3175>

wavparse: Fix crash that occurs in push mode when header chunks are corrupted
in certain ways.

In the case that a test is provided for, the size of the `fmt ` chunk is
changed from 16 bytes to 18 bytes (bytes 17 - 20 below):
```
$ hexdump -C corruptheadertestsrc.wav
00000000  52 49 46 46 e4 fd 00 00  57 41 56 45 66 6d 74 20  |RIFF....WAVEfmt |
00000010  12 00 00 00 01 00 01 00  80 3e 00 00 00 7d 00 00  |.........>...}..|
00000020  02 00 10 00 64 61 74 61                           |....data|
00000028
```

(Note that the original file is much larger.  This was the smallest sub-file
I could find that would generate the crash.)

Note that, while the same issue doesn't cause a crash in pull mode, there's a
different issue in that the file is processed successfully as if it was a .wav
file with zero samples.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3173>

vp9parse: The show_existing_frame buffer should not be decode only.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3155>

vp9parse: Correct the pts for frames inside a super frame.

When the alignment is "FRAME" and the parse is likely connecting to
a decoder, the current PTS setting for VP9 frames inside a super
frame is not very correct.

For example, the super frame may begin with non-displayed frames and
end with a displayed frame. The current way will assign the PTS to
the first non-displayed frame, which is a decode-only frame and the
PTS will be discarded in the video decoder. While the last displayed
frame has invalid PTS, and so the video decoder needs to guess its
PTS based on the frame rate and previous frame's PTS. This is not a
decent and robust way. And more important, when the previous frames
provide DTS, the video decoder will also guess the PTS based on the
previous frames' DTS and trigger the warning like:

  gstvideodecoder.c:3147:gst_video_decoder_prepare_finish_frame: \
  <vavp9dec0> decreasing timestame

It sets the reordered_output and makes the decoder in free run mode.

We should correct the PTS for a super frame, let the non-displayed
frames have no PTS while set the correct PTS to the displayed one.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3155>

avfvideosrc: Allow specifying crop coordinates during screen capture

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3153>

rtsp-server: Free client if no connection could be created

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3164>

mxfdemux: Add support for Canon XF-HEVC

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1495

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3163>

mxfdemux: Don't leak index table segments on failures

The segment was freed ... but not the contents :)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3163>

rtsp-server: Add since marker for adjust_error_code

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3157>

meson: Update libsoup.wrap

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3162>

videorate: Do not close segment when getting a same segment twice

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3059>

videorate: Handle closing segment on EOS right after caps event

The scenario is what we try in the tests:
- we have a segment with .stop set
- some frame(s) flow
- we get a CAPS event
- we get an EOS (before getting buffers after the CAPS event)

in that case, without that patch, the segment is not properly closed
which is not correct. In this patch we keep track of previous caps until
a new buffer arrives, this way in that situation we set previous caps
again, and close the segment with the previous buffer.

Fixes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1352
in this specific case

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3059>

validate: Plug some leaks

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3059>

oss4: Fix debug category initialization

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1456

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3158>

webrtc: Move GST_WEBRTC_ERROR_TYPE_ERROR at the end of the enum to keep ABI compatibility

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3154>

tests/webrtc: Add test for 'add-turn-server' action signal

It just checks return value of the action signal.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3131>

sctpenc: Prohibit sending of interleaved message parts

Apparently we cannot start sending messages from another datachannel
before the previous message was completely sent. usrsctplib will
complain about being locked on another stream id and set
errno=EINVAL.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2454>

opengl: Fix usage of eglCreate/DestroyImage

The implementation was inconsistent between create and destroy. EGLImage
creation and destruction is requires for EGL 1.5 and up, while
otherwise the KHR version is only available if EGL_KHR_image_base
feature is set. Not doing these check may lead to getting a function
pointer to a stub, which is notably the case when using apitrace.

Fixes #1389

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2925>

rtsp-server: Add support for adjusting request response on pipeline errors

The idea is to give the application the possibility to adjust the error
code when responding to a request. For that purpose the pipeline's bus
messages are emitted to subscribers through a signal handle-message.
The subscribers can then check those messages for errors and adjust
the response error code by overriding the virtual method
adjust_error_code().

Fixes #1294

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2972>

splitmuxsrc: don't queue data on unlinked pads

Once a pad has returned NOT_LINKED, the part reader shouldn't let its
corresponding data queue run full and eventually (after 20 seconds)
stall playback.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3145>

rtpsource: Don't do probation for RTX sources

Disable probation for RTX sources as packets will arrive very
irregularly and waiting for a second packet usually exceeds the deadline
of the retransmission.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/issues/181

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3112>

rtp: examples: client-rtpaux: Provide correct caps by payload type and RTX pt map by session

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3112>

tests/check/rtpsession: extend test_internal_sources_timeout

to verify that rtx SSRCs do not BYE after timeout

See https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/issues/360

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3112>

rtpsession: Remember the corresponding media SSRC for RTX sources

This allows timing out the RTX source and sending BYE for it when the
actual media source belonging to it is timed out.

This change only applies to sending sources from this session.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/issues/360

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3112>

rtpsource: Rename rtp_source_update_caps to rtp_source_update_send_caps

To make it clear that this is only used for sending RTP sources.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3112>

rtpsession: Rename gst_rtp_session_sink_setcaps to gst_rtp_session_setcaps_recv_rtp

to make it clearer that this is for setting receiver caps and to make it
more consistent with gst_rtp_session_setcaps_send_rtp.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3112>

decodebin3: allow to call "dispose" multiple times

https://docs.gtk.org/gobject/concepts.html#reference-counts-and-cycles

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3135>

decklink: reset calculation of gst_decklink_video_src_update_time_mapping on no_signal. When the HDMI cable was disconnected for a long time, the calculation took too much time. SDI cable works fine.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3132>

ci: Fix website regen on push

Don't make the integrate stage manual, we need it to regen the website

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3134>

aggregator: fix input buffering

We need to be able to buffer at least the aggregator latency +
upstream latency, which is the value used to compute the aggregator
deadline.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3138>

Meson: Fix libxml2 fallback

The variable xml2lib_dep does not exist. The correct name is already in
the wrap file.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3136>

rtspsrc: Retry SETUP with non-compliant URL resolution on "Bad Request" and "Not found"

Various RTSP servers/cameras assume base and control URL to be simply
appended instead of being resolved according to the relative URL
resolution algorithm as mandated by the RTSP specification.

To work around this, try using such a non-compliant control URL if the
server didn't like the URL used in the first SETUP request.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1447
Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/issues/922

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3127>

webrtcbin: Fix pointer dereference before null check

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3129>

decklink: Add new persistent-id property and sort devices by persistent ID

The order of the devices iterator from the SDK is undefined and can
randomly change.

Keep the device-number property for backwards compatibility and
simplicity but prefer the persistent-id property and also use it for the
device provider implementation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3078>

decodebin3: fix mutex leaks

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3126>

transcodebin: Implement support for upstream stream selection

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3128>

decodebin3: Do not try to plug a decoder on raw formats

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3123>

videosink: Don't return unknown end-time from get_times()

... in case of reverse playback. Otherwise basesink will not
wait for clock

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3121>

avauddec: address regression with WMA files ..

By outputting lead-in samples that FFmpeg now would like us to ignore,
and discarding trailing samples that it would now like us to output.

See https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1474

Fixes: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1348
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3117>

validate:launcher: Cleanup test uuid when copying it

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3122>

validate: launcher: Add a argument to avoid rereuning flaky tests

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3122>

validate: launcher: Keep variable framerate from input when possible

But disable it if forcing a framerate for some reason

Fixing our support for variable framerate in the encoding profile
serialization format.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3122>

validate: launcher: Fix some pep8 issues

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3122>

validate: Handle testfiles that need an HTTP server

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3122>

validate:ssim: Show original/nok/diff images on error images

This way the user can easily see how off the image is very easily

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3122>

validate:scenario: Fix setting iterating variable from arrays

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3122>

d3d11: Fix debug build on UWP

GstDXGIGetDebugInterface() is unused when targeting UWP. We directly
call DXGIGetDebugInterface1() in that case.

Fixes build failure:

../gst-libs/gst/d3d11/gstd3d11device.cpp(271): error C2440: '=': cannot convert from 'HRESULT (__cdecl *)(UINT,const IID &,void **)' to 'DXGIGetDebugInterface_t'
../gst-libs/gst/d3d11/gstd3d11device.cpp(271): note: This conversion requires a reinterpret_cast, a C-style cast or function-style cast

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3118>

webrtc/nice: Make sure to return NULL when validating turn server fails

It affects 'add-turn-server' signal action and 'turn-server' property
of webrtcbin.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3120>

webrtc: return error when sending on non-open datachannel

According to W3C
specification (https://w3c.github.io/webrtc-pc/#datachannel-send) we
should return InvalidStateError exception when trying to send when the
channel is not open. In the world of C/glib/gstreamer we don't have
exceptions but have to rely on gboolean/GError instead. Introducing
these calls for a change in function signature of the action signals
used to send data on the datachannel. Changing the signature of the
existing "send-string" and "send-data" signals would mean an immediate
breaking change so instead we deprecate them. Furthermore, there is no
way to express GError** as an argument to an action signal in a way
that fits language bindings (pointer-to-pointer simply does not work)
and we have to use regular functions instead.

Therefore we introduce gst_webrtc_data_channel_send_data_full() and
gst_webrtc_data_channel_send_string_full() while deprecating the old
functions and corresponding signals.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1958>

eglimage: fix Y412_LE DMABuf import support

This fixed the error below:

GST_GL_PLATFORM=egl gst-launch-1.0 videotestsrc ! msdkvpp !
"video/x-raw(memory:DMABuf),format=Y412_LE" ! glimagesink

0:00:00.063737691 576978 0x557c11266920 ERROR             gleglimage
gsteglimage.c:577:_drm_rgba_fourcc_from_info: Unsupported format for
DMABuf.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1389>

eglimage: add Y410 DMABuf import support

This fixed the error below:

GST_GL_PLATFORM=egl gst-launch-1.0 videotestsrc ! msdkvpp ! \
"video/x-raw(memory:DMABuf),format=Y410" ! glimagesink

0:00:00.040553883 576732 0x55be9b35f920 ERROR             gleglimage
gsteglimage.c:577:_drm_rgba_fourcc_from_info: Unsupported format for
DMABuf.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1389>

eglimage: add Y212_LE / Y212_BE DMABuf import support

This fixed the error below:

GST_GL_PLATFORM=egl gst-launch-1.0 videotestsrc ! msdkvpp \
! "video/x-raw(memory:DMABuf),format=Y212_LE" ! glimagesink

0:00:00.148086281 575148 0x562a6d36d920 ERROR             gleglimage
gsteglimage.c:555:_drm_rgba_fourcc_from_info: Unsupported format for
DMABuf.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1389>

eglimage: add Y210 DMABuf import support

This fixed the error below:

GST_GL_PLATFORM=egl gst-launch-1.0 videotestsrc ! msdkvpp ! \
"video/x-raw(memory:DMABuf),format=Y210" ! glimagesink

0:00:00.039881627 571365 0x56245eba6920 ERROR             gleglimage
gsteglimage.c:549:_drm_rgba_fourcc_from_info: Unsupported format for
DMABuf

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1389>

gstbin: Use g_queue_clear_full()

Use g_queue_clear_full() to release the child list
instead of iterating over the list twice.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3101>

gstbin: Fix a potential leak in gst_bin_do_deep_add_remove()

If a child element of a bin is unparented while
gst_bin_do_deep_add_remove() is iterating the children,
don't leak a ref to it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3101>

gstiterator: Don't use gst_object_unref() for a GObject

The owner of a GstIterator is a plain GObject. Don't unref it
with gst_object_unref(), or it will be logged in tracer
info incorrectly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3101>

h265parse: fix typo in member of GstH265SPS

Rename sps_extnsion_params to sps_extension_params

Fix comment about vui_parameters_present_flag

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3072>

Back to development

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3115>

Release 1.21.1

Update ChangeLogs for 1.21.1

playsink: Hold a reference to the soft volume element

Always hold a reference to the soft volume element
provided by the playsinkaudioconvert bin helper, the
same as when volume is provided by a sink element,
or the soft volume element gets unreffed too soon.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3108>

xvimagesink: Don't leak XvShmImage

If allocating an XvShmImage generates an X error, but
still returns some allocated memory, make sure to free it
so it doesn't leak.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3100>

xvimagesink: Zero initialize mask array

Clear the stack array before setting bits to pass to
XISelectEvents(). Fixes spurious crash from an X error:

X Error of failed request:  BadValue (integer parameter out of range for operation)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3100>

xvimagesink: Don't leak temporary

Use a stack variable instead of a temporarily
malloced variable that wasn't being freed
properly. Fixes a small leak of a GstXvTouchDevice

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3100>

decodebin3: Make sure event is writable before modifying

Make sure we're operating on a private copy of an event when
modifying it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3107>

urisourcebin: Make sure event is writable before modifying.

Make sure we're operating on a private copy of an event before
modifying it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3107>

decodebin3: Don't lose a ref on EOS event

Make sure not to give away the ref on the final EOS
event for which the probe handler is returning GST_PAD_PROBE_REMOVE
when pushing the event manually.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3107>

queue2: Hold the lock when modifying sinkresult

As it's done elsewhere. Avoids a potential race of the field being modified in
the meantime.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3110>

qtdemux: Don't stop task when resetting

This is a regression that was introduced in
cca2f555d14b5751f7f9d466b66127544dad5138 (yes, 9 years ago).

The only place where a demuxer streaming thread should be stopped is when the
sinkpad is deactivated from pull mode (i.e. PAUSED->READY).

Attempting to stop the task in this function would cause this to happen when a
FLUSH_STOP or STREAM_START event is received... which can cause deadlocks.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3109>

avauddec: fix unnecessary reconfiguration if the audio layout isn't specified

It would constantly want to renegotiate (and spam the debug log) even
though the channel layout hasn't actually changed. We use the same
fallback in gst_ffmpegauddec_negotiate() already.

This happens with WMA files for example.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3103>

splitmuxsrc: don't consider unlinked pads when deactivating part

If splitmuxsrc exposes multiple pads, but only one is linked, part pads
will never see an EOS event. This shouldn't prevent the part from being
eventually deactivated.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3099>

pbutils: descriptions: add meta/x-klv

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3094>

pbutils: descriptions: fix gst_pb_utils_get_caps_description_flags()

And add a little unit test.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3094>

xmptag: Call gst_tag_register_musicbrainz_tags during init

We need to call this to register the MusixBrainz tags before we use
them in an XMP schema.

Fixes this critical when attempting to run jpegparse on a JPEG
containing MusicBrainz XMP tags:

  GStreamer-CRITICAL **: 20:41:07.885: gst_tag_get_type: assertion 'info != NULL' failed

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3092>

test-effect-switch: Fix some memory leaks and make effect element ownership clearer

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3091>

gaussblur: Fix memory leak when setting caps multiple times

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3091>

v4l2codecs: Warn user when no media devices are found

Currently if the user is not able to access the devices under /dev/media*,
either due to no media devices present on the system or simply no permission
to access the device, v4l2codecs initialises with no features or debug messages.

Since calling `GST_DEBUG="v4l2*:7" gst-inspect-1.0 v4l2codecs` is a typical way
to diagnose why element(s) failed to enumerate, we should be more verbose here
when the user is not able to access any /dev/media* device. So print a simple
debug message in this case to aid debugging.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3088>

buffer: drop parent meta in deep copy/foreach_metadata

The purpose of a deep buffer copy is to be able to release the source
buffer and all its dependencies. Attaching the parent buffer meta to
the newly created deep copy needlessly keeps holding a reference to the
parent buffer.

The issue this solves is the fact you need to allocate more
buffers, as you have free buffers being held for no reason. In the good
cases it will use more memory, in the bad case it will stall your
pipeline (since codecs often need a minimum number of buffers to
actually work).

Fixes #283

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2928>

tsdemux: Don't trigger a program change when falling back to ignore-pcr behaviour

Since commit a79a756b79aa1675e we could change to ignore-pcr automatically at 500ms
into a live stream when no PCR is seen by then. However the stream counting in
program change detection was wrongly considering ignore-pcr programs to have a
separate PCR PID, even though we are actually ignoring the PCR PID completely,
resulting in an erroneous program switch getting triggered from the different
stream count. This in turn would send an EOS and switch out the pads for what
actually is still the same program, while we intended to simply apply a
workaround for broken encoders.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3060>

gst-docs: update macOS new project instructions

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3074>

meson: Do not modify default value of -Dtools for subprojects

See: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1136
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2166>

meson: Remove some dead code

We no longer create any symlinks.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2166>

examples: Support multiple video streams in JS webrtc sendrecv

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3079>

rtspsrc: Fix usage of IPv6 connections in SETUP

If the SETUP request returns an IPv6 server address in the Transport
field, we would generate an incorrect URI, and multiudpsink would fail
to initialize:

```
     rtspsrc gstrtspsrc.c:9780:dump_key_value:<source>    key: 'Transport', value: 'RTP/AVP;unicast;source=fe80::dc27:25ff:fe5e:bd13:8080;client_port=62696-62697;server_port=4000-4001'
...
     rtspsrc gstrtspsrc.c:4595:gst_rtspsrc_stream_configure_udp_sinks:<source> configure RTP UDP sink for fe80::dc27:25ff:fe5e:bd13:8080:4000
...
multiudpsink gstmultiudpsink.c:1229:gst_multiudpsink_configure_client:<udpsink0> error: Invalid address family (got 23)
```

We can't look at stream->is_ipv6 because we can't rely on the server
returning the right value there. In the issue reported about this,
server reported itself as `KuP RTSP Server/0.1`, and the SDP was:

```
c=IN IP4
m=video 54608 RTP/AVP 96
a=rtpmap:96 H264/90000
```

So we need to parse the string value and figure out the family
ourselves.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1058

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1819>

meson: Remove builtin variable datadir from pkgconfig variables

Fixes warning with meson 0.62:

gst-plugins-bad| subprojects/gst-plugins-bad/meson.build:546: WARNING:
Project targets '>= 0.62' but uses feature deprecated since '0.62.0':
pkgconfig.generate variable for builtin directories. They will be
automatically included when referenced

and more.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3086>

decodebin3: Fix memory issues with active selection list

This had a couple of issues:
* The backing strings (from GstStream) could disappear
* The actual list wasn't properly reset/freed when decodebin3 was re-used

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3083>

qtdemux: guard against timestamp calculation overflow in gap event loop

Could possibly cause an endless loop.

Fixes #1400.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3084>

dashdemux2: fix mpd unit test expectations

Update unit test for some mpd cases that were reporting
timestamps including the period start time, while
dashdemux2 expects that it needs to add the period
start time itself.

Fix the tests to not expect the period start time
to be included.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3025>

dashdemux2: Set timestamp relative to period start

These values will be referred to as timestamp relative to period start
so need to subtract period start time from the values.

Fixes a problem with determining the start position when playing Live content
with SegmentTimeline, presentationTimeOffset and a non-0 period start time.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3025>

gsturi: When setting the same string again do nothing

Otherwise code like gst_uri_set_host(uri, gst_uri_get_host(uri)) would
first free the string, then create a copy of the freed string and then
assigned that.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3076>