Merge branch 'upstream/1.24.10' into tizen_gst_1.24.10

Change-Id: I913bfe3f95a197a96283ab35506a437d04871357

Merge branch 'upstream/1.24.9' into tizen_gst_1.24.9

Change-Id: I20e783a9005d385a154a2e166ad9721c29a92e4a

Release 1.24.10

qtdemux: Add size check for parsing SMI / SEQH atom

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-244
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3853

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Check for invalid atom length when extracting Closed Caption data

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-243
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3849

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Actually handle errors returns from various functions instead of ignoring them

Ignoring them might cause the element to continue as if all is fine despite the
internal state being inconsistent. This can lead to all kinds of follow-up
issues, including memory safety issues.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-245
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3847

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Make sure there are enough offsets to read when parsing samples

While this specific case is also caught when initializing co_chunk, the error
is ignored in various places and calling into the function would lead to out of
bounds reads if the error message doesn't cause the pipeline to be shut down
fast enough.

To avoid this, no matter what, make sure enough offsets are available when
parsing them. While this is potentially slower, the same is already done in the
non-chunks_are_samples case.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-245
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3847

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Fix error handling when parsing cenc sample groups fails

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-238, GHSL-2024-239, GHSL-2024-240
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3846

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Fix length checks and offsets in stsd entry parsing

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-242
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3845

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Make sure enough data is available before reading wave header node

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-236
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3843

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Make sure only an even number of bytes is processed when handling CEA608 data

An odd number of bytes would lead to out of bound reads and writes, and doesn't
make any sense as CEA608 comes in byte pairs.

Strip off any leftover bytes and assume everything before that is valid.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-195
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3841

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Check sizes of stsc/stco/stts before trying to merge entries

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-246
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3854

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Don't iterate over all trun entries if none of the flags are set

Nothing would be printed anyway.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Fix debug output during trun parsing

Various integers are unsigned so print them as such. Also print the actual
allocation size if allocation fails, not only parts of it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

qtdemux: Fix integer overflow when allocating the samples table for fragmented MP4

This can lead to out of bounds writes and NULL pointer dereferences.

Fixes GHSL-2024-094, GHSL-2024-237, GHSL-2024-241
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3839

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8060>

h265parser/bitwriter: add some comments for ceil_log2 use

Validate that the length of field must be calculate with
ceil_logs2 and not bit storage.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7905>

h264parse: use of ceil_log2 instead of bit_storage

According to the specification:
The length of the slice_group_id[ i ] syntax element is Ceil( Log2(
num_slice_groups_minus1 + 1 ) ) bits

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7905>

utils: improve gst_util_ceil_log2

According to the following comparison of algorithms, the value
for 0 and 1 was giving an incorrect result.

https://gist.github.com/ceyusa/6061b33ac109a68bcd222f6919968c9a

More information here:
https://github.com/rofrol/codeforces/blob/master/ceil_log2.c

Use a different algorithm which offers better result and keep the
performance.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7905>

matroskademux: Put a copy of the codec data into the A_MS/ACM caps

The original codec data buffer is owned by matroskademux and does not
necessarily live as long as the caps.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-280
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3894

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8058>

matroskademux: Skip over zero-sized Xiph stream headers

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-251
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3867

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8058>

matroskademux: Skip over laces directly when postprocessing the frame fails

Otherwise NULL buffers might be handled afterwards.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-249
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3865

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8058>

matroskademux: Don't take data out of an empty adapter when processing WavPack frames

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-249
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3865

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8058>

matroskademux: Check for big enough WavPack codec private data before accessing it

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-250
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3866

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8058>

matroskademux: Fix off-by-one when parsing multi-channel WavPack

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8058>

matroskademux: Only unmap GstMapInfo in WavPack header extraction error paths if previously mapped

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-197
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3863

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8058>

avisubtitle: Fix size checks and avoid overflows when checking sizes

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-262
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3890

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8055>

wavparse: Check size before reading ds64 chunk

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-261
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3889

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8054>

wavparse: Fix clipping of size to the file size

The size does not include the 8 bytes tag and length, so an additional 8 bytes
must be removed here. 8 bytes are always available at this point because
otherwise the parsing of the tag and length right above would've failed.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-260
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3888

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8054>

wavparse: Check that at least 32 bytes are available before parsing smpl chunks

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-259
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3887

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8054>

wavparse: Check that at least 4 bytes are available before parsing cue chunks

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8054>

wavparse: Fix parsing of acid chunk

Simply casting the bytes to a struct can lead to crashes because of unaligned
reads, and is also missing the endianness swapping that is necessary on big
endian architectures.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8054>

wavparse: Make sure enough data for the tag list tag is available before parsing

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-258
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3886

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8054>

wavparse: Check for short reads when parsing headers in pull mode

And also return the actual flow return to the caller instead of always returning
GST_FLOW_ERROR.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-258, GHSL-2024-260
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3886
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3888

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8054>

gdkpixbufdec: Check if initializing the video info actually succeeded

Otherwise a 0-byte buffer would be allocated, which gives NULL memory when
mapped.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-118
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3876

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8053>

oggstream: review and fix per-format min_packet_size

This addresses all manually detected invalid reads in setup functions.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8050>

vorbis_parse: check writes to GstOggStream.vorbis_mode_sizes

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-117 Fixes gstreamer#3875

Also perform out-of-bounds check for accesses to op->packet

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8050>

discoverer: Don't print channel layout for more than 64 channels

64+ channels are always unpositioned / unknown layout.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-248
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3864

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8046>

jpegdec: Directly error out on negotiation failures

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-247
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3862

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8052>

subparse: Check for NULL return of strchr() when parsing LRC subtitles

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-263
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3892

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8051>

opusdec: Set at most 64 channels to NONE position

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-116
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3871

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8049>

ssaparse: Don't use strstr() on strings that are potentially not NULL-terminated

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8048>

ssaparse: Search for closing brace after opening brace

Otherwise removing anything between the braces leads to out of bound writes if
there is a closing brace before the first opening brace.

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-228
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3870

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8048>

vorbisdec: Set at most 64 channels to NONE position

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-115
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3869

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8047>

id3v2: Don't try parsing extended header if not enough data is available

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-235
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3842

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8045>

qtdemux: Avoid integer overflow when parsing Theora extension

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-166
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3851

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8044>

allocator: Avoid integer overflow when allocating sysmem

Thanks to Antonio Morales for finding and reporting the issue.

Fixes GHSL-2024-166
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3851

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8044>

level: Fix integer overflow when filling LevelMeta

The level in GstAudioLevelMeta is represented as a signed 8bit value from 0 to
127 (with 127 meaning silence). When converting from double, make sure to clip
the value, this also prevent integer overflow in the conversion. This fixes an
issue where a lower then -127db is reported and random level with near silent
streams (due to integer overflow).

Fixes #4068

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8029>

kmssink: Add mediatek auto-detection

Add MediaTek display controller into list of
auto-detected modules.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8027>

rtp: Fix precision loss in gst_rtcp_ntp_to_unix()

Without this patch the UNIX timestamp resulting from the translation from NTP
would be off by one nano-second.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8022>

systemclock: Don't keep the clock entry locked while getting the time from the clock

gst_clock_get_time() will take the clock mutex, which would then result in a lock
order violation and possible deadlocks. If both mutexes are to be locked, the
clock must always be locked first.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8019>

systemclock: Get rid of conditional unlocking of the clock entries

At every point it is known whether the entry needs to be unlocked or not.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8019>

systemclock: Remove confusing conditional unlock

At this point the entry is always locked and needs to be unlocked.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8019>

systemclock: Use a flag while waiting for the async thread to start

Otherwise there can be spurious wakeups.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8019>

mxfmux: Fix off-by-one in the month when generating a timestamp for now

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8018>

shout2send: Unref event at the end of the event function

The function takes ownership of it and should get rid of it at the end.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8006>

rtspsrc: Optionally timestamp RTP packets with their receive times in TCP/HTTP mode

Until https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6509
this was accidentally done inside rtpjitterbuffer for many years, and
doing so potentially solves problems on some streams while introducing
problems on others.

Make this configurable on rtspsrc and default to not set timestamps.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8004>

pad: Never push sticky events in response to a FLUSH_STOP

FLUSH_STOP is meant to clear the flushing state of pads and elements
downstream, not to process data. Hence, a FLUSH_STOP should not
propagate sticky events. This is also consistent with how flushes are a
special case for probes.

Currently this is almost always the case, since a FLUSH_STOP is
__usually__ preceded by a FLUSH_START, and events (sticky or not) are
discarded while a pad has the FLUSHING flag active (set by FLUSH_START).

However, it is currently assumed that a FLUSH_STOP not preceded by a
FLUSH_START is correct behavior, and this will occur while autoplugging
pipelines are constructed. This leaves us with an unhandled edge case!

This patch explicitly disables sending sticky events when pushing a
FLUSH_STOP, instead of relying on the flushing flag of the pad, which
will break in the edge case of a FLUSH_STOP not preceded by a
FLUSH_START.

If sticky events are propagated in response to a FLUSH_STOP, the
flushing thread can end up deadlocked in blocking code of a downstream
pad, such as a blocking probe. Instead, those events should be
propagated from the streaming thread of the pad when handling a
non-flushing synchronized event or buffer.

This fixes a deadlock found in WebKit with playbin3 when seeks occur
before preroll, where the seeking thread ended up stuck in the blocking
probe of playsink:
https://github.com/WebPlatformForEmbedded/WPEWebKit/issues/1367

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8003>

gstreamer: prefix debug dot node names to prevent splitting

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7998>

streamsynchronizer: Only send GAP events out of source pads

If streamsynchronizer is waiting on the stream's sinkpad and srcpad at the same
time, it can happen that the GAP event is otherwise sent out of the sinkpad,
which is in the wrong direction.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7997>

ci: docker: use cargo install --locked for cargo-c installation

Backported from commit 2263caff

Should fix the following error on image regeneration:

  package `cargo-util v0.2.15` cannot be built because it requires rustc 1.80
  or newer, while the currently active rustc version is 1.76.0

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7989>

Fix build error due to upgrade to ffmpeg 7.1

All patches are from gst-libav-1.24.9

Change-Id: I02fdb4e79ad3a5a47948a67f1164a763dcc41b42

discoverer: Also copy missing element details when copying a GstDiscovererInfo

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7986>

discoverer: Make sure the missing elements details array is NULL-terminated in a thread-safe way

And also avoid a possible out-of-bounds read if the array contains no elements.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7986>

discoverer: Fix typos

discovererd / discoverered -> discovered

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7986>

v4l2videodec: fix freeze race condition

This fixes a possible deadlock between gst_v4l2_video_dec_change_state
and gst_v4l2_video_dec_loop on the buffer pool.

When stopping capture, the flushing state of the v4l2 capture buffer
pool gets reverted in the processing loop after it was set via
gst_v4l2_object_unlock (self->v4l2capture) (in
gst_v4l2_video_dec_change_state). As a result, gst_v4l2_video_dec_loop
does not return and consequently, gst_pad_stop_task gets stuck waiting
for the GST_PAD_STREAM_LOCK. To circumvent this, skip acquiring the
buffer pool if stopping capture.

Suggested-by: Nicolas Dufresne <nicolas.dufresne@collabora.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7987>

x265: Allow building with x265-4.1

In x265-4.1 masteringDisplayColorVolume is changed from a pointer to a
character array embedded in struct x265_param.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7983>

meson: Don't unconditionally invoke the libsoup subproject

fallback: kwarg will invoke the specified subproject even if required:
false, which is not what we want here.

Reported at https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4045#note_2674340

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7982>

webrtc: don't crash on invalid bundle id

If the bundle id forwarded to connect_rtpfunnel is not valid,
the assertion fails and crashes the program. This is now instead
an error.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7966>

appsink: fix timeout logic for gst_app_sink_try_pull_sample

In case of gst_app_sink_try_pull_object returns an object instead of a sample,
the whole process must be restarted with the reduced timeout, otherwise requested
timeout could be easily exceeded.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7980>

qt(6)/material: ensure that we always update the context in setBuffer()

Scenario is that there are two (or more) GstGLContext's wrapping Qt's GL
context from either multiple qml(6)glsink or qml(6)glsrc elements.  Call flow is this:

1. material 1 setBuffer()
2. material 1 bind()
3. material 2 setBuffer()
4. material 2 bind()

If the call to setBuffer() reuses the same buffer as previous call, then the
qt context is not updated in the material.  If however the previously used qt
context by the material had been deactivated or freed, then bind() would fail
and could result in a critical like so:

gst_gl_context_thread_add: assertion 'context->priv->active_thread == g_thread_self ()' failed

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7978>

ccutils: Fix a typo in cc_buffer_take_cea608_field2

There are no users of cc_buffer_take_cea608_field2, so this never was a
problem.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7973>

ccutils: Fix a typo in max_buffer_time handling

All users set max_buffer_time to GST_CLOCK_TIME_NONE, effectively
infinite, so this never was a problem.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7973>

ccutils: Remove broken branch

This branch was added in dd00dab5e9e8650f3f00660c2e611f81f1e8cd5b but is
never actually taken, as it requires `cc_data` to be null but
`cc_data_len` to be non-null. It would then dereference the null
`cc_data`.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7973>

v4l2object: Fixed incorrect maximum value for int range

There are objects where maximum is not multiplication of the step,
e.g. there was a combination where max was 65535 with step 2.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7961>

tracers: unlock leaks tracer if already tracking

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7960>

splitmuxsrc: Convert part reader to a bin with a non-async bus

A pipeline always has an async bus, which involves allocating an fd pair. As
splitmuxsrc only uses the bus' sync handler, this is not required and can easily
cause splitmuxsrc to exceed the fd limit for no good reason.

The other features of GstPipeline are also not needed here, e.g. clock selection.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7957>

decodebin3: Unify collection switching checks

We only want to switch to a selection of an output collection if all streams are
present.

This was previously only done in one place (when triggering by new incoming
streams) but not when triggered by user/application.

Avoid this by moving the check to handle_stream_switch()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7956>

meson: Fix libva.wrap meson build

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7955>

discoverer: fix segfault in race condition adding a new uri's

There is a race condition adding new uri's right after receiving
the `discovered` event. We must wait until we have cleaned-up
the last discovery to start processing the new one

Fix #3758

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7934>

meson: Don't use libdrm_dep in cc.has_header()

It can't handle subproject dependencies. Pointed out in
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4033#note_2665974

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7932>

sdp: Add debug categories for message and mikey modules

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7926>

libav: avviddec: log the old resolution when updating the pool

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7925>

libav: viddec: provide details if meta has the wrong resolution

Will produce a more useful assert error message.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7925>

Revert "avviddec: Unlock video decoder stream lock temporarily while finishing frames"

This reverts commit 6b18f0233900ae4ecdbcde0ce76ab55fd85286f4.

This commit introduced a regression, see
https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4032

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7921>

jpegparse: Change plugin rank GST_RANK_NONE to GST_RANK_SECONDARY

- In order to use v4l2jpegdec, plugin need a parsed:true caps.
  During autoplugin, it is GST_RANK_NONE, so parsebin can't find
  jpegparse and therefore can't use v4l2jpegdec
- Increase the rank of the jpegparse plugin to use v4l2jpegdec

Change-Id: Icd89c177836580d91d0b730e31731cc8b383c23a

gstdeviceprovider: fix leaking hidden providers

Free list of hidden providers upon stop and dispose.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7915>

rtpmanager: don't map READWRITE in twcc header ext

There is no need to map the buffer as writable, as there is
only a read performed on the mapped buffer. This is in line
with other header extensions, as no other extensions maps
it as readwrite.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7910>

playbackutils: Fix caps leak in get_n_common_capsfeatures()

The gst_static_caps_get() return value is transfer-full.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7909>

glupload: Upgrade uploader selection message to INFO

This allow using INFO log to quickly see what uploader we
picked. This is very useful as the DEBUG category has became
very noisy.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7907>

glupload: dmabuf: Fix emulated tiled import

The added check for LINEAR modifier broke dmabuf import for
NV12 emulation (R8/RG88) of all sort (linear and tiled).

Fixes !5461

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7907>

audiostreamalign: Don't report disconts for every buffer if alignment-threshold is too small

If it is too small, the maximum allowed diff becomes 0 samples which would then
trigger if there is no discontinuity at all.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7906>

mpegtsmux: Mux timestampless buffers immediately

Instead of leaving them queued indefinitely, or until we're timing out
and it's the only buffer queued.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7901>

mpegtsmux: Don't time out in live mode if no timestamped next buffer is available

The muxer can only advance the time if it has a timestamped buffer that can be
output, otherwise it will just busy-wait and use up a lot of CPU.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/3912

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7901>

flvmux: Mux timestampless buffers immediately

Instead of leaving them queued indefinitely, or until we're timing out
and it's the only buffer queued.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7902>

flvmux: Don't time out in live mode if no timestamped next buffer is available

But also don't wait for a buffer on both pads, which might take forever in case
of gaps in one of the streams.

The muxer can only advance the time if it has a timestamped buffer that can be
output, otherwise it will just busy-wait and use up a lot of CPU.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7902>

udpsrc: protect cancellable from unlock/unlock_stop race

Protect cancellable from simultaneous unlock and unlock_stop calls from
basesrc class.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7897>

ci: Don't use a massive image for the cerbero trigger

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/4024

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7896>

mpegvideoparse: do not set delta unit flag on unknown frame type

When encoding an image to mpeg2 video, with something like:
  gst-launch-1.0 encodebin name=e profile=mpegpsmux:video/mpeg,mpegversion=2,systemstream=false ! \
                 filesink location=sample.mpg filesrc num-buffers=1 blocksize=$(stat -c%s sample.png) \
                 location=sample/dts.png ! pngdec ! e.
The only frame's type is set to an invalid value 0
The consequence is that mpegvideoparse sets the delta unit flag on the buffer because
it is not an I frame, then decodebin3 drops this only frame because the delta
unit flag is set and the decoder receives eos before it was able to receive any
encoded data

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7876>

ptp: use ip_mreq instead of ip_mreqn for macOS

To join a multicast the macOS still uses the interface address
from the ip_mreq instead of the ip_mreqn unlike other Linux systems.

So add a new conditional block for macOS to use ip_mreq for IP_ADD_MEMBERSHIP
and ip_mreqn for IP_MULTICAST_IF

This is similar to the fix in the glib for multicast join/leave
operation on macOS
https://gitlab.gnome.org/GNOME/glib/-/merge_requests/4333

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7864>

ccconverter: Don't override in_fps_entry when trying to take output

This allows to handle CDP streams where the framerate is not provided by the
caps and generally gives preference to the framerate inside the CDP packets over
the one in the caps.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7862>