The earlier size of 2 MB was set back in 2009, it doesn't
seem unreasonable to raise it to 8 MB these days. The use
case at hand is matroskademux containing both a video stream
with a very low amount of compression but no decoding latency,
and a H265 stream.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1538>
It is an extremely common mistake on Windows to have incorrect PATH
values when loading a plugin, and the error from g_module_error()
(which just calls FormatMessageW()) is very confusing in this case:
The specified module could not be found.
https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--0-499-#ERROR_MOD_NOT_FOUND
It implies the plugin itself could not be found. The actual issue is
that a DLL dependency could not be found. We need to detect this case
and print a more useful error message.
We should still print the error fetched from FormatMessage() so that
people are able to google for it.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1540>
GstAudioRingBufferSpec::segsize has been configured by using
device period but GstWasapi2RingBuffer was referencing the
buffer size returned by IAudioClient::GetBufferSize()
which is most likely larger than device period.
Fixing to sync them.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1533>
If the `area_surface` got unmapped when changing to the `READY` or
`NULL` state, we currently don't remap it when playback resumes and
`wp_viewporter` is supported. Without `wp_viewporter` we do remap
it, but rather unintentionally and also when not wanted.
On Weston this has not been a big problem as it so far wrongly maps
subsurfaces of unmapped surfaces anyway - i.e. only the black
background was missing on resume. On other compositors and future
Weston this prevents the `video_surface` to get remapped.
Shuffle things around to ensure `area_surface` is mapped in the
right situations and do some minor cleanup.
See also https://gitlab.freedesktop.org/wayland/weston/-/issues/426
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1483>
Latest ffmpeg has removed avcodec_get_context_defaults(), and its
documentation says a new AVCodecContext should be allocated for this
purpose. The pointer returned by avcodec_find_decoder() is now
const-qualified so we also need to adjust for it. And, AVCOL_RANGE_MPEG
is now rejected with strict_std_compliance > FF_COMPLIANCE_UNOFFICIAL.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1531>
The later, doing damage in surface coordinates instead of buffer
coordinates, has been deprecated. The reason for that is that it
is more prone to bugs, both on the client and the compositor side,
especially when paired with buffer scale, `wp_viewporter` or
buffer transforms.
Unfortunately, on Weston this risks running into
https://gitlab.freedesktop.org/wayland/weston/-/issues/446
(which causes trouble for several other projects as well). However,
that bug only affects cases where we run in sync mode, i.e. only
during resizes. In practise I haven't been able to observe the
issue.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>
Each time we call `wl_surface_damage()` we want to do full surface
damage. Like Mesa, just use `G_MAXINT32` to ensure we always do
full damage, reducing the need to track the right dimensions.
`window->video_rectangle` is now unused, but we keep it around for
now as we may need it again in the future.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>
From the spec:
> This request is used to describe the regions where the pending
> buffer is different from the current surface contents
We currently also call `wl_surface_damage()` on surfaces without
new or still compositor-hold buffers, e.g. when resizing the window.
In that case we call it on `area_surface_wrapper`, even though it
gets resized via `wp_viewport_set_destination()`, in which case
the compositor is in charge of repainting the area on screen.
Doing so is currently not forbidden by the spec, however it might
be in the future, see
https://gitlab.freedesktop.org/wayland/wayland/-/issues/267
Thus lets stay close to the spec and only call `wl_surface_damage()`
when we just attached a buffer.
Right now this prevents runtime assertions in Mutter.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>
`gst_wl_window_set_opaque` does not get called on window resizes,
potentially leaving opaque regions too small.
According to the spec opaque regions can be bigger than the surface
size - parts that fall outside of the surface will get ignored.
Thus we can can simply use `G_MAXINT32` and be sure that the whole
surfaces will always be covered.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1446>
If the VANC track does contain packets, but we skip over all packets, just
treat it the same as if there hadn't been any packets at all and send a
GAP event instead of erroring out with "Failed to handle essence element".
We would error out because when we reach the end of the loop without having
found a closed caption packet the flow return variable is still FLOW_ERROR
which is what it has been initialised to.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1518>
Though the profiles[0] is inited as GST_H265_PROFILE_INVALID in the
gst_h265_profile_tier_level_get_profile(), the profile detecting may
change its content later. So the return of profiles[0] may not be an
invalid profile even the len is 0.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1517>
The previous code was mistakenly trying to compute a cc_type out
of the first byte in the byte triplet, whereas it is to be interpreted
as:
> Bit b7 of the LINE value is the field number (0 for field 2; 1 for field 1).
> Bits b6 and b5 are 0. Bits b4-b0 form a 5-bit unsigned integer which
> represents the offset
The same mistake was made when creating padding packets.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1496>
Sometimes we can't output anything because we don't have enough
incoming frames. In that case, the resampler was trying to call
do_quantize() and do_resample() in a loop forever because there would
never be samples to output (so chain->samples would always be NULL).
Fix this by not calling chain->make_func() in a loop -- seems
completely unnecessary since calling it over and over won't change
anything if the make_func() can't output samples.
Also add some checks for the input and / or output being NULL when
doing conversion or quantization. This will happen when we have
nothing to output.
We can't bail early, because we need resampler->samples_avail to be
updated in gst_audio_resampler_resample(), so we must call that and
no-op everything along the way.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1461>
When the image is opaque but the output ProRes format has an alpha
component (4 component, 32 bits per pixel), Apple requires that we
signal that it should be ignored by setting the depth to 24 bits per
pixel. Not doing so causes the encoded files to fail validation.
So we set that in the caps and qtmux sets the depth value in the
container, which will be read by demuxers so that decoders can skip
those bytes entirely. qtdemux does this, but vtdec does not use this
information at present.
The sister change was made in qtmux and qtdemux in:
https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/merge_requests/1061
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1489>
Old "application/*" are now as per RFC8081 deprecated in favor of
new "font/*" mime types. Some new encoders are already using the
updated mime types. We need to also add them to the support list
in order for assrender to correctly identify them as fonts.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1481>
If there is no jitterbuffer stats we should not attempt to store them in the
global stats structure.
Also add a g_return_if_fail in _gst_structure_take_structure() about this
because it is a programmer error to pass an invalid pointer address there.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1479>
Add gst_dep to gst_rtsp_server_deps, in the context of buildroot, this
will avoid the following build failure, because the correct girdir
location will be retrieved from gstreamer-1.0.pc:
/home/giuliobenetti/autobuild/run/instance-3/output-1/host/riscv32-buildroot-linux-gnu/sysroot/usr/bin/g-ir-compiler gst/rtsp-server/GstRtspServer-1.0.gir --output gst/rtsp-server/GstRtspServer-1.0.typelib --includedir=/usr/share/gir-1.0
Could not find GIR file 'Gst-1.0.gir'; check XDG_DATA_DIRS or use --includedir
error parsing file gst/rtsp-server/GstRtspServer-1.0.gir: Failed to parse included gir Gst-1.0
If the above error message is about missing .so libraries, then setting up GIR_EXTRA_LIBS_PATH in the .mk file should help.
Typically like this: PKG_MAKE_ENV += GIR_EXTRA_LIBS_PATH="$(@D)/.libs"
Fixes:
- http://autobuild.buildroot.org/results/04af6b22cfa0cffb6a3109a3b32b27137ad2e0b0
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1460>
Prior to this patch, we considered that a stream was blocking
whenever a pad probe was triggered for either the RTP pad or
the RTCP pad.
This led to situations where we subsequently unblocked and expected
to find a segment on the RTP pad, which was racy.
Instead, we now only consider that the stream is blocking when
the pad probe for the RTP pad has triggered with a blockable object
(buffer, buffer list, gap event).
The RTCP pad is simply blocked without affecting the state of the
stream otherwise.
Fixes#929
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1452>
Instead of a sequence of if statements, declare a table to map profile
idc with profiles and traverse it.
Also, first add the profile from the parsed profile idc and later add,
into the profile array, the profile from the compatibility flags.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1440>
It's possible a HEVC stream to have multiple profiles given the
compatibility bits. Instead of returning a single profile, internal
gst_h265_profile_tier_level_get_profiles() returns an array with all
it possible profiles.
Profiles are appended into the array only if the generated profile
is not invalid.
gst_h265_profile_tier_level_get_profile() is rewritten in terms of
gst_h265_profile_tier_level_get_profiles(), returning the first
profile found the array.
And gst_h265_get_profile_from_sps() is also rewritten in terms of
gst_h265_profile_tier_level_get_profiles(), but traversing the array
verifying if the proposed profile is actually valid by Annex A.3.x of
the specification.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1440>
BT.2020 color primaries are designed to cover much wider range of
CIE chromaticity than BT.709, and also it's used for both SDR and HDR
contents. So, the incorrect assumption (i.e., BT.709 as a BT.2020)
is risky and resulting image color tends to be visually very wrong.
Unless there's obvious clue, don't consider color space of high resolution
video stream as BT.2020
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1445>
* Add fec / red encoders as direct children of webrtcbin, instead
of providing them to rtpbin through the request-fec-encoder signal.
That is because they need to be placed before the rtpfunnel, which
is placed upstream of rtpbin.
* Update configuration of red decoders to set a list of RED payloads
on them, instead of setting the pt property.
That is because there may be one RED pt per media in the same session.
* Connect to request-fec-decoder-full instead of request-fec-decoder,
in order to instantiate FEC decoders according to the payload type
of the stream.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1429>
When multiple streams are bundled together, there may be more
than one red payload type to handle.
In addition, as the red decoder works by filling in gaps in
the seqnums, there needs to be one rtp_history queue per sequence
domain.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1429>
In redenc, when input buffers have a header for the TWCC extension,
we now add one to our wrapper buffers.
In ulpfecenc we add one in that case to our protection buffers.
This makes TWCC functional when UlpRed is used in webrtcbin.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1414>
In order to allow "level-asymmetry-allowed" we now handle a new
"profile" field, which as the same semantics as the "profile" field in
H.264 stream so that we can force payloaded stream to have the right
format when using the `gst_sdp_media_get_caps_from_media` to set caps
filter after the payloader. This allows a simple negotiation in standard
RTP negotiation based on SDPs (like webrtc) for that particular case,
closely respecting the specs.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1410>
The ["level-asymmetry-allowed"] field states that the peer wants the
profile specified in the "profile-level-id" fields but doesn't care
about the level. To express this in GStreamer caps term, we add a
"profile" field in the caps, which reuses the usual "profile" semantics
for H.264 streams and, and remove "profile-level-id" and
"level-asymmetry-allowed" fields.
["level-asymmetry-allowed"]: https://www.iana.org/assignments/media-types/video/H264
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1410>
We are querying supported swapchain colorspace via
CheckColorSpaceSupport() but it doesn't seem to be reliable.
Use only tested full-range RGB formats which are:
- sRGB
- BT709 primaries with linear RGB
- BT2020 primaries with PQ gamma
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1433>
When using playbin3, it seems that the alpha decode is always first to
push caps and run an allocation query. As the format change from sink
and alpha were not synchronized, the allocation query could endup
being run before the caps are pushed. That may lead to failing query,
which makes the decoder thinks there is no GstVideoMeta downstream and
most likely CPU copy the frame.
This patch implements a format cookie to track and synchronize the
format changes on both pads fixing the racy performance issue.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>
This adds the alignment field to the template caps. Without this field
set, the auto-plugger will see fixed caps and will use
gst_caps_is_subset() against the caps produced by the parser. This is a
challenge for all cases where a parser can do conversion. This is fixed
by adding alignment field, which makes the auto-pluggers do an
intersection of the caps as it gets unfixed caps after intersection now.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1439>
XNextEvent() blocks indefinitely in absence of X11 events, which can
prevent the pipeline from stopping.
This can cause problems when ximagesrc is used in "remote desktop"
scenarios and the GStreamer application itself, through which the user
is viewing and controlling the machine, is the only source of input
events.
Replace the call with non-blocking XCheckTypedEvent().
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1438>
since `gst_caps_replace()` and `gst_pad_set_caps()` both ref the caps and neither of them takes the ownership of the caps -> it must be unreffed in `gst_multi_file_src_set_property()`
to test the leak (on Unix): `echo coucou > /tmp/file.txt && GST_TRACERS=leaks GST_DEBUG="GST_TRACER:7" gst-launch-1.0 multifilesrc location=/tmp/file.txt caps='txt' ! fakesink`
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1436>
When detecting the remote time has been reset which may occur if remote
device providing the clock server has been power reset, then clock is
no longer synced. Setting clock state will trigger a signal to client
informing on sync lost making it possibility to take appropriate action.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/975>
Due to a copy paste bug, the bitdepth was never set and that was leading
to requesting sizeimage of 0. Previously that worked since the driver
would in that case pick a size for us. But now the we bumped the minimum
to 4KB, the driver happily allocate 4KB of bitstream which lead to
decoding error.
As MPEG2 have a fixed bitdeph of 8, use a define instead of the run-time
variable.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1415>
vaapidecode is used in vaapidecodebin and it exposes all the
theoretically supported caps, but that slows down autoplug. With this
autplug is negotiated faster, giving more option to decodebin to select
other decoder.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1405>
The V4L2 uAPI uses pic_num for both PicNum and ShortTermPicNum. It also
doe the same for both FrameNum and LongTermFrameIdx. This change does
not change the fluster score, but fixed a visual corruption noticed
with some third party streams.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1387>
The current code does not set the copied memory correctly when it is popped
from the surface cache pool.
1. We forget to ref the allocator, which causes the allocator to be freed
unexpected, and we get a crash later because of the memory violation.
2. We forget to add ref_mems_count, which causes the surface leak because
the surface can not be pushed back to the cache pool again.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1373>
Set minimum sizeimage such that there is enough space for any overhead
introduced by the codec.
Notably fix a vp9 issue in which a small image would not have a
bitstream buffer large enough to accomodate it.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1012>
Rework gstvp9{decoder|statefulparser} to optionally parse compressed headers.
The information in these headers might be needed for accelerators
downstream, so optionally parse them if downstream requests it.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1012>
Also ignore 0x0 sizes in the fallback case and assume the size can be
anything between 1x1 and MAXxMAX.
This fixes the case where a width=0, height=0 caps are created. Whith
this patch the caps will contain width=[1,MAX], height=[1,MAX].
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1396>
When an extmap is defined twice for the same ID, firefox complains and
errors out (chrome is smart enough to accept strict duplicates).
To work around this, we deduplicate extmap attributes, and also error
out when a different extmap is defined for the same ID.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1383>
There are a lot of info in the mpeg2's sequence(also including ext
display_ext and scalable_ext). We need to notify the subclass about
its change, but not all the changes should trigger a drain(), which
may change the output picture order. For example, the matrix changes
in sequence header does not change the decoder context and so no need
to trigger a drain().
Fixes: #899
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1375>
Follow-up on 97d83056b3, only check
for intersection with the current srccaps when checking if a sinkpad
can accept caps.
I must have been lucky in my firefox testing then, and always entered
the code path with audio getting negotiated first, thus not failing
the is_subset check when srccaps had been negotiated as
application/x-rtp, and an accept-caps query was made for the video
caps with a defined extmap.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1384>
This is a requirement for GstPlayer when using the default overlay interface
provided by the pipeline. The GstPlayerWrappedVideoRenderer requires a valid
pipeline, but that's available only after the GstPlay thread has successfully
started.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1345>
Dynamically registered elements (hardware element in most cases)
may or may not be available on a system and properties may be different
per system.
This new API will make documentation skipping possible in programmable way.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1360>
ShowWindow() could be blocked while doing gst_d3d11_window_win32_unprepare
when external window handle provided to d3d11videosink in multi-threaded
environment.
The condition that issue happened is, UI thread is waiting for a
background thread that changes d3d11videosink state to NULL, and the
background thread would try to send a window message to the queue.
The queue is already occupied by the UI thread, so the background
thread will be blocked.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1366>
Instead of defining a sized array for function signature, use it
unsized (a pointer alias, basically). In this way clang warning is
silenced:
warning: ‘fill_profiles’ accessing 64 bytes in a region of size 12 [-Wstringop-overflow=]
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1357>
If qt asks us to redraw before we have both set a buffer and caps we
would attempt to use the new caps with the old buffer which could result
in bad things happening.
Only update caps from new_caps once the buffer has actually been set.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1346>
Make sure the EGLImage we're rendering to the GL memory stays alive long enough,
until the the GL memory has been destroyed.
This change fixes tearing and black flashes artefacts that were happening
because the EGLImage was sometimes destroyed before the sink actually rendered
the associated texture.
Fixes#889
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1354>
This takes a plain message string and not a format string, and as a
result doesn't have to be passed through vasprintf() and lead to further
unnecessary allocations. It can also contain literal `%` because of
that.
The new function is mostly useful for bindings that would have to pass a
full string to GStreamer anyway and would do formatting themselves with
language-specific functionality.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1356>
There's a potential race condition with this sort of pipelines on
certain systems (depends on the processing load):
GST_DEBUG_DUMP_DOT_DIR=/tmp \
gst-launch-1.0 uridecodebin3 uri=file://stream.mp4 ! glupload ! \
glimagesink --gst-debug=*:4
Right after the pipeline passes from PAUSED to READY, bin_to_dot_file
dumps uridecodebin3 properties, but current uri and suburi might be
already freed, causing a potential use-after-freed.
This patch makes NULL the current item right after all the play items
are freed.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1353>
A previous patch has caused rtpfunnel to output twcc-related
information downstream, however this leaked into upstream
negotiation (through funnel->srccaps), causing payloader to
negotiate twcc caps even when not prompted to do so by the user.
Fix this by only enforcing that upstream sends us application/x-rtp
caps as was the case originally.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1278>
Negative composition time offsets are only allowed with version 1 of the
box, however we parse it as a signed value also for version 0 boxes as
unfortunately there are such files out there and it's unlikely to have
(valid) huge composition offsets.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1294>
Sometimes the resampler has enough space to store all the incoming
samples without outputting anything. When this happens,
gst_audio_resampler_get_out_frames() returns 0.
In that case, the resampler should consume samples and just return.
Otherwise, we get a segfault when gst_audio_resampler_resample() tries
to resample into a NULL 'out' pointer.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1343>
g_sequence_remove_range's end iter is exclusive, so if one
wants to remove that item as well, it should be called with
the next iter.
This could in theory fix an issue where:
* The sequence isn't entirely trimmed, with an old item lingering
* Following FEC packets are immediately discarded because they
arrived later than corresponding media packets, long enough for
seqnums to wrap around
* We now try to reconstruct a media packet with a completely obsolete
FEC packet, chaos ensues.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1341>
We need to hold onto the last buffer until the next buffer arrives.
Before, if a caps change comes we would remove the currently rendering
buffer. if Qt asks use to render something, we would render the dummy
black texture.
Fixes a period of black output when upstream is e.g. changing resolution
as in hls adaptive bitrate scenarios.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1338>
The `gst_v4l2_buffer_pool_dqbuf` function contains this ominous comment:
/* get our GstBuffer with that index from the pool, if the buffer was
* outstanding we have a serious problem.
*/
outbuf = pool->buffers[group->buffer.index];
Unfortunately it is common for buffers in _output_ buffer pools to be
both queued and outstanding at the same time. This can happen if the
upstream element keeps a reference to the buffer, or in an encoder
element itself when it keeps a reference to the input buffer for each
frame.
Since the current code doesn't handle this case properly we can end up
with crashes in other elements such as:
(gst-launch-1.0:32559): CRITICAL **: 17:33:35.740: gst_video_frame_map_id: assertion 'GST_IS_BUFFER (buffer)' failed
and:
(gst-launch-1.0:231): GStreamer-CRITICAL **: 00:16:20.882: write map requested on non-writable buffer
Both these crashes are caused by a race condition related to releasing
the same buffer twice from two different threads. If a buffer is queued
and outstanding this situation is possible:
**Thread 1**
- Calls `gst_buffer_unref` decrementing the reference count to zero.
- The core GstBufferPool object marks the buffer non-outstanding.
- Calls the V4L2 release buffer function.
- If the buffer is _not_ queued:
- Release it back to the free pool (containing non-queued buffers).
**Thread 2**
- Dequeues the queued output buffer.
- Marks the buffer as not queued.
- If the buffer is _not_ outstanding:
- Calls the V4L2 release buffer function.
- Release it back to the free pool (containing non-queued buffers).
If both of these threads run at exactly the same time there is a small
window where the buffer is marked both not outstanding and not queued
but before it has been released. In this case the buffer will be freed
twice causing the above crashes.
Unfortunately the variable recording whether a buffer is outstanding is
part of the core `GstBuffer` object and is managed by `GstBufferPool` so
it's not as straightforward as adding a mutex. Instead we can fix this
by additionally recording the buffer state in `GstV4l2BufferPool`, and
handle "internal" and "external" buffer release separately so we can
detect when a buffer becomes not outstanding.
In the new solution:
- The "external" buffer pool release and the "dqbuf" functions
atomically update the buffer state and determine if a buffer is still
queued or outstanding.
- Subsequent code and a new
`gst_v4l2_buffer_pool_complete_release_buffer` function can proceed to
release (or not) a buffer knowing that it's not racing with another
thread.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1010>
The pipeline flow for receiving looks like this:
rtpsession ! rtpssrcdemux ! session_fec_decoder ! rtpjitterbuffer ! \
rtpptdemux ! stream_fec_decoder ! ...
There are two places where a fec decoder could be placed.
1. As requested from the 'request-fec-decoder' signal: after rtpptdemux
for each ssrc/pt produced
2. after rtpssrcdemux but before rtpjitterbuffer: added for the
rtpst2022-1-fecenc/dec elements,
However, there was some cross-contamination of the elements involved and
the request-fec-decoder signal was also being used to request the fec
decoder for the session_fec_decoder which would then be cached and
re-used for subsequent fec decoder requests. This would cause the same
element to be attempted to be linked to multiple elements in different
places in the pipeline. This would fail and cause all kinds of havoc
usually resulting in a not-linked error being returned upstream and an
error message being posted by the source.
Fix by not using the request-fec-decoder signal for requesting the
session_fec_decoder and instead solely rely on the added properties for
that case.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1300>
The average_period should always represent the time between two
events. The specification defines the event time as the time
between audio samples, video frame sync, video line sync, etc.
In case of one timestamp per PDU the timestamp_interval identifies
the amount of events between the timestamp of one PDU and the
timestamp of the next PDU.
As described in IEEE 1722-2016 chapter
"10.4.12 timestamp_interval field" timestamp_interval shall be
nonzero.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1076>
The failure conditions can be overidden by subclasses, and a boolean
return value is provided to the caller whether adding/removing the child
element has actually worked. The caller can then handle this
accordingly but flooding stderr with this is not very useful.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1320>
Upstream caps might for example be
application/x-rtp,media=audio,encoding-name={OPUS, X-GST-OPUS-DRAFT-SPITTKA-00, multiopus}
and while that is not fixed caps it is enough to match it with a media.
Only caps structures that have the correct structure name and that have
the media and encoding-name field are preserved, but if both are present
then these caps are used as "codec preferences".
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1291>
Currently reading extension relies on the fact that everything after the
last"." character is a file extension. Whereas that works fine for most
of the cases, it breaks when the URI contains a query part.
E.g.: `http://url.com/file.mp4?param=value` returns `mp4?param=value`
instead of `mp4`.
In this commit we use URI parser to read the path of the URI (in the example
above, that is `/file.mp4`) and read extension from that path.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1305>
Some decoding APIs support delayed output for performance reasons.
One example would be to request decoding for multiple frames and
then query for the oldest frame in the output queue.
This also increases throughput for transcoding and improves seek
performance when supported by the underlying backend.
Introduce support in the mpeg2 base class, so that backends that
support render delays can actually implement it.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1013>
Downstream might need the start code offset when decoding.
Previously this computation would be scattered in multiple sites. This
is error prone, so move it to the base class. Subclasses can access
slice->sc_offset directly without computing the address themselves
knowing that the size will also take the start code into account.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1013>
The GstV4l2CodecAllocator dispose function clears `self->decoder` but
the finalize function then tries to use it if the allocator has no been
detached yet.
Fix by detaching in the dispose function before we clear
`self->decoder`.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1220>
Standard interlace handling:
* If we have interlace-mode=interleaved and the field order, we just
set it when creating the session
* If we have interlace-mode=(interleaved|mixed) and no field order, we
set the field order on the first buffer
The encoder session does not support changing the FieldDetail after it
has started encoding frames, so we cannot support mixed streams
correctly.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1214>
Delay decoders downstream negotiation just before an output frame
needs to be allocated.
This is required, are least for H.264 and H.265 decoders, since
codec_data might trigger a new sequence before finishing upstream
negotiation, and sink pad caps need to set before setting source pad
caps, particularly to forward HDR fields. The other decoders are
changed too in order to keep the same structure among them.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1257>
Using GstBaseDec hack to access the parent_object of each element in
the element itself is a bit fragile. It would be better to keep its
own parent object as the usual global variable. It would make it
resistant to code changes.
The GstBaseDec macro to access the parent object now it's internal to
base decoder.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1257>
We were checking if the start time of the gap event was
GST_CLOCK_TIME_NONE, which is superfluous because that cannot happen,
and then not checking if it was NONE after gst_segment_to_running_time,
which caused a crash if an identity received a gap event fully or
partially outside the current segment.
This patch was done in cooperation with:
Jan Alexander Steffens (heftig) <jan.steffens@ltnglobal.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1269>
.. if a current direction has already been set
When `webrtcbin` has created an offer based on codec_preferences,
it might not have received caps on its sinkpads by the time a
remote description is set, in which case we want to connect the
input stream upon actual reception of the caps instead.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1233>
With mpeg4videoparse drop=false config-interval=N|-1 we might be
trying to insert a config before we have actually received one,
in which case we'll try to map a NULL buffer which will generate
lots of criticals.
Fixes#855
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1265>
gst_va_fixate_format() will iterate all othercaps' structures to find
the one with less information lost at color conversion. If a structure
with same color format is found, the iteration stops. It's like a
smart truncation. Then, this function also will choose the caps
feature.
Later this structure is used fixate its size and no further truncation
is needed.
Don't intersect at fixate, since it kills possible resizing.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1261>
If the pad does not have a current caps, get_pad() returns the query
caps which can be ANY. In such case the caps does not have any structure
resulting in a critical warning when calling gst_caps_get_structure().
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1254>
Detected while reading the code, cccombiner must set
self->current_video_buffer to NULL *after* emitting selected-samples
in order for the application to get a useful return when peeking
the next video sample.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1252>
When schedule is true (as is the case by default), we insert padding
when no caption data is present in the schedule queue, and previously
weren't checking whether the caption pad had gone EOS, leading to
infinite scheduling of padding after EOS on the caption pad.
Rectify that by adding a "drain" parameter to dequeue_caption()
In addition, update the captions_and_eos test to push valid cc_data
in: without this cccombiner was attaching padding buffers it had
generated itself, and with that patch would now stop attaching
said padding to the second buffer. By pushing valid, non-padding
cc_data we ensure a caption buffer is indeed attached to the first
and second video buffers.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1252>
Also refactor various internals of the monitor code:
- Don't allow starting twice but just return directly when starting a
second time.
- Don't end up in an inconsistent state if call start() a second time
while the monitor is starting up.
- Remove complicated cookie code: it was not possible to add/remove
filters while the monitor was started anyway so this was only useful
in the very small time-window while starting the monitor or while
getting the devices. Instead disallow adding/removing filters while
the monitor is starting, and when getting devices work on a snapshot
of providers/filters.
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/667
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1189>
At gst_va_dmabuf_allocator_setup_buffer_full, static code analysis tool
does not know number of objects in descriptor is always larger than 0 if
export_surface_to_dmabuf succeeds. Thus, the tool will assume buf is
allocated with mem but not released when desc.num_objects equals to 0
and raise a mem leak issue.
For gst_va_dambuf_memories_setup, we should also inform the tool that
n_planes will be larger than 0 by checking the value at very beginning.
Then, the defect similar to above will not be raised during static analysis.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1241>
gst_base_parse_reset() does not reset data_bytecount to 0, so
gst_base_parse_update_bitrates() uses a wrong value to calculate
the average bitrate on subsequent pipeline starts. This leads to an
excessive amount of "tag" events being pushed. These events include
very high "bitrate" values that diminish over time, and are produced
until the average bitrate is back to sane values.
Fixes#840
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1228>
This will only affect individual/tarball module builds, as the
options yield to the parent project which was set to gpl=disabled
by default already. We kept it as auto in the original commit
to accommodate the need to update cerbero as well, which had to
be done separately after the initial commit.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1217>
Previously gst_structure_has_name was used to get a string to compare with supported mimetypes.
This is incorrect as above function returns a user defined structure name which is
not the structure mimetype value.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1206>
