A race condition can occur in `srtpdec` during the READY -> NULL transition:
an RTCP buffer can make its way to `gst_srtp_dec_chain` while the element is
partially stopped, resulting in the following critical warning:
> Got data flow before segment event
The problematic sequence is the following:
1. An RTCP buffer is being handled by the chain function for the
`rtcp_sinkpad`. Since, this is the first buffer, we try pushing the sticky
events to `rtcp_srcpad`.
2. At the same moment, the element is being transitioned from PAUSED to READY.
3. While checking and pushing the sticky events for `rtcp_srcpad`, we reach the
Segment event. For this, we try to get it from the "otherpad", in this case
`rtp_srcpad`. In the problematic case, `rtp_srcpad` has already been
deactivated so its sticky events have been cleared. We won't be pushing any
Segment event to `rtcp_srcpad`.
4. We return to the chain function for `rtcp_sinkpad` and try pushing the
buffer to `rtcp_srcpad` for which deactivation hasn't started yet, hence the
"Got data flow before segment event".
This commit:
- Adds a boolean return value to `gst_srtp_dec_push_early_events`: in case the
Segment event can't be retrieved, `gst_srtp_dec_chain` can return an error
instead of calling `gst_pad_push`.
- Replaces the obsolete `gst_pad_set_caps` with `gst_pad_push_event`. The
additional preconditions checked by previous function are guaranteed here
since we push a fixed Caps which was built in the same function.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4860>
glfilter will unref input buffer after _transform() call immidiately,
but gpu may still reading input buffer for rendering because gl
api is executed async. Need hold reference for input buffer by
adding parent meta to output buffer.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4846>
Appsink will unref prev sample in dispose function. Which is later
when V4L2 video decoder link with appsink as V4L2 video decoder
will close V4L2 device fd during GST_STATE_CHANGE_READY_TO_NULL.
If the video buffer return to V4L2 video decoder after the decoder
closed V4L2 device fd, V4L2 can't release the video frame buffer
which allocated with MMAP mode as application can't call
VIDIOC_REQBUFS 0 to release the video frame buffer by V4L2 driver.
The memory of the video frame will leak.
Unref the gstbuffer in stop() function, so V4L2 video decoder
can received all video frame buffers and release it before close
V4L2 device fd.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4820>
The qt5 and qt6 plugins will now correctly error out if you enable the
option, and you can also now explicitly ensure that wayland, x11,
eglfs support is actually functional by enabling the options. It was
too easy to build non-functional support for these.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4776>
Without this, the plugin cannot be loaded in a devenv because the
RPATH is not added to the plugin dylib. This RPATH will be stripped on
install, which is what we want.
When deploying apps, people are supposed to use `macdeployqt` to
create an AppBundle that bundles Qt for you and sets the RPATHs
correctly to point to that bundled Qt.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4776>
Configuration of our debugging system is possible before init, and in
fact is necessary too, otherwise the settings won't apply to logging
that happens during init.
For instance, since you cannot register a log function before you call
init in python, there is no way for you to log errors during init to
whatever logging service your app uses.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4763>
Due to the alpha value being inserted with _BEFORE, we were ending up
with ARGB instead of RGBA, thus displaying completely wrong colours.
According to libpng's manual, "to add an opaque alpha channel, use filler=0xff
or 0xffff and PNG_FILLER_AFTER which will generate RGBA pixels".
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4759>
self->eos was never reset after streamsynchronizer has sent EOS
(except on explicit flush or switching back to PAUSED).
As a result, synchronization was broken if new streams were pushed later
as gst_stream_synchronizer_wait() does not wait if self->eos is set.
Fix this by reseting self->eos on STREAM_START as that means a new
stream is being sent upstream and so a new EOS will follow later on.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4755>
The `switch (n_rear)` supports up to 5 rear channels, but our channel
set only had space for 3. Size the set properly to fix this.
This didn't actually cause any memory unsafety as `PUSH_CHAN` would stop
incrementing `n_rear` if the channel set is already full.
Thanks to @alatiera for noticing this.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4740>
Since c0bf793c05 ("flvmux: Set PTS based on
running time") the timestamp of the output buffer is already in running
time. So using that for 'srcpad->segment.position' does not work correctly
because gst_aggregator_simple_get_next_time() will convert it again with
gst_segment_to_running_time().
This means that the timestamp returned by
gst_aggregator_simple_get_next_time() may be incorrect. For example, if
flvmux is added to a already runinng pipeline then the timestamp is too
small and gst_aggregator_wait_and_check() returns immediately. As a result,
buffers may be muxed in the wrong order.
To fix this, use the PTS of the incoming buffer instead of the outgoing
buffer. Also add the duration as get_next_time() is supposed to return the
timestamp of the next buffer, not the current one.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4734>
There are broken(?) mjpeg videos that are incorrectly detected as
interlaced. This happens because 'info.height > height' (e.g. 1088 > 1080).
In the interlaced case info.height is approximately 'height * 2' but not
exactly because height is a multiple of DCTSIZE. Make the check more
restrictive but take the rounding effect into account.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4717>
For interlaced jpeg, gst_jpeg_dec_decode_direct() is called twice, once for each
field. In this case, stride[n] is plane_stride[n] * 2 to ensure that only every
other line is written. So the loop must stop at height / num_fields.
If the frame is really interlaced then continuing beyound this, is not harmful,
because jpeg_read_raw_data() will do nothing and return 0, so am info message is
printed.
However, if the frame is not actually interlaced, just misdetected as interlaced
then there is still data available from the second half of the frame. Now
line[0][j] is set to the scratch buffer. If the scratch buffer is not allocated
(because the height is a multiple of v_samp[0] * DCTSIZE) then the result is a
segfault due to a null-pointer dereference.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4717>
When the alignment contains nothing, all its fields are 0 and always
can be satisfied. So there is no need to validate it in this case.
And there are a lot of places just setting this alignment to default
all zero value, this validation generates lots of warnings.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4704>
Add d3d11 conversion path to make gst_video_convert_sample() work
for GstD3D11Memory.
Note that just adding "d3d11download" to the exisitng code is
suboptimal from GstD3D11 point of view because:
* d3d11convert element can support crop/colorspace-conversion/scale
all at once while existing software pipeline needs intermediate steps
for the conversion
* "Process everything on GPU then download it to CPU memory" would be likely
faster than "download GPU memory to CPU then processing it on CPU"
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4689>
ptpd is defaulting to the hybrid mode, and was sending invalid multicast
PTP messages in that configuration until ce96c742a88792a8d92deebaf03927e1b367f4a9.
While this commit was made in 2015 there was no release in the meantime.
Work around this by detecting this case and defaulting to the default
values for the given intervals as given by the PTP standard.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4683>
Drivers may signal end of sequence using an empty buffer and LAST buffer
set, or just an empty buffer on certain legacy implementation. When this
occured, we'd send GST_V4L2_FLOW_LAST_BUFFER were the code expected
GST_FLOW_EOS. Stop abusing GST_FLOW_EOS and port all the code to the new
GST_V4L2_FLOW_LAST_BUFFER.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4671>
when play rtsp stream with playbin3 enabled, there are some critical logs:
g_object_get_is_valid_property: object class 'GstPlayBin3' has no property named 'n-video'
g_object_get_is_valid_property: object class 'GstPlayBin3' has no property named 'n-audio'
g_object_get_is_valid_property: object class 'GstPlayBin3' has no property named 'n-text'
self->collection could be NULL when READY->PAUSED if the pipeline
is live, then it will fallback to query playbin2's property,
we can call gst_play_streams_info_create_from_collection
directly, it will check self->collection internal.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4666>
Make splitmuxsrc deal better with stream reordering by
making the largest observed PTS contiguous in the
next fragment. Previously, it selected DTS, but then
aligned that with the segment start of the next fragment,
which holds PTS values - leading to glitches in
streams that don't have PTS = DTS at the start.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4660>
When reconfigure_output_stream entry missing decoder path,
requested_selection should been update with what is really
active/selected immdiately with SELECTION_LOCK hold. So
use an optional message return from reconfigure_output_stream
and post it after release SELECTION_LOCK. This can make sure
other thread call to check_slot_reconfiguration will got
a correct requested_selection.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4656>
The index is already incremented by 3 every iteration so multiplying it
by 3 additionally on each array access is doing it twice and does not
work.
This caused invalid files to be created if there's more than one CEA608
triplet in a buffer, and out of bounds memory reads.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4645>
When transitioning from state PAUSED to READY, the sctpenc element
could previously be stuck in an endless loop trying to resend data
in case the underlying sctp stream was in the process of
resetting. usrsctp_sendv() would repeatedly return EAGAIN with the
result that 0 bytes were sent and then sctpenc would retry forever.
To bring sctpenc out of the resend loop we just need to inform the
sink pad that it is flushing, which is already done for the associated
data queue, but we also need to set the bools associated with the
sinkpads that are used as the loop criterion.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4636>
Current implementation can in some cases detect
that all data is sent but in reality it is not,
leading to a push to an unlinked pad.
This is a race between the probe used to track data sent and a
call to close.
This patch sends an EOS before starting the close procedure
and then waits for the EOS event to come through to the
src pad before commencing with tear down.
This ensures that any queued data before EOS is flushed.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4633>
Invoking gst_osx_video_sink_osxwindow_destroy() can currently cause a deadlock
because showFrame() keeps trying to get the same lock as well. Moving the lock
closer to where it's actually needed seems to be enough to fix the issue for now.
Reported-by: Alexande B <abobrikovich@gmail.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4627>
This is a fix for a data race leading to:
> GLib-CRITICAL: g_hash_table_foreach:
> assertion 'version == hash_table->version' failed
Identified sequence:
* `rtp_session_on_timeout` acquires the lock on `session` and proceeds with its
processing.
* `rtp_session_process_rtcp` is called (debug log : received RTCP packet) and
attempts to acquire the lock on `session`, which is still held by
`rtp_session_on_timeout`.
* as part of an hash table iterator, `rtp_session_on_timeout` transitively
invokes `source_caps` which releases the lock on `session` so as to call
`session->callbacks.caps`.
* Since `rtp_session_process_rtcp` was waiting for the lock to be released, it
succeeds in acquiring it and proceeds with `rtp_session_process_rr` which
transitively calls `g_hash_table_insert` via `add_source`.
* After `source_caps` re-acquires the lock and gives the control flow back to
`rtp_session_on_timeout`, the hash table iterator is changed, resulting in the
assertion failure.
This commits copies `sess->ssrcs[sess->mask_idx]` and iterates on the copy so
the iterator is not affected by a concurrent change due to the lock being
released in the `source_caps` callback.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4585>
On very quick start/stop, the mainloop may never be run. As a side
effect, our idle stop function is not really being ran, so we can't rely
on that to free the main loop. Simply unref the mainloop when the
thread have completely stop.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4539>
If we don't do that, clients can rely on this signal to see the final pad
topology but it won't be the real one as some of them will disappear after
emitting that signal. This can happen after injecting a different init segment.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4557>
Since c2f890ab, element properties are gathered from the parse-launch
line and passed at object construction.
This caused the following issue to happen in videoflip:
* videoflip installed a CONSTRUCT property named method, now deprecated
* videoflip now also overrides that property with a video-direction
property
GObject construction causes method to be set first at construct time,
with the user-provided value, then video-direction with the default
value.
The user-provided value was thus overridden, causing a regression.
Fix by not installing the properties as CONSTRUCT, and explicitly
implementing constructed() instead in order to ensure that we do still
call gst_video_flip_set_method() at least once during construction.
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/2529
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4551>
A blocking pad probe is added on new sink pads, it's usually removed after the
caps have been negotiated or the signaling state switched to stable, but if that
never happens and the pad is released we kept the pad probe active, leaving the
pad blocked, preventing clean disposal.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4533>
While testing the [implementation for insertable streams] in `webrtcsink` &
`webrtcsrc`, I encountered critical warnings, which turned out to result from
two race conditions in `rtpsession`. Both race conditions produce:
> GLib-CRITICAL: g_hash_table_foreach:
> assertion 'version == hash_table->version' failed
This commit fixes one of the race conditions observed.
In its simplest form, the test consists in 2 pipelines and a Signalling server:
* pipelines_sink: audiotestsrc ! webrtcsink
* pipelines_src: webrtcsrc ! appsrc
1. Set `pipelines_sink` to `Playing`.
2. The Signalling server delivers the `producer_id`.
3. Initialize `pipelines_src` to establish a session with `producer_id`.
4. Set `pipelines_src` to `Playing`.
5. Wait for a buffer to be received by the `appsrc`.
6. Set `pipelines_src` to `Null`.
7. Set `pipelines_sink` to `Null`.
The race condition happens in the following sequence:
* `webrtcsink` runs a task to periodically retrieve statistics from `webrtcbin`.
This transitively ends up executing `rtp_session_create_stats`.
* `pipelines_sink` is set to `Null`.
* In `Paused` to `Ready`, `gst_rtp_session_change_state()` calls
`rtp_session_reset()`.
* The assertion failure occurs when `rtp_session_reset` is called while
`rtp_session_create_stats` is executing.
This is because `rtp_session_create_stats` acquires the lock on `session` prior
to calling `g_hash_table_foreach`, but `rtp_session_reset` doesn't acquire the
lock before calling `g_hash_table_remove_all`.
Acquiring the lock in `rtp_session_reset` fixes the issue.
[implementing insertable streams support]: https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1176
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4532>
Unfortunately streamoff does not flush the events, and this can cause all
sort of issues. Flush events on capture queue. We also return
GST_V4L2_FLOW_RESOLUTION_CHANGE in case a resolution change was seen.
This allow skipping streamon(capture) on flush, which could lead to a
configuration miss-match, or failure if the buffers aren't of the right
size.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4525>
