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/3181>
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/3174>
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/3170>
Due to some interaction between the latest meson version and the
subproject cache on the image, the build fails:
File "/usr/local/lib/python3.7/site-packages/mesonbuild/mesonmain.py", line 153, in run
return options.run_func(options)
File "/usr/local/lib/python3.7/site-packages/mesonbuild/msubprojects.py", line 679, in run
results = loop.run_until_complete(asyncio.gather(*tasks))
File "/usr/lib64/python3.7/asyncio/base_events.py", line 587, in run_until_complete
return future.result()
File "/usr/lib64/python3.7/concurrent/futures/thread.py", line 57, in run
result = self.fn(*self.args, **self.kwargs)
File "/usr/local/lib/python3.7/site-packages/mesonbuild/msubprojects.py", line 126, in run
result = self.run_method()
File "/usr/local/lib/python3.7/site-packages/mesonbuild/msubprojects.py", line 411, in update
self.wrap.update_hash_cache(self.wrap_resolver.dirname)
File "/usr/local/lib/python3.7/site-packages/mesonbuild/wrap/wrap.py", line 228, in update_hash_cache
with open(self.get_hashfile(subproject_directory), 'w', encoding='utf-8') as file:
FileNotFoundError: [Errno 2] No such file or directory: './subprojects/pango-1.48.11/.meson-subproject-wrap-hash.txt'
https://gitlab.freedesktop.org/tpm/gstreamer/-/jobs/29467932
The error is in a different wrap each time. Restricting the meson
version to an older one fixes it, which is fine for a stable branch.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3139>
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/3113>
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/3105>
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/3095>
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/3090>
Since commit a79a756b79 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/3089>
Check that `self` and `self->callback` are defined. `self` can be set to
`NULL` in `remove_listener`, and `self->callback` can be set to `NULL`
inside `gst_amc_surface_texture_jni_set_on_frame_available_callback`.
This can cause a segfault since the Java object can outlive the C
object, and call the callback after `remove_listener` is called.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3056>
Check back pressure of a stream transport before popping buffer from its backlog.
If the stream transport is not experiencing back pressure, the buffer can be popped from backlog and pushed to client.
Fixes:#1298
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2998>
Timers for RTX packets are dealt with later in update_rtx_timers(), and
timers for non-RTX packets would potentially also be unscheduled a
second time from there so avoid that.
Also don't shadow the timer variable from the outer scope but instead
make use of it directly.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2996>
The address/port is pre-defined by the caller of the function, so
retrying is only going to loop forever.
Ideally the multicast address should be checked after allocating but
this doesn't happen currently, so it's better to error out cleanly then
to loop forever trying the same address.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2997>
If read_one or write_one was called but the stream closed before it could
read/write a whole packet, read_one/write_one would hang indefinitely,
consuming 100% CPU. This commit fixes that by treating a short read/write
as an error.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2994>
