The logic did not follow the documented behaviour and that caused skipping of
some audio in the loop and in the leftover buffer.
Example command line which should produce a smooth sine wave for the whole
duration of the output:
ffmpeg -f lavfi -i "sine=r=48000:f=480:d=4" -af "aloop=loop=4:start=48000:size=48000" out.wav
Fixes ticket #11283.
Signed-off-by: Marton Balint <cus@passwd.hu>
(cherry picked from commit fe18ed3f2a)
If the audio loop stops inside an audio frame, the leftover buffer contains the
end of the frame, which is not looped. The length supposed to be the part which
was not written to the loop buffer, so we need to drain exactly that number of
bytes from the leftover buffer.
Signed-off-by: Marton Balint <cus@passwd.hu>
(cherry picked from commit b33a594160)
realign_frame called av_pix_fmt_count_planes with incorrect parameter.
Signed-off-by: James Almer <jamrial@gmail.com>
(cherry picked from commit 00cf3df03f)
And not just the linesizes. Use the extra align bytes allocated for this purpose.
Signed-off-by: James Almer <jamrial@gmail.com>
(cherry picked from commit f15fc27db5)
Note1: when the EOF pts is not accurate enough, the last frame
can be dropped by vf_fps with default rounding.
Note2: vf_scale use framesync since e82a3997cd,
so this is a very commonplace scenario.
For example:
./ffprobe -f lavfi testsrc=d=1,scale,fps -of flat \
-count_frames -show_entries stream=nb_read_frames
Before:
streams.stream.0.nb_read_frames="24"
After:
streams.stream.0.nb_read_frames="25"
Signed-off-by: Anton Khirnov <anton@khirnov.net>
(cherry picked from commit de976eaf30)
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Push descriptors are in theory slightly faster, but come with
limitations for which we have to check.
Either way, they're not difficult to implement, so even though
no one should be using peasant-tier descriptors, do it anyway.
This permits:
- The use of Vulkan filtering on many more devices
- Better debugging due to lack of descriptor buffer support in layers
Much of the changes here are due to a requirement that updates to
descriptors must happen between the command buffer being waited on,
and the pipeline not being bound.
We routinely did it the other way around, by updating only after
we bind the pipeline.
The channelmap_init function was returning success even on error after
7dc81d33c2 due to shadowing of the
outer ret variable.
Fixes CID1619297 Logically dead code
Ironically, despite being introduced to make YUVJ unnecessary, the new
YUV negotiation logic failed to actually negotiate YUVJ formats
themselves correctly, leading to errors when passing YUVJ frames into
a filter graph. (They were effectively treated like RGB or Grayscale
formats, rather than as forced-full-range YUV, and hence did not have
their colorspace matrix correctly negotiated)
Fix this by splitting off the YUVJ check from ff_fmt_is_regular_yuv().
Obviously, we can trivially undo this change again once YUVJ is actually
deleted from the codebase.
Fixes: #11179
Should fix memory leaks show in fate-filter-join and fate-filter-crazychannels.
Reviewed-by: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
Signed-off-by: James Almer <jamrial@gmail.com>
