Font sizes are relative to the subtitle frame dimensions. If the
expected frame dimensions are not known, the font sizes will most
likely be incorrect.
Signed-off-by: Philip Langdale <philipl@overt.org>
Style flags were only being turned on. If the default was on and style
record turned off, style flag remained on.
Signed-off-by: Philip Langdale <philipl@overt.org>
It's not necessary to walk the style record list twice per subtitle
character. style records are in order and do not overlap.
Signed-off-by: Philip Langdale <philipl@overt.org>
It represents the relationship between them more naturally and will be
useful in the following commits.
Allows significantly more frames in fate-h264-attachment-631 to be
decoded.
Those SEIs refer to the currently active SPS. However, since the SEI
NALUs precede the coded picture data in the bitstream, the active SPS is
in general not known when we are decoding the SEI.
Therefore, store the content of the picture timing SEIs and actually
parse it when the active SPS is known.
Such errors are not necessarily fatal and decoding might still be
possible, e.g. it happens for MVC streams where we do not handle the
subset SPS thus failing to parse its corresponding PPS.
Currently the frame pool used by the default get_buffer2()
implementation is a single struct, allocated when opening the decoder.
A pointer to it is simply copied to each frame thread and we assume that
no thread attempts to modify it at an unexpected time. This is rather
fragile and potentially dangerous.
With this commit, the frame pool is made refcounted, with the reference
being propagated across threads along with other context variables. The
frame pool is now also immutable - when the stream parameters change we
drop the old reference and create a new one.
Specifically, between the user-facing one and the first frame thread
one.
This is fragile and dangerous, allocate separate private data for each
per-thread context.
The current design, where
- proper init is called for the first per-thread context
- first thread's private data is copied into private data for all the
other threads
- a "fixup" function is called for all the other threads to e.g.
allocate dynamically allocated data
is very fragile and hard to follow, so it is abandoned. Instead, the
same init function is used to init each per-thread context. Where
necessary, AVCodecInternal.is_copy can be used to differentiate between
the first thread and the other ones (e.g. for decoding the extradata
just once).
This patch adds functional changes to support the
PPT marker. This patch fixes bug ticket #4610.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: left shift of negative value -2
Fixes: 20303/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_FLAC_fuzzer-5096829297623040
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 23170 * 95058 cannot be represented in type 'int'
Fixes: 20295/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MJPEG_fuzzer-5800212870463488
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 9223372036854775775 + 128 cannot be represented in type 'long'
Fixes: 20054/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_FFWAVESYNTH_fuzzer-5686385113825280
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
throughout vp3_decode_frame the error code was being captured (ret) but never returned.
Signed-off-by: Peter Ross <pross@xvid.org>
Reviewed-by: Anton Khirnov <anton@khirnov.net>
struct v4l2_selection contains reserved bytes which should be set to
zero before the ioctl call.
Fixes valgrind error:
Syscall param ioctl(VKI_V4L2_S_SELECTION) points to uninitialised byte(s)
Signed-off-by: Andriy Gelman <andriy.gelman@gmail.com>
Currently the driver's frame period is incorrectly set to the frame
rate. This is fixed in the commit.
Signed-off-by: Ming Qian <ming.qian@nxp.com>
Signed-off-by: Andriy Gelman <andriy.gelman@gmail.com>
Fixes: out of array read
Fixes: 21286/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_HCA_fuzzer-5683183715876864
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Up until now, it was completely unspecified what the content of the
destination packet dst was on error. Depending upon where the error
happened calling av_packet_unref() on dst might be dangerous.
This commit changes this by making sure that dst is blank on error, so
unreferencing it again is safe (and still pointless). This behaviour is
documented.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
av_packet_ref() mostly treated the destination packet dst as uninitialized,
i.e. the destination fields were simply overwritten. But if the source
packet was not reference-counted, dst->buf was treated as if it pointed
to an already allocated buffer (if != NULL) to be reallocated to the
desired size.
The documentation did not explicitly state whether the dst will be treated
as uninitialized, but it stated that if the source packet is not refcounted,
a new buffer in dst will be allocated. This and the fact that the side-data
as well as the codepath taken in case src is refcounted always treated the
packet as uninitialized means that dst should always be treated as
uninitialized for the sake of consistency. And this behaviour has been
explicitly documented.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Validate the value of ST field in the TLM marker of JPEG2000.
Throw an error when ST takes value of 0b11.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Resolution/format changes lead to re-initialization of hardware
accelerations(vaapi/dxva2/..) with new hwaccel_priv_data in
the worker-thread. But hwaccel_priv_data in user context won't
be updated until the resolution changing frame is output.
A termination with "-vframes" just after the reinit will lead to:
1. memory leak in worker-thread.
2. double free in user-thread.
Update user context in ff_frame_thread_free with the last thread
submit_packet() was called on.
To reproduce:
ffmpeg -hwaccel vaapi(dxva2) -v verbose -i
fate-suite/h264/reinit-large_420_8-to-small_420_8.h264 -pix_fmt nv12
-f rawvideo -vsync passthrough -vframes 47 -y out.yuv
Signed-off-by: Linjie Fu <linjie.fu@intel.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
The comments for some of the markers were incorrect.
This patch fixes the comments associated with the markers.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
This commit fixes the names and constants of the reserved NAL units
with nal_unit_type 22 resp. 23. They were "IRAP_IRAP_VLC2x", but are
actually "RSV_IRAP_VLC2x".
This also required a change to cbs_h265_syntax_template.c.
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Surprisingly neither GCC nor Clang did this transformation on their own.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Up until now, the name of every AVBSFContext for logging purposes was
"AVBSFContext", so that the default logging callback produced output
like "[AVBSFContext @ 0x55813bae92c0] Extradata". This has been changed
to "[trace_headers @ 0x60a000000700] Extradata" by adding an item_name-
function to the AVClass for bitstream filters.
Furthermore, the correct category has been set so that the introductory
part before the actual message (everything before "Extradata" in the
above examples) are displayed in a different colour than the rest.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: James Almer <jamrial@gmail.com>
This happened for AVBSFContext.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: James Almer <jamrial@gmail.com>
This fixes mpeg2video stream copies to mpeg muxer like this:
ffmpeg -i xdcamhd.mxf -c:v copy output.mpg
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Fixes: signed integer overflow: 685813396 + 1803454769 cannot be represented in type 'int'
Fixes: 21073/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_SIREN_fuzzer-5744900508483584
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 1302123111085380114 - -8319005078741256972 cannot be represented in type 'long'
Fixes: 20991/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_FFWAVESYNTH_fuzzer-5148554161291264
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
as well as includes of libavutil/timer.h.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
This version is able to output multiple coefficients at a time and
is able to altogether remove actual golomb code parsing.
Its also able to partially recover the last coefficient in case
the packet is incomplete.
Total decoder performance gain for 8bit 420 1080p lossless: 40%.
Total decoder performance gain for 10bit 420 1080p lossless: 40%.
clang was able to vectorize the loop much better than
my handwritten assembly, but gcc was very naive and didn't.
Lookup table is a rewritten version of vc2hqdecode.
