avcodec/rscc: Avoid returning frames that have nearly no undamaged pixels in them

Fixes: Timeout
Fixes: 12192/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_RSCC_fuzzer-6279038004363264
Before: clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_RSCC_fuzzer-6279038004363264 in 15423 ms
After:  clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_RSCC_fuzzer-6279038004363264 in 190 ms

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec: Add discard_sample_percentage

Suggested-by: BBB
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

tools/target_dec_fate.list: add entries upto 1214

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/huffyuvdec: Check slice_offset/size

Fixes: out of array access
Fixes: 12447/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_HYMT_fuzzer-5201623956062208
Fixes: 12458/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_HYMT_fuzzer-5705567736168448
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/ilbcdec: Fix undefined integer overflow lsf2poly()

The addition is moved up into the context where the variable is unsigned avoiding
the undefined behavior

Fixes: runtime error: signed integer overflow: 2147481972 + 4096 cannot be represented in type 'int'
Fixes: 12444/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ILBC_fuzzer-5755706244857856
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/ilbcdec: Fix integer overflow in construct_vector()

webrtc contains explicit code to ignore the undefined behavior (RTC_NO_SANITIZE / OverflowingAddS32S32ToS32())

Probably fixes: Integer overflow (unreproducable here)
Probably fixes: 12215/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ILBC_fuzzer-5767142427852800

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

MAINTAINERS: add AVS2 section

Signed-off-by: hwren <hwrenx@126.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

vaapi_encode_h265: Ensure that ref pics are always in the RPS

When making a new P-frame when B-frames are present the previous P-frame
is normally in the DPB because it will be referred to by subsequent
B-frames.  However, this is not true if there are no B-frames, or in edge
cases where a GOP ends with two P-frames.  Fix this by adding the direct
ref pics to the RPS explicitly.

Fixes #7699.

Tested-by: Ullysses A Eoff <ullysses.a.eoff@intel.com>

lavf/subviewerdec: Skip leading BOM.

Fixes ticket #7661.

avfilter/buffersrc: print relevant info when skipping filter reinit

The timestamp of the changed input frame as well as its relevant
properties can be examined by the user. Only applicable when
reinit_filter is disabled on the input stream.

lavf/cafdec: Do not fail for unknown atoms with negative size.

The specification requires the demuxer to only read the data
atom up to its given size, this is necessary as atoms are
allowed at the end of the file.
This patch duplicates the behaviour of the QuickTime player.

avcodec: add ARBC decoder

Thanks Kostya for great help in reversing binary.

avcodec/vp6: select idct based (loosely) on number of coefficients decoded

The VP3/4/5/6 reference decoders all use three IDCT versions: one for the
DC-only case, another for blocks with more than 10 coefficients, and an
optimised one for blocks with up to 10 AC coefficents. VP6 relies on the
sparse 10 coefficient version, and without it, IDCT drift occurs.

Fixes: https://trac.ffmpeg.org/ticket/1282
Signed-off-by: Peter Ross <pross@xvid.org>

avcodec/vp6: use ff_vp3dsp_[hv]_loop_filter_12

Partially fixes: https://trac.ffmpeg.org/ticket/1282

Signed-off-by: Peter Ross <pross@xvid.org>

avcodec/vp6: use rounded shift for chroma motion vector calculation

Partially fixes: https://trac.ffmpeg.org/ticket/1282

Signed-off-by: Peter Ross <pross@xvid.org>

avcodec/vp3dsp: add 10 coefficient version of the VP3 IDCT

This version of the IDCT is used by the VP6 decoder.

Signed-off-by: Peter Ross <pross@xvid.org>

avcodec/vp3dsp: add 12 pixel loop filter functions

Signed-off-by: Peter Ross <pross@xvid.org>

avcodec/vp3dsp: move vp3 init loop filter function to vp3dsp

This is also used by the VP6 decoder.

Signed-off-by: Peter Ross <pross@xvid.org>

lavf/vaapi_deinterlace: return error if mode unsupported

Reviewed-by: Mark Thompson <sw@jkqxz.net>
Signed-off-by: Fuwei Tang <fuweix.tang@intel.com>
Signed-off-by: Zhong Li <zhong.li@intel.com>

lavc/qsvenc: enable QVBR mode

QVBR mode is to use the variable bitrate control algorithm
with constant quality.
mfxExtCodingOption3 should be supported to enable QVBR mode.

It is neccesary to specify a max_rate for QVBR, else it may be ICQ mode.
Example usage: ffmpeg -hwaccel qsv -c:v h264_qsv -i input.mp4 -c:v
h264_qsv -global_quality 25 -maxrate 2M test_qvbr.mp4 -v verbose

Clip QVBR quality range to be [0, 51] as Mark's commments.
It is similar to qp range of CQP but possibly should be updated when VP8/VP9
encoding can be supported.

Reviewed-by: Mark Thompson <sw@jkqxz.net>
Signed-off-by: Zhong Li <zhong.li@intel.com>

avfilter/afade+acrossfade: allow skipping fade on inputs

New fade curve value 'nofade' passes audio samples as-is.
Primarily useful in carrying out acrossfade without fades.

avcodec/libx265: add support for ROI-based encoding

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

lavf/supenc: Remove some unneeded casts.

lavf/movenc: Mention video_track_timescale as option to fix the timescale.

Smarter improvements for this error message were suggested in the past
but this is certainly an improvement.

libavfilter: add transpose_vaapi filter

Swap width and height when do clock/cclock rotation
Add reversal/hflip/vflip options

ffmpeg -hwaccel vaapi -vaapi_device /dev/dri/renderD128
-hwaccel_output_format vaapi -i input.264 -vf "transpose_vaapi=clock_flip"
-c:v h264_vaapi output.h264

Signed-off-by: Zachary Zhou <zachary.zhou@intel.com>
Signed-off-by: Mark Thompson <sw@jkqxz.net>

vaapi_encode_vp9: Support more complex reference structures

vaapi_encode_h265: Support more complex reference structures

The reference picture sets are now constructed directly from the DPB
information.

vaapi_encode_h264: Support more complex reference structures

vaapi_encode: Let the reconstructed frame pool be sized dynamically

No supported encode driver requires the pool to be fixed-size, so just
remove this constraint.

vaapi_encode: Convert to send/receive API

This attaches the logic of picking the mode of for the next picture to
the output, which simplifies some choices by removing the concept of
the picture for which input is not yet available.  At the same time,
we allow more complex reference structures and track more reference
metadata (particularly the contents of the DPB) for use in the
codec-specific code.

It also adds flags to explicitly track the available features of the
different codecs.  The new structure also allows open-GOP support, so
that is now available for codecs which can do it.

vaapi_encode: Allocate picture-private data in generic code

cbs_h2645: Avoid memcpy when splitting fragment #2

Now memcpy can be avoided for NAL units containing escapes, too.

Particularly improves performance for files with hardcoded black bars.
For such a file, time spent in cbs_h2645_split_fragment went down from
369410 decicycles to 327677 decicycles. (It were 379114 decicycles when
every NAL unit was copied.)

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@googlemail.com>

h2645_parse: Make ff_h2645_packet_split reference-compatible

This is in preparation for a patch for cbs_h2645. Now the packet's
rbsp_buffer can be owned by an AVBuffer.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@googlemail.com>

lavc/g723_1enc: Set the default bitrate to 6300.

lavf/hlsproto: Replace a wrong eol comma with a semicolon.

avformat/img2enc: mention -frames:v in error message

Signed-off-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Kieran O Leary <kieran.o.leary@gmail.com>

avformat/concatdec: always re-calculate start time and duration

This allows the underlying files to change their duration on subsequent
avformat context opens.

An example use case where this matters:

ffconcat version 1.0
file dummy.mxf
file dummy.mxf

ffmpeg -re -stream_loop -1 -i dummy.ffconcat -f sdl2 none

The user can seamlessly change the input by atomically replacing dummy.mxf.

v2: Set ConcatFile duration in read_header for all segments with known
durations because from now on we always recalculate the start time in
open_file, and an instant seek could have caused unset ConcatFile durations.

Signed-off-by: Marton Balint <cus@passwd.hu>

avformat/concatdec: fix cur_dts based duration calculation with nonzero stream start_time

Signed-off-by: Marton Balint <cus@passwd.hu>

avformat/concatdec: factorize the duration calculating function

Signed-off-by: Marton Balint <cus@passwd.hu>

lavc/qsvenc: set BRCParamMultiplier to aviod BRC overflow

Fix ticket #7663

Reviewed-by: Carl Eugen Hoyos <ceffmpeg@gmail.com>
Reviewed-by: Hendrik Leppkes <h.leppkes@gmail.com>
Signed-off-by: Zhong Li <zhong.li@intel.com>

avcodec/mips: [loongson] optimize put_hevc_qpel_bi_hv_8 with mmi.

Optimize put_hevc_qpel_bi_hv_8 with mmi in the case width=4/8/12/16/24/32/48/64.
This optimization improved HEVC decoding performance 11.4%(2.01x to 2.24x, tested on loongson 3A3000).

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/mips: [loongson] optimize put_hevc_qpel_hv_8 with mmi.

Optimize put_hevc_qpel_hv_8 with mmi in the case width=4/8/12/16/24/32/48/64.
This optimization improved HEVC decoding performance 11%(1.81x to 2.01x, tested on loongson 3A3000).

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/tests: Add codec_desc to .gitignore

The compiled libavcodec/tests/codec_desc was left out of that dir's
.gitignore when the test was added, so it shows up in 'git status'
as an untracked file if it's been built.

Signed-off-by: FeRD (Frank Dana) <ferdnyc@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/dashenc: Added documentation for $ext$ identifier in filenames

avformat/dashenc: Format xs:datetime in millisecond precision

For low latency streaming even milliseconds matter!

avcodec/prosumer: Error out if decompress() stops reading data

if 0 is encountered in the LUT then decompress() will continue to output 0 bytes but never read more data.
Without a specification it is impossible to say if this is invalid or a feature.
None of the valid prosumer files tested cause a 0 to be read, so it is likely
not a intended feature.

Fixes: Timeout
Fixes: 11266/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PROSUMER_fuzzer-5681827423977472
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/prosumer: Reduce lut size

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/prosumer: Simplify code slightly in decompress()

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/tiff: Check for 12bit gray fax

Fixes: Assertion failure
Fixes: 11898/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_TIFF_fuzzer-5759794191794176
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avutil/imgutils: Optimize memset_bytes() by using av_memcpy_backptr()

This is strongly based on code by Marton Balint, and depends on the previous commit

Fixes: Timeout
Fixes: 11502/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_WCMV_fuzzer-5664893810769920
Before: Executed clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_WCMV_fuzzer-5664893810769920 in 11209 ms
After:  Executed clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_WCMV_fuzzer-5664893810769920 in  4104 ms

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Marton Balint <cus@passwd.hu>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avutil/mem: Optimize fill32() by unrolling and using 64bit

Reviewed-by: Marton Balint <cus@passwd.hu>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/mips: [loongson] optimize put_hevc_pel_bi_pixels_8 with mmi.

Optimize put_hevc_pel_bi_pixels_8 with mmi in the case width=8/16/24/32/48/64.
This optimization improved HEVC decoding performance 2%(1.77x to 1.81x, tested on loongson 3A3000).

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

lavc/hls: Cosmetics: Fix indentation for free_segment_list

Commit 673d8cfd518 missed the indent

Signed-off-by: Jun Zhao <mypopydev@gmail.com>

lavu/frame: Fix typo.

avcodec/libx264: add support for ROI-based encoding

This patch just enables the path from ffmpeg to libx264,
the more encoders can be added later.

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

avutil: add ROI (Region Of Interest) data struct and bump version

The encoders such as libx264 support different QPs offset for different MBs,
it makes possible for ROI-based encoding. It makes sense to add support
within ffmpeg to generate/accept ROI infos and pass into encoders.

Typical usage: After AVFrame is decoded, a ffmpeg filter or user's code
generates ROI info for that frame, and the encoder finally does the
ROI-based encoding.

The ROI info is maintained as side data of AVFrame.

Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

avfilter/vf_paletteuse: don't constantly free and realloc internal frames

Reviewed-by: Clément Bœsch <u@pkh.me>
Signed-off-by: James Almer <jamrial@gmail.com>

avfilter/extractplanes: add support for 12-bit YUVA formats

At present, 16-bit auto-scaled format results in incorrect alpha
extraction.

avformat/hcom: check probe buffer size

avcodec/dirac_arith: Treat overread as error

Fixes: Timeout
Fixes: 11663/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_DIRAC_fuzzer-5636791864918016
Before:Executed clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_DIRAC_fuzzer-5636791864918016 in 26006 ms
After: Executed clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_DIRAC_fuzzer-5636791864918016 in 106 ms

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/diracdec: Propagate errors from dirac_get_arith_uint()

Testcase: 11663/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_DIRAC_fuzzer-5636791864918016

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/ac3dec: Optimize frame start search

Fixes: Timeout
Fixes: 11619/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_AC3_FIXED_fuzzer-5632398021099520
Fixes: 11620/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_AC3_fuzzer-5711996515778560
Fixes: 11658/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_EAC3_fuzzer-5701006524940288
Before: Executed clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_AC3_FIXED_fuzzer-5632398021099520 in 20338 ms
After:  Executed clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_AC3_FIXED_fuzzer-5632398021099520 in 11825 ms

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/lzw: Check for end of input

Fixes: Timeout
Fixes: 11873/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_TIFF_fuzzer-5093495044308992
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

lavd/iec61883: Fix the include path for poll.h.

lavf/rtpproto: Use the correct patch when including poll.h

Fixes a warning using musl:
In file included from libavformat/rtpproto.c:43:0:
/usr/local/musl/include/sys/poll.h:1:2: warning: #warning redirecting incorrect #include <sys/poll.h> to <poll.h>

avcodec/hcom: get rid of single line brackets

avformat: add HCOM demuxer

avcodec: add HCOM decoder

ffmpeg_opt: Print a warning if more than one -vf/-af option was specified.

Fixes ticket #4184.

lavf/mov: Do not fail hard for more invalid atoms.

This is what several other players do and what FFmpeg already does for the sidx atom.

Fixes ticket #7679.

avcodec/tests/rangecoder: initialize array to avoid valgrind warning

Found-by: jamrial
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/gdv: Optimize and factorize scaling loops

Fixes: Timeout
Fixes: 11067/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_GDV_fuzzer-5686623711264768
Before change: Executed clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_GDV_fuzzer-5686623711264768 in 34386 ms
After  change: Executed clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_GDV_fuzzer-5686623711264768 in 24327 ms

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

api-h264-slice-test: fix arguments and help

This program only takes 2 arguments
Remove comment that was never right

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/vc1: fix B predictor validity for 4-MV MBs

The B predictor for 4-MV MBs in interlace field pictures is not used
for blocks 0 and 2 when the picture is 1 MB wide.

Signed-off-by: Jerome Borsboom <jerome.borsboom@carpalis.nl>

avcodec/vp3: ref_frame/ref_frames are only required when HAVE_THREADS=1

checkasm/af_afir: relax the max allowed absolute difference

Should fix failures on x86_32.

Signed-off-by: James Almer <jamrial@gmail.com>

avfilter/af_anlmdn: add timeline support

avfilter: add maskfun filter

lavc: Allow very high bitrates in AVCPBProperties after next version bump.

lavc/tiff: Support CMYK images.

Fixes ticket #3459.

lavc/psd: Support CMYK images.

Based on a05635e by Michael Niedermayer.

Fixes ticket #6797.

avcodec/h264_slice: Fix integer overflow in implicit_weight_table()

Fixes: signed integer overflow: 2 * 2132811760 cannot be represented in type 'int'
Fixes: 11156/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_H264_fuzzer-6237685933408256
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/exr: set layer_match in all branches

Otherwise it is left to the value from the previous iteration

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/exr: Check for duplicate channel index

Fixes: Out of memory
Fixes: 11582/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_EXR_fuzzer-5730204559867904
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/vc1: fix decoding of old WMV3 format

The position of the second MV predicitor candidate is slightly different
for the old WMV3 format indicated by RES_RTM_FLAG. This patch fixes
decoding of niceday.wmv on the samples server.

Fixes: #6641
Signed-off-by: Jerome Borsboom <jerome.borsboom@carpalis.nl>

avcodec/vc1: shuffle calculation of MV predictor candidates

The B predictor for 4-MV macroblocks is only out of bounds when
the A predictor is also out of bounds.

Signed-off-by: Jerome Borsboom <jerome.borsboom@carpalis.nl>

avfilter/af_anlmdn: rework how denoising strength is used

Make into account patch size.

avcodec/dstdec: use appropriate alignment

this was a typo in my original dst decoder. there is no requirement for
64-byte alignment here. the change does not affect decoder performance.

Signed-off-by: Peter Ross <pross@xvid.org>

lavc/hevc_parser: report detailed log when missing picture occurs

Report the detailed log with buf_size in parse_nal_units to provide
more information when picture could not be found.

Match the behaviour in h264_parser.

Signed-off-by: Linjie Fu <linjie.fu@intel.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/nuv: add FF_CODEC_CAP_INIT_CLEANUP

Fixes: memleak
Fixes: 12244/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_NUV_fuzzer-5099447273390080
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/af_anlmdn: do not output more samples than received

dstdec: big-endian compatiblity

intreadwrite: add AV_RL64A, AV_WL64A

macros for reading and writing 64-bit aligned little-endian values.

these macros are used by the DST decoder and give a performance boost
on platforms that where the compiler must guard against unaligned
memory access.

avfilter/af_anlmdn: drain samples at end

avfilter/af_anlmdn: do not trim first samples

Instead denoise from start, assumming missing samples are silence.

avfilter/af_anlmdn: log used parameters

avfilter/af_anlmdn: use lut table to calculate weights

avfilter: add anlmdn filter x86 SIMD optimizations

lavfi/f_select: Cosmetics, move a function.

Silences a warning when only aselect was enabled:
libavfilter/f_select.c:421:12: warning: 'query_formats' defined but not used

lavfi/f_select: Fix aselect filter standalone compilation.