lavc/htmlsubtitles: handle colors starting with many '#'

lavc/htmlsubtitles: improve line breaks handling

lavc/tests: add htmlsubtitles

lavc/htmlsubtitles: improve handling broken garbage

This commit switches off forced correct nesting of tags and only keeps
it for font tags. See long explanations in the code for the rationale.

This results in various FATE changes which I'll explain here:

- various swapping in font attributes, this is mostly noise due to the
  old reverse stack way of printing them. The new one is more correct as
  the last attribute takes over the previous ones.

- unrecognized tags disappears

- invalid tags that were previously displayed aren't anymore (instead,
  we have a warning). This is better for the end user

The main benefit of this commit is to be more tolerant to error, leading
to a better handling of badly nested tags or random wrong formatting for
the end user.

avcodec/diracdsp: fix integer overflow

Fixes: runtime error: signed integer overflow: 11 * 225726413 cannot be represented in type 'int'
Fixes: 2764/clusterfuzz-testcase-minimized-5382561922547712

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: Check weight_log2denom

Fixes: runtime error: shift exponent -1 is negative
Fixes: 2742/clusterfuzz-testcase-minimized-5724322402402304
Fixes: 2744/clusterfuzz-testcase-minimized-4672435653705728
Fixes: 2749/clusterfuzz-testcase-minimized-5298741273690112

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

avcodec/dirac_vlc: Fix invalid shift

Fixes: runtime error: shift exponent 65 is too large for 64-bit type 'residual' (aka 'unsigned long')
Fixes: 2737/clusterfuzz-testcase-minimized-4968639147016192

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

avcodec/atrac3p: use float_dsp in ff_atrac3p_power_compensation

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

doc/filters.texi: add another lut2 example

Signed-off-by: Paul B Mahol <onemda@gmail.com>

sws/tests/pixdesc_query: fix use of free() instead of av_free()

Fix CID 1415949

sws/tests/pixdesc_query: replace rgb based pix fmts with endianess agnostic names

Fixes ticket #6554

sws/tests/pixdesc_query: sort pixel formats

sws/tests/pixdesc_query: save every pix fmts in a list

This will be required for the next commit.

avdevice/decklink_dec: set field order via codecpar

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

lavfi/vf_displace: move to "activate" design.

lavfi/vf_hysteresis: move to "activate" design.

lavfi/vf_lut2: move to "activate" design.

lavfi/vf_maskedclamp: move to "activate" design.

lavfi/vf_mergeplanes: move to "activate" design.

lavfi/vf_maskedmerge: move to "activate" design.

lavfi/vf_midequalizer: move to "activate" design.

lavfi/vf_premultiply: move to "activate" design.

lavfi/vf_remap: move to "activate" design.

lavfi/vf_threshold: move to "activate" design.

Also fix missing dependency.

lavfi/vf_stack: move to "activate" design.

lavfi/framesync2: implement "activate" design.

With this helper API, filters that used the first framesync
helper API can easily be changed to use the new and more
extensible design for filters with a single activate() callback.

lavfi: add outlink helper functions.

These wrappers cost nothing, they make the namespace more
consistent and they will be useful if/when locking becomes
necessary.

lavfi: make FFERROR_NOT_READY available to filters.

I am not entirely sure that this return code is useful,
but having and using it makes no harm.

lavfi/framesync2: rename all conflicting symbols.

lavfi: copy framesync into framesync2.

framesync2 will be the base for the version using activate.
Most of the logic will be the same, but the code cannot be shared.
Copying the file initially without change will make the diff
easier to read.

Revert "Revert "lavfi/buffersrc: push the frame deeper if requested.""

This reverts commit 04aa09c4bcf2d5a634a35da3a3ae3fc1abe30ef8
and reintroduces 0ff5567a30be6d7c804e95997ae282d6bacd76c3 that
was temporarily reverted due to minor regressions.

It also reverts e5bce8b4ce7b1f3a83998febdfa86a3771df96ce that fixed FATE refs.
The fate-ffm change is caused by field_order now being set
on the output format because the first frame arrives earlier.
The fate-mxf change is assumed to be the same.

mdct15: add inverse transform postrotation SIMD

2.5ms frames:
Before   (c):  2638 decicycles in postrotate, 2097040 runs,    112 skips
After (sse3):  1467 decicycles in postrotate, 2097083 runs,     69 skips
After (avx2):  1244 decicycles in postrotate, 2097085 runs,     67 skips

5ms frames:
Before   (c):  4987 decicycles in postrotate, 1048371 runs,    205 skips
After (sse3):  2644 decicycles in postrotate, 1048509 runs,     67 skips
After (avx2):  2031 decicycles in postrotate, 1048523 runs,     53 skips

10ms frames:
Before   (c):  9153 decicycles in postrotate,  523575 runs,    713 skips
After (sse3):  5110 decicycles in postrotate,  523726 runs,    562 skips
After (avx2):  3738 decicycles in postrotate,  524223 runs,     65 skips

20ms frames:
Before   (c): 17857 decicycles in postrotate,  261866 runs,    278 skips
After (sse3): 10041 decicycles in postrotate,  261746 runs,    398 skips
After (avx2):  7050 decicycles in postrotate,  262116 runs,     28 skips

Improves total decoding performance for real world content by 9% with avx2.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

lavc/makefile: Add missing file dependencies.

ac3dsp.c uses tables from ac3.c
ac3.c uses tables from ac3tab.c
hevc_ps uses tables from hevc_data.c
intrax8.c uses tables from msmpeg4data.c

Signed-off-by: Matt Oliver <protogonoi@gmail.com>

lavf/os_support: Use existing WinRT config value.

Signed-off-by: Matt Oliver <protogonoi@gmail.com>

win32_dlfcn: Support WinRT/UWP.

This only enables dlls that are packaged with the application to be
loaded. Due to the limitations of WinRT/UWP it is not allowed to load
external/system dlls so this cannot be used as a complete replacement
for normal win32 dll loading.

Signed-off-by: Matt Oliver <protogonoi@gmail.com>

examples/hw_decode: Add a HWAccel decoding example.

Works with VAAPI, VDPAU, DXVA2 and D3D11VA.

Signed-off-by: Liu, Kaixuan <kaixuan.liu@intel.com>
Signed-off-by: Jun Zhao <jun.zhao@intel.com>
Reviewed-by: Steven Liu <lingjiujianke@gmail.com>
Signed-off-by: Mark Thompson <sw@jkqxz.net>

avcodec/dirac_dwt: Fix multiple integer overflows in COMPOSE_DD97iH0()

Fixes: runtime error: signed integer overflow: 9 * 335544320 cannot be represented in type 'int'
Fixes: 2739/clusterfuzz-testcase-minimized-6737297955356672

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: Fix integer overflow in divide3()

Fixes: runtime error: signed integer overflow: -1073746548 * 21845 cannot be represented in type 'int'
Fixes: 2729/clusterfuzz-testcase-minimized-5902915464069120

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

avcodec/takdec: Fix integer overflow in decode_subframe()

Fixes: runtime error: signed integer overflow: -536870912 - 1972191120 cannot be represented in type 'int'
Fixes: 2711/clusterfuzz-testcase-minimized-4975142398590976

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

avformat/rtmppkt: Convert ff_amf_get_field_value() to bytestream2

Fixes: out of array accesses

Found-by: JunDong Xie of Ant-financial Light-Year Security Lab
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/rtmppkt: Convert ff_amf_tag_size() to bytestream2

Fixes: out of array accesses
Fixes: crash-9238fa9e8d4fde3beda1f279626f53812cb001cb-SEGV

Found-by: JunDong Xie of Ant-financial Light-Year Security Lab
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libavcodec/mips: Improve avc dequant-idct luma dc msa function

Signed-off-by: Kaustubh Raste <kaustubh.raste@imgtec.com>
Reviewed-by: Manojkumar Bhosale <Manojkumar.Bhosale@imgtec.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/diracdec: Fix integer overflow in signed multiplication in UNPACK_ARITH()

Fixes: runtime error: signed integer overflow: 1073741823 * 4 cannot be represented in type 'int'
Fixes: 2729/clusterfuzz-testcase-minimized-5902915464069120

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

avcodec/pixlet: Simplify nbits computation

Fixes multiple integer overflows
Fixes: runtime error: signed integer overflow: 1 + 2147483647 cannot be represented in type 'int'

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>

pthread_frame: revert 2e664b9c1e73c80aab91070c1eb7676f04bdd12d.

The patch does not fix the tsan warning it was intended to fix.
Reverting the patch moves the av_log() back to the outside of the lock.

Signed-off-by: Wan-Teh Chang <wtc@google.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

avcodec/h264_slice: don't sync default_ref[] between threads.

default_ref[] is unconditionally initialized in h264_initialise_ref_list()
(called from ff_h264_build_ref_list(), called from h264_slice_init()).

This fixes the following tsan warning when running fate-h264:

WARNING: ThreadSanitizer: data race (pid=31070)
  Write of size 8 at 0x7bbc000082a8 by thread T1 (mutexes: write M1628):
    #0 memcpy /work/release-test/final/llvm.src/projects/compiler-rt/lib/tsan/../sanitizer_common/sanitizer_common_interceptors.inc:655:5
(ffmpeg+0x10de9d)
    #1 h264_initialise_ref_list ffmpeg/libavcodec/h264_refs.c:214:29 (ffmpeg+0x1186b3f)
    #2 ff_h264_build_ref_list ffmpeg/libavcodec/h264_refs.c:306 (ffmpeg+0x1186b3f)
    #3 h264_slice_init ffmpeg/libavcodec/h264_slice.c:1900:11 (ffmpeg+0x1191149)
[..]
  Previous read of size 8 at 0x7bbc000082a8 by main thread (mutexes:
write M1630):
    #0 memcpy /work/release-test/final/llvm.src/projects/compiler-rt/lib/tsan/../sanitizer_common/sanitizer_common_interceptors.inc:655:5
(ffmpeg+0x10de9d)
    #1 ff_h264_update_thread_context ffmpeg/libavcodec/h264_slice.c:411:5 (ffmpeg+0x118b7dc)

Signed-off-by: Wan-Teh Chang <wtc@google.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

doc/APIChanges: add missing entry for ICC Profile side data type.

avcodec/dnxhdenc: enable frame threading

Signed-off-by: Paul B Mahol <onemda@gmail.com>

fate: update pixfmt_best test to check for endianness

Signed-off-by: Tobias Rapp <t.rapp@noa-archive.com>

avcodec/dnxhddec: Move mb height check out of non hr branch

Fixes: out of array access
Fixes: poc.dnxhd

Found-by: Bingchang, Liu@VARAS of IIE
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

pngdec: fix potential memory leak

Fixes CID1412026.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

avcodec/jpeg2000dec: Fix h/vden typo

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

avformat/s337m: fix potentially undefined pointer arithmetic

Use integer position instead of pointer for loop variable. Also only
skip header fields after header has been fully validated.

avcodec/dolby_e: fix potentially undefined pointer arithmetic

Avoid undefined behavior in skip_input() by checking that enough data is
available before incrementing input pointer.

Check return values of parse_key() and skip_input() and exit early with
error if there is not enough data.

avcodec/dolby_e: add AV_CODEC_CAP_CHANNEL_CONF capability

avcodec/dolby_e: add 'f' suffixes to floating point literals

lavc/htmlsubtitles: simplify 1-char tags case insensitive test

avcodec/jpeg2000dec: Fix division by zero in jp2_find_codestream()

Fixes: 2707/clusterfuzz-testcase-minimized-5179636394754048

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

avcodec/hevc_ps: fix integer overflow in log2_parallel_merge_level_minus2

Fixes: runtime error: signed integer overflow: -2147483647 - 2 cannot be represented in type 'int'
Fixes: 2702/clusterfuzz-testcase-minimized-4511932591636480

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

avcodec/webp: add support for ICCP chunks

Export the raw data as ICC Profile frame side data.

Reviwed-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>

ffprobe: add support for ICC Profile frame side data

Print the name metadata entry and the buffer size.

Reviwed-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>

avutil/frame: add ICC Profile to av_frame_side_data_name

Reviwed-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>

avcodec/rdft: reorder calculation

old:
 165188 decicycles in  rdft,   65536 runs,      0 skips
 165865 decicycles in irdft,   65536 runs,      0 skips
new:
 142487 decicycles in  rdft,   65536 runs,      0 skips
 141498 decicycles in irdft,   65536 runs,      0 skips

Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>

avformat/oggparsecelt: Do not re-allocate os->private

Fixes: double free
Fixes: clusterfuzz-testcase-minimized-5080550145785856

Found-by: ClusterFuzz
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libavcodec/mips: Optimize avc idct 4x4 for msa

Removed memset call and improved performance.

Signed-off-by: Kaustubh Raste <kaustubh.raste@imgtec.com>
Reviewed-by: Manojkumar Bhosale <Manojkumar.Bhosale@imgtec.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

mdct15: simplify prereindexing and forward transform postrotation

Equivalent.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

pngdec: decode and expose iCCP chunks as side data

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

lavu/frame: add new side data type for ICC profiles

Many image formats support embedding of ICC profiles directly in
their bitstreams. Add a new side data type to allow exposing them to
API users.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

lavfi/testsrc2: fix completely transparent alpha.

avcodec/mpegvideo_enc: fix mixed declarations and code warning

avformat/hlsenc: improve hls encrypt get key file operation

get key file only once time is ok, no need more times.
Ticket-id: #6545

Found-by: JohnPi
Signed-off-by: Steven Liu <lq@onvideo.cn>

avcodec/ylc: Fix shift overflow

Fixes: runtime error: shift exponent 32 is too large for 32-bit type 'unsigned int'
Fixes: 2698/clusterfuzz-testcase-minimized-4713541443518464

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

avcodec/aacps: Fix multiple integer overflow in map_val_34_to_20()

Fixes: avcodec/aacps.c:511:40: runtime error: signed integer overflow: 1509077651 + 758068176 cannot be represented in type 'int'
Fixes: 2678/clusterfuzz-testcase-minimized-4702787684270080

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

lavf/movenc: support GPMF track (gpmd) remuxing

See https://github.com/gopro/gpmf-parser for more information on the
data stream itself.

avformat/hlsenc: fix hls fmp4 extention name bug

ticket-id: #6541
when use hls fmp4 muxer, the extention name is not .m4s, this
code can fix it.

Found-by: JohnPi
Signed-off-by: Steven Liu <lq@onvideo.cn>

avcodec/mpegvideo_enc: Use intra/inter scantable matching mb type in quantization

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

avcodec/aacdec_fixed: fix: left shift of negative value -1

Fixes: 2699/clusterfuzz-testcase-minimized-5631303862976512

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

avcodec/dirac_vlc: Fix undefined shift

Fixes: runtime error: shift exponent 64 is too large for 64-bit type 'residual' (aka 'unsigned long')
Fixes: 2674/clusterfuzz-testcase-minimized-4999700518273024

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

avfilter/w3fdif: do not write to line before start line

That line has been written by previous job.
Fix tsan warning.

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>

doc/filters: typo in frei0r

Signed-off-by: Brice Waegeneire <brice.wge@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/cfhd: Fix decoding regression due to height check

Fixes: Ticket6546

Regression since: 54aaadf648073149f1ac34f56cbde4e6c5aa22ef

Reviewed-by: Muhammad Faiz <mfcc64@gmail.com>
Reviewed-by: Kieran Kunhya <kierank@obe.tv>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/dcaadpcm: fix use of uninitialized variable

Fixes CID 1409924.

avcodec/dcaadpcm: check for av_malloc() failure

Fixes CID 1409915.

pixdesc: Reindent after previous commit

pixdesc: Explicitly handle invalid arguments to av_find_best_pix_fmt_of_2()

ffprobe: Print color properties from show_frames

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

hevc: Make sure to update the current frame transfer characteristic

Otherwise the first decoded frame will still be tagged with the
original transfer instead of the alternative one.

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

zscale: Factor out graph building

zscale: Factor out format initialization

avcodec/diracdec: Check dimensions which are closer to what is allocated in alloc_sequence_buffers()

Fixes OOM
Fixes: 2674/clusterfuzz-testcase-minimized-4999700518273024

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

doc: Add initial documentation explaining undefined behavior and SUINT

Requested-by: Kieran Kunhya <kierank@obe.tv>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libavutil/mips: Updated msa generic macros

Reduced msa load-store code.
Removed inline asm of GP load-store for 64 bit.
Updated variable names in GP load-store macros for naming consistency.
Corrected macro descriptions.

Signed-off-by: Kaustubh Raste <kaustubh.raste@imgtec.com>
Reviewed-by: Manojkumar Bhosale <Manojkumar.Bhosale@imgtec.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

MAINTAINERS: add myself

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

avcodec/dca: remove GetBitContext usage from avpriv_dca_parse_core_frame_header()

This prevents potential ABI issues with GetBitContext.

Reviewed-by: foo86 <foobaz86@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>

avcodec/h264: Declare the local variable decode_chroma as const.

ff_h264_decode_mb_cabac() and ff_h264_decode_mb_cavlc() are very long
functions. Declaring decode_chroma as const makes it clear the variable
doesn't change after initialization.

Signed-off-by: Wan-Teh Chang <wtc@google.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/hevcdec: hevc_await_progress: declare |y| only if used.

hevc_await_progress() uses the variable |y| only inside the "if" block.
So |y| only needs to be declared and initialized in that block.

Signed-off-by: Wan-Teh Chang <wtc@google.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/x86/cavsdsp: Delete #include "libavcodec/x86/idctdsp.h".

This file already has #include "idctdsp.h", which is resolved to the
idctdsp.h header in the directory where this file resides by compilers.
Two other files in this directory, libavcodec/x86/idctdsp_init.c and
libavcodec/x86/xvididct_init.c, also rely on #include "idctdsp.h"
working this way.

Signed-off-by: Wan-Teh Chang <wtc@google.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

opus_rc: fix CDF encoding of zeroes

0th CDF value is where probabilities are.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

fate: add libavdevice/reverse.c to fate-source

avdevice: make ff_reverse available on shared builds

Should fix compilation failures introduced by 9b937958907daaddade139c36ce33c6eac269631.