hevc: Move hevc_decode_extradata before frame decoding

Avoids a forward-declaration in the following commit.

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffmpeg: fix width/height overrides for sub2video processing

Merge commit '722ec3eb35bc152ce91d0a4502eca0df1c0086d0'

* commit '722ec3eb35bc152ce91d0a4502eca0df1c0086d0':
  avconv: decouple configuring filtergraphs and setting input parameters

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

ppc: pixblockdsp: do unaligned block accesses correctly again

This was broken by the following Libav commit:
4c387c7 ppc: dsputil: do unaligned block accesses correctly

The following tests fail due to this:
fate-checkasm
fate-vsynth1-dnxhd-2k-hr-hq fate-vsynth1-dnxhd-edge1-hr
fate-vsynth1-dnxhd-edge2-hr fate-vsynth1-dnxhd-edge3-hr
fate-vsynth1-dnxhd-hr-sq-mov fate-vsynth1-dnxhd-hr-hq-mov
fate-vsynth2-dnxhd-2k-hr-hq fate-vsynth2-dnxhd-edge1-hr
fate-vsynth2-dnxhd-edge2-hr fate-vsynth2-dnxhd-edge3-hr
fate-vsynth2-dnxhd-hr-sq-mov fate-vsynth2-dnxhd-hr-hq-mov
fate-vsynth3-dnxhd-2k-hr-hq fate-vsynth3-dnxhd-edge1-hr
fate-vsynth3-dnxhd-edge2-hr fate-vsynth3-dnxhd-edge3-hr
fate-vsynth3-dnxhd-hr-sq-mov fate-vsynth3-dnxhd-hr-hq-mov

Fixes trac ticket #5508.

Reviewed-by: Carl Eugen Hoyos <ceffmpeg@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avformat: close parser if codec changed

The parser depends on the codec and thus must not be used with a different one.
If it is, the 'avctx->codec_id == s->parser->codec_ids[0] ...' assert in
av_parser_parse2 gets triggered.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

crystalhd: Update high level description

We don't need to document the horrible hacks that we removed.

crystalhd: Simplify output frame handling

The old code had to retain a partial frame across two calls in
the case of separate interlaced fields. Now, we know that we'll
get both fields within the same receive_frame call, and so we
don't need to manage the frame as private state any more.

crystalhd: Loop for a frame internally where possible.

It's not possible to return EAGAIN when we've passed input EOF and are
in draining mode. If do return EAGAIN, we're saying there's no way to
get any more output - which isn't true in many cases.

So let's handled these cases in an internal loop as best we can.

crystalhd: Keep NOPTS_VALUE so we know it's not there.

crystalhd: Remove h.264 parser

Now that we don't need to do ridiculous things to work out if a
frame is interlaced or not, we don't need an extra h.264 parser.

crystalhd: We don't need the track the last picture number anymore

This was needed to detect an interlaced failure case that doesn't
happen with the new decode api.

crystalhd: Remove trust_interlaced heuristic

It seems that without all the other 1:1 heuristics, we don't have
a fundamental problem trusting the interlaced flag on output
pictures. That's a relief.

crystalhd: Revert back to letting hardware handle packed b-frames

I'm not sure why, but the mpeg4_unpack_bframes bsf is not
interacting well with seeking. Looking at the code, it should be
ok, with possibly one warning shown, but I see it getting stuck
for an extended period of time after a seek where a packed frame
is cached to be shown later.

So, I gave up on that and went back to making the old hardware
based path work. Turns out that it wasn't broken except that some
samples have a 6 byte drop packet which I wasn't accounting for.

Now it works again and seeks are good.

crystalhd: Switch to new decode API and remove the insanity

The new decode API allows for m:n decode patterns, which is what
you need to use this hardware in a sane way. There are so many
situations where 1:1 doesn't happen naturally that it's a miracle
I got it working as well as I did.

With this change, we can throw all of the crazy heuristics and
sleeps(!) out, and things work correctly.

crystalhd: Fix up the missing first sample

Why on earth the hardware returns garbage for the first sample of
a decoded picture is anyone's guess. The simplest reasonable way
to patch it up is to copy the first sample of the second line. This
should result in the correct chroma values (because the data was
original 4:2:0 upsampled to 4:2:2) even if the luma is isn't.

avformat/matroskaenc: use display aspect ratio for DisplayWidth and DisplayHeight when possible

This avoids potential rounding errors and guarantees the source aspect
ratio is preserved.
Keep writing pixel values when Stereo 3D Mode is enabled and for WebM,
as the format doesn't support anything else.

This fixes ticket #5743, implementing the suggestion from ticket #5903.

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

avformat/matroskaenc: support writing Chroma Location elements

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

avformat/mpegtsenc: Add option to mark stream begin as discontinuous

This avoids continuity check failures in concatenated streams

Reviewed-by: Steven Liu <lingjiujianke@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

lavc: Add hevc main10 profile to ffmpeg cli

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

avcodec/sunrast: Fix input buffer pointer check

Fixes: out of array read
Fixes: poc.dat

Found-by: Bingchang, Liu @VARAS of IIE
Tested-by: bc L <l.bing.chang.bc@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

tests/fate/avformat: add segment.c tests

lavf/matroskaenc: don't try to modify the header when live-streaming

lavf/matroskaenc: fix uninitialized read

lavf/segment: fix autobsf

lavc/hapenc: Use the correct printf length modifier for size_t arguments.

Fixes the following warning:
libavcodec/hapenc.c:122:20: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 4 has type ‘size_t’ [-Wformat]

Based on a patch by Diego Biurrun.

Revert "img2 encoder: allow %t in filename, based on patch from Yuval Adam"

breaks API

Found-by: jamrial
This reverts commit 1a956c64c8eff5edecb004fc7aafd21207e6485c.

vf_colorspace: Add support for film primaries

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

vf_colorspace: Add support for smpte 431/432 (dci/display p3) primaries

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

vf_colorspace: Add support for ycgco color space

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

img2 encoder: use more descriptive vsync names

Signed-off-by: rogerdpack <rogerpack2005@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

img2 encoder: allow %t in filename, based on patch from Yuval Adam

Signed-off-by: rogerdpack <rogerpack2005@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/dnxhdenc: Fix alignment of edge_buf*

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

fate: add streamcopy test for apng

Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

apngdec: use side data to pass extradata to the decoder

Fixes remuxing apng streams coming from the apng demuxer.
This is a regression since 940b8908b94404a65f9f55e33efb4ccc6c81383c.

Found-by: James Almer <jamrial@gmail.com>
Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

mov: immediately return from mov_fix_index without old index entries

If there are no index entries, e_old = st->index_entries is only one
byte large, since it was created by av_realloc called with size 0.

Thus accessing e_old[0].timestamp causes a heap buffer overflow.

Reviewed-by: Sasi Inguva <isasi@google.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

Fix build warnings due to misleading indentation

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

fate filter metadata : add test for aphasemeter - Test a mono file (in phase) -> 1. as result - Test a out of phase 1000 Hz -> -1. as result

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

lavfi/mergeplanes: Fix little endian yuv formats >8 bit and <16bit.

Fixes remaining cases of ticket #5916.

lavfi/mergeplanes: Fix >8 bit for big endian formats and yuv4xxp16le.

Fixes part of ticket #5916.

lavf/mov: Only search for invalid moov in free if compliance < STRICT.

interplayacm: increase bitstream buffer size by AV_INPUT_BUFFER_PADDING_SIZE

This fixes out-of-bounds reads by the bitstream reader.

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avcodec/smc:  Check side data size before use

Fixes out of array read

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

avcodec/tscc:  Check side data size before use

Fixes out of array read

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

avcodec/rscc: Fix constant

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

vf_colorspace: Add support for iec61966-2.4 (xvYCC) transfer

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/qsv: remove MFX_EXTBUFF_CODING_OPTION3

4th generation Intel CPUs don't support MFX_EXTBUFF_CODING_OPTION3.

This patch fixes bug #5324.

qsvdec: Avoid probing with qsv decoders

Set the AV_CODEC_CAP_AVOID_PROBING flag on all of the qsv decoders.

qsv: Merge libav implementation

Merged as-at libav 398f015, and therefore includes outstanding
skipped merges 04b17ff and 130e1f1.

All features not in libav are preserved, and no options change.

mpegvideo: Return correct coded frame sizes from parser

hevc: Return stream format information from parser

vc1: Return stream format information from parser

lavf/mov.c: Use the correct timescale when seeking for audio.

Signed-off-by: Sasi Inguva <isasi@google.com>
Reviewed-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

configure: add '-uninstalled' to uninstalled .pc files

pkg-config(1) expects uninstalled pc files to follow the
blah-uninstalled.pc naming convention and the behavior
of the program is impacted by it. Without this fix
overriding PKGP_CONFIG_LIBDIR is required to ensure
uninstalled files are preferred (overkill), instead of
just adding pc-uninstalled/ to the utility's search path
by setting PKG_CONFIG_PATH accordingly.

Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>

avcodec/rawdec: Check side data size before use

Fixes out of array read

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

avcodec/rscc: Check side data size before use

Fixes out of array read

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

avcodec/msvideo1: Check side data size before use

Fixes out of array read

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

avcodec/qpeg:  Check side data size before use

Fixes out of array read

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

avcodec/qtrle:  Check side data size before use

Fixes out of array read

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

avcodec/msrle:  Check side data size before use

Fixes out of array read

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

avcodec/cuvid: Don't claim to decode h.263 (it doesn't)

Turns out cuvid doesn't support h.263.

interplayacm: validate number of channels

The number of channels is used as divisor in decode_frame, so it must
not be zero to avoid SIGFPE crashes.

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

interplayacm: check for too large b

This fixes out-of-bounds reads.

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avcodec/kmvc:  Check side data size before use

Fixes out of array read

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

avcodec/idcinvideo: Check side data size before use

Fixes out of array read

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

openssl: Allow newer TLS versions than TLSv1

The use of TLSv1_*_method() disallows newer protocol versions; instead
use SSLv23_*_method() and then explicitly disable the deprecated
protocol versions which should not be supported.

Fixes ticket #5915.

avcodec/cinepak: Check side data size before use

Fixes out of array read

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

avcodec/8bps: Check side data size before use

Fixes out of array read

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

doc: fix spelling errors

Reviewed-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

configure: add missing fork() dependency to http_multiclient example

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

avfilter/avf_showcqt: add bar_t option

custom bargraph transparency

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

crystalhd: Reorder mspeg4 decoder after software decoders

This avoids it getting picked by default, which is generally
undesirable and can break test runs.

configure: make sure LTO does not optimize out the test functions

Fixes trac ticket #5909

Bud-Id: https://bugs.gentoo.org/show_bug.cgi?id=598054
Acked-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

fate: add apng encoding/muxing test

Also test the fallback to png creation for a single frame.

Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avformat/flvdec: Fix regression loosing streams

Fixes: unknown_video.flv

Found-by: Thierry Foucu <tfoucu@google.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

apng: use side data to pass extradata to muxer

This fixes creating apng files, which is broken since commit
5ef19590802f000299e418143fc2301e3f43affe.

Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avisynth: fix Planar RGB output

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

lavfi/vf_overlay: support NV12 and NV21

Tested-by: Michael on x86-32/64 linux, mingw, mips/arm qemu linux

avcodec: remove missing incompatible_libav_abi references

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

configure: remove missing incompatible_libav_abi references

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

avformat/matroskaenc: fix cue relative position values when CRC32 is enabled

The dynamic buffer does not contain the CRC32 element so calls to avio_tell()
don't take it into account. This resulted in CueRelativePosition values being
six bytes short.
This is a regression since 6724525a1576ca334d2ffdc085620bb44aea7394

Instead of adding yet another custom check for CRC32 to fix a size or an offset,
remove the existing ones and reserve the six bytes in the dynamic buffer.

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

RELEASE: Update for past 3.2 branch

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

Changelog: Add back next marker

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

Bump minor versions after 3.2 branchpoint to seperate release

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

Changelog: Add 3.2

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

doc/APIchanges: add 3.2 Cut marker

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

Bump minor versions for 3.2

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

Revert "avformat/utils: Discard huge timestamps which would cause overflows if used in basic computations"

Some people seem to oppose this the patch seems to have been missed for a month on the ML
lets restart the discussion and solve this after the release

This reverts commit e936c8d176efd1a0a41e22df24564b1178c79ea9.

avcodec/dvdsubdec: Fix off by 1 error

Fixes out of array read

Found-by: Thomas Garnier using libFuzzer
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/utils: Discard huge timestamps which would cause overflows if used in basic computations

Allowing larger timestamps makes it impossible to calculate basic things like the
difference of 2 timestamps or their sum without checking each individual computation for
overflow.
This should avoid a significant number of overflow checks

Fixes Ticket5136

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

avformat/isom: Fix old API regression with exporting max bitrate

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

doc/patchwork: Document the patchwork states

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

lavu: remove comma at final enumeration items to fix pedantic warnings

swresample/rematrix: Fix float part of swr_set_matrix()

Fixes Ticket #5897.

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

configure: Remove --enable-incompatible-libav-abi from the help output

As it does not work anymore since 655b6dcb34b25d591e15ede17673ea6cb8074711

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

avcodec/dvdsubdec: Fix buf_size check

Fixes out of array access

Found-by: Thomas Garnier using libFuzzer
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

doc/APIchanges: Fill in some missing things

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

lavf/flvdec: init AVPacket::pos to FLVTAG offset

Current code doesn't initialize AVPacket::pos. Made it point to FLVTAG so flv_read_packet can decode from pos

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

fate: Add MXF D10/DNXHD/DV25 probe tests

Signed-off-by: Tobias Rapp <t.rapp@noa-archive.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/scale_npp: fix passthrough mode

Reviewed-by: Timo Rothenpieler <timo@rothenpieler.org>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

lavc/utvideoenc: Set bits_per_coded_sample for rgba.

Allows to write correct value for biBitCount into BITMAPINFOHEADER.
Before, ff_put_bmp_header() always wrote "24" as biBitCount
for utvideo because bits_per_coded_sample was never set by the
encoder.