avutil/hwcontext_cuda: check for missing pixel format

avfilter/vf_scale_npp: check ff_set_common_formats return value

avfilter/vf_scale_npp: move aspect ratio correction after av_frame_copy_props

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>

avcodec/nvenc: always reduce DAR width and height

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>

avcodec/nvenc: Delay identification of underlying format of cuda frames

When input surfaces are cuda frames, we will not know what the actual
underlying format (nv12, p010, etc) is at surface allocation time.

On the other hand, we will know when the input frames are actually
registered and associated with a surface.

So, let's delay format discovery until registration time, which is
actually how we handle other frame properties, such as dimensions.

By itself, this change doesn't allow for transcoding of 10bit
content from cuvid, but it reduces the problem to the hardcoding of
the sw format in ffmpeg_cuvid.c

Signed-off-by: Philip Langdale <philipl@overt.org>
Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>

ffserver: Add client requested urls to the status page

Fixes Ticket3791

Reviewed-by: "Reynaldo H. Verdejo Pinochet" <reynaldo@osg.samsung.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avidec: fix leaking extradata

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

matroskadec: prevent access of elements after freeing

Using the decode interrupt feature of ffmpeg may cause crashes by
accessing previously freed pointers in matroska_read_close.

To prevent this reset nb_elem to zero after freeing the elements,
because ffmpeg normally tests for nb_elem.

Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avcodec/flac_parser: Update nb_headers_buffered

Fixes infinite loop
Fixes: fuzz.flac

Found-by: Frank Liberato <liberato@google.com>
Reviewed-by: Frank Liberato <liberato@google.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

idroqdec: fix leaking pkt on failure

The code calls av_new_packet a few lines above and the allocated memory
has to be freed in case of an error.

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

fate: update after addition of gray10 to lavc/raw

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

tests/ffserver-regression: Check the status page too

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

ffsrever: Make the status page bitexact if any stream is bitexact

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

ffserver_config: Setup codecpar in add_codec()

fixes segfault in the status page code

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

tests/ffserver.regression.ref: update ffserver checksums

Reviewed-by: "Reynaldo H. Verdejo Pinochet" <reynaldo@osg.samsung.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

tests/ffserver-regression.sh: give wget a timeout and prevent retries

Reviewed-by: "Reynaldo H. Verdejo Pinochet" <reynaldo@osg.samsung.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/idroqdec: Check chunk_size for being too large

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

avfilter/vf_masked*: add gray10 and gray12 support

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

avfilter/vf_bitplanenoise: add gray10 and gray12 suppport

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

avfilter/vf_extractplanes: add extract support for 10bit and 12bit formats

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

avfilter/vf_waveform: add gray10 and gray12 support

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

avcodec/raw: add gray10 support in nut

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

avcodec/msrledec: implement vertical offset in 4-bit RLE

The delta escape (2) is supposed to work the same in 4-bit RLE as in
8-bit RLE.  This is documented in the MSDN Bitmap Compression page:
https://msdn.microsoft.com/en-us/library/windows/desktop/dd183383(v=vs.85).aspx

The unchecked modification of line is safe, since the loop condition
(line >= 0) will check it before any pixel data is written.

Fixes ticket #5153 (output now matches ImageMagick for the provided sample).

Signed-off-by: Daniel Verkamp <daniel@drv.nu>

tests: Fix running ffserver under qemu

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

tests/Makefile: Fix ffservertest dependencies

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

ffserver: Remove last use of AVStream size

Reviewed-by: "Reynaldo H. Verdejo Pinochet" <reynaldo@osg.samsung.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffserver: Remove some deprecated API use related to codec/codecpar

Reviewed-by: "Reynaldo H. Verdejo Pinochet" <reynaldo@osg.samsung.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffserver: Remove use of AVStream as a intermediate to store parameters

Reviewed-by: "Reynaldo H. Verdejo Pinochet" <reynaldo@osg.samsung.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffserver: drop FeedData, its unused

Reviewed-by: "Reynaldo H. Verdejo Pinochet" <reynaldo@osg.samsung.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

tests/ffserver-regression.sh: Fix file truncation introduced in 508826f961caf662cadb7c253e3c0e7d75104bdd

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

libvpxenc: Report encoded VP9 level

Report the actual level of the encoded output if a level is
targeted or the level is passively tracked with a target of 0.

vf_colorspace: Forbid odd dimensions

This prevents writing past bounds.

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

ffmpeg_opt: Fix starttime with ffm in bitexact mode

This fixes some differences between runs of the ffserver tests
(in my local tree 2 runs gave the same result with this but i had other
 changes too)

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

pngdec: check if previous frame exists instead of trusting sequence_number

This fixes a segmentation fault caused by calling memcpy with NULL as
second argument in handle_p_frame_apng.

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

tests/ffserver.conf: Force bitexactness in the ffmpeg command

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

avformat/rmenc: Check framerate before storing

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

lavfi/f_ebur128: relicense to LGPL

All copyright holders have agreed to the relicensing.

Approved-by: Andreas Cadhalpun <andreas.cadhalpun@googlemail.com>
Approved-by: David Sedacca <sedacca@comcast.net>
Approved-by: Ganesh Ajjanagadde <gajjanag@mit.edu>
Approved-by: Jean First <jeanfirst@gmail.com>
Approved-by: Kyle Swanson <k@ylo.ph>
Approved-by: Michael Niedermayer <michael@niedermayer.cc>
Approved-by: Nicolas George <george@nsup.org>
Approved-by: Paul B Mahol <onemda@gmail.com>
Approved-by: Thilo Borgmann <thilo.borgmann@mail.de>

avcodec/me_cmp: Fix median_sad size

Fixes out of array read
Fixes: COV1396255

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

ffserver: Remove extract_mpeg4_header()

This should not be needed, our AVParsers should do this
I do not have a testcase though, please help testing this and please
add fate tests if you can.

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

Revert "ffserver: use AVStream.codecpar in open_input_stream()"

Fixes null pointer dereference

Testcase is simply a ffmpeg instance sending a stream to ffserver while another ffmpeg reads from it

This reverts commit 6f0a1710d77dde0d803861506a2157a23f08c14c.

configure: check for stdatomic.h

Since this is a C11 feature, it requires -std=c11.

Not actually used for anything yet, that will be added in the following
commits.

This merges libav commit 13f5d2bf75b95a0bfdb9940a5e359a719e242bed.

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

avfilter/vf_zscale: export approximate gamma option and enable it by default

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

avformat/utils: Fix type mismatch

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

lavf: always forward codec_whitelist in avformat_find_stream_info

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

dxva2_vc1: support multiple slices

vc1dec: support multiple slices in frame coded images with hwaccel

Based on a patch by Jun Zhao <mypopydev@gmail.com>

refine avformat/flvdec set bit_rate

avformat/flvdec: move set bit_rate from parse AMF OBJECT to
 create_stream

before patch:
Stream #0:0: Video: h264 (High), yuv420p(progressive), 1280x714 [SAR 1:1 DAR 640:357], 25 fps, 25 tbr, 1k tbn, 50 tbc

after patch:
Stream #0:0: Video: h264 (High), yuv420p(progressive), 1280x714 [SAR 1:1 DAR 640:357], 2576 kb/s, 25 fps, 25 tbr, 1k tbn, 50 tbc

Signed-off-by: Steven Liu <lq@chinaffmpeg.org>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/flvenc: fix ticket 5976 and use old commit

mythtv have problem with non-seekable dont write duration and filesize
and there have problem with some other server and player with 0 value
duation and filesize.
So add a flv flags to fix the ticket and make a choose for users.

Signed-off-by: Steven Liu <lq@chinaffmpeg.org>

swresample/soxr: fix invalid use of linear_interp

give very bad quality for soxr resampler.
linear_interp is intended for  using linear interpolation
between filter bank so quality will be better.

i guess this is misunderstood as 'do not use filter bank,
but directly interpolate linearly between samples'.

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

avcodec/aac_adtstoasc_bsf: validate and forward extradata if the stream is already ASC

Fixes ticket #5973

Reviewed-by: Hendrik Leppkes <h.leppkes@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>

mss2: only use error correction for matching block counts

This fixes a heap-buffer-overflow in ff_er_frame_end when decoding mss2
with coded_width/coded_height larger than width/height.

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

avcodec/nvenc: Remove aspect-ratio decompensation logic

This dubious behaviour in nvenc was finally removed by nvidia, and
as we refuse to run on anything older than 7.0, we don't need to
keep it around for old versions.

avcodec/mpeg4audio: correctly propagate meaningful error values

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

lavf/flvdec: Set avg_frame_rate for video streams.

Requested-by: qw, applemax82 at 163 dot com
Reviewed-by: Steven Liu

MAINTAINERS: Add Vittorio to "Developers with git write access who are currently not maintaining any specific part"

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

MAINTAINERS: Add developers who have git write access but are otherwise not listed

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

softfloat: decrease MIN_EXP to cover full float range

floats are not necessarily normalized, so a normalized softfloat needs
MIN_EXP lowered by 23 to cover that range.

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

fate/exr: add test for pxr24 with half and uint32 channel

Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

libavcodec/exr: add support for uint32 channel decoding with pxr24

Doesn't decode the uint32 layer, but decodes the half part of the file.

Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

libopusdec: default to stereo for invalid number of channels

This fixes an out-of-bounds read if avc->channels is 0.

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

flvdec: require need_context_update when changing codec id

Otherwise the codec context and codecpar might disagree on the codec id,
triggering asserts in av_parser_parse2.

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

swresample/resample: optimize exact_rational=on:linear_interp=on case

separate dsp.resample to dsp.resample_common and dsp.resample_linear
and choose to call faster resample_common even when linear_interp=on
when c->frac and c->dst_incr_mod are both zero

speed up resampling when exact_rational and linear_interp are both
enabled because exact_rational force c->frac and c->dst_incr_mod to
be zero when soft compensation does not happen

benchmark on exact_rational=on:linear_interp=on
        old     new
real    8.432s  5.097s
user    7.679s  4.989s
sys     0.125s  0.107s

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

fate/swresample: add resample exact_lin and exact_lin_async test

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

avutil/tests: add cpu_init to .gitignore and tests/fate

This is a follow-up to commit d84a21207ea83055dc9b6dc1cd6a379f2ea756e7,
which added the libavutil/tests/cpu_init.c.

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

avfilter/tests/.gitignore: add integral

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

cuda: check for cuda.h when enabled

Fixes make checkheaders on systems without the Cuda Toolkit, which
was broken after the dynlink changes.

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

ffmpeg_filter: fix several logic failures

Move global thread variables to better place.
Use correct variable for simple and complex filtergraphs.

This makes number of threads set per filter work again.

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

pgssubdec: only set w/h/linesize when allocating data

Rects with positive w/h/linesize but no data are invalid.

Reviewed-by: Petri Hintukainen <phintuka@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

lavfi/pan: allow negative gain parameters also for other inputs than the first named

Expands the parser to also accept the separator '-' in addition to
'+', and take the negative sign into consideration.

The optional sign for the first factor in the expression is already
covered by parsing for an integer.

Signed-off-by: Moritz Barsnick <barsnick@gmx.net>
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

lavc/vaapi_hevc: fix scaling list duplicate transfer issue.

scaling list is already transfered to raster scan during head parsing,
so no need to transfer it again.

And after this fix, FATE test SLIST_A_Sony_4/SLIST_B_Sony_8/
SLIST_C_Sony_3/SLIST_D_Sony_9 will pass in i965/Skylake.

Signed-off-by: Wang, Yi A <yi.a.wamg@intel.com>
Signed-off-by: Jun Zhao <jun.zhao@intel.com>
Signed-off-by: Mark Thompson <sw@jkqxz.net>

avutil/tests: Add cpu_init.c to check whether the one-time initialization in av_get_cpu_flags() has data races.

Co-author: Dmitry Vyukov of Google

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

avutil/cpu: remove the |checked| static variable

Remove the |checked| variable because the invalid value of -1 for
|flags| can be used to indicate the same condition. Also rename |flags|
to |cpu_flags| because there are a local variable and a function
parameter named |flags| in the same file.

Co-author: Dmitry Vyukov of Google

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

avcodec/cuvid: Restore initialization of pixel format in init()

I moved this into the handle_video_sequence callback because that's
the earliest time you can make an accurate decision as to what the
format should be.

However, transcoding requires that the decision between using
the accelerated PIX_FMT_CUDA vs a normal pix format happen at init()
time. There is enough information available to make that decision
and things work out with the underlying format only being discovered
in the sequence callback.

avfilter/vf_zscale: add support for some recent new additions

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

fate: update fate-source ref file

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

doc/examples/muxing: Fix av_frame_make_writable usage

This patch moves the av_frame_make_writable() call from fill_yuv_image
to get_video_frame so that its argument can be the actual frame that
will be sent to the encoder.

This fixes data corruption issues in codecs that keep references on
one or several previous frames.

Signed-off-by: Sam Hocevar <sam@hocevar.net>
Reviewed-by: wm4
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/options_table: make channel_layouts uint64

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

avutil/opt: Add AV_OPT_TYPE_UINT64

Requested-by: wm4 ([FFmpeg-devel] [PATCH] avutil/opt: Support max > INT64_MAX in write_number() with AV_OPT_TYPE_INT64)
Requested-by: ronald ([FFmpeg-devel] [PATCH] avutil/opt: Support max > INT64_MAX in write_number() with AV_OPT_TYPE_INT64)
Reviewed-by: Andreas Cadhalpun <andreas.cadhalpun@googlemail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

sbgdec: prevent NULL pointer access

Reviewed-by: Josh de Kock <josh@itanimul.li>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

rmdec: validate block alignment

This fixes division by zero crashes.

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

smacker: limit recursion depth of smacker_decode_bigtree

This fixes segmentation faults due to stack-overflow caused by too deep
recursion.

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

avformat/mpeg: Adjust vid probe threshold to correct mis-detection

Fixes: _ij.mp3

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

mxfdec: fix NULL pointer dereference in mxf_read_packet_old

Metadata streams have priv_data set to NULL.

Reviewed-by: Josh de Kock <josh@itanimul.li>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

libvpxenc: Support targeting a VP9 level

Levels are specified at https://www.webmproject.org/vp9/levels/

avcodec/cuvid: Add support for P010/P016 as an output surface format

The nvidia 375.xx driver introduces support for P016 output surfaces,
for 10bit and 12bit HEVC content (it's also the first driver to support
hardware decoding of 12bit content).

The cuvid api, as far as I can tell, only declares one output format
that they appear to refer to as P016 in the driver strings. Of course,
10bit content in P016 is identical to P010, and it is useful for
compatibility purposes to declare the format to be P010 to work with
other components that only know how to consume P010 (and to avoid
triggering swscale conversions that are lossy when they shouldn't be).

For simplicity, this change does not maintain the previous ability
to output dithered NV12 for 10/12 bit input video - the user will need
to update their driver to decode such videos.

libavutil/hwcontext_cuda: Support P010 and P016 formats

CUVID is now capable of returning 10bit and 12bit decoded content
in P010/P016. Let's support transfering those formats.

avutil: add P016 pixel format

P016 is the 16-bit variant of NV12 (planar luma, packed chroma), using
two bytes per component.

It may, and in fact is most likely to, be used in situations where
there are less than 16 bits of data. It is the responsibility of
the writer to zero out any unused LSBs.

avcodec/libx264: fix forced_idr logic

Currently, it forces IDR frames for both true and false.
Not entirely sure what the original idea behind the tri-state bool
option is.

Reviewed-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

avcodec/cuvid: allow setting number of used surfaces

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>

avcodec/nvenc: better surface allocation alghoritm, fix rc_lookahead

User selectable surfaces are not working correctly, if you set number of
surfaces on cmdline, it will always use minimum 32 or 48 depends on
selected resolution, but in nvenc it is not necessary to use so many
surfaces.

So from now you can define as low as 1 surface and nvenc will still
work, it will ofcourse lower GPU memory usage by 95% and async_delay to zero

That was the easy part, now littlebit more...

Next part of this patch is to always prefer rc_lookahead to be more
important for number of surfaces, than user defined surfaces value.
Maximum rc_lookahead from nvidia documentation is 32, but could increase
in future generations so there is no limit for this yet. Value
async_depth is still accepted and prefered over rc_lookahead.

There were also bug when you request more than rc_lookahead > 31, it
will always set maximum 31, because surface numbers recalculation was
after setting lookahead, which is now fixed.

Results:
If you set -rc_lookahead 32 and -bf 3 it will now use only 40 surfaces
and lower GPU memory usage by 20%, also it will now increase PSNR by 0.012dB

Two more comments:

1. from my internal test, i don't understand addition of 4 more surfaces
when lookahead is calculated, i didn't used this and everything works as
with those 4 more extra surfaces, does anybody know what is going on
there? I looks like it was used for B frames which are calculated
separately, because B frames maximum is 4.

2. rc_lookahead is defined default to -1, but in test condition if
(ctx->rc_lookahead) which sets lookahead it will be always true, i don't
know if this is intended behavior, so in default behavior is lookahead
always on!

This is default condition when rc_lokkahead is -1 (not defined on
cmdline), whis is maybe something that is not intended:
ctx->encode_config.rcParams.enableLookahead = 1;
ctx->encode_config.rcParams.lookaheadDepth  = 0;
ctx->encode_config.rcParams.disableIadapt   = 0;
ctx->encode_config.rcParams.disableBadapt   = 0;

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>

avcodec/nvenc: maximum usable surfaces are limited to maximum registered frames

Maximum usable surfaces is limited to MAX_REGISTERED_FRAMES constant in
nvenc.h

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>

configure: cuda is no longer nonfree, enable and autodetect by default

avfilter/vf_hwupload_cuda: check ff_formats_ref for errors

avfilter/vf_hwupload_cuda: use new hwdevice allocation API

avcodec/nvenc: use dynamically loaded CUDA

avfilter/vf_scale_npp: use dynamically loaded CUDA

avcodec/cuvid: use dynamically loaded CUDA/CUVID

And remove the now obsolete compat headers.

avutil/hwcontext_cuda: use dynamically loaded CUDA

compat/cuda: add dynamic loader