avcodec/dca: simplify condition

Reviewed-by: Hendrik Leppkes <h.leppkes@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/dca: fix av_cold placement in declarations

Reviewed-by: Hendrik Leppkes <h.leppkes@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter: add ciescope filter

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

lavfi/drawutils: Add some missing GBRP pix_fmts.

Fixes for example extractplanes with GBRAP16.

avformat/yuv4mpegdec: fix seeking for partial files

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

lavc/videotoolboxenc: remove *_NULLABLE annotations; fixes pre-10.11 build

These macros were added in OS X 10.11, and the file compiles without warnings
on both 10.10 and 10.11 with them removed.

Thanks to mark4o on IRC for pointing out the failure and testing the patch.

fate: add filter-hls

ffmpeg: remove hardcoded 'now' creation_time support

Every date parsing routine now uses av_parse_time which handles 'now' and
provides greater precision as well. This change also enables the segmenter
muxer to set the proper 'now' creation time at the beginning of each segment.

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

avformat: use ff_standardize_creation_time for formats writing all format string metadata

Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Marton Balint <cus@passwd.hu>

avformat/utils: add a function to standardize creation time

This can be used for formats which write all format metadata as string to
files, therefore non-standard creation times such as 'now' will be parsed.

The standardized creation time is UTC ISO 8601 with microsecond precision.

Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Marton Balint <cus@passwd.hu>

lavc: add VideoToolbox H.264 Encoder

Autodetected by default. Encode using -codec:v h264_videotoolbox.

Signed-off-by: Rick Kern <kernrj@gmail.com>
Signed-off-by: wm4 <nfxjfg@googlemail.com>

lavf/utils: Add ff_get_packet_palette()

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

avformat/cache: Fix memleak of tree entries

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

fate/api: add missing FLV dependency to fate-api-seek

avfilter/vf_vectorscope: add 9 & 10 bit depth input & output support with alpha

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

doc/utils: fix typo for min() description

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

avfilter/vf_vectorscope: make background opacity customizable

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

avfilter/vf_vectorscope: add graticule

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

avfilter/vf_vectorscope: avoid crash by explicitly checking for 8-bit depth

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

lavc/aacenc_utils: replace sqrtf(Q*sqrtf(Q)) by precomputed value

It makes no sense whatsoever to do this at each function call; we
already have a table for this.

Yields a 2x improvement in find_min_book (x86-64, Haswell+GCC):
ffmpeg -i sin.flac -acodec aac -y sin.aac
find_min_book
old
    605 decicycles in find_min_book, 8388453 runs,    155 skips.9x
    606 decicycles in find_min_book,16776912 runs,    304 skips.9x
    607 decicycles in find_min_book,33553819 runs,    613 skips.2x
    607 decicycles in find_min_book,67107668 runs,   1196 skips.3x
    607 decicycles in find_min_book,134215360 runs,   2368 skips3x

new
    359 decicycles in find_min_book, 8388552 runs,     56 skips.3x
    360 decicycles in find_min_book,16777112 runs,    104 skips.1x
    361 decicycles in find_min_book,33554218 runs,    214 skips.4x
    361 decicycles in find_min_book,67108381 runs,    483 skips.5x
    361 decicycles in find_min_book,134216725 runs,   1003 skips5x

and more importantly a non-negligible speedup (~ 8%) to overall AAC encoding:
old:
ffmpeg -i sin.flac -acodec aac -strict -2 -y sin_new.aac  6.82s user 0.03s system 104% cpu 6.565 total
new:
ffmpeg -i sin.flac -acodec aac -strict -2 -y sin_old.aac  6.24s user 0.03s system 104% cpu 5.993 total

This also improves accuracy of the expression by ~ 2 ulp in some cases.

Reviewed-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
Reviewed-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Signed-off-by: Ganesh Ajjanagadde <gajjanag@gmail.com>

fate: add pipe and cache test

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

avformat/seek-test: Support passing options to demuxers and protocols

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

lavf/mov: Set display aspect ratio for avid dv.

Fixes ticket #5271.

sdp: fix opus sprop-stereo fmtp syntax

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

Move the |die| member of FrameThreadContext to PerThreadContext.

This fixes a data race warning by ThreadSanitizer.
FrameThreadContext.die is read by all the worker threads but is not
protected by any mutex. Move it to PerThreadContext so that each worker
thread reads its own copy of |die|, which can then be protected with
PerThreadContext.mutex.

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

sws/aarch64: add {nv12,nv21,yuv420p,yuv422p}_to_{argb,rgba,abgr,rgba}_neon

vc2enc: fix packet underallocation and minimum bitrate with interlacing

This was a regression introduced by commit e7345abe052 which
enabled full use of the allocated packet but due to the overhead of
using field coding the buffer was too small and triggered warnings and
crashes.

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

vc2enc: simplify slice cost caching

The fact that now all quantization indices costs are cached justifies
storing 20 more integers in a structure already allocated on heap.

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

vc2enc: remove redundant put_padding() and use skip_put_bytes() instead

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

lavfi: add bench and abench filters

avformat/rtpdec_h264: Ignore invalid sprop-parameter-sets missing PPS

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

lavu/rational: add more info regarding floor(x+0.5) usage

Add some more verbose info regarding why the imprecise and slow floor(x+0.5) hack
is used; helpful for future maintenance.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanag@gmail.com>

fate: add libavcodec utils test

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

avcodec: Add utils test

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

hls: Add and use a memebr of AVIOInternal rather than abuse opaque

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

fate: fix fate-libavformat target

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

Merge commit 'cae448cfbf31d492cba782bc64fc4eed556ed83d'

* commit 'cae448cfbf31d492cba782bc64fc4eed556ed83d':
  aviobuf: add a private data struct for avio_open()ed contexts

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

avformat/protocols: Fix ff_urlcontext_child_class_next()

This fixes -read_ahead_limit

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

avformat/msf: Also check the codec tag in probing

Fixes probing failure

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

Merge commit '832a202c47a246ed15e3edc6b05dfcfa7d82c4b2'

* commit '832a202c47a246ed15e3edc6b05dfcfa7d82c4b2':
  protocols: make the list of protocols static

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

avformat: Remove async from TESTPROGS

It is current disabled.

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

Merge commit '7d61dc95d741ca134d59b1f34c4e10c4c4e36f56'

* commit '7d61dc95d741ca134d59b1f34c4e10c4c4e36f56':
  lavf: move urlcontext_child_class_next() to protocols.c

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

Merge commit '0fa00d05911aa8043ecad8dead4a73cab7faadf6'

* commit '0fa00d05911aa8043ecad8dead4a73cab7faadf6':
  lavf: move avio_enum_protocols() to protocols.c

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

Merge commit '2758cdedfb7ac61f8b5e4861f99218b6fd43491d'

This commit also disables the async fate test, because it
used internal APIs in a non-kosher way, which no longer
exists.

* commit '2758cdedfb7ac61f8b5e4861f99218b6fd43491d':
  lavf: reorganize URLProtocols

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

x86/vc1dsp: Split the file into MC and loopfilter

Merge commit '225e84e74544062706c0159ec0737b0e1d40915f'

* commit '225e84e74544062706c0159ec0737b0e1d40915f':
  hls: disallow opening nested files in child demuxers

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

Merge commit 'e192cd9ce2b51c2e6919f2a78b1ce53e0024e728'

* commit 'e192cd9ce2b51c2e6919f2a78b1ce53e0024e728':
  smoothstreamingenc: do not open the files as read+write

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

Merge commit 'd082078a88da3b3e926197d0d2aa9fa322123b76'

* commit 'd082078a88da3b3e926197d0d2aa9fa322123b76':
  dashenc: eliminate ffurl_* usage

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

Merge commit '7fbb3b5b9857276b4cd17b2a530c7e0880d2bc0a'

* commit '7fbb3b5b9857276b4cd17b2a530c7e0880d2bc0a':
  lavf: use the io_open callbacks for files opened from open_input() as well

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

Merge commit '5efd91284e56d444139ed811671c59a129bbb92f'

This commit is no-op. The affected piece of code does not
exist in FFmpeg.

* commit '5efd91284e56d444139ed811671c59a129bbb92f':
  avprobe: do not call avio_close() on a custom context

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

Merge commit 'dc6527ed908e4d330738f139074455ffbe56a2de'

FATE tests have been updated to patch. They do not differ in
any meaningful way.

* commit 'dc6527ed908e4d330738f139074455ffbe56a2de':
  nutenc: do not use AVCodecContext.frame_size

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

lavf: Add pcx auto-detection.

lavc/pcxenc: Update format description link.

lavf/img2dec: Skip SOS when auto-detecting jpeg.

Improves jpeg auto-detection.

build: add --install-name-dir=DIR Darwin option

This option is typically useful when cross-compiling dynamic libraries
for iOS, with something such as --install-name-dir=@rpath

avfilter/avf_showcqt: optimize draw routines

optimize draw_bar_yuv (slightly faster)
optimize draw_axis (about 2x faster)

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

fate/mpeg4: add mpeg4-es with codec timestamps (vlc ticket 7571)

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

fate/vp8: add test for Ticket2451

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

avformat/utils: Be slightly more tolerant with fps vs. stream timebase

Fixes regression with ticket2451

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

fate/gif: add Test for Ticket3052

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

fate/filter-video: add test for Ticket1578

reference test sample is tiny

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

tests/fate/lossless-video: Add test for ticket4119

The reference file is very small, and there was no previous test for the -vsync values

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

avcodec/libzvbi-teletextdec: Remove unused variable

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

avfilter/avf_showvolume: support unknown channel layouts too

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

ffmpeg_vdpau: Remove unused variable

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

avformat/mov: do not leak memory on ffio_read_size failure

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

avformat/mov: merge mov_read_custom functions

This also fixes reading gapless metadata when the entries do not start with the
mean atom. Such samples can be found here:
https://hydrogenaud.io/index.php/topic,93310.0.html

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

avformat/dashenc: Enable dash output to work when the output isn't a local file

Use avpriv_io_move instead of ff_rename to support more than only
the file protocol.

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

avformat/file: enable file_move() without unistd.h

it only requires the rename function from os_support.h.

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

lavf/mov: downgrade sidx errors to non-fatal warnings; fixes trac #5216

lavf/mov: fix sidx with edit lists

vc2enc: redistribute leftover bytes

This commit redistributes the leftover bytes amongst the top 150 slices
in terms of size (in the hopes that they'll be the ones pretty bitrate
starved).

A more perceptual method would probably need to cut bits off from slices
which don't need much, but that'll be implemented later.

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

vc2enc: allocate the DWT context with the current plane size

Previously used the luma size only.

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

2enc: clip and warn when user bitrate set too low

The encoder crashed on verly low bitrates since there wasn't enough
space allocated.

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

vc2enc: clip and warn when user bitrate set too low

The encoder crashed on verly low bitrates since there wasn't enough
space allocated.

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

vc2enc: carry over quantization index across frames as a starting point

Previously a global average was used. Using the previous quantizer
resulted in a fairly significant speedup as slice size selection settled
down quicker.

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

vc2enc: use 32 bits for quantized coefficients LUT

16 bits were definitely not enough and caused artifacts to appear on
images at barely compressed images.

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

vc2enc: cache bits per quantizer, calculate wasted bits

Needed for following commits, also a speed increase.

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

avfilter/vf_vectorscope: add color5 mode, mode like color but with higher saturation

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

avcodec/mjpegdec: Fix decoding slightly odd progressive jpeg

Fixes: ebd58db6-dc86-11e5-91c2-59daeddf50c7.jpg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

LICENSE: Thorough editing

The changes in "Incompatible libraries" section are strictly based on
what our configure script says.

x86/vf_blend: Add SSE2 optimization for divide

 4.5x faster than C float version with autovectorization
10  x faster than C int version
25  x faster than C float version without autovectorization

lavf/mp3: Properly check return values of seeks and reads while reading the header

Fixes large amounts of seeking past EOF, which could be extremely
slow over a network.

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

avformat/Makefile: Fixed missing rawutils.o reference (required by movenc.c)

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

Document and validate AVFrame plane pointers.

Check that the required plane pointers and only
those are set up.
Currently does not enforce anything for the palette
pointer of pseudopal formats as I am unsure about the
requirements.

Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>

mjpegdec: Do not assume unused plane pointer are NULL.

We do neither document nor check such a requirement
and for application-provided get_buffer2 they could
contain the result of a malloc(0) or whatever value
they had previously.
This fixes a use-after-free in e.g. MPlayer:
https://trac.mplayerhq.hu/ticket/2262
We might want to consider changing the (documented)
API in addition though.

Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>

mss2: Fix buffer overflow.

Reported as https://trac.mplayerhq.hu/ticket/2264 but have
not been able to reproduce with FFmpeg-only.
I have no idea what coded_height is used for here exactly,
so this might not be the best fix.
Fixes the following chain of events:
ff_mss12_decode_init sets coded_height while not setting height.
ff_mpv_decode_init then copies coded_height into MpegEncContext height.
This is then used by init_context_frame to allocate the data structures.
However the wmv9rects are validated/initialized based on avctx->height, not
avctx->coded_height.
Thus the decode_wmv9 function will try to decode a larger video that we
allocated data structures for, causing out-of-bounds writes.

Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>

tests/lavf-regression: Add mp4 timecode test

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

movenc: Timecode in MP4 Although MP4 does not have a concrete specification to store timecode information, the following technical note from Apple describes a way to achieve this via timecode track, similar to how it is done for MOV files.

https://developer.apple.com/library/mac/technotes/tn2174/_index.html

- Enabled creation of timecode tracks for MP4 in the same way as MOV.
- Used nmhd as media information header of timecode track of MP4 instead
  of gmhd used in MOV, thus avoiding tcmi also, as recommended above.
- Bypassed adding source reference field for MP4, as suggested above.

Issue: https://trac.ffmpeg.org/ticket/4704

Signed-off-by: Syed Andaleeb Roomy <andaleebcse@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

configure: add direct detection of libopencv

The pkg-config file contains all opencv libraries, not only the
neccessary ones.

This change makes it possible to use the libopencv-imgproc-dev Debian
package instead of libopencv-dev, saving about 200 MB of useless
build-dependencies.

In particular one doesn't need to install the parts of opencv that
depend on ffmpeg libraries.

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

avutil/imgutils: Assert that the 2nd av_image_fill_linesizes() call in av_image_fill_linesizes() still succeeds

Fixes CID1271742

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

lavc/aac_ac3_parser: reindent

lavc/aac_ac3_parser: avoid zeroing codec parameters if we haven't read a frame

This caused issues when seeking in some unusual MPEGTS files

probe TrueHD in MPEGTS

lavc/Makefile: dnxhd demuxer depends on dnxhddata.o

vf_copy: exclude hwaccel formats

This does obviously not work.

(Additions based on vf_crop.c.)

avfilter/avf_showwolume: add orientation and step option

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

lavfi/ass: fix version check for sub_text_format option

avcodec/cfhd: Add support for 12-bit RGBA.

Plays all known samples

cfhd: reallocate internal buffers on format change.

Fixes some, but not all, of the threading fuzz crashes