Merge commit 'b4a82e740743744136e07054525560351e4892ae'

* commit 'b4a82e740743744136e07054525560351e4892ae':
  h264: Print the complete user message

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

Merge commit '8487987b87a252963b27b6adfd82887ea5933918'

* commit '8487987b87a252963b27b6adfd82887ea5933918':
  h264: Print user data SEI under normal debug verbosity

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

Merge commit '1720791e36f9cc24c05efea5bb275ab52156ce50'

* commit '1720791e36f9cc24c05efea5bb275ab52156ce50':
  png: read and write stereo3d frame side data information

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

Merge commit '00b62968d079e63bf22028f253ac297292436ebe'

* commit '00b62968d079e63bf22028f253ac297292436ebe':
  os_support: Don't try to return the service name as a string in getnameinfo

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

Merge commit '567ca142952c5be57e52c149c815dfe5d6ac6d41'

* commit '567ca142952c5be57e52c149c815dfe5d6ac6d41':
  configure: Add -D_CRT_NONSTDC_NO_WARNINGS when building with msvc

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

avformat/segafilm: set video stream duration

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

avformat/segafilm: set audio stream duration

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

avformat/segafilm: implement seeking

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

avformat: unref packet after storing it in internal packet queue

Fixes a memory leak when using genpts

avformat: always unref the packet after parsing

This fixes a memory leak when side-data is present.

avformat/ipmovie: add context to av_log()

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

avformat/ipmovie: put video decoding_map_size into packet and use it in decoder

The size of decoding map can differ from one calculated
internally, producing artifacts while decoding video.

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

vf_boxblur: Templatize blur{8,16}

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

avutil/eval: minor typo

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

pixblockdsp: Use AV_COPY128U for get_pixels_16_c

Before:
  15543 decicycles in get_pixels, 4193214 runs,   1090 skips
After:
   5713 decicycles in get_pixels, 8387564 runs,   1044 skips

avcodec/h264_slice: Disable slice threads if there are multiple access units in a packet

Fixes null pointer dereference
Fixes Ticket4977

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

avformat/icecast: Initialize .str based default for AV_OPT_TYPE_STRING types

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

avformat/http: Initialize .str based default for AV_OPT_TYPE_STRING types

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

avcodec/libopenh264enc: use AV_OPT_TYPE_BOOL for allow_skip_frames option

avutil/x86/intmath: Add missing header.

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

avcodec/nvenc: fix potential profile error when encoding yuv444p

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

avutil/opencl_internal: add av_warn_unused_result

clSetKernelArg can return an error due to lack of memory (for instance):
https://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clSetKernelArg.html.
Thus this error must be propagated.

Currently should not trigger warnings, but adds robustness.
Untested.

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

avfilter/vf_deshake: replace qsort with AV_QSORT

qsort is called indirectly in filter_frame, suggesting its performance
criticality. AV_QSORT is substantially faster due to the inlining of the
comparison callback. Thus, the increase in performance should be worth
the increase in binary size.

This optimization is just a low hanging fruit. The trac ticket 1430 is
a request for an improved deshake filter.

Sample benchmark (x86-64, Haswell, GNU/Linux):
File: original from https://trac.ffmpeg.org/ticket/1430
command: ffmpeg -stream_loop 8 -i file.webm -vf deshake=rx=64:ry=64 -f null -

Timer truncated at 1024 runs.
new:
28260 decicycles in qsort,       1 runs,      0 skips
35570 decicycles in qsort,       2 runs,      0 skips
39010 decicycles in qsort,       4 runs,      0 skips
46897 decicycles in qsort,       8 runs,      0 skips
40442 decicycles in qsort,      16 runs,      0 skips
41611 decicycles in qsort,      32 runs,      0 skips
40345 decicycles in qsort,      64 runs,      0 skips
38967 decicycles in qsort,     128 runs,      0 skips
38647 decicycles in qsort,     256 runs,      0 skips
40238 decicycles in qsort,     512 runs,      0 skips
39676 decicycles in qsort,    1024 runs,      0 skips

old:
1740280 decicycles in qsort,       1 runs,      0 skips
 923560 decicycles in qsort,       2 runs,      0 skips
 511330 decicycles in qsort,       4 runs,      0 skips
 309720 decicycles in qsort,       8 runs,      0 skips
 194900 decicycles in qsort,      16 runs,      0 skips
 142686 decicycles in qsort,      32 runs,      0 skips
 112516 decicycles in qsort,      64 runs,      0 skips
  98166 decicycles in qsort,     128 runs,      0 skips
  88147 decicycles in qsort,     256 runs,      0 skips
  88706 decicycles in qsort,     512 runs,      0 skips
  86783 decicycles in qsort,    1024 runs,      0 skips

Reviewed-by: Nicolas George <george@nsup.org>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

avutil/x86/intmath: Use tzcnt in place of bsf.

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

gdigrab: grab right desktop size if DPI in use, based on patch from Alexander Brotzge

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

avutil/rational: use frexp rather than ad-hoc log to get floating point exponent

This simplifies and cleans up the code.
Furthermore, it is much faster due to absence of the slow log computation.

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

avdevice/caca: change AV_OPT_TYPE_STRING to set .str default

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

avcodec/libopenh264enc: AV_OPT_TYPE_STRING should use .str default

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

avcodec/libopenh264enc: Fix project name

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

avcodec/libopenh264enc: set .i64 for AV_OPT_TYPE_INT

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

avfilter/avf_showcqt: Replace all fmin* and fmax* by FFMIN/FFMAX

Should fix build on x86_32-msvc2012

The alternative of emulating fmin/fmax* turns out to be non trivial

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

avfilter/showcqt: fix dependency with avformat

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

ffprobe: add support for printing packet strings metadata as packet tags

ffprobe.xsd already contains the tag element.

Reviewed-by: Stefano Sabatini <stefasab@gmail.com>
Signed-off-by: Marton Balint <cus@passwd.hu>

all: fix enum definition for large values

ISO C restricts enumerator values to the range of int. Thus (for instance) 0x80000000
unfortunately does not work, and throws a warning with -Wpedantic on
clang 3.7.

This fixes it by using alternative expressions that result in identical
values but do not have this issue.

Tested with FATE.

Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

avutil/wchar_filename: add av_warn_unused_result

Current code is fine, this just adds robustness.

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

avutil/mathematics: correct documentation for av_gcd

av_gcd is now always defined regardless of input. This documents this
change in the "documented API". Two benefits (closely related):
1. The function is robust, and there is no need to worry about INT64_MIN, etc.

2. Clients of av_gcd, like av_reduce, can now be made fully correct. Currently,
av_reduce can trigger undefined behavior if e.g num is INT64_MIN due to
integer overflow in the FFABS. Furthermore, this undefined behavior is
completely undocumented, and could be a fuzzer's paradise. The FFABS was needed in the past as
av_gcd was undefined for negative inputs. In order to make av_reduce
robust, it is essential to guarantee that av_gcd works for all int64_t.

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

h264: Print the complete user message

Previously the message was cut off at 256th byte.
This will allow dumping the complete x264 encode info when needed.

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

h264: Print user data SEI under normal debug verbosity

Drop the need of setting -debug bugs since it's not a bug, and the
message is already under a AV_LOG_DEBUG log level. Instead only print
it when there is an actual string in it.

png: read and write stereo3d frame side data information

Use optional sTER chunk defining side-by-side stereo pair
within "Extensions to the PNG 1.2 Specification", version 1.3.0.

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

ffmpeg_filter: check encoder before using it to set frame size.

Fix a segfault when no encoder is found for a default codec
with a complex filter graph.

os_support: Don't try to return the service name as a string in getnameinfo

Some systems may be lacking getservbyport; the previous ifdef wasn't
quite enough since it still assumed that struct servent was defined,
as pointed out by Clément Gregoire.

Simply remove the possibility to return non-numeric services in
getnameinfo; no caller of getnameinfo within libavformat
currently try to use getnameinfo for retrieving the port number without
NI_NUMERICSERV, and falling back on getservbyport may be non-threadsafe.

Signed-off-by: Martin Storsjö <martin@martin.st>

vf_lut: Add support for RGB48 and RGBA64.

Signed-off-by: Steven Robertson <steven@strobe.cc>
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/3dostr: Remove redundant ;

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

avformat/oggparseogm: Enable parser for mpeg4

Fixes regression with iJi.ogv

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

lavc/proresdec2: Fix slice_count for very high resolutions.

QT ignores the value written in the frame header.
Issue reported by forum user Koracas.

lavc/cdg: Add transparency support.

avutil/mathematics: make av_gcd more robust

This ensures that no undefined behavior is invoked, while retaining
identical return values in all cases and at no loss of performance
(identical asm on clang and gcc).
Essentially, this patch exchanges undefined behavior with implementation
defined behavior, a strict improvement.

Rationale:
1. The ideal solution is to have the return type a uint64_t. This
unfortunately requires an API change.
2. The only pathological behavior happens if both arguments are
INT64_MIN, to the best of my knowledge. In such a case, the
implementation defined behavior is invoked in the sense that UINT64_MAX
is interpreted as INT64_MIN, which any reasonable implementation will
do. In any case, any usage where both arguments are INT64_MIN is a
fuzzer anyway.
3. Alternatives of checking, etc require branching and lose performance
for no concrete gain - no client cares about av_gcd's actual value when
both args are INT64_MIN. Even if it did, on sane platforms (e.g all the
ones FFmpeg cares about), it produces a correct gcd, namely INT64_MIN.

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

avfilter/avf_showvolume: use log10 instead of log()/M_LN10

This is likely more precise and conveys the intent better.

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

ffmpeg: use log10 instead of log()/log(10)

This is more concise and conveys the intent better.
Furthermore, it is likely more precise as well due to lack of floating
point division.

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

avfilter/vf_ssim: use log10 instead of log()/log(10)

This is likely more precise and conveys the intent better.

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

avfilter/avf_showspectrum: use log10 instead of log()/...

This is likely more precise and conveys the intent better.

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

avfilter/vf_psnr: use log10 instead of log()/log(10)

This is likely more precise and conveys the intent better.

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

swscale: rename sws_rgb2rgb_init to ff_sws_rgb2rgb_init

It is an internal swscale function and thus should not be exported.

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

swscale: rename sws_context_class to ff_sws_context_class

It is an internal swscale symbol and thus should not be exported.

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

avformat: stop exporting ffurl_read_complete, ffurl_seek and ffurl_size

They are not in public headers and not used outside of libavformat.

Reviewed-by: Hendrik Leppkes <h.leppkes@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avfilter/af_volume: use log10 instead of log()/M_LN10

This is likely more precise and conveys the intent better.

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

avcodec/zmbvenc: use log2 instead of log()/M_LN2

This is likely more precise and conveys the intent better.

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

avcodec/nellymoserenc: use log2 instead of log()/M_LN2

This is likely more precise and conveys the intent better.

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

avcodec/snowenc: use log2 instead of log() / log(2...)

This is likely more precise and conveys the intent better.
The expression has also been accordingly simplified.

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

doc/general.texi: fix SDX2 description

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

doc/general.texi: move Screenpresso to right section

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

avformat: add DC STR demuxer

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

avcodec: add ADPCM AICA decoder

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

avformat: add 3dostr demuxer

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

avfilter/tremolo: fix wavetable buffer size

Signed-off-by: Kyle Swanson <k@ylo.ph>
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/asrc_sine: fix options typos

Signed-off-by: Kyle Swanson <k@ylo.ph>
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

configure: Add -D_CRT_NONSTDC_NO_WARNINGS when building with msvc

This silences warnings like this one:
libavformat/file.c(62) : warning C4996: 'read': The POSIX name for this
item is deprecated. Instead, use the ISO C++ conformant name: _read.
See online help for details.

Signed-off-by: Martin Storsjö <martin@martin.st>

Merge commit '8edaf625f3c38c695c33745822182e94e17d6e1b'

* commit '8edaf625f3c38c695c33745822182e94e17d6e1b':
  libopenh264enc: Count and warn about the number of skipped frames

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

Merge commit '9e14a992409348630ea96521411127418a92e8a3'

* commit '9e14a992409348630ea96521411127418a92e8a3':
  libopenh264enc: Add an option for controlling the frame skipping

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

Merge commit 'b8deb7c34f755d5e3eee0b5930c3a6ad2dda96bc'

* commit 'b8deb7c34f755d5e3eee0b5930c3a6ad2dda96bc':
  opus: Do not call vector_fmul_scalar on zero samples

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

Merge commit '6dcd82290438eb810936de4febd0558b38a08ec5'

* commit '6dcd82290438eb810936de4febd0558b38a08ec5':
  opus: Buffer the samples from the correct offset

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

Merge commit 'f3aff31e5f66a4f1c4e34ce4944f4a402aca61ed'

* commit 'f3aff31e5f66a4f1c4e34ce4944f4a402aca61ed':
  avpicture: Deprecate the single fields

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

Merge commit 'cd0e08813a0484002b5defbf557c859f123953ae'

* commit 'cd0e08813a0484002b5defbf557c859f123953ae':
  avconv: support infinite loop for the loop option

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

Merge commit 'a9a60106370f862e191dea58e748626da6a8fe97'

* commit 'a9a60106370f862e191dea58e748626da6a8fe97':
  avpacket: Provide an alloc and a free function for the struct

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

Merge commit '9b56d5c11488656254e9aed8d964ef2b7c2ff5e6'

* commit '9b56d5c11488656254e9aed8d964ef2b7c2ff5e6':
  avpacket: Deprecate av_dup_packet

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

Merge commit 'd584533cf38141172e20bae5436629ee17c8ce50'

* commit 'd584533cf38141172e20bae5436629ee17c8ce50':
  avformat: Rework add_to_pktbuf

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

vp9: update timestamps in ref files using multiple invisible frames.

For the 10-show-existing-frame, the source file indeed has a timestamp
of 3 (or 100/33) for the second visible frame, so the fix appears to
work correctly. For the other, only the timebase is fixed, but again
appears to be correct now.

lavc/mjpegdec: honor skip_frame option

lavc/pngdec: honor skip_frame option

avfilter/showcqt: fix compilation with libswscale disabled

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

avutil/audio_fifo: add av_warn_unused_result

This one should not trigger any warnings, but will be useful for future
robustness.

Strictly speaking, one could check the size after the call by examining
the structure instead of the return value. Such a use case is highly
unusual, and this commit may be easily reverted if there is a legitimate
need of such use.

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

vp9_parser: don't overwrite cached timestamps with nopts.

avcodec/motion_est_template: replace qsort with AV_QSORT

This code is in a performance critical section. AV_QSORT is
substantially faster due to the inlining of the comparison callback.
Thus, the increase in performance is worth the increase in binary size.

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

avutil/ripemd: make rol macro more robust by adding parentheses

This ensures that the macro remains correct in the sense of allowing
expressions for value and bits, by placing the value and bits expressions within
parentheses.

Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

doc/encoders: fix "the their" typo

Signed-off-by: Lou Logan <lou@lrcd.com>

libopenh264enc: Count and warn about the number of skipped frames

Signed-off-by: Martin Storsjö <martin@martin.st>

libopenh264enc: Add an option for controlling the frame skipping

This allows enabling the frame skipping, which is required for the
encoder to properly hit the target bitrate.

Signed-off-by: Martin Storsjö <martin@martin.st>

doc/filters: Remove article redundancy

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

swr: do not reject channel layouts that use channel 63

Channel layouts are essentially uint64_t, and every value is valid.

opus: Do not call vector_fmul_scalar on zero samples

The x86 variant of this function crashes in that specific case.

CC: libav-devel@libav.org

opus: Buffer the samples from the correct offset

When not all the opus stream have the same amount of decoded samples
process the least amount and store what is left from the other streams.

Bug-Id: 909
CC: libav-stable@libav.org
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

lavu/aes: test CBC functionality

avcodec/png: read and write stereo3d frame side data information

Use optional sTER chunk defining side-by-side stereo pair
within "Extensions to the PNG 1.2 Specification", version 1.3.0.

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

lavu/aes: align AVAES struct members

lavu/aes: add runtime dispatch for crypt function

lavu/aes: move AVAES to separate internal header

lavu: add AESNI CPU flag

ffprobe: Remove abort()

Found-by: Ganesh Ajjanagadde and wm4
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avutil/avstring: add av_warn_unused_result

This does not trigger any warnings, but adds robustness.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

avfilter/avf_showcqt: Fix ;;

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