lavfi: add new iteration API

Signed-off-by: Josh de Kock <josh@itanimul.li>

checkasm/Makefile: add EXTRALIBS-libavformat

Signed-off-by: Josh de Kock <josh@itanimul.li>

arm: hevcdsp: Add commas between macro arguments

When targeting darwin, clang requires commas between arguments,
while the no-comma form is allowed for other targets.

Since Xcode 9.3, the bundled clang supports altmacro and doesn't
require using gas-preprocessor any longer.

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

arm: hevcdsp: Avoid using macro expansion counters

Clang supports the macro expansion counter (used for making unique
labels within macro expansions), but not when targeting darwin.

Convert uses of the counter into normal local labels, as used
elsewhere.

Since Xcode 9.3, the bundled clang supports altmacro and doesn't
require using gas-preprocessor any longer.

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

arm: swscale: Only compile the rgb2yuv asm if .dn aliases are supported

Vanilla clang supports altmacro since clang 5.0, and thus doesn't
require gas-preprocessor for building the arm assembly any longer.

However, the built-in assembler doesn't support .dn directives.

This readds checks that were removed in d7320ca3ed10f0d, when
the last usage of .dn directives within libav were removed.

Alternatively, the assembly could be rewritten to not use the
.dn directive, making it available to clang users.

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

avfilter/avf_showvolume: let fade be also exactly 0

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

doc/avfilter/showvolume : fix doc for orientation param

possible value are h or v.

avfilter/showvolume : indent after prev commit

avfilter/showvolume : calculate fade only if fade < 1.

avfilter/showvolume : add comment for orientation condition

doc/encoders: list missing options for x265

Add entries for crf and profile in libx265 section

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

Merge commit '3a7b4ae62c798edbd82bcd8fef863c74ed2acd4a'

* commit '3a7b4ae62c798edbd82bcd8fef863c74ed2acd4a':
  arm: Produce .const_data instead of .section .rodata for Mach-O

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

Merge commit 'ab05d3934de8e932dbd77979a687e6598e67535c'

* commit 'ab05d3934de8e932dbd77979a687e6598e67535c':
  arm: vc1dsp: Add commas between macro arguments

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

Merge commit 'e744281c49496b0e0a357e9f84c37fbf99215e20'

* commit 'e744281c49496b0e0a357e9f84c37fbf99215e20':
  configure: Revert some incorrect uses of check_cc()

This commit is a noop. The changes reverted here were never applied
to begin with.

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

Merge commit 'ea3320bb828553182fb34e164826f95df5743522'

* commit 'ea3320bb828553182fb34e164826f95df5743522':
  libaomenc: fix profile setting
  avcodec: rename the AV1 profiles

This commit is a noop, see 99cc3cf7a26cd4113e5047034577027d8c6f07a2

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

fate: add test for eac3 dependant stream

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

arm: Produce .const_data instead of .section .rodata for Mach-O

This is the same combination of .section directives as used in
aarch64/asm.S.

Since Xcode 9.3, the bundled clang supports altmacro and doesn't
require using gas-preprocessor any longer.

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

arm: vc1dsp: Add commas between macro arguments

When targeting darwin, clang requires commas between arguments,
while the no-comma form is allowed for other targets.

Since Xcode 9.3, the bundled clang supports altmacro and doesn't
require using gas-preprocessor any longer.

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

avcodec/mpc8: get frame output buffer right before it is actually needed

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

avcodec/mpc8: check for overread earlier and abort decoding frame

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

fftools/ffmpeg: Remove an unused variable.

Fixes a warning:
fftools/ffmpeg.c:2201:22: warning: unused variable 'fg'

fate: add eac3_core bitstream filter test

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

avcodec/ac3: fix out of array access introduced previously

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

configure: Revert some incorrect uses of check_cc()

avcodec/mpc8: check for overread first

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

movtextenc: fix handling of utf-8 subtitles

See the earlier fix for movtextdec for details. The equivalent bug is
present on the encoder side as well.

We need to track the text length in 'characters' (which seems to really
mean codepoints) to ensure that styles are applied across the correct
ranges.

avcodec/libaomenc: use av_assert0()

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

Merge commit 'a2fc8dbae85339d1b418d296f2982b6c04c53c57'

* commit 'a2fc8dbae85339d1b418d296f2982b6c04c53c57':
  Add Haivision SRT protocol

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

Merge commit '2124a97a4998413c7e81539b16b01ef6ac570ea9'

* commit '2124a97a4998413c7e81539b16b01ef6ac570ea9':
  configure: Drop unused helper function test_cflags_cpp()

This commit is a noop. The function was replaced in
af7a75cb517141f649cbbe0a9dcdb4854359b740

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

Merge commit '434b44cd6fb4bb9a2bf2bb29ef55ce1a315314b8'

* commit '434b44cd6fb4bb9a2bf2bb29ef55ce1a315314b8':
  configure: Simplify vararg check

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

Merge commit '8c7554e6a9b126bd6ee5bf80dae9e11e056db2f1'

* commit '8c7554e6a9b126bd6ee5bf80dae9e11e056db2f1':
  configure: Add check_x86asm() helper function to simplify some expressions

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

Merge commit '17ee5b0c13bc17465b71bc9ca1cde9f0eed8b3ff'

* commit '17ee5b0c13bc17465b71bc9ca1cde9f0eed8b3ff':
  configure: Use indirection for the -o assembler flag also for x86asm

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

Merge commit 'b9ea301e02472d0982b0fa0f80294bd95885bde8'

* commit 'b9ea301e02472d0982b0fa0f80294bd95885bde8':
  configure: Use a more sensible suffix for x86 assembly tempfiles

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

Merge commit '5292e97c42b05db7ad4e51c1ea756b12fdf721ff'

* commit '5292e97c42b05db7ad4e51c1ea756b12fdf721ff':
  configure: Document available options for the --toolchain parameter

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

Merge commit '86499771d1228d8303c8eb6509e20c0caaa02da5'

* commit '86499771d1228d8303c8eb6509e20c0caaa02da5':
  qsv: align surface width/height to 16.

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

Merge commit 'ea2f72a2c14c67a3b35dac6426d1e3c0fae33fd5'

* commit 'ea2f72a2c14c67a3b35dac6426d1e3c0fae33fd5':
  configure: Don't assume a 16 byte aligned stack on BSDs on i386

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

Merge commit 'deefca02c275ce4bc5ccbee690463ffef81a18b8'

* commit 'deefca02c275ce4bc5ccbee690463ffef81a18b8':
  qsvenc: add the Access Unit Delimiter NAL Unit support

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

Merge commit '3d6e76b953afd36e23ef8532b81aea58a6338931'

* commit '3d6e76b953afd36e23ef8532b81aea58a6338931':
  qsvenc: Fix a typo of FrameRateExtD/FrameRateExtN

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

Merge commit '847190ebd99ffd57dc89bd568a33bf2d5c424129'

* commit '847190ebd99ffd57dc89bd568a33bf2d5c424129':
  configure: Don't assume an aligned stack on clang on windows

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

avcodec/wmalosslessdec: Fix null pointer dereference in decode_frame()

Fixes: 2018_03_23_poc.wav
Found-by: GwanYeong Kim <gy741.kim@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/libaomenc: minor cosmetics

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

avformat/mpc8: do not return error on stream end

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

libaomenc: fix profile setting

Main Profile is yuv420p 8 and 10 bit
High Profile is yuv444p 8 and 10 bit
Professional Profile is yuv422p 8, 10, and 12 bit, plus every other pixfmt at 12 bit

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

avcodec: rename the AV1 profiles

Use the proper names instead of numbers

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

lavfi/silencedetect: Fix missing log at eos

Fixes ticket #6968

lavfi/silencedetect: Fix silence_end accuracy

lavfi/silencedetect: Fix silence_start accuracy

lavfi/silencedetect: Update test parameters

Set relevant filter parameters such that the result can easily be
checked with a waveform editor.
In particular, it makes it clear the silence_start is not accurate in
the current code.

lavfi/silencedetect: Fix when silence_start=0

0 is a valid value for silence_start

lavfi/silencedetect: Add mono mode

In mono mode, silence is detected in any single channel instead of all
of them simultaneously

doc/general.texi: fix warning

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

avcodec: add eac3_core bitstream filter

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

avcodec/eac3: add support for dependent stream

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

libavcodec/libaomenc: add support for transfer characteristics and color primaries

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

libavcodec/libaomdec: add support for transfer characteristics and color primaries

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

libavcodec/libaomenc: fix size specifier in an av_log call

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

libavcodec/libaomdec: use the matrix coefficients value from aom_image

"color_space" is no longer a sytax element in libaom.

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

avcode/profiles: add AV1 profiles

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

Merge commit '43778a501f1bfbceeddc8eaeea2ea2b3506beeda'

* commit '43778a501f1bfbceeddc8eaeea2ea2b3506beeda':
  Support AV1 encoding using libaom

This contains some extra changes taken from the libvpx encoder
wrapper, most of them contained in the set_pix_fmt() function.

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

avcodec/libaomdec: remove references to yuv440p pixfmt

While the enums are defined in the libaom headers, it's not officially
supported.

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

avcodec/tableprint_vlc: Fix build failure with --enable-hardcoded-tables

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

avfilter/af_pan: reject expressions referencing the same channel multiple times

Fixes parsing of expressions like c0=c0+c0 or c0=c0|c0=c1.  Previously no
error was thrown and for input channels, only the last gain factor was used,
for output channels the source channel gains were combined.

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

ffmpeg: do not finish output streams manually on eof even if no input is provided

The generic code should be able to finish the streams just fine initializing
and flushing the filters and codecs properly.

Fixes the following command:

ffmpeg -f lavfi -i "testsrc=d=0.1[out0];aevalsrc=0:d=0[out1]" -af apad -shortest -f framecrc -

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

ffmpeg: fallback to codecpar parameters on input filter eof

Fixes ticket #6854 and the following simpler case:

ffmpeg -f lavfi -i testsrc=d=1 -f lavfi -i testsrc=d=0 -filter_complex overlay -f null none

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

Merge commit 'c438899a706422b8362a13714580e988be4d638b'

* commit 'c438899a706422b8362a13714580e988be4d638b':
  Add AV1 video decoding support through libaom

This contains some extra changes taken from the libvpx decoder
wrapper, most of them contained in the set_pix_fmt() function.

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

Merge commit '44a1731011e87fbf4180d026aefb8bfe85d8c7dc'

* commit '44a1731011e87fbf4180d026aefb8bfe85d8c7dc':
  ivf: Support VP9 and AV1 as well

See e8f0a463b0d27a4d74c0d9f1857e72a8033d1069

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

Merge commit 'dd7e63af93b2430b5d42b87a966160c66736342c'

* commit 'dd7e63af93b2430b5d42b87a966160c66736342c':
  configure: Restore original endianness test

The test was not changed in 67e8f476b7d3c21686a2d453d052818ac92688b3
Merging only the explanation.

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

Merge commit '80a4e6a46f21256e9bf508ead686563616945ad5'

* commit '80a4e6a46f21256e9bf508ead686563616945ad5':
  amf: Replace writer_id option with LIBAV_AMF_WRITER_ID constant

This commit is a noop, see 9ea6607d294526688ab1b1342cb36ee159683e88

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

Merge commit '31a53ab34e22fe1eec902f79ec1f19ab828a7a0c'

* commit '31a53ab34e22fe1eec902f79ec1f19ab828a7a0c':
  configure: Add check_as() helper function to simplify some expressions

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

Merge commit '18dc1ff0fb4572b1d50a44905aa1e76bc3bbb0ad'

* commit '18dc1ff0fb4572b1d50a44905aa1e76bc3bbb0ad':
  configure: Add check_ld() helper function to simplify some expressions

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

Merge commit '9c37d765ef28b027414f86b0088b0c282a3c46d8'

* commit '9c37d765ef28b027414f86b0088b0c282a3c46d8':
  configure: Add check_cc/require_cc helper functions to simplify some expressions

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

Add Haivision SRT protocol

The protocol requires libsrt (https://github.com/Haivision/srt) to be
installed

Signed-off-by: Sven Dueking <sven.dueking@nablet.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

avformat/dashdec: Support signaling of last segment number

Last segment indicated by mpd is not parsed.
Example stream: http://dash.akamaized.net/dash264/TestCasesIOP41/LastSegmentNumber/1/manifest_last_segment_num.mpd

This patch supports parsing of Supplemental Descriptor with @schemeIdUri set to http://dashif.org/guide-
lines/last-segment-number with the @value set to the last segment number.

Reviewed-by: Steven Liu <lq@onvideo.cn>

hwcontext_d3d11: Fix crash with valid adapter but no device

This crash was introduced by 8bbf2dacbfb4ead1535dea411035994f507f517d,
which could incorrectly overwrite the failure result from creating the
device.

Fixes ticket #7108.

Reviewed-by: wm4 <nfxjfg@googlemail.com>

doc/examples/hw_decode: Remove logically dead code in decode_write()

Fixes CID1415951

Reviewed-by: Jun Zhao <mypopydev@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

doc/examples/hw_decode: Remove useless NULL check

Reviewed-by: Jun Zhao <mypopydev@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

doc/APIchanges: fix lavu version for the AVEncryptionInfo addition

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

avcodec/avpacket: remove unnecessary check in av_packet_make_writable()

Zero sized packets are already handled below in the function.
This is more in line with av_packet_ref().

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

configure: Drop unused helper function test_cflags_cpp()

avformat/hlsenc: initialize saveptrs

av_strtok calls strspn on a non-NULL *saveptr, so not NULL initializing
it is an issue.
Fixes CID #1428568

Reviewed-by: Karthick Jeyapal <kjeyapal@akamai.com>
Signed-off-by: Steven Liu <lq@onvideo.cn>
Signed-off-by: Steven Liu <lq@chinaffmpeg.org>

kmsgrab: add category for kmsgrab

Makes kmsgrab visible in "ffmpeg -devices".

Signed-off-by: Jun Zhao <mypopydev@gmail.com>
Signed-off-by: Mark Thompson <sw@jkqxz.net>

avfilter/vf_avgblur_opencl: fix error when clSetKernelArg fails

Fixes Coverity CID 1430382.

vf_avgblur_opencl: Don't run kernel on pixels outside the image

The output frame size is larger than the image containing a subsampled
plane - use the actual size of the image being written rather than the
dimensions of the intended output frame.

Reviewed-by: Dylan Fernando <dylanf123@gmail.com>

libavfilter: Add OpenCL convolution filter

Behaves like existing convolution filter.

Signed-off-by: Mark Thompson <sw@jkqxz.net>

avformat/segafilm - fix keyframe detection and set packet flags

Streams from a Segafilm cpk file can't be streamcopied because
keyframe flag isn't correctly set in stream index and
said flag is never conveyed to the packet

Fixes #7091

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

configure: Simplify vararg check

configure: Add check_x86asm() helper function to simplify some expressions

configure: Use indirection for the -o assembler flag also for x86asm

Similar indirections are used for the -o compiler/assembler flag to
account for differences in compiler/assembler syntax. For x86asm half
the infrastructure for doing the same currently exists unused.
Finish and use that infrastructure for consistency.

configure: Use a more sensible suffix for x86 assembly tempfiles

configure: Document available options for the --toolchain parameter

avcodec/avdct: use the proper function to free AVCodecContext

Fixes ticket #7074

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

avformat/rtpenc_chain: use the proper function to free AVFormatContext

Fixes ticket #7075

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

avcodec/mpeg4videodec: Use more specific error codes

Forward error codes where possible.

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

avformat/mov: Move +1 in check to avoid hypothetical overflow in add_ctts_entry()

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

avcodec/get_bits: Make sure the input bitstream with padding can be addressed

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

avfilter/af_mcompand: make error message more helpful

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

doc/filter.texi: fix some spotted typos

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

movtextdec: fix handling of UTF-8 subtitles

Subtitles which contained styled UTF-8 subtitles (i.e. not just 7 bit
ASCII characters) were not handled correctly. The spec mandates that
styling start/end ranges are in "characters". It's not quite clear what
a "character" is supposed to be, but maybe they mean unicode codepoints.

FFmpeg's decoder treated the style ranges as byte idexes, which could
lead to UTF-8 sequences being broken, and the common code dropping the
whole subtitle line.

Change this and count the codepoint instead. This also means that even
if this is somehow wrong, the decoder won't break UTF-8 sequences
anymore. The sample which led me to investigate this now appears to work
correctly.

avcodec: add a subcharenc mode that disables UTF-8 check

This is for applications which want to explicitly check for invalid
UTF-8 manually, and take actions that are better than dropping invalid
subtitles silently. (It's pretty much silent because sporadic avcodec
error messages are so common that you can't reasonably display them in a
prominent and meaningful way in a application GUI.)

avformat/hlsenc: use stream's maximum bit rate as fall-back advertised rate

Enables having proper bit rate values being written into the master
playlist in case of hard-constrained VBR where the maximum bit
rate utilized is known before hand.

Does the same thing as movenc.c, for example.

Signed-off-by: Jan Ekström <jan.ekstrom@aminocom.com>