Merge commit '7c57a582a03fb473091a88737ab92b9f2a5bb87a'

* commit '7c57a582a03fb473091a88737ab92b9f2a5bb87a':
  jpeg2000dec: don't use deprecated PIX_FMT values

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'ffba2053edfc177d217bf4a95edf51cd0fc40753'

* commit 'ffba2053edfc177d217bf4a95edf51cd0fc40753':
  lavfi: fix compatibility code for old vf_scale options syntax

Conflicts:
libavfilter/avfilter.c

No change as the buggy code does not exist in FFmpeg

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '096696ef0dd391d9430376d1444c1a3cde9171fd'

* commit '096696ef0dd391d9430376d1444c1a3cde9171fd':
  avfiltergraph: simplify inserting conversion filters.
  Clarify output of av_get_bits_per_pixel

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '7cbc4cb4424430586ce2d5c0dba81174c6d1db01'

* commit '7cbc4cb4424430586ce2d5c0dba81174c6d1db01':
  fate: update the mmf reference
  mmf: Write metadata into an Optional Data chunk
  jpeg2000dec: Use correct printf length modifier for pointer differences

Conflicts:
libavformat/mmf.c
tests/ref/lavf/mmf

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '5d9d8461fbb87046a7ca220d3676a045bd7ab46e'

* commit '5d9d8461fbb87046a7ca220d3676a045bd7ab46e':
  vc1dec: Don't apply the loop filter on fields

Conflicts:
tests/ref/fate/vc1_sa10143

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'fe06795d783785a76c6cbcaa2cd3012e6deb5fa5'

* commit 'fe06795d783785a76c6cbcaa2cd3012e6deb5fa5':
  vc1: implement frame interlaced b-frame header parsing

Conflicts:
libavcodec/vc1.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '46430fd47c6239ef8742d0a34f9412d5060fa798'

* commit '46430fd47c6239ef8742d0a34f9412d5060fa798':
  vc1dec: Don't attempt error concealment on field pictures
  vc1dec: fieldtx is only valid for interlaced frame pictures
  aacenc: Fix erasure of surround channels
  aacenc: Fix target bitrate for twoloop quantiser search

Conflicts:
libavcodec/vc1dec.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '7c71ada4cad3c6aab5fa24155c379465c41cfd76'

* commit '7c71ada4cad3c6aab5fa24155c379465c41cfd76':
  aacenc: Fix a rounding bug in aacpsy channel bitrate computation
  yadif: correct strides in filter_edges_16bit
  yadif: restore speed of the C filtering code

Conflicts:
libavfilter/x86/vf_yadif_init.c

No change as these commits where already in FFmpeg

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '252c0bfdc014c1fb6ad4fe06242c7beca58a6b41'

* commit '252c0bfdc014c1fb6ad4fe06242c7beca58a6b41':
  lavc: Fix msvc compilation of utils.c with -WX (warnings treated as errors)

Merged-by: Michael Niedermayer <michaelni@gmx.at>

jpeg2000dec: don't use deprecated PIX_FMT values

lavfi: fix compatibility code for old vf_scale options syntax

Currently it would incorrectly trigger on a string that contains a '='
but does not contain a ':', e.g. flags=<flags>.

avfiltergraph: simplify inserting conversion filters.

There is now no need to explicitly pass 0:0 as width/height to scale,
those are the defaults.

Clarify output of av_get_bits_per_pixel

Signed-off-by: Anton Khirnov <anton@khirnov.net>

lavfi/dctdnoiz: fix 10l typo in pr_height computation.

ffmpeg: free threads on error conditions.

Fixes Ticket2562

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

lavfi: add dctdnoiz filter.

avcodec/avpacket: Refactoring copy_side_data into a separate function

Refactoring copy_side_data into a separate function so that it can be called
in cases where side data needs to be duplicated.

Signed-off-by: Vignesh Venkatasubramanian <vigneshv@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

libx264: zero x264_picture before use.

This prevents use of uninitialized memory by ffmpeg later
(i_qpplus1)

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

libavfilter/Makefile: unbreak make alltools

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

tools: add zmqsend tool, useful to test the zmq filters

lavfi: add zmq filters

lavu/hash: Fix name of SHA224

Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

lavfi/blend: timeline support

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

swscale/output: silence compiler warnings about uninitialized variables

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

fate: update the mmf reference

Using Optional Data chunk introduced changes the hash.

doc/outdevs: fix typos (devices -> device)

lavfi: cosmetics: fix vertical alignment for pads in some filters

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

mmf: Write metadata into an Optional Data chunk

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

jpeg2000dec: Use correct printf length modifier for pointer differences

Also eliminate an unnecessary cast. Fixes the following warning:
jpeg2000dec.c:1200:20: warning: format ‘%lX’ expects argument of type ‘long unsigned int’, but argument 5 has type ‘long long unsigned int’

avformat/srtdec: Fix pointer corruption

This fixes use of uninitialized memory and possible out of array access

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

vc1dec: Don't apply the loop filter on fields

Fixes read of uninitialized memory.

Based on a patch by Michael Niedermayer.

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

vf_colorchannelmixer: round lut entries to nearest

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

fate: change colorchannelmixers coefficients to be less unstable

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

vc1: implement frame interlaced b-frame header parsing

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

vc1dec: Don't attempt error concealment on field pictures

This is not implemented and doesn't work.

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  configure: Disable PIC for cygwin just as for win32/mingw

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '1399931d07f0f37ef4526eb8d39d33c64e09618a'

* commit '1399931d07f0f37ef4526eb8d39d33c64e09618a':
  x86: dsputil: Rename dsputil_mmx.h --> dsputil_x86.h

Conflicts:
libavcodec/x86/dsputil_mmx.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

vc1dec: fieldtx is only valid for interlaced frame pictures

The fieldtx_plane is not cleared for interlaced fields.

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

aacenc: Fix erasure of surround channels

This was due to a miscomputation of s->cur_channel, which led to
psy-based encoders using the psy coefficients for the wrong channel.

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

aacenc: Fix target bitrate for twoloop quantiser search

This fixes a case where multichannel bitrate isn't accurately
targetted by psy model alone, never achieving the target bitrate.

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

aacenc: Fix a rounding bug in aacpsy channel bitrate computation

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

Merge commit '245b76a108585b6fb52eebc2626c472d6fa530dc'

* commit '245b76a108585b6fb52eebc2626c472d6fa530dc':
  x86: dsputil: Split inline assembly from init code

Conflicts:
libavcodec/x86/dsputil_mmx.c

Note, the author attribution is left in place and not removed
as it is in the merged commit.

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '46bb456853b197f4562de7acf5d42abf11ded9be'

* commit '46bb456853b197f4562de7acf5d42abf11ded9be':
  x86: dsputil: Refactor pixels16 wrapper functions with a macro

Conflicts:
libavcodec/x86/hpeldsp_avg_template.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

yadif: correct strides in filter_edges_16bit

The C code treats the data as arrays of uint16_t so strides must not
be in bytes but in pixels.

Signed-off-by: Anton Khirnov <anton@khirnov.net>

yadif: restore speed of the C filtering code

Always use the special filter for the first and last 3 columns (only).

Changes made in 64ed397 slowed the filter to just under 3/4 of what it
was.  This commit restores the speed while maintaining identical output.

For reference, on my Athlon64:
1733222 decicycles in old
2358563 decicycles in new
1727558 decicycles in this

Signed-off-by: Anton Khirnov <anton@khirnov.net>

lavc: Fix msvc compilation of utils.c with -WX (warnings treated as errors)

This fixes https://bugzilla.libav.org/show_bug.cgi?id=514.

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

doc/filters: add some examples for the fps filter

Signed-off-by: burek <burek021@gmail.com>

md5enc: don't reuse priv_class in two formats

Apparently, this is not allowed and can trip up all sorts of things.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Add a generic hash API.

Also use this API in framemd5.

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

Add 128 bit murmur3 hash function.

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

avfilter: Add 'drawgrid' video filter

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

configure: Disable PIC for cygwin just as for win32/mingw

If this is explicitly disabled for win32/mingw, it should also
be disabled for cygwin, for consistency and for the same reasons
as for win32/mingw.

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

Better handling for MXF essence reading reaching EOF.

If no metadata follows MXF essence, that is essence ends with EOF,
"Operation not permitted" error was produced and a non-zero code
was returned from ffmpeg executable.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/lcldec: Check that dimensions are a multiple of the subsample factors

Other dimensions would not work correctly currently,
also ask for a sample for files that fail this check.

This fixes an integer overflow leading to out of array
accesses.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

lavu/common: make FF_CEIL_RSHIFT faster when shift is constant.

See "[PATCH] lavfi/lut: use FF_CEIL_RSHIFT for chroma w/h rounding."
thread for more information.

fate: fix filter-colorchannelmixer by adding bitexact & accurate flags

Idea by durandal11707
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

imgconvert: silence "incompatible pointer type" warning

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

cdgraphics: initialize buffer

Prevent use of uninitialized memory

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

vorbisdec: check codebook entry count

Fixes assertion failure

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

tools/ffeval: use av_dynarray2_add()

Simplify, increment robustness.

lavu/mem: add av_dynarray2_add()

Based on a patch by Clément Bœsch.

See thread:
From: Clément Bœsch <ubitux@gmail.com>
Subject: [FFmpeg-devel] [PATCH 1/5] lavu: add av_dynarray_alloc_elem().
Date: Sun, 14 Apr 2013 03:07:54 +0200

lavu: define FF_MEMORY_POISON and use it

Allow single-place definition of constant used to fill poisoned memory.

Merge remote-tracking branch 'cehoyos/master'

* cehoyos/master:
  Do not read strd chunk in avi files as H264 extradata.

Merged-by: Michael Niedermayer <michaelni@gmx.at>

fate: add colorchannelmixer test

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

Do not read strd chunk in avi files as H264 extradata.

Fixes ticket #2561.

tests/lavf-regression: fix gbrp10 dpx test on big endian

For some reason only for that pix format decoder picks native format.

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  configure: x86: Only enable cpunop on i686

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'f54b55058a429c4eea5bae7e5bcb49bd29b34199'

* commit 'f54b55058a429c4eea5bae7e5bcb49bd29b34199':
  configure: Rename cmov processor capability to i686

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'e08c946c6860a78b0c479551d5f6735361160cbd'

* commit 'e08c946c6860a78b0c479551d5f6735361160cbd':
  configure: Explicitly disable PIC when targeting win32/mingw

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'b1803c79dcd6d0a345fa1cbe18dd8e2149717121'

* commit 'b1803c79dcd6d0a345fa1cbe18dd8e2149717121':
  configure: Use enable_weak when enabling pic

Merged-by: Michael Niedermayer <michaelni@gmx.at>

pngenc: Add 'dpi', 'dpm' options

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

fate: increase coverage for dpx encoder

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

dpxenc: simplifiy code using AVPixFmtDescriptor

Update rgb24 fate test as it no longer writes image in big endian
format.

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

dpxenc: do not set coded_frame

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

x86: dsputil: Rename dsputil_mmx.h --> dsputil_x86.h

The header is not (anymore) MMX-specific.

x86: dsputil: Split inline assembly from init code

Also remove some pointless comments.

x86: dsputil: Refactor pixels16 wrapper functions with a macro

sws/output: remove offset/coef for alpha in rgba64

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

swscale: RGBA64 output

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

configure: x86: Only enable cpunop on i686

The assembler may insert nopl instructions for cpunop, which are not
(universally) supported on i586 CPUs.

configure: Rename cmov processor capability to i686

The goal is to make the capapility slightly more general and have it
cover the availability of the nopl instruction in addition to cmov.

aacenc: Fix ticket #1784: erasure of surround channels

This was due to a miscomputation of s->cur_channel, which led to
psy-based encoders using the psy coefficients for the wrong channel.
Test sample attached on the bug tracker had the peculiar case of all
other channels being silent, so the error was far more noticeable.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

configure: Explicitly disable PIC when targeting win32/mingw

These platforms do not have any notion of PIC. On some compilers,
enabling pic produces a number of warnings.

This avoids trying to produce PIC loads in the ARM assembly - there
are no relocation types in PE/COFF that correspond to
BFD_RELOC_32_PCREL (R_ARM_REL32 in ELF).

As a side-effect, this avoids enabling PIC on mingw64, getting rid
of the warnings about PIC not having any effect on that platform.

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

configure: Use enable_weak when enabling pic

This allows it to be overridden, either by the user on the command
line, or by other sections of the configure script.

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

sws/packed_16bpc_bswap: fix handling of negative stride

Revert "sws: support endianness conversion for AV_PIX_FMT_RGBA64"

This reverts commit 23f7b14f2851d1d1a969c12b7e337826f0c74a43.

This commit broke fate

redspark: fix memleak in redspark_read_header

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

fate: add exr tests

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

sws: support endianness conversion for AV_PIX_FMT_RGBA64

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

doc/filters: split item options which are not aliases.

fate/filter: test gradfun timeline.

fate/filter: test hqdn3d timeline.

lavfi/hqdn3d: restore timeline feature.

lavfi/hqdn3d: use macros instead of hardcoded indexes.

lavfi: replace passthrough_filter_frame with a flag.

With the introduction of AVFilterContext->is_disabled, we can simplify
the custom passthrough mode in filters.

This commit is technically a small compat break, but the timeline was
introduced very recently.

Doxy by Stefano Sabatini.

lavfi/vflip: use FF_CEIL_RSHIFT for chroma w/h rounding.

lavfi/lut: use FF_CEIL_RSHIFT for chroma w/h rounding.

lavfi/select: unbreak compatibility with older PICT_TYPE syntax

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

Merge remote-tracking branch 'jamrial/adpcm'

* jamrial/adpcm:
  ADPCM IMA Radical decoder
  RedSpark demuxer
  RSD demuxer
  adpcm_thp: Allow the use of extradata for the adpcm table
  ADP demuxer
  ADPCM DTK decoder

Merged-by: Michael Niedermayer <michaelni@gmx.at>

ffv1enc: fix 1000l typo

Thanks to ubitux for spotting that
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>