h264: make sure the current picture is not made a long ref multiple times

Fixes possible invalid reads, once one of those refs is freed, but the
others remain.
CC: libav-stable@libav.org

avconv: do not stop processing the input packet on decoding error

We still want to flush the filters on EOF and possibly apply streamcopy.

avconv: factor out flushing the filters

This also ensures this is always done, avoiding infinite loops if an
error occurs at the end of the input.

h264: make sure the slices do not overlap during slice threading

Based on a patch by Michael Niedermayer <michaelni@gmx.at>.
CC: libav-stable@libav.org
Found-by: Kieran Kunhya <kierank@obe.tv>

avcodec/jpeg2000dec: Check for mismatching per expected termination

not just at the end

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

lavf/brstm: if the file lies about the last block's size, correct it

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

configure: Check for x265_api_get

Any other x265 symbol may not exported, e.g. if the build is a
multilib (10-bit and 8-bit in one) build.

This is the only symbol we directly call, and is available in the
build number we check for.

Fixes the configure check on multilib x265 builds.

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

configure: Check for x265_api_get

Any other x265 symbol may not exported, e.g. if the build is a
multilib (10-bit and 8-bit in one) build.

This is the only symbol we directly call, and is available in the
build number we check for.

Fixes the configure check on multilib x265 builds.

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

avcodec/jpeg2000dec: Limit poc end fields to whats available in jpeg2000_decode_packets()

Fixes: Ticket4677
Solution based on information from
https://code.google.com/p/openjpeg/issues/detail?id=80

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

avcodec/utils: Replace ENOTSUP by AVERROR_PATCHWELCOME

ENOTSUP is not available on all platforms

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

avcodec: Add flag for experimental HWAccels and use it for VDPAU/HEVC

This HWAccel isn't really usable right now due to an nvidia driver bug,
so we don't want it selected by default.

HWAccels have a capabilities field and there's a comment about flags,
but no flags exist today, so let's add one for experimental hwaccels.

fate/api-tests: Move api-flac-test to API tests directory.

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

fate/api-tests: Add directory and Makefile for API tests

The intention of this change is to allow separation of API tests from the
existing tests, and also to have a place for the API test source/executable
files so they're not mixed in with the actual library code.

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

Merge commit '92fdc80cab2acad9f171ba38a08aa89b392bdadd'

* commit '92fdc80cab2acad9f171ba38a08aa89b392bdadd':
  nvenc: Fix NV12 input

Conflicts:
libavcodec/nvenc.c

Not merged, the changed code is not in FFmpeg

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

Merge commit '072756cdd2f949462520041e357f52f15d8c274d'

* commit '072756cdd2f949462520041e357f52f15d8c274d':
  vc1dec: use get_bits_long and limit the read bits to 32

See: 1f1e0a2971b2a01f275bb5088c2e36166514be64
Merged-by: Michael Niedermayer <michaelni@gmx.at>

s302m: fix arithmetic exception

If nb_samples is zero, the bit_rate calculation results in a division by
zero.

Since ff_get_buffer fails if frame->nb_samples is zero, this can be
fixed by moving the bit_rate calculation after that function call.

That also makes it possible to reuse the already calculated
frame->nb_samples value.

Reviewed-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avcodec: Add support for per-frame AFD output in h264

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

nvenc: Fix NV12 input

avcodec/jpeg2000dec: Fix tp_index for POC

Fixes Ticket2586

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

avcodec/jpeg2000dec: Print what is found in place of EPH if EPH is not found

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

libvpx: disable unused function prototypes

Fixes make checkheader

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

avcodec/jpeg2000dec: Support progression order changes

Fixes Ticket4657

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

avcodec/jpeg2000dec: Parse POCs

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

vc1dec: use get_bits_long and limit the read bits to 32

get_bits should not be used with more than 25 bits.

Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

vc1dec: use get_bits_long and limit the read bits to 32

get_bits should not be used with more than 25 bits.

Reviewed-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

mpegaudiodec: copy AVFloatDSPContext from first context to all contexts

This fixes a segfault when decoding multi-channel MP3onMP4 files.

This is similar to commit cb72230d for MPADSPContext.

Reviewed-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

fate: test ea vp6 with alpha stream

Signed-off-by: Peter Ross <pross@xvid.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/hls: do not iterate to next sequence number if interruption is requested

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

avutil/log: modify AV_LOG_MAX_OFFSET for AV_LOG_TRACE

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

avcodec/jpeg2000dec: iterate in tile sample space for CPRL & RPCL

Thats what the spec says should be done
aka iterate in the wrong space and use special cases to patch that up.
It sometimes can result in different order

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

ffmpeg_opt: Fix sync_ist

This code was lost in 2663540821ee010edc7a47316b00a4745aaf973c

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

avcodec/jpeg2000dec: Add missing \n to av_log()

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

avcodec/jpeg2000dec: iterate over positions with the special cases from jpeg2000

The order in j2k is not the simple and logic one

Fixes Ticket4670

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

avcodec/jpeg2000dec: Add coords to Jpeg2000Tile

These will be needed in subsequent commits

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

avcodec/vdpau: Support for VDPAU accelerated HEVC decoding

This change introduces basic support for HEVC decoding through vdpau.
Right now, there are problems with the nvidia driver/library implementation
that mean that frames are incorrectly laid out in memory when they are
returned from the decoder, and it is normally impossible to recover the
complete decoded frame due to loss of data from alignment inconsistencies.

I obviously hope that nvidia will be fixing it in due course - I've verified
the problems exist with their example application.

As such, this support is not useful for any real world application, but I
believe that it is correct (with the caveat that the mangled frames may hide
problems) and will work properly once the nvidia problem is fixed.

Right now it appears that any file encoded by x265 or nvenc is decoded
correctly, but that's because these files don't use a bunch of HEVC
features.

Quick summary:

Features that seem to work:

1) Short Term References
2) Scaling Lists
3) Tiling

Features with known problems:

1) Long Term References

It's hard to tell what's going on here. After I read the nvidia example
app that does not set the IsLongTerm flag on LTRs, and changed my code,
a bunch of frames using LTR started to display correctly, but there
are still samples with glitches that are related to LTRs.

In terms of real world files, both x265 and nvenc only use short term
refs from this list. The divx encoder seems similar.

Signed-off-by: Philip Langdale <philipl@overt.org>

avcodec/jpeg2000: Remove CBLK limit

This also reduces the amount of memory needed
Fixes Ticket4672

The new code seems slightly faster as well, probably due to better cache usage

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

ffmpeg: Use 2 instead of STDERR_FILENO

STDERR_FILENO is not available on windows

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

avcodec/evrc: make it possible to disable postfilter

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

lavc+doc: adjust names to reflect ADPCM THP not being GameCube-only

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

lavf/brstm: expose the loop point when present

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

configure: Fix ranlib failure with older cross toolchains

Cross-compile toolchains without support for ranlib -D would fail.
This fixes the configure script to test the cross ranlib rather than the native ranlib.

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

modify exit message when signaled

Signed-off-by: rogerdpack <rogerpack2005@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

electronicarts: demux alpha stream

Electronic Arts VP6 files may contain two video streams: one for the
primary video stream and another for the alpha mask. The file format
uses identical data structures for both streams.

Signed-off-by: Peter Ross <pross@xvid.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

electronicarts: move video stream properties into dedicated structure

This is required for the alpha stream demux patch.

Signed-off-by: Peter Ross <pross@xvid.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

ffmpeg: log error message when shutting down from too many signals

write() suggested by wm4
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

vp9: fix profile check for intraonly frames.

Reviewed-by: James Zern <jzern@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

libavformat: Add H264 API test

Result differs in pkt_duration and time_base.den for some reason.
Right now it tests only one example (adjusted to match the output).

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

avcodec/jpeg2000dec: Limit x/y step to 1<<30

Fixes infinite loop
Fixes Ticket4663

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

avcodec/jpeg2000: Improve right bottom precinct positions

Fixes Ticket4656

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

avcodec/jpeg2000dec: Check PPx / PPy values

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

ffserver: make statement block unconditional by exiting early

Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>

ffserver: make statement block unconditional by exiting early

Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>

ffserver: unify comment formating & drop unneeded braces

Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>

ffserver: factor out stream params printing

Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>

ffserver: simplify URLContext cleanup

Drop unneeded var and avoid checking for NULL twice as
ffurl_closep() already does this.

Signed-off-by: Reynaldo H. Verdejo Pinochet <reynaldo@osg.samsung.com>

libvpxdec: report colorspace

Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Signed-off-by: James Zern <jzern@google.com>

libvpxdec: report profile for vp9

Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Signed-off-by: James Zern <jzern@google.com>

Merge commit '5dbd491eb38efab1d1313d4129ed76ab2e98176d'

* commit '5dbd491eb38efab1d1313d4129ed76ab2e98176d':
  TextureDSP: fix erroneous condition which produced blocky output for DXT

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

Merge commit '4d55484391112a30d16812597b9ad81605af0c24'

* commit '4d55484391112a30d16812597b9ad81605af0c24':
  dds: Fix palette mode on big endian

Conflicts:
libavcodec/dds.c

See: 21d2e3d602dc5f173901ba51f6a237b9dfce9e8e
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'ded5957d75def70d2f1fc1c1eae079230004974b'

* commit 'ded5957d75def70d2f1fc1c1eae079230004974b':
  segafilm: Remove deplanarization hack

Conflicts:
libavformat/segafilm.c
libavformat/version.h

See: 42b8f5fba1f067c55231791039283e41b5167247
Merged-by: Michael Niedermayer <michaelni@gmx.at>

avcodec: Add codec_desc and remapping for AV_CODEC_ID_PCM_S16BE_PLANAR_DEPRECATED

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

Merge commit '5a79bf0284a37855636390e620d189a964214c9e'

* commit '5a79bf0284a37855636390e620d189a964214c9e':
  PCM signed 16-bit big-endian planar decoder

Conflicts:
doc/general.texi
libavcodec/Makefile
libavcodec/allcodecs.c
libavcodec/avcodec.h
libavcodec/pcm.c
libavcodec/version.h
libavformat/nut.c

See: 9ba41ae63e079198ef4a9e83b5fa147e20966bdb
Merged-by: Michael Niedermayer <michaelni@gmx.at>

brstm: fix use of uninitialized variables

Fixes valgrind complaining about "Conditional jump or move depends on uninitialised value(s)"

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

avcodec/jpeg2000: change flags array type to 16bit

reduce memory needed for flags

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

vcodec/jpeg2000: Increase cblk size limit to 128

Fixes Ticket4649

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

avcodec/jpeg2000dec: Use 32x32->64bit for 9/7i dequantization

Like the 5/3 case this is needed to avoid overflows and similarly for 16bpp
output pure 32bit operations are insufficient if high quality is wanted

Note, this code-path is only used in bitexact mode, so this should not
affect the speed of any real use-case

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

avcodec/jpeg2000dec: Add 5/3 de-quantization special case for the lossless case

This avoids a multiplication and 64bits

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

avcodec/jpeg2000dec: Use 32x32->64bit for 5/3 dequantization

This fixes overflows, using fewer bits would impact high bit depth quality

Fixes Ticket4654

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

avfilter: add ssim filter

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

TextureDSP: fix erroneous condition which produced blocky output for DXT

dds: Fix palette mode on big endian

Found-By: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

segafilm: Remove deplanarization hack

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

PCM signed 16-bit big-endian planar decoder

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

lavf: set is_connected flag so url can be properly closed

Reviewed-by: Lukasz Marek <lukasz.m.luki2 at gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

ffmpeg: Do not use the data/size of a bitstream filter after failure

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

Merge commit '41740ef8be6ec409f7eff3932ddba9a9eeec27b1'

* commit '41740ef8be6ec409f7eff3932ddba9a9eeec27b1':
  libvpx: fix test for VPX_IMAGE_ABI_VERSION

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

Merge commit 'a88e21f5b78ef5e75b073ecbfd4e92921002a703'

* commit 'a88e21f5b78ef5e75b073ecbfd4e92921002a703':
  Fix missing symbols when libvpx has only vp8 enabled

Not merged, the FFmpeg libvpx-vp8 encoder does not use anything from libvpx.c

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

Merge commit 'c2de2cf0d2927f3f584dab6d54276fbda92a0a71'

* commit 'c2de2cf0d2927f3f584dab6d54276fbda92a0a71':
  arm64: constify src in h264qpel dsp function definitions

See: e16b7338d8777618936ff6ab9133e19f4db9069c
Merged-by: Michael Niedermayer <michaelni@gmx.at>

libvpx: fix test for VPX_IMAGE_ABI_VERSION

Fix missing symbols when libvpx has only vp8 enabled

Signed-off-by: Janne Grunau <janne-libav@jannau.net>

arm64: constify src in h264qpel dsp function definitions

avformat/mp3dec: Remove the ID3v1 tag removial code

The code is simply broken, the read packets are not aligned to
the mp3 frames, the file end or the id3 tag thus this simply
cannot reliably find the ID3v1 tag to remove it

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

avcodec/mpegaudio_parser: Discard ID3v1 tag at the end

Ideally this should be discarded by the demuxer but this is not
possible without fully parsing which would be then very similar
to this. The current ID3v1 discard code in the demuxer does not work
and will be removed in a subsequent commit

The discard code could be adjusted if needed to also discard tags at
other locations than the end or to limit this possibly to input
from the mp3 demuxer or even to move the discarding to the
decoder.

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

lavf/libssh: implement move and delete callbacks

Reviewed-by: Lukasz Marek <lukasz.m.luki2 at gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

lavf/libssh: read empty path from url as /

Reviewed-by: Lukasz Marek <lukasz.m.luki2 at gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

lavf/libssh: implement directory listing callbacks

Signed-off-by: Lukasz Marek <lukasz.m.luki2 at gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

lavf/libsmbclient: implement move and delete callbacks

Reviewed-by: Lukasz Marek <lukasz.m.luki2@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

lavf/brstm: add FATE tests for BFSTM and BCSTM files

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

avcodec/jpeg2000dwt: Replace /2 by >>1

Divisions can be slow if the compiler fails to replace them by shifts

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

swscale/x86/rgb2rgb_template: fix signedness of v in shuffle_bytes_2103_{mmx,mmxext}

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

swscale/x86/rgb2rgb_template: don't call emms on sse2/avx functions

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

swscale/x86/rgb2rgb_template: add missing xmm clobbers

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

avcodec/texturedspenc: Add () to protect macro / argument evaluation order

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

vaapi_h264: fix RefPicList[] field flags.

Use new H264Ref.reference field to track field picture flags. The
H264Picture.reference flag in DPB is now irrelevant here.

This is a regression from git commit d8151a7, and that affected
multiple interlaced video streams.

Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>

vda: unlock the pixel buffer base address.

The pixel buffer base address is never unlocked this causes
a bug with some pixel format types that are produced natively
by the hardware decoder: the first buffer was always used.
Unlock the pixel buffer base address fixes the issue.

movtextdec.c: Improve upon dynarrays and text_to_ass

Signed-off-by: Niklesh <niklesh.lalwani@iitb.ac.in>

swscale/rgb2rgb_template: Fix signedness of v in shuffle_bytes_2103_c()

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

swscale/rgb2rgb_template: Implement shuffle_bytes_0321_c and fix shuffle_bytes_2103_c on BE

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

avcodec/texturedsp: Add protective () to RGBA() macro

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

avformat/mpegts: Use STREAM_TYPE_PRIVATE_DATA instead of 6

Suggested-by: Wolfgang Lorenz <wl-chmw@gmx.de>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>