avcodec/h264: Fix undefined shifts in pack16to32() and pack8to16()

Reported-by: Thierry Foucu <tfoucu@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/mpegvideo: Fix undefined shifts in ff_init_block_index()

Reported-by: Thierry Foucu <tfoucu@google.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/012v: Check dimensions more completely

Fixes division by 0

Found-by: Thomas Lindroth <thomas.lindroth@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avfilter/aeval: Fixed a memory leak in EvalContext::channel_values

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

Force -D__EXTENSIONS__ on Solaris.

This fixes compilation of libavformat/udp.o on some installations,
regression since 3a5cbc91
Reported by Cristian Dustinta for SunOS 5.10

lavc/hevcdsp: Fix compilation for arm with --disable-neon.

lavd/avfoundation: Silence warnings when compiling for iOS.

Reviewed-by: Thilo Borgmann

lavd/avfoundation: Silence c99 warnings when using gcc.

Reviewed-by: Thilo Borgmann

lavd/qtkit: Silence deprecation warnings when using clang.

Reviewed-by: Thilo Borgmann

Merge commit '448c8cfe4c53e9e806effd8505b46d57fa707061'

* commit '448c8cfe4c53e9e806effd8505b46d57fa707061':
  movenc: Support setting fragment_index before the moov atom is written

Conflicts:
libavformat/movenc.c

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

Merge commit '0c5e380c2c266d2e8a13c000cc527529db837f10'

* commit '0c5e380c2c266d2e8a13c000cc527529db837f10':
  movenc: Don't rely on the fragment index for vc1 info gathering

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

Merge commit 'cf402d6fa88acd647cdff993429583bec8a34fdc'

* commit 'cf402d6fa88acd647cdff993429583bec8a34fdc':
  rtpenc_mpegts: Set chain->rtp_ctx only after avformat_write_header succeeded

Reviewed-by: Andreas Cadhalpun <andreas.cadhalpun@googlemail.com>
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'c83dd2d2a458075a58895c384372f57c1ec26276'

* commit 'c83dd2d2a458075a58895c384372f57c1ec26276':
  rtpenc_mpegts: Free the right ->pb in the error path in the init function

Reviewed-by: Andreas Cadhalpun <andreas.cadhalpun@googlemail.com>
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '96a06dbaf278e8152487e08772946f63bd2a3843'

* commit '96a06dbaf278e8152487e08772946f63bd2a3843':
  FATE: add support for testing hwaccels

Conflicts:
tests/Makefile
tests/fate-run.sh

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

Merge commit 'dc7536ca3d2dbe47f40cc0fcd0fc2555a84d5f56'

* commit 'dc7536ca3d2dbe47f40cc0fcd0fc2555a84d5f56':
  avconv: do not abort immediately if initializing hwaccel fails

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

movenc: Support setting fragment_index before the moov atom is written

This way, the caller doesn't need to coordinate setting the option
after the moov atom has been written. The downside is that it is
no longer possible to use the option for checking whether the moov
atom already has been written, but a caller is able to keep track
of that by other means anyway.

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

movenc: Don't rely on the fragment index for vc1 info gathering

The previous use of the mov->fragments field, for determining whether
written packets were part of the first fragment or not, didn't
work as intended when using the empty_moov flag.

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

rtpenc_mpegts: Set chain->rtp_ctx only after avformat_write_header succeeded

By making sure we at each time only have one pointer set, either a
local variable or one in the context, we avoid potential double frees
in the cleanup routines. If chain->rtp_ctx is set, it is closed by
calling avformat_write_trailer, but that shouldn't be called unless
avformat_write_header succeeded.

This issue was pointed out by Andreas Cadhalpun.

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

rtpenc_mpegts: Free the right ->pb in the error path in the init function

This fixes a typo from 8e32b1f096.

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

FATE: add support for testing hwaccels

avconv: do not abort immediately if initializing hwaccel fails

exit_program() will try to free the decoders, which is not a good idea
from within get_format().
Return an error instead.

avfilter/vf_fftfilt: increase RDFT length by 10%

This fixes artifacts when the dimensions are exact powers of 2

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

libavformat/mxfenc: write package name metadata

Previous-version-reviewed-by: Tomas Härdin <tomas.hardin@codemill.se>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

asfenc: fix leaking asf->index_ptr on error

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

roqvideoenc: set enc->avctx in roq_encode_init

So far it is only set in roq_encode_frame, but it is used in
roq_encode_end to free the coded_frame. This currently segfaults if
roq_encode_frame is not called between roq_encode_init and
roq_encode_end.

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

avcodec/libvpxenc: Clear twopass_stats.sz on deallocation

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

Merge commit '93f7948136fcda8ddbbc44a6c24418f11ca829b8'

* commit '93f7948136fcda8ddbbc44a6c24418f11ca829b8':
  libvpx: Fix mixed use of av_malloc() and av_reallocp()

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

avformat/siff: Fix checks and variable types

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

Merge commit 'ad94c6ca0b86c463f476b26606259a2041dcddc9'

* commit 'ad94c6ca0b86c463f476b26606259a2041dcddc9':
  siff: Use the correct type for packet size variables

Conflicts:
libavformat/siff.c

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

Merge commit '9f25a109922da43c1f81273a431d3b40cb5a785a'

* commit '9f25a109922da43c1f81273a431d3b40cb5a785a':
  matroskaenc: Also validate chapter end time

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

Merge commit 'bfeb83a8b7d3fcf09a54d8dbc9c521e10bb17530'

* commit 'bfeb83a8b7d3fcf09a54d8dbc9c521e10bb17530':
  rtpdec_hevc: Drop extra sanity check for size of input packet

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

avcodec/hevc_parser: use avpriv_find_start_code in hevc_split()

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

avcodec/options_table: remove extradata_size from the AVOptions table

allowing access to the size but not the extradata itself is not useful
and could lead to potential problems if writing happens through this field

Reviewed-by: Andreas Cadhalpun <andreas.cadhalpun@googlemail.com>
Reviewed-by: Lukasz Marek <lukasz.m.luki2@gmail.com>
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

tests/fate-run: Print more details on failure of stddev / filesize compares

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

ffmdec: limit the backward seek to the last resync position

If resyncing leads to the same position as previously, it will again
lead to a resync attempt, resulting in an infinite loop.

Thus don't seek back beyond the last syncpoint.

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

Add dependencies to configure file for vf_fftfilt

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

libvpx: Fix mixed use of av_malloc() and av_reallocp()

This buffer is resized when vpx_codec_get_cx_data() returns a
VPX_CODEC_STATS_PKT packet.

CC: libav-stable@libav.org
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

siff: Use the correct type for packet size variables

The avio functions used here return an unsigned value.
Also reduce a variable scope.

CC: libav-stable@libav.org
Bug-Id: CID 1258461

matroskaenc: Also validate chapter end time

This prevents it to be written as unsigned. Also add an error message.

CC: libav-stable@libav.org
Bug-Id: CID 1265717

rtpdec_hevc: Drop extra sanity check for size of input packet

In this case len is always at least 3, since it is checked against
RTP_HEVC_PAYLOAD_HEADER_SIZE + 1 before entering the switch block.

Bug-Id: CID 1238784

ffmdec: make sure the time base is valid

A negative time base can trigger assertions.

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

Merge commit '46d4d8575979a24a8d026d9805039b724e0e3e5f'

* commit '46d4d8575979a24a8d026d9805039b724e0e3e5f':
  movenc: Avoid writing separate flags for the first sample if not necessary

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

Merge commit '00d751d4fc20ec88d2cc2c9f39ec8b9e9c8cdeba'

* commit '00d751d4fc20ec88d2cc2c9f39ec8b9e9c8cdeba':
  movenc: Set tfhd default sample flags based on actual samples, if possible

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

doc: Clarify documentation for fade duration.

doc: Improve video creation examples (from image2).

The examples used to drop frames in most cases which
was probably not what users expected.

Fix the intra_dc_precision API doc.

The mpegvideo decoder has set intra_dc_precision since forever.

lavc/ffv1enc: Auto-select -coder 1 for >8bit also for yuv.

Reported-by: Christoph Gerstbauer

movenc: Avoid writing separate flags for the first sample if not necessary

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

movenc: Set tfhd default sample flags based on actual samples, if possible

This avoids assuming that e.g. audio samples are marked as
sync samples.

This allows omitting the sample flags from trun, if the default
flags happen to be right for all the samples.

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

ffmdec: fix infinite loop at EOF

If EOF is reached, while skipping bytes, avio_tell(pb) won't change
anymore, resulting in an infinite loop.

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

ffmdec: initialize f_cprv, f_stvi and f_stau

They are used in a switch statement, but it is not guaranteed that the
COMM case (where they are set to 0) is reached before the other cases.

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

Merge commit '0ce3a0f9d9523a9bcad4c6d451ca5bbd7a4f420d'

* commit '0ce3a0f9d9523a9bcad4c6d451ca5bbd7a4f420d':
  utvideodec: Handle slice_height being zero

See: 3881606240953b9275a247a1c98a567f3c44890f
Merged-by: Michael Niedermayer <michaelni@gmx.at>

utvideodec: Handle slice_height being zero

Fixes out of array accesses.

CC: libav-stable@libav.org
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Bug-Id: CVE-2014-9604
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

Merge commit '7b9cb7b36543c2a0e46d99b0e41824b9e7dd0c8f'

* commit '7b9cb7b36543c2a0e46d99b0e41824b9e7dd0c8f':
  avprobe: Export coded_{width,height} in -show_streams

Conflicts:
avprobe.c

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

avcodec/tiff: move bpp check to after "end:"

This ensures that all current and future code-pathes get bpp checked

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

Merge commit 'ae5e1f3d663a8c9a532d89e588cbc61f171c9186'

* commit 'ae5e1f3d663a8c9a532d89e588cbc61f171c9186':
  tiff: Check that there is no aliasing in pixel format selection

See: e1c0cfaa419aa5d320540d5a1b3f8fd9b82ab7e5
Merged-by: Michael Niedermayer <michaelni@gmx.at>

RELEASE: update to 2.6.git

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

ffserver: remove stale debug stanza

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

ffserver: move assignment outside condition eval

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

ffserver: break some too long lines

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

doc: avoid the incorrect phrase 'allow to'

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

avprobe: Export coded_{width,height} in -show_streams

avfilter/vf_edgedetect: Change enums to int, which are accessed via AVOption as int

This fixes depending on implementation defined behavior

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

avfilter/vf_curves: Change enums to int, which are accessed via AVOption as int

This fixes depending on implementation defined behavior

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

avfilter/vf_drawtext: Change enums to int, which are accessed via AVOption as int

This fixes depending on implementation defined behavior

fate: add test for vp90-2-trac4359.webm

Regression test for the bug from trac ticket #4359 fixed in commit efff3854

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

Merge commit '6448f15af02f2c3cf0df8cb8237957e426041f2d'

* commit '6448f15af02f2c3cf0df8cb8237957e426041f2d':
  mxfdec: Fix the error handling for when strftime fails

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

Merge commit '5f5b78aca35d07c771f5c4c73a984be9fe04a0b8'

* commit '5f5b78aca35d07c771f5c4c73a984be9fe04a0b8':
  xcbgrab: Free the host string after checking the connection

Conflicts:
libavdevice/xcbgrab.c

No change as the bug was not in FFmpeg

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

Merge commit '0025f7408a0fab2cab4a950064e4784a67463994'

* commit '0025f7408a0fab2cab4a950064e4784a67463994':
  vorbis: Check the vlc value in setup_classifs

Conflicts:
libavcodec/vorbisdec.c

See: ae038c0914460646503be083e30e3971093239a1
See: 709cae2bcbc0ea2c5d46c932b3d8301cf8f98e6b
Merged-by: Michael Niedermayer <michaelni@gmx.at>

tiff: Check that there is no aliasing in pixel format selection

Fixes possible issues with unexpected bpp/bppcount values.

CC: libav-stable@libav.org
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Bug-Id: CVE-2014-8544
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

Merge commit '62de77ffcaebd42fe685b8426da56b89b2532318'

* commit '62de77ffcaebd42fe685b8426da56b89b2532318':
  vorbis: Use a local codebook variable

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

mxfdec: Fix the error handling for when strftime fails

The str variable is a char ** here.

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

doc/filters: Add some random examples for fftfilt

Reviewed-by: Timothy Gu <timothygu99@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

vp9: included uses_2pass member in vp9_ref_frame().

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

xcbgrab: Free the host string after checking the connection

Prevent an use after free in the error message.

Bug-Id: CID 1274040
CC: libav-stable@libav.org

vorbis: Check the vlc value in setup_classifs

The valid returned values are always at most 11bit.
Remove the previous check that assumed larger values plausible and
use a signed integer to check get_vlc2 return values.

CC: libav-stable@libav.org

vorbis: Use a local codebook variable

Makes the code a little simpler.

vp9: split segmentation map / mvpair references.

This prevents a memcpy if segmentation.update_map == false.

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

avfilter/af_astats: redo zero channel handling

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

Merge commit '0f9f7969ef4df3661131cede3e8cc770e1ea3db8'

* commit '0f9f7969ef4df3661131cede3e8cc770e1ea3db8':
  rmdec: Improve message for demux error

See: 3880b4541ace2697f380ae1f43cb2299efeb2cc7
Merged-by: Michael Niedermayer <michaelni@gmx.at>

avfilter/af_astats: Avoid Zero division in print part.

rmdec: Improve message for demux error

Use correct context, reduce log level, don't assume it is a video stream,
and print the tag of the unknown stream.

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

avcodec/vc1_mc: optimize get_chroma_mv, and add get_luma_mv

Test results:

before:
  fate-suite/vc1/SA10091.vc1: 1062 decicycles in ff_vc1_mc_4mv_chroma's get mv, 4089 runs, 7 skips
  fate-suite/vc1/SA10143.vc1: 1112 decicycles in ff_vc1_mc_4mv_luma's get mv, 4096 runs, 0 skips
  fate-suite/vc1/SA20021.vc1: 991 decicycles in ff_vc1_mc_4mv_chroma get mv, 8192 runs, 0 skips

after:
  fate-suite/vc1/SA10091.vc1: 940 decicycles in ff_vc1_mc_4mv_chroma's get mv, 4096 runs, 0 skips
  fate-suite/vc1/SA10143.vc1: 786 decicycles in ff_vc1_mc_4mv_luma's get mv, 4096 runs, 0 skips
  fate-suite/vc1/SA20021.vc1: 876 decicycles in ff_vc1_mc_4mv_chroma get mv, 8189 runs, 3 skips

Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/mpegvideo_motion: Fix off by 1 error in MV bounds checking in qpel, chroma_4mv and 8x8

No testcase available

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

avcodec/mpegvideo_motion: Fix off by 1 error in MV bounds checking

Fixes Ticket4299

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

avcodec/opusdec: Fix delayed sample value

Fixes out of array access
Fixes: ffmpeg_opus_crash1.ogg
This solution is likely not optimal in terms of error concealment but
its simple and fixes the out of array access.

Found-by: Thomas Lindroth <thomas.lindroth@gmail.com>
Tested-by: Thomas Lindroth <thomas.lindroth@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/opusdec: Clear out pointers per packet

This is safer than to assume that all error pathes cleared them and
nothing will use uncleared pointers.

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

avcodec/utils: Align YUV411 by as much as the other YUV variants

Fixes out of array accesses
Fixes: ffmpeg_mjpeg_crash2.avi
Found-by: Thomas Lindroth <thomas.lindroth@gmail.com>
Tested-by: Thomas Lindroth <thomas.lindroth@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/vc1_mc: change the type of s_rndtblfield to uint8_t

Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/vc1_mc: Simplify v_edge_pos and src_y correction when fieldmv is true

Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/golomb: cleanup

Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

ChangeLog: Remove the redundant VP9 RTP entry

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

Merge commit 'b77e335e441040a40fc6156b8e4a134745d10233'

* commit 'b77e335e441040a40fc6156b8e4a134745d10233':
  configure: Move the .object_arch check to the right place

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

Merge commit 'dcae2e32f7d8a1ca5fb8c1e4aa81313be854dd73'

* commit 'dcae2e32f7d8a1ca5fb8c1e4aa81313be854dd73':
  arm: Suppress tags about used cpu arch and extensions

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

configure: Move the .object_arch check to the right place

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

avcodec/vc1_mc: remove useless variables

Signed-off-by: Zeng Zhaoxiu <zhaoxiu.zeng@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

arm: Suppress tags about used cpu arch and extensions

When all the codepaths using manually set .arch/.fpu code is
behind runtime detection, the elf attributes should be suppressed.

This allows tools to know that the final built binary doesn't
strictly require these extensions.

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

avfilter: Add FFT domain filter.

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

mips/asmdefs: change include guard to read AVUTIL_ instead of AVCODEC_

Signed-off-by: James Cowgill <james410@cowgill.org.uk>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Revert "avutil/opencl: is_compiled flag not being cleared in av_opencl_uninit"

Fixed build with opencl enabled
Found-by: WJ Liu
This reverts commit 0f2359b86926ed33da4bd64ca76d84d03d5ad380.