lavu/parseutils: allow MM:SS format for duration in av_parse_time()

Fix trac ticket #2258.

Signed-off-by: Senthilnathan M <senthilnathan.maadasamy@gmail.com>
Signed-off-by: Stefano Sabatini <stefasab@gmail.com>

Improve QTRLE encoding performance, no change to output file size/content.

Avoid searching for the lowest bulk cost for each pixel that isn't a repeat/skip. Instead store the lowest cost as we go along each pixel, and use it as needed.

Signed-off-by: Malcolm Bechard <malcolm.bechard@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avutil/mem: do a small set of checks for memalign hack before freeing.

These can detect some kinds of memory and or pointer corruptions

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

exr: shorter channel buffer names

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

exr: remove superfluous check

get_header_variable_length() already check for
out of array read and returns 0 in such case.

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

exr: cosmetics: reindent

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

exr: check if line offset table is complete only once per frame

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

exr: support more line orders

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

exr: fix overflow check in lineoffset processing

Also read data size for raw compressions too and
make sure its value is sane.

Remove code that fills missing blocks with zeroes.
It is marginally useful and make implementation
of actually useful features harder.

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

exr: merge multiple if () blocks

Remove useless comments.

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

exr: check version too

Separate version from flags.

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

doc/filters: apply minor spacing fixes to boxblur docs

lavfi/boxblur: add support to named options

aac: reconfigure output on pop

Fixes Ticket1918

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

cafenc: do not leak caf->pkt_sizes

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

aacdec: dont pop random configurations

Fixes out of array accesses

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

mxg: do not leak mxg->buffer

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  x86: h264: Don't use redzone in AVX h264_deblock on Win64

No code change, as the commit has been already integrated
merge just for metadata

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

dxva2: move _WIN32_WINNT handling to public header

This removes duplicate code

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

libavcodec/mips/aacsbr_mips: fix ;;

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

mips: optimization for float aac decoder (ps module)

Signed-off-by: Mirjana Vulin <mvulin@mips.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

mips: optimization for float aac decoder (sbr module)

Signed-off-by: Mirjana Vulin <mvulin@mips.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

configure: Use libiconv on systems that lack built-in iconv

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

Fix Win64 AVX h264_deblock by not using redzone on Win64

Thanks-to: "Ronald S. Bultje" <rsbultje@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

h264: ensure that get_format() is called when changing format but not otherwise.

Fixes Ticket2288

Tested-by: Stefano Pigozzi <stefano.pigozzi@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

matroska: fix missing ,

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

avcodec_decode_audio4: check got_frame_ptr before handling initial skip

Fixes out of array accesses

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

blend filter

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

h264: move next/last picture init to decode_slice_header

This is a regression introduced from the h264/mpegvideo split
Fixes out of array reads

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

ec: check that reference pictures have matching parameters before using them.

Fixes out of array reads

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

h264: check that luma and chroma depth match

Fixes out of array access

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

x86: h264: Don't use redzone in AVX h264_deblock on Win64

This fixes crashes in chromium on win64 on machines with AVX
(crashes that apparently aren't triggered by fate).

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

exr: constify s in rle_uncompress()

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

exr: reset compression in decode_frame()

This fixes decoding of images if compression changes.
Regression since b040ffc84c5e.

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  dxva2: Add missing #define to make header compile standalone
  arm: vp8: Add missing #includes for header to compile standalone
  doc: filters: Correct BNF FILTER description

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

ffmpeg: set resample_pix_fmt for sub2video.

Forgotten when the resample_* fields were added.
Fix assert failures.

lavfi/buffersrc: set channel layout if it is known.

If buffersrc was configured for frames with an unknown layout,
the incoming frames will have channel_layout = 0.
If the format negotiation has selected a known (and compatible)
channel layout for the link, the frame is assumed to have
that layout, the field must be set before injecting the frame
in the filters.

Use SMPTE standard notation when showing mxf Universal Labels.

vf_idet: adapt default scores due to bugfix

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

vf_idet: Fix macro arguments sideeffect

Fixes valgrind errors
should improve detection scores

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

fate/subtitles: add MicroDVD remux test.

lavf/microdvd: fix muxing.

This was broken since 1f265f52.

lavf/swfdec: remove 8-bit audio FIXME.

Implemented since 624fb5f9.

brender_pix: use a grayscale palette as fallback instead of a uninitialized one.

Improves: 3-0-CORPSE.PIX

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

mjpegdec: check for packed rgb mjpeg.

Fixes null pointer dereference

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

Fix bits_per_coded_sample when encoding png with frame-level multithreading.

Fixes ticket #2290.

lavf/swfdec: support 8-bits PCM audio.

lavf/swfdec: factorize the creation of a new stream.

This also makes the changes of a3949fe11 applicable in both cases.

h264: Fix handling of changing reference counts between slices

Fixes out of array accesses

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

h264: call alloc_scratch_buffers() only once linesize is known

Fixes out of array accesses

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

lavfi/subtitles: fix recently introduced warning about discarded const qualifier.

lavfi/overlay: add format option

In particular, fix misbehavior in case main and overlay input formats
mismatch (e.g. YUV420 and YUV444).

lavfi/overlay: apply minor consistency fixes

lavfi/overlay: implement shortest option

Force termination when the overlay stream ends. Simplify scripting logic,
for example when an infinite source is used to generate a background for
a composite video.

doc/filters: rework introductory description for decimate

In particular, dedicate first paragraph to deliver a short description of
the filter.

lavfi/decimate: use named options

doc/filters: apply misc fixes to drawbox documentation

mov: fix self referencing timecode tracks

Fixes read after free.

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

dxva2: Add missing #define to make header compile standalone

arm: vp8: Add missing #includes for header to compile standalone

avfilter: fix segfault with setsar/dar

Reported-by: durandal_1707
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  intreadwrite: tomi: Add missing attributes.h #include for av_always_inline

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

lavf/matroskaenc: avoid assert failure in case of cuepoints with duplicated PTS

Avoid to write more than one cuepoint per track and PTS in
mkv_write_cues(). This avoids a later assertion failure on "(bytes >=
needed_bytes)" in put_ebml_num() called from end_ebml_master(), in case
there are several cuepoints per track with the same PTS.

This may happen with files containing packets with duplicated PTS in the
same track.

Merge commit '8a11ce43d08352f7a290355ebb5b29c495ad9609'

* commit '8a11ce43d08352f7a290355ebb5b29c495ad9609':
  build: Ensure that output directories for header objects are created
  h264: Get rid of unnecessary casts

Conflicts:
common.mak

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

Merge commit 'fae6fd5b873911cfcd7f7b3d72de6598af5e003a'

* commit 'fae6fd5b873911cfcd7f7b3d72de6598af5e003a':
  h264/svq3: Stop using draw_edges

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

Merge commit '7ebfb466aec2c4628fcd42a72b29034efcaba4bc'

* commit '7ebfb466aec2c4628fcd42a72b29034efcaba4bc':
  h264: Don't store intra pcm samples in h->mb
  get_bits: Return pointer to buffer that is the result of the alignment

Conflicts:
libavcodec/h264_mb_template.c

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

doc: filters: Correct BNF FILTER description

Signed-off-by: Diego Biurrun <diego@biurrun.de>

lavfi/hue: reindent after previous commit

lavfi/hue: simplify parsing by making use of option shorthands

Merge commit '2ed008204d5467be03a0a3af1e293b2f7038d0a0'

* commit '2ed008204d5467be03a0a3af1e293b2f7038d0a0':
  h264: Add add_pixels4/8() to h264dsp, and remove add_pixels4 from dsputil

Conflicts:
libavcodec/dsputil.c
libavcodec/h264.c
libavcodec/h264_mb_template.c
libavcodec/h264addpx_template.c
libavcodec/h264dsp.c
libavcodec/h264dsp.h

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

Merge commit 'e5ffffe48d20642acc079166f0fa7d93a6a9f594'

* commit 'e5ffffe48d20642acc079166f0fa7d93a6a9f594':
  h264chroma: Remove duplicate 9/10 bit functions
  x86: Use simple nop codes for <= sse (rather than <= mmx)
  vp56: Remove clear_blocks call, and clear alpha plane U/V DC only

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

cinepak: More correct Cinepak decoder.

change the treatment of the strip y coordinates which previously did
not follow the description (nor did it behave like the binary decoder
on files with absolute strip offsets).

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

cinepak: simplify

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

cinepak: Use correct colorspace.

The new code is also faster and more robust.

As for the performance:

old decoder + conversion to rgb:         fps = 2618
old decoder, without converting to rgb:  fps = 4012

new decoder, producing rgb:              fps = 4502

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

avcodec_align_dimensions2: Ensure cinepak has large enough buffers.

This is partly redundant with the following patches, but its safer

Found-by: u-bo1b@0w.se
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

h264: add videodsp dependency.

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

intreadwrite: tomi: Add missing attributes.h #include for av_always_inline

build: Ensure that output directories for header objects are created

cook: check js_subband_start for validity

Fixes out of array read

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

h264: Get rid of unnecessary casts

This gets rid of a number of warnings about casts discarding
qualifiers from the pointer target, present since 7ebfb466a.

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

shorten: Fix signedness of comparission

Fixes out of array accessed

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

vc1dec: dont decode slices when the slice header failed to decode.

Fixes assertion failure

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

h264/svq3: Stop using draw_edges

Instead, only extend edges on-demand when the motion vector actually
crosses the visible decoded area using ff_emulated_edge_mc(). This
changes decoding time for cathedral from 8.722sec to 8.706sec, i.e.
0.2% faster overall. More generally (VP8 uses this also), low-motion
content gets significant speed improvements, whereas high-motion content
tends to decode in approximately the same time.

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

h264: Don't store intra pcm samples in h->mb

Instead, keep them in the bitstream buffer until we read them verbatim,
this saves a memcpy() and a subsequent clearing of the target buffer.
decode_cabac+decode_mb for a sample file (CAPM3_Sony_D.jsv) goes from
6121.4 to 6095.5 cycles, i.e. 26 cycles faster.

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

get_bits: Return pointer to buffer that is the result of the alignment

This allows more transparent mixing of get_bits and whole-byte access
without having to touch get_bits internals.

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

h264: Add add_pixels4/8() to h264dsp, and remove add_pixels4 from dsputil

These functions are mostly H264-specific (the only other user I can
spot is bink), and this allows us to special-case some functionality
for H264. Also remove the 16-bit-coeff with >8bpp versions (unused)
and merge the duplicate 32-bit-coeff for >8bpp (identical).

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

h264chroma: Remove duplicate 9/10 bit functions

These functions do the same thing in 16 bit space and don't need
any depth specific clipping.

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

x86: Use simple nop codes for <= sse (rather than <= mmx)

The "CentaurHauls family 6 model 9 stepping 8" family of CPUs
(flags: fpu vme de pse tsc msr cx8 sep mtrr pge mov pat mmx fxsr sse
up rng rng_en ace ace_en) SIGILLs on long nop codes.

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

vp56: Remove clear_blocks call, and clear alpha plane U/V DC only

The non-alpha and alpha-Y planes are cleared in the idct_put/add()
calls. For the alpha U/V planes, we only care about the DC for entropy
context prediction purposes, the rest of the data is unused.

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

configure: fix libavfilter pkgconfig description

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

tiff: check bppcount

Fixes division by 0

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

bmp: check available space when reading palette

Fixes out of array read

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

h264: integrate clear_blocks calls with IDCT.

The non-intra-pcm branch in hl_decode_mb (simple, 8bpp) goes from 700
to 672 cycles, and the complete loop of decode_mb_cabac and hl_decode_mb
(in the decode_slice loop) goes from 1759 to 1733 cycles on the clip
tested (cathedral), i.e. almost 30 cycles per mb faster.

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

svq3: fix decoding residual blocks of b-frames.

The residual block data of 16x16 blocks was ignored for b-frames, which
leads to easy-to-identify artifacts. After this patch, the artifacts are
gone. Sample video: svq3_watermark.mov. (Fate results unaffected.)

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

avfilter/ff_merge_formats: only merge if doing so does not loose chroma or alpha

Fixes Ticket1280

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

x86/h263_loopfilter: Fix author attribution after code has been moved/splited around

    Reference:
    commit 3615e2be846f6382aa42360d02d16b25f443af9f
    Author: Michael Niedermayer <michaelni@gmx.at>
    Date:   Tue Dec 2 22:02:57 2003 +0000

        h263_h_loop_filter_mmx

        Originally committed as revision 2553 to svn://svn.ffmpeg.org/ffmpeg/trunk

    commit 359f98ded95ec65f7109b25f3cfaa9189c4fab9c
    Author: Michael Niedermayer <michaelni@gmx.at>
    Date:   Tue Dec 2 20:28:10 2003 +0000

        h263_v_loop_filter_mmx

        Originally committed as revision 2552 to svn://svn.ffmpeg.org/ffmpeg/trunk
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  x86: dsputil: Fix h263 loop filter link error in some configurations

Conflicts:
libavcodec/x86/dsputil.asm

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

Merge commit '7a03145ed7cb4f1ce794b5126559dd6f38029243'

* commit '7a03145ed7cb4f1ce794b5126559dd6f38029243':
  x86: dsputil: int --> ptrdiff_t for ff_put_pixels16_mmxext line_size param

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

Merge commit '202b5f6deb65e405b07b9b5c20f97c8cb925cf49'

* commit '202b5f6deb65e405b07b9b5c20f97c8cb925cf49':
  doc: Fix some obsolete references to av* tools as ff* tools

Conflicts:
doc/demuxers.texi

No code changes, merge for metadata only

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

tmv: initialize unused pallete entries with 0

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