checkasm: clip vp9 loopfilter test pixels inside allowed bitdepth range.

Merge commit 'b2417ee6d1ee0c5e9b170a642d73bdf68908966f'

* commit 'b2417ee6d1ee0c5e9b170a642d73bdf68908966f':
  dxv: Improve error message

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit '1bcd4a476ba45a7fdf59d1701b8f0e274418cc32'

* commit '1bcd4a476ba45a7fdf59d1701b8f0e274418cc32':
  dxv: Support RAW intermediate compression

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'bbf71d46db3417b43bcbd745cbf235e8e2ff69ae'

* commit 'bbf71d46db3417b43bcbd745cbf235e8e2ff69ae':
  dxv: Print texture information after header parsing

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'fb2889691cb7720d2680e188eb6036a35afa2392'

* commit 'fb2889691cb7720d2680e188eb6036a35afa2392':
  dxv: Support the original first version

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit '588a5619da0d041e55b365f63d0fa9c72bdbd4d3'

* commit '588a5619da0d041e55b365f63d0fa9c72bdbd4d3':
  dxv: Parse ancillary encoder information

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'a9b8c638cfe2f82191db65e3e3a39f3b35df81f5'

* commit 'a9b8c638cfe2f82191db65e3e3a39f3b35df81f5':
  mmal: Fix AVBufferRef usage

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit '49623f531972be5dc2dd8c1b4b8748cad7c424ff'

* commit '49623f531972be5dc2dd8c1b4b8748cad7c424ff':
  mmal: Remove setting extradata on input format

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

Merge commit 'f290e48d86e10f34b5ddc519127636bcebec7c43'

* commit 'f290e48d86e10f34b5ddc519127636bcebec7c43':
  mmal: drop the h264 BSF

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

AAC encoder: fix OOB access in search_for_pns

Fix out of bounds access caused by wrongful usage
of swb_offset constants when computing scalefactor
positions.

tests/checkasm: make randomize_buffers a function for easier debugging

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

ffserver: drop unneeded branching

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

ffserver: drop error counting when parsing ACL row

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

ffserver: rm whitespace c&p leftovers from macro

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

dnxhddec: indicate colorspace

It is supposed to only old BT.709 colorspaces.

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

avformat/iff: check for possible overflow in 2nd argument of av_new_packet

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

dnxhddata: correct weight tables

CID 1260 (as evidenced by incorrect decoding of a sample from ticket
4876) seems to use incorrect weight tables. It appears those tables
were not zigzag-scanned.

Apply zigzag on weight tables for new CIDs 1258, 1259, and 1260, and
fix an incorrect chroma table for CID 1256.

Fixes last issue from ticket #4876.

Found-by: Christophe Gisquet <christophe.gisquet@gmail.com>
Signed-off-by: Christophe Gisquet <christophe.gisquet@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

dnxhddec: decode and use interlace mb flag

This bit is 1 in some samples, and seems to coincide with interlaced
mbs and CID1260. 2008 specs do not know about it, and maintain qscale
is 11 bits. This looks oversized, but may help larger bitdepths.

Currently, it leads to an obviously incorrect qscale value, meaning
its syntax is shifted by 1. However, reading 11 bits also leads to
obviously incorrect decoding: qscale seems to be 10 bits.

However, as most profiles still have 11bits qscale, the feature is
restricted to the CID1260 profile.

The encoder writes 12 bits of syntax, last and first bits always 0,
which is now somewhat inconsistent with the decoder, but ends up with
the same effect (progressive + reserved bit).

Partially fixes ticket #4876.

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

avformat/hls: Update Cookies response with Setcookie

Context cookies must be updated when a playlist response return Setcookie header.

Signed-off-by: wm4 <nfxjfg@googlemail.com>

swscale: fix ticket #4877

avcodec/ffv1dec: Fix >8bps error concealment

Fixes: 03_cave_girls_h264_slc24_slicecrc1_fuzzed.avi

Found-by: Dinah Handel <dinahrhandel@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/rangecoder: Check e

Fixes hang.nut

Found-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

AAC encoder: tweak PNS usage to be more aggressive

This patch tweaks search_for_pns to be both more
aggressive and more careful when applying PNS. On
the one side, it will again try to use PNS on zero
(or effectively zero) bands. For this, both zeroes
and band_type have to be checked (some ZERO bands
aren't marked in zeroes). On the other side, a more
accurate rate-distortion measure avoids using PNS
where it would cause audible distortion.

Also fixed a small bug in the computation of freq
that caused PNS usage on low-frequency bands during
8-short windows. This allows re-enabling PNS during
8-short.

avcodec/ffv1: seperate slice_count from max_slice_count

Fix segfault with too large slice_count
Fixes Ticket4879

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

Return EOF for ICO when the end is reached

tests/checkasm/vp9dsp: Revert first hunk of bddcf758d3a68ac0bcc3bc4fc4aa7156e05245d4

The change was wrong, also add a comment explaining it

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

avfilter/generate_wave_table: clean up extra newlines

Signed-off-by: Kyle Swanson <k@ylo.ph>

dxv: Improve error message

dxv: Support RAW intermediate compression

Sample-Id: boombox64.mov
Reported-by: Aarni Koskela
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

dxv: Print texture information after header parsing

dxv: Support the original first version

DXV 1.0 provided DXT1 with LZF only.

Sample-Id: Orange-Strings-07_640_DXV.mov
Reported-by: Carl Eugen Hoyos
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

dxv: Parse ancillary encoder information

Header does not contain the number of channels, but rather the encoder
version, so rename the variable, and read the additional byte.

mmal: Fix AVBufferRef usage

AVBufferRef.data and AVPacket.data don't need to have the same value.
AVPacket could point anywhere into the buffer. Likewise, the sizes
don't need to be the same.

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

mmal: Remove setting extradata on input format

This works only for extradata sizes up to 128 bytes. Additionally, I
could never actually see it doing anything. The new code using
MMAL_BUFFER_HEADER_FLAG_CONFIG now takes care of this.

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

mmal: drop the h264 BSF

We can send mp4-style data directly. But for some reason, this requires
sending the extradata as buffer with MMAL_BUFFER_HEADER_FLAG_CONFIG
set. Reuse the infrastructure for sending AVPackets to do this.

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

avfilter/af_tremolo: make it bit-exact with sox effect of same name

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

avcodec/ac3enc: use long long after switch to 64 bit bitrate

Commit 7404f3bdb switched bitrate to 64 bits.
This triggers -Wabsolute-value on clang, e.g
http://fate.ffmpeg.org/log.cgi?time=20150917122742&log=compile&slot=x86_64-darwin-clang-3.7-O3.
Therefore, usage of abs is changed to llabs, which is available on all of the platforms.
Unfortunately, LLONG_MAX is not always available, so INT64_MAX is used instead.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/x86/cavsdsp: silence -Wunused-variable on --disable-mmx

This silences -Wunused-variable when compiled with --disable-mmx, e.g
http://fate.ffmpeg.org/log.cgi?time=20150919094617&log=compile&slot=x86_64-archlinux-gcc-disable-mmx.
The alternative of header guards will make it far too ugly.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

fate: add chroma position scale test

The sample position is made weird and non-nominal to force catching
such issues as default values or specialized operations hiding
issues in corner cases.

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

vf_scale: conditionally override chroma position

For yuv420p, the chroma position is unilaterally overriden, even
if ffmpeg's command-line explicitly set it. To fix this, override
only if the value is the default one.

Regression since 1515bfb3.

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

avformat/brstm: fix overflow

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

avfilter/vf_chromakey: Add chromakey video filter

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>

avcodec/aacenctab: Make aac_maxval_cb const

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

avcodec/libstagefright: fix Stagefright_decode_frame() failing to exit when the source is done

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

doc/filters.texi: add rubberband documentation

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

avfilter/af_rubberband: rename duplicate option name

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

AAC encoder: tweak rate-distortion logic

This patch modifies the encode frame function to
retry encoding the frame when the resulting bit count
is too far off target, but only adjusting lambda
in small, incremental step. It also makes the logic
more conservative - otherwise it will contend with
bit reservoir-related variations in bit allocation,
and result in artifacts when frame have to be truncated
(usually at high bit rates transitioning from low
complexity to high complexity).

AAC encoder: simplify and speed up find_min_book

Trivial change to simplify the small but hot
find_min_book function. The new form is easier to
understand and faster.

configure: Force mingw's ld to keep the reloc section

Binutils will always strip the relocation information from executable
files even if it needs it (dynamicbase/ASLR).  We can work around this
by using the pic-executable flag combined with setting the correct entry
point since apparently ld forgets what that should be.  This problem
affects both 32 and 64-bit mingw-w64.

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

avcodec/x86/mpegaudiodsp: silence -Wunused-variable on --disable-mmx

This silences -Wunused-variable when compiled with --disable-mmx, e.g
http://fate.ffmpeg.org/log.cgi?time=20150919094617&log=compile&slot=x86_64-archlinux-gcc-disable-mmx.
The alternative of header guards will make it far too ugly.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/x86/rv40dsp_init: silence -Wunused-variable on --disable-mmx

This silences -Wunused-variable when compiled with --disable-mmx, e.g
http://fate.ffmpeg.org/log.cgi?time=20150919094617&log=compile&slot=x86_64-archlinux-gcc-disable-mmx.
The alternative of header guards will make it far too ugly.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/hermite: fix "libavfilter/hermite.h:19:15: error: no previous prototype for hermite_interpolation"

Fix build

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

avfilter: add agate filter

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

avcodec/mpegvideo_enc: Avoid fine lambda steps in VBV retry code when RD is not in use

Reduces the number of times the vbv retry code is used and should have no
effect on quality

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

hls: only seek if there is an offset

If there is no #EXT-X-BYTERANGE specified, there is no need to seek.
Seeking fails anyway for rtmp, because this protocol does not support
url_seek.

This fixes CNN.m3u from trac ticket 4797 (i.e. Debian bug #798189).

Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avcodec/dvdsubdec: don't use a NULL log context

avcodec/dvdsubdec: reject some broken packets

If cmd_pos is broken, this would just keep accumulating packets in the
reassembly buffer, until it fails and flushes the buffer on overflow.
Since packets are usually rather small, this will take a lot of subtitle
packets. The perceived effect is that subtitles are not displayed
anymore after the faulty packet was passed to the decoder.

I'm not terribly sure about this, but on the other hand this code is
active only when fragmented packets need to be reassembled.

Fixes sample file in trac issue #4872.

avcodec/dvdsub: fix partial packet assembly

Assuming the first and second packets are partial, this would append the
reassembly buffer (ctx->buf) to itself with the second
append_to_cached_buf() call, because buf is set to ctx->buf.

I do not know a valid sample file which triggers this, and do not know
if packets can be split into more than 2 sub-packets, but it triggered
with a (differently) broken sample file in trac issue #4872.

avformat/hls: fix some cases of HLS streams which require cookies

Broken by commit ba12ba859aabfa7153ba397d869db13acdaba340. This only
happens with HLS streams which use encryption and require preserving
cookies sent by the server.

Fixes trac issue #4846.

avfilter: add tremolo filter

Signed-off-by: Kyle Swanson <k@ylo.ph>

Update my email address

My old address no longer works.

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

hevc: fix wpp threading deadlock.

Fixes ticket 4258.

vp9: fix loopfilter test code to address Hendrik's comments.

(I forgot to actually merge them into the patch I just pushed.)

x86/vp9dsp: fix local header include

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

x86/vp9dsp: add missing header include

Fixes make checkheaders

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

avcodec/dvdsubdec: fix indentation

avformat/assdec: do not drop duplicate subtitles

Duplicates are perfectly legitimate with this format. A correct
renderer will typically show them stacked.

avformat/vobsub: do not attempt to check duplicate subtitles

Some .idx files actually contain duplicate subtitle events:

timestamp: 00:07:52:600, filepos: 00004e800
timestamp: 00:07:52:600, filepos: 00004f800

The second will be dropped, because it has same pts, duration, and text
(the text is just a dummy empty string; the real data is retrieved when
actually reading vobsub subtitle packets).

Dropping this is probably not intended/safe, so avoid it.

See trac issue #4872 for a sample. This patch doesn't fix decoding of
the sample, though.

avformat/subtitles: make dropping duplicate events optional

avformat/vobsub: compare correct packet stream IDs

The stream ID is essentially an arbitrary number defined by the .idx
file headers. They have to match the IDs in the .sub stream. The vobsub
demuxer assumed the IDs would just start from 0, increassing by 1 for
each stream. This is not correct. In the sample I had, the IDs were
starting from 1, leading to no subtitles being displayed at all.

Fix this by using the correct stream ID.

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

configure: silence error if tput not found

If tput is not found for colorizing, error message should be squashed.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

doc: mention libavcodec can decode Opus natively

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

mpjpeg: CRLF terminating a sequence of MIME headers should not cause an error

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

avcodec/snowenc: Do not write into const AVFrame

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

avcodec/snowenc: Remove unused new_picture field

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

swscale: fix ticket 4850

avdevice/libdc1394: add const to suppress "assignment discards const qualifier from pointer target type" warnings

See: http://fate.ffmpeg.org/log.cgi?time=20150919100330&log=compile&slot=x86_64-archlinux-gcc-enableshared

Found-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/hevc_mp4toannexb_bsf: silence -Wdiscarded-qualifiers

*poutbuf is non-const, so this casts it explicitly.
This suppresses -Wdiscarded-qualifiers seen in e.g
http://fate.ffmpeg.org/log.cgi?time=20150919100330&log=compile&slot=x86_64-archlinux-gcc-enableshared.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

tests/checkasm: fix stack smash in check_loopfilter

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

lavf/matroska: ignore ChapCountry ID for now

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

avfilter: add rubberband wrapper

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

lavfi: remove FF_LINK_FLAG_REQUEST_LOOP.

It has no longer any effect.

lavfi: drop the requirement that request_frame returns a frame.

It requires a loop in filters or the framework,
that makes the scheduling less efficient and more complex.
This is purely an internal change since the loop is now
present in buffersink.
Note that no filter except buffersink did rely on the requirement.

lavfi/vf_idet: reindent after last commit.

lavfi/vf_idet: remove the loop in request_frame().

It is not necessary due to the use of FF_LINK_FLAG_REQUEST_LOOP.

lavfi/buffersink: loop over ff_request_frame().

Do not assume that ff_request_frame() returning success
implies a frame has arrived in the FIFO.
Instead, just loop until a frame is in the FIFO.
It does not change anything since the same loop is present
in ff_request_frame(), confirmed by an assertion.

x86/vp9mc: fix string concatenation of fullpel function names

Fixes compilation with NASM

Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>

avcodec/internal: Use do {} while() for ff_tlog()

Avoids problems when used without braces

Found-by: Clément Bœsch <u@pkh.me>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

tests/checkasm/vp9dsp: Add () to protect macro arguments

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

checkasm: add VP9 loopfilter tests.

The randomize_buffer() implementation assures that "most of the time",
we'll do a good mix of wide16/wide8/hev/regular/no filters for complete
code coverage. However, this is not mathematically assured because that
would make the code either much more complex, or much less random.

checkasm: add jpeg2000dsp rct_int tests

Reviewed-by: Henrik Gramner <henrik@gramner.com>
Signed-off-by: James Almer <jamrial@gmail.com>

avcodec/x86/hpeldsp_rnd_template: silence -Wunused-function on --disable-mmx

This silences some of the -Wunused-function warnings when compiled with --disable-mmx, e.g
http://fate.ffmpeg.org/log.cgi?time=20150919094617&log=compile&slot=x86_64-archlinux-gcc-disable-mmx.
Header guards are too brittle and ugly for this case.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/format: silence -Wdiscarded-qualifiers

lpd.buf is non-const and discards the const qualifier of zerobuffer.
This fixes -Wdiscarded-qualifiers observed with a variety of compilers, including GCC 5.2.
Note that this does not change the type of zerobuffer, and merely makes the intent explicit.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

swscale/output: fix undefined left shifts of negative numbers

This fixes -Wshift-negative-value reported with clang 3.7+, e.g
http://fate.ffmpeg.org/log.cgi?time=20150918181527&log=compile&slot=x86_64-darwin-clang-polly-vectorize-stripmine-3.7.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/mlpdec: fix a undefined left shift of negative number

This fixes a -Wshift-negative-value reported with clang 3.7+, e.g
http://fate.ffmpeg.org/log.cgi?time=20150918181527&log=compile&slot=x86_64-darwin-clang-polly-vectorize-stripmine-3.7.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

hevc: properly handle no_rasl_output_flag when removing pictures from the DPB

Fixes ticket #4185.

Reviewed-By: Mickael Raulet <Mickael.Raulet@insa-rennes.fr>
Signed-off-by: Hendrik Leppkes <h.leppkes@gmail.com>

avcodec/motion_est_template: fix undefined left shift of negative number

This fixes -Wshift-negative-value reported with clang 3.7+, e.g
http://fate.ffmpeg.org/log.cgi?time=20150918181527&log=compile&slot=x86_64-darwin-clang-polly-vectorize-stripmine-3.7.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avresample/resample: remove unused variable

This fixes a -Wunused-variable, see e.g
http://fate.ffmpeg.org/log.cgi?time=20150919162338&log=compile&slot=x86_64-archlinux-gcc-threads.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/x86/mpegvideoenc: silence -Wunused-function on --disable-mmx

This silences -Wunused-function when compiled with --disable-mmx, e.g
http://fate.ffmpeg.org/log.cgi?time=20150919094617&log=compile&slot=x86_64-archlinux-gcc-disable-mmx.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/x86/hpeldsp_init: silence -Wunused-function on --disable-mmx

This silences some of the -Wunused-function warnings when compiled with --disable-mmx, e.g
http://fate.ffmpeg.org/log.cgi?time=20150919094617&log=compile&slot=x86_64-archlinux-gcc-disable-mmx.

Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>