Merge commit '11f87ca71e9c7b917f594194f827fd040d1df5ca'

* commit '11f87ca71e9c7b917f594194f827fd040d1df5ca':
  vf_frei0r: also set AVFilterLink.frame_rate

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

Merge commit '7d12cba95ca15198a930c05458dc414ac00c578b'

* commit '7d12cba95ca15198a930c05458dc414ac00c578b':
  vf_framepack: Check and update frame_rate

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

Merge commit '65e73bc60f98dc7b26c687e145dfb755d3f2ccfa'

* commit '65e73bc60f98dc7b26c687e145dfb755d3f2ccfa':
  vf_interlace: implement frame rate

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

Merge commit '018bdaed37d2f1735dbecfc58309a1a164abadd5'

* commit '018bdaed37d2f1735dbecfc58309a1a164abadd5':
  setpts: add FRAME_RATE constant

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

Merge commit '5e91a5c5cf1db88f254b4c358eb1b06ff6ca274f'

* commit '5e91a5c5cf1db88f254b4c358eb1b06ff6ca274f':
  testsrc: set output framerate

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

Merge commit '9df477e03ef74068f3de130adc4dd34349a16ef2'

* commit '9df477e03ef74068f3de130adc4dd34349a16ef2':
  yadif: update frame rate

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

Merge commit '1062880d69a4fdc8d8929dd5c22bb447182f1c41'

* commit '1062880d69a4fdc8d8929dd5c22bb447182f1c41':
  vf_fps: set frame_rate.

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

Merge commit '61fb67dcb2e71a268c422fc19d366040e59fb337'

* commit '61fb67dcb2e71a268c422fc19d366040e59fb337':
  buffersrc: accept the frame rate as argument.

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

Merge commit '48ff6683ba5d40b629428673b1028e8ec542a9fa'

* commit '48ff6683ba5d40b629428673b1028e8ec542a9fa':
  lavfi: add a frame_rate field to AVFilterLink.

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

Merge commit '79f5347a983342e2711ca8ba19ec3d8d151183f0'

* commit '79f5347a983342e2711ca8ba19ec3d8d151183f0':
  avcodec: fix doxy placement

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

Merge commit '8de1ee9f725aa3c550f425bd3120bcd95d5b2ea8'

* commit '8de1ee9f725aa3c550f425bd3120bcd95d5b2ea8':
  lavf: deprecate compute_pkt_fields2

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

avformat/utils: re-factor freeing AVStreams

This matches the Libav signature and simplifies merging future changes.

avformat/rsd: XMA2 is actually stored, not XMA1

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

ffmpeg_filter: remove redundant null ptr check

Fixes: CID1338331
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/avf_showcqt: Fix uninitialized return code

Fixes CID1322329

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

avcodec/webvttdec: Fix uninitialized use of variable "again"

Fixes CID1338336
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

dvdsubdec: validate offset2 similar to offset1

If it is negative, it causes segmentation faults in decode_rle.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avutil/x86/intmath: Disable use of tzcnt on older intel compilers.

ICC versions older than atleast 12.1.6 dont have the tzcnt intrinsics.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Matt Oliver <protogonoi@gmail.com>

avformat/mov: Add option to ignore chapters during parsing

Chapter-indexing can be expensive since chapters may be interspersed
throughout the entire file and may require many seeks - especially
costly when consuming a video over a remote protocol like http.
Furthermore it is often unnecessary, especially when only trying to get
video info (e.g. via ffprobe).

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

avformat/mxfenc: Only store user comment related tags when needed

Also support disabling them as they seem to cause problems to some
Users. They are also not allowed in IRT D-10 thus the default for
mxf_d10 is not to write them

This also decreases the filesize when no user comment are stored

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

avcodec/gifdec: skip the data lzw consumed

this fixes the return code of avcodec_decode_video2 for gif decoding
and the gif frame data buffer is skipped properly

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

avcodec/lzw: ff_lzw_decode_tail() now returns the consumed bytes in lzw decompress

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

brstm: reject negative sample rate

A negative sample rate causes assertion failures in av_rescale_rnd.

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avcodec/wmaprodec: XMA1 actually have packet sequence numbers

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

aviflter/vf_waveform: speed up non-color filters

Remove loops, output is also more interesting.

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

avformat/wavdec: allow data chunk to be before fmt/xma2 chunk

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

avformat/wavdec: parse XMA2 tag

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

Merge commit 'aeda1121c91395d1f0d434bc7f227423aa2ff97d'

* commit 'aeda1121c91395d1f0d434bc7f227423aa2ff97d':
  lavf: factor out freeing an AVStream

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

Merge commit '3efd71b4d0b4a73ccbbbdc092e6bbd54d92633f4'

* commit '3efd71b4d0b4a73ccbbbdc092e6bbd54d92633f4':
  avconv: set packet duration for CFR video streams

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

Merge commit 'a0c71a575109f123978e345fa7eb4ac03cd4d3c3'

* commit 'a0c71a575109f123978e345fa7eb4ac03cd4d3c3':
  lavf: initialize cur_dts to AV_NOPTS_VALUE for muxing

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

Merge commit 'cc96018c79cd4dc823e0cef938da2131e4660fe5'

* commit 'cc96018c79cd4dc823e0cef938da2131e4660fe5':
  libopenh264enc: Set AVOption data type

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

Merge commit 'fcc49924db6c5530a3346c45a6999244d92eaf48'

* commit 'fcc49924db6c5530a3346c45a6999244d92eaf48':
  vdpau: Remove a spurious CONFIG_H263_VDPAU_HWACCEL

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

Merge commit '563e6d860391bac0511984e5c0842320b5c94d2d'

* commit '563e6d860391bac0511984e5c0842320b5c94d2d':
  segafilm: drop the "song and dance" for cinepak

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

Merge commit '1e791ee3aae63122afb10cbabe86a747aea58992'

* commit '1e791ee3aae63122afb10cbabe86a747aea58992':
  segafilm: set video and audio stream duration

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

Merge commit 'c012c6f1a8b34828a7870dc1854422934f14b79a'

* commit 'c012c6f1a8b34828a7870dc1854422934f14b79a':
  segafilm: implement seeking

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

Merge commit '2ec112f71cd03ccab1b6f9a00d29199a57bcc7a5'

* commit '2ec112f71cd03ccab1b6f9a00d29199a57bcc7a5':
  vf_pad: fix x, y option expression evaluation

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

Merge commit 'e2854e731f843906d9a9a5b882bed872341999fd'

* commit 'e2854e731f843906d9a9a5b882bed872341999fd':
  avresample: Document avresample_open() a little better

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

Merge commit 'f128b8e19ac7f702adae899ab91cc1e80f238761'

* commit 'f128b8e19ac7f702adae899ab91cc1e80f238761':
  mov: detect cover art pictures by content

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

Merge commit '50d2a3b5f34e6f99e5ffe17f2be5eb1815555960'

* commit '50d2a3b5f34e6f99e5ffe17f2be5eb1815555960':
  flashsv: Initialize the block array

Not merged, blocks are already zeroed when appropriate.

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

Merge commit 'de41b555cdea2dcacbe98ee9edc83a8c15c73c4c'

* commit 'de41b555cdea2dcacbe98ee9edc83a8c15c73c4c':
  truemotion2: Fix the buffer check

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

Merge commit '4a326923381f2bc0fb7d960c8f8366a5456bc4ab'

* commit '4a326923381f2bc0fb7d960c8f8366a5456bc4ab':
  mimic: Always return on failure

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

Merge commit '452d659aa4a742106c70ffe9ef3df47ef5b81ea6'

* commit '452d659aa4a742106c70ffe9ef3df47ef5b81ea6':
  msnwc_tcp: Use the correct return values

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

Merge commit 'f8d10511fef0bb3265f2d7bdfd38f90f60731d80'

* commit 'f8d10511fef0bb3265f2d7bdfd38f90f60731d80':
  cosmetics: msnwc_tcp: Reformat

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

Merge commit '4dfbc7a7559ccab666a8fd39de4224eb4b02c768'

* commit '4dfbc7a7559ccab666a8fd39de4224eb4b02c768':
  msnwc_tcp: Correctly report failure

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

Merge commit '60f50374f1955442dc987abc4a6c61c2109620c2'

* commit '60f50374f1955442dc987abc4a6c61c2109620c2':
  rpza: Check the blocks left before processing one

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

Merge commit '0b699920f3c0b2e30969b38ce59226653f9d80ec'

* commit '0b699920f3c0b2e30969b38ce59226653f9d80ec':
  lagarith: Correctly compute hash_shift

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

Merge commit '1e7ff5ac6923996f7292c82f102c68384fbc9d97'

* commit '1e7ff5ac6923996f7292c82f102c68384fbc9d97':
  nut: Use the correct codec_tag when multiple are available

Not merged since ffnutenc handles the codec_tag differently

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

avutil/softfloat: use abort() instead of av_assert0(0)

Fixes compilation of host tool aacps_fixed_tablegen.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: James Almer <jamrial@gmail.com>

avcodec/g729postfilter: Avoid function calls in FFMAX() arguments

This avoid double calling functions

Found-by: Muhammad Faiz <mfcc64@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

mmaldec: send only a single EOS packet on flushing

Fixes apparent mmal_port_disable() freezes in ffmmal_stop_decoder() when
calling ffmmal_decode() with flush semantics a large number of times in
a row.

avcodec/aacsbr_fixed: Try to initialize sum[0..1] differently to fix build with VS2012

Found-by: Hendrik Leppkes <h.leppkes@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat: Add v210[x] demuxer

Allows one to do:

    ffmpeg -s 1920x1080          -i blah.v210 ...
    ffmpeg -s 1920x1080 -f v210x -i blah.yuv10 ...

Fixes #1869.

Reviewed-by: James Almer <jamrial@gmail.com>
Reviewed-by: Kieran Kunhya <kierank@obe.tv>

tests/fate/avformat: Fix fate-lavf

The CMP variable seems to have been inherited from fate-api-seek which set it to null

the mxf reference needed a change due to c7e14a279fa7348db10ec824bb2d67858cb1c1ca

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

aacdec: don't return frames without data from aac_decode_er_frame

This is similar to commit ec38a1b for aac_decode_frame_int.

Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>

aacps: avoid division by zero in stereo_processing

This fixes a SIGFPE crash in the aac_fixed decoder.

Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>
Reviewed-by: Rostislav Pehlivanov <atomnuker@gmail.com>

ffserver: fix incorrect strlcpy usage

Somewhat ironic that this "safe" interface is actually being used
unsafely here. This fixes the usage preventing potential null pointer
dereference, where the old code was doubly broken: ctime can return
NULL, and ctime can return an arbitrarily long buffer.

Reviewed-by: Mark Harris <mark.hsj@gmail.com>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

swresample/resample: speed up Blackman Nuttall filter

This may be a slightly surprising optimization, but is actually based on
an understanding of how math libraries compute trigonometric functions.
Explanation is given here so that future development uses libm more effectively
across the codebase.

All libm's essentially compute transcendental functions via some kind of
polynomial approximation, be it Taylor-Maclaurin or Chebyshev.
Correction terms are added via polynomial correction factors when needed
to squeeze out the last bits of accuracy. Lookup tables are also
inserted strategically.

In the case of trigonometric functions, periodicity is exploited via
first doing a range reduction to an interval around zero, and then using
some polynomial approximation.

This range reduction is the most natural way of doing things - else one
would need polynomials for ranges in different periods which makes no
sense whatsoever.

To avoid the need for the range reduction, it is helpful to feed in
arguments as close to the origin as possible for the trigonometric
functions. In fact, this also makes sense from an accuracy point of view:
IEEE floating point has far more resolution for small numbers than big ones.

This patch does this for the Blackman-Nuttall filter, and yields a
non-negligible speedup.

Sample benchmark (x86-64, Haswell, GNU/Linux)
test: fate-swr-resample-dblp-2626-44100
old:
18893514 decicycles in build_filter (loop 1000),     256 runs,      0 skips
18599863 decicycles in build_filter (loop 1000),     512 runs,      0 skips
18445574 decicycles in build_filter (loop 1000),    1000 runs,     24 skips

new:
16290697 decicycles in build_filter (loop 1000),     256 runs,      0 skips
16267172 decicycles in build_filter (loop 1000),     512 runs,      0 skips
16251105 decicycles in build_filter (loop 1000),    1000 runs,     24 skips

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

swresample/resample: speed up upsampling by precomputing sines

When upsampling, factor is set to 1 and sines need to be evaluated only
once for each phase, and the complexity should not depend on the number
of filter taps. This does the desired precomputation, yielding
significant speedups. Hard guarantees on the gain are not possible, but gains
themselves are obvious and are illustrated below.

Sample benchmark (x86-64, Haswell, GNU/Linux)
test: fate-swr-resample-dblp-2626-44100
old:
29161085 decicycles in build_filter (loop 1000),     256 runs,      0 skips
28821467 decicycles in build_filter (loop 1000),     512 runs,      0 skips
28668201 decicycles in build_filter (loop 1000),    1000 runs,     24 skips

new:
14351936 decicycles in build_filter (loop 1000),     256 runs,      0 skips
14306652 decicycles in build_filter (loop 1000),     512 runs,      0 skips
14299923 decicycles in build_filter (loop 1000),    1000 runs,     24 skips

Note that this does not statically allocate the sin lookup table. This
may be done for the default 1024 phases, yielding a 512*8 = 4kB array
which should be small enough.
This should yield a small improvement. Nevertheless, this is separate from
this patch, is more ambiguous due to the binary increase, and requires a
lut to be generated offline.

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

doc/ffmpeg: Clarify that the sdp_file option requires an rtp output.

Signed-off-by: Simon Thelen <ffmpeg-dev@c-14.de>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

ffmpeg: Don't try and write sdp info if none of the outputs had an rtp format.

Fixes a segfault when trying to write nonexistent rtp information.

Signed-off-by: Simon Thelen <ffmpeg-dev@c-14.de>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/cache: Avoid int-overflow in cache compare function

cache protocol indexes its cache using AVTreeNodes which require a cmp
function for inserting and searching new cache-entries. This cmp
function expects a 32-bit int return value (negative, zero, or positive)
but the cache cmp function returns an int64_t which can overflow the
int, giving negative numbers for when it should be positive, vice versa.
This manifests itself only for very large files (e.g. 4GB+)

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

avcodec/nvenc: update nvenc default parameters

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

avutil/x86/intmath: Correct intrinsic headers for older compilers.

Signed-off-by: Matt Oliver <protogonoi@gmail.com>

avformat/rsd: add XMA support

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

vf_frei0r: also set AVFilterLink.frame_rate

(cherry picked from ffmpeg commit 353cf95f948ef7c6139c8ead79e9eeb9eb8d2e6e)
Signed-off-by: Anton Khirnov <anton@khirnov.net>

vf_framepack: Check and update frame_rate

(cherry picked from ffmpeg commit a46a23d30fea9c8a5570e07ec4d9c9b4eaa6eb4f)
Signed-off-by: Anton Khirnov <anton@khirnov.net>

vf_interlace: implement frame rate

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
(cherry picked from ffmpeg commit 227b4458fb434875b127a0c2306fa3b899a770bf)
Signed-off-by: Anton Khirnov <anton@khirnov.net>

setpts: add FRAME_RATE constant

Useful for dealing with constant frame-rate video.

(cherry picked from ffmpeg commit 722762f7e1212b4d02142778fabfdd73375722e0)

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

testsrc: set output framerate

(cherry picked from ffmpeg commit 7cd5fa35fecc2717223459d04822faed6bbd40fa)

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

yadif: update frame rate

(cherry picked from ffmpeg commit 31619584556466e4beab98e9b04ed4c5ba0db178)
Signed-off-by: Anton Khirnov <anton@khirnov.net>

vf_fps: set frame_rate.

(cherry picked from ffmpeg commit 0d249316043fb69a3972029bff3a2969689ba8b6)

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

buffersrc: accept the frame rate as argument.

(cherry picked from ffmpeg commit 9ca440679dc535b31edd569393d8d3dda59db90e)

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

lavfi: add a frame_rate field to AVFilterLink.

(cherry picked from ffmpeg commit 7b42036b3b23c85f473bf9369e37fa8da22eaf93)

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

avcodec: fix doxy placement

lavf: deprecate compute_pkt_fields2

All encoders set pts and dts properly now (and have been doing that for
a while), so there is no good reason to do any timestamp guessing in the
muxer.

The newly added AVStreamInternal will be later used for storing all the
private fields currently living in AVStream.

swresample/resample: improve bessel function accuracy and speed

This improves accuracy for the bessel function at large arguments, and this in turn
should improve the quality of the Kaiser window. It also improves the
performance of the bessel function and hence build_filter by ~ 20%.
Details are given below.

Algorithm: taken from the Boost project, who have done a detailed
investigation of the accuracy of their method, as compared with e.g the
GNU Scientific Library (GSL):
http://www.boost.org/doc/libs/1_52_0/libs/math/doc/sf_and_dist/html/math_toolkit/special/bessel/mbessel.html.
Boost source code (also cited and licensed in the code):
https://searchcode.com/codesearch/view/14918379/.

Accuracy: sample values may be obtained as follows. i0 denotes the old bessel code,
i0_boost the approach here, and i0_real an arbitrary precision result (truncated) from Wolfram Alpha:
type "bessel i0(6.0)" to reproduce. These are evaluation points that occur for
the default kaiser_beta = 9.

Some illustrations:
bessel(8.0)
i0      (8.000000) = 427.564115721804739678191254
i0_boost(8.000000) = 427.564115721804796521610115
i0_real (8.000000) = 427.564115721804785177396791

bessel(6.0)
i0      (6.000000) = 67.234406976477956163762428
i0_boost(6.000000) = 67.234406976477970374617144
i0_real (6.000000) = 67.234406976477975326188025

Reason for accuracy: Main accuracy benefits come at larger bessel arguments, where the
Taylor-Maclaurin method is not that good: 23+ iterations
(at large arguments, since the series is about 0) can cause
significant floating point error accumulation.

Benchmarks: Obtained on x86-64, Haswell, GNU/Linux via a loop calling
build_filter 1000 times:
test: fate-swr-resample-dblp-44100-2626

new:
995894468 decicycles in build_filter(loop 1000),     256 runs,      0 skips
1029719302 decicycles in build_filter(loop 1000),     512 runs,      0 skips
984101131 decicycles in build_filter(loop 1000),    1024 runs,      0 skips

old:
1250020763 decicycles in build_filter(loop 1000),     256 runs,      0 skips
1246353282 decicycles in build_filter(loop 1000),     512 runs,      0 skips
1220017565 decicycles in build_filter(loop 1000),    1024 runs,      0 skips

A further ~ 5% may be squeezed by enabling -ftree-vectorize. However,
this is a separate issue from this patch.

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

swresample: allow double precision beta value for the Kaiser window

Kaiser windows inherently don't require beta to be an integer. This was
an arbitrary restriction. Moreover, soxr does not require it, and in
fact often estimates beta to a non-integral value.

Thus, this patch allows greater flexibility for swresample clients.
Micro version is updated.

Reviewed-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>

softfloat: handle INT_MIN correctly in av_int2sf

Otherwise v=INT_MIN doesn't get normalized and thus triggers av_assert2
in other functions.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

softfloat: assert when the argument of av_sqrt_sf is negative

The correct result can't be expressed in SoftFloat.
Currently it returns a random value from an out of bounds read.

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>

avfilter: add anoisesrc

Signed-off-by: Kyle Swanson <k@ylo.ph>
Signed-off-by: Paul B Mahol <onemda@gmail.com>

avutil/softfloat: Include negative numbers in cmp/gt tests

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

avutil/softfloat: Fix av_gt_sf() with large exponents try #2

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

avutil/softfloat: Add test for av_gt_sf()

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

avutil/softfloat: Extend the av_cmp_sf() test to cover a wider range of exponents

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

avutil/softfloat: Fix overflows in shifts in av_cmp_sf() and av_gt_sf()

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

avutil/softfloat: Add test for av_cmp_sf()

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

avutil/softfloat: Add tests for exponent underflows

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

avutil/softfloat: Fix exponent underflow in av_div_sf()

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

avutil/softfloat: Fix exponent underflow in av_mul_sf()

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

avutil/softfloat: Fix typo in av_mul_sf() doxy

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

Revert "avutil/softfloat: Check for MIN_EXP in av_sqrt_sf()"

This case should not be possible if the input has a exponent within
the valid range

This reverts commit 0269fb11e3de17375f86d9120599af8c87cdfa0a.

avutil/softfloat: Check for MIN_EXP in av_sqrt_sf()

Otherwise the exponent could eventually underflow

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

avutil/softfloat: Correctly set the exponent for 0.0 in av_sqrt_sf()

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

lavf: factor out freeing an AVStream

It will be needed in other functions.

avconv: set packet duration for CFR video streams

lavf: initialize cur_dts to AV_NOPTS_VALUE for muxing

The reasoning for setting it to zero only applies to demuxing.

avcodec/aacsbr: Use FLOAT_0

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

avutil/softfloat: FLOAT_0 should use MIN_EXP

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

Add pixblockdsp checkasm tests

pixblockdsp: x86: Condense diff_pixels_* to a shared macro

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