Merge commit 'd0bf20a4f25ac5de021c860a0c8ad05638ee2078'

* commit 'd0bf20a4f25ac5de021c860a0c8ad05638ee2078':
  ppc: vsx: Implement diff_pixels and get_pixels

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

avutil/ppc/float_dsp_init: Disable duplicate functions

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

Merge commit 'eecd29b3fd7fee221580798346d6582b75c7ade4'

* commit 'eecd29b3fd7fee221580798346d6582b75c7ade4':
  ppc: vsx: Implement float_dsp

Conflicts:
libavutil/ppc/float_dsp_init.c

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

configure: ldbrx seems not working on qemu power8, disable for now

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

Merge commit '7d07ee5a9bd170a06d26fd967cf8de5d3b1ce331'

* commit '7d07ee5a9bd170a06d26fd967cf8de5d3b1ce331':
  ppc: cpu: Add support for VSX and POWER8 extensions

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

Merge commit 'da60b99a8857d5ca236f32c1799a066e0135a866'

* commit 'da60b99a8857d5ca236f32c1799a066e0135a866':
  ppc: Restrict some Altivec implementations to Big Endian

Conflicts:
libavcodec/ppc/huffyuvdsp_altivec.c
libavcodec/ppc/me_cmp.c

Not merged, our PPC64LE fate clients passed all tests last time they did run

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

Merge commit 'f22ebd2555d15b59e109a5b630dd71374b076d0b'

* commit 'f22ebd2555d15b59e109a5b630dd71374b076d0b':
  ppc: configure: Support ISA 2.06 and later

Conflicts:
configure

See: ab12373956e92b865bced9b05b12971f062cfd3e and others
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '3058872c293e239e3b51e86fe18cfbe720aadff1'

* commit '3058872c293e239e3b51e86fe18cfbe720aadff1':
  ppc: Clarify and extend the cpuid check

Conflicts:
libavutil/ppc/cpu.c

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

avutil/ppc/cpu: Check if unistd.h is available before including it

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

Merge commit 'baa94563fede8959a638b0fa132dd2124acd93e8'

* commit 'baa94563fede8959a638b0fa132dd2124acd93e8':
  ppc: linux: Check altivec using the auxv

Conflicts:
libavutil/ppc/cpu.c

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

Merge commit '7014b65995b1fe6188fb53447bf61b08e3963355'

* commit '7014b65995b1fe6188fb53447bf61b08e3963355':
  ppc: pixblockdsp: Use the abriged vector types

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

Merge commit '72cebae0d981dde144340cf51f3c323f01e215e5'

* commit '72cebae0d981dde144340cf51f3c323f01e215e5':
  ppc: avutil: Use the abriged vector types

Conflicts:
libavutil/ppc/float_dsp_altivec.c
libavutil/ppc/util_altivec.h

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

Merge commit '254eb5b6faebb7bcfc3cefc1edced6652fe9d5be'

* commit '254eb5b6faebb7bcfc3cefc1edced6652fe9d5be':
  ppc: avutil: Drop a potentially dangerous workaround

See: 9564375deeaf22f8e8eaa4fcbc33115d104f7138
Merged-by: Michael Niedermayer <michaelni@gmx.at>

avformat/libquvi: fix error handling

avoid calling cleanup functions on uninitialized variables

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

avfilter/vf_mpdecimate: add comment about 8x8 SAD

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

avfilter/vf_mpdecimate: support more pixel formats, including GBRP

Add some planar formats we can handle without distorting the meaning of
hi and lo.

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

avfilter/vf_mpdecimate: debug logging of threshold stats

It's often really helpful to know why a frame wasn't decimated (lo or
hi), and how much you exceeded the threshold by, if you're trying to
tweak the thresholds to get what you want.

mpdecimate already prints a line per input frame with -v debug, this
just makes it longer.

Signed-off-by: Peter Cordes <peter@cordes.ca>

avfilter/vf_mpdecimate: pass the same AVClass context as use elsewhere to av_pixelutils_get_sad_fn()

Signed-off-by: Peter Cordes <peter@cordes.ca>

tests: add some ASS/SSA/MKS remux and transcode tests

ppc: vsx: Implement diff_pixels and get_pixels

Use a macro to abstract the endianness.

ppc: vsx: Implement float_dsp

ppc: cpu: Add support for VSX and POWER8 extensions

ppc: Restrict some Altivec implementations to Big Endian

In Little Endian the vec_ld/vec_st operations work as
expected only for byte-vectors.

ppc: configure: Support ISA 2.06 and later

POWER 7 and POWER 8 support VSX and ldbrx.

POWER 8 supports additional extended VSX instructions introduced
with ISA 2.07.

ppc: Clarify and extend the cpuid check

Add POWER entries.

ppc: linux: Check altivec using the auxv

Should prevent trying to use altivec when it is disabled by the kernel.

ppc: pixblockdsp: Use the abriged vector types

ppc: avutil: Use the abriged vector types

ppc: avutil: Drop a potentially dangerous workaround

The compiler is free to optimize such expressions in any sort of way.

configure: Check for SecItemImport for securetransport as well

This fixes the default build on iOS; eventually I should come up with
a better solution for that platform.

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

Merge commit '881b80b3294483696a21b21a69fa5eee30f33037'

* commit '881b80b3294483696a21b21a69fa5eee30f33037':
  network: prevent SIGPIPE on OSX

See: c41e0aedf5fc2a92502bc400010e64b8fea89d4e
Merged-by: Michael Niedermayer <michaelni@gmx.at>

network: prevent SIGPIPE on OSX

OSX does not know MSG_NOSIGNAL. BSD (which OSX is based on) has got
the socket option SO_NOSIGPIPE (even if modern BSDs also support
MSG_NOSIGNAL).

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

avfilter/vf_fade: Force alpha pixel format if alpha is set to 1

Fixes Ticket4321

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

avcodec/parser: Print an error in case of reallocation fails in ff_combine_frame()

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

postproc: Avoid floats in maxClipped computation

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

postproc/postprocess_template: Compute packedYScale and QPCorrecture without floats

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

network: prevent SIGPIPE on OSX

OSX does not know MSG_NOSIGNAL, and provides its own non-standard
mechanism instead. I guess Apple hates standards.

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

Merge commit '0181ae9af2de1526464d23209b82e6674d362f5d'

* commit '0181ae9af2de1526464d23209b82e6674d362f5d':
  h264: Make sure reinit failures mark the context as not initialized

Conflicts:
libavcodec/h264_slice.c

See: e8714f6f93d1a32f4e4655209960afcf4c185214
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '5ecabd3c54b7c802522dc338838c9a4c2dc42948'

* commit '5ecabd3c54b7c802522dc338838c9a4c2dc42948':
  msrle: Use FFABS to determine the frame size in msrle_decode_pal4

Conflicts:
libavcodec/msrledec.c

See: f7e1367f58263593e6cee3c282f7277d7ee9d553
Merged-by: Michael Niedermayer <michaelni@gmx.at>

libopenjpegdec: check existence of image component data

libopenjpeg can return images with components without data.

This fixes segmentation faults.

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

doc/ffmpeg: Document xerror

Previous version reviewed-by: Lou Logan <lou@lrcd.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/movenc: Add loction/xyz support to allow copying data between mov files

Fixes Ticket4337

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

h264: Make sure reinit failures mark the context as not initialized

Bug-Id: CVE-2015-3417
CC: libav-stable@libav.org

msrle: Use FFABS to determine the frame size in msrle_decode_pal4

As done in msrle_decode_8_16_24_32.

Bug-Id: CVE-2015-3395
CC: libav-stable@libav.org

avformat/mov: Also export loci altitude

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

avformat/mov: Append place name instead of overwriting for loci

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

lavf/tls_securetransport: add missing copyright notice

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

avcodec/dvbsubdec: Fix buf_size check in dvbsub_parse_display_definition_segment()

Fixes Ticket4326

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

avformat/h264dec: Add ff_tlog() with trace information about h264 probing

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

avformat/h264dec: Fix detection of invalid h264 with 0x100 startcodes

Fies Ticket4325

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

Merge commit 'b2f0f37d242f1194fe1f886557cf6cefdf98caf6'

* commit 'b2f0f37d242f1194fe1f886557cf6cefdf98caf6':
  rtmpdh: Generate the whole private exponent using av_get_random_seed() with nettle/gmp

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

rtmpdh: Generate the whole private exponent using av_get_random_seed() with nettle/gmp

Don't use a PRNG for generating it; that defies the intended use
within the cryptograhic handshake.

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

lavf/tls: Support Secure Transport

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

configure: Fix showcqt fft dependancy

Found-by: Muhammad Faiz <mfcc64@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

build: add configure option pkgconfigdir

This allows the user to override the directory for the installation
of the pkg-config files (from the default LIBDIR/pkgconfig).

It follows the usual behaviour of Makefiles generated by automake.

Signed-off-by: Ingo Brückl <ib@wupperonline.de>
Reviewed-by: Andreas Cadhalpun <andreas.cadhalpun@googlemail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

libavutil: Cosmetic changes to fixed_dsp file.

Names of functions vector_fmul_window_fixed_c and
vector_fmul_window_fixed_scaled_c are changed by removing "_fixed"
from the name since it is redundant.

Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

mov: fix DTS calculation for samples with negative stts duration

A negative sample duration is invalid according to the spec, but there
are samples that use it for the DTS calculation, e.g.:
http://files.1f0.de/samples/mp4-negative-stts-problem.mp4

These currently get out of A/V sync.

Also change the logging type to AV_LOG_WARNING, because decoding the
sample can continue.

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

avcodec/mpegvideo: Use FFSWAP to exchange pointers

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

avcodec/ac3_parser: Avoid floats in bitrate computation

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

avformat/movenc: Check that track_width_1616 fits within the available 32bit before storing it

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

avformat/movenc: avoid floats in width/height/aspect ratio computations

This avoids the possibility for rounding/precision differences between platforms

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

avformat/concatdec: Enable auto_convert by default

Users have no means to find out from a failure how to make it work
or is it preferred to check and print a warning for h264 concat without auto_convert ?

Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/mov: Avoid float usage in yuv_to_rgba()

This avoids the possibility for rounding/precision differences between platforms

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

Merge commit 'e9e86d9ef637f5a600c76b352ffe5a82b71b25d1'

* commit 'e9e86d9ef637f5a600c76b352ffe5a82b71b25d1':
  rtmpdh: Create sufficiently long private keys for gcrypt/nettle

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

Merge commit '8016a1bd3b60e917e1b12748dd80c06c3462c286'

* commit '8016a1bd3b60e917e1b12748dd80c06c3462c286':
  rtmpdh: Remove an unnecessary check in the gcrypt/nettle dh_compute_key

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

Merge commit '063f7467e4d14ab7fe01b2845dab60cc75df8b53'

* commit '063f7467e4d14ab7fe01b2845dab60cc75df8b53':
  rtmpdh: Add fate test for the DH handshake routine

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

Merge commit '0508faaa11bf7507ffdd655aee57c9dc5a8203f4'

* commit '0508faaa11bf7507ffdd655aee57c9dc5a8203f4':
  rtmpdh: Pass the actual buffer size of the output secret key

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

Merge commit '9f1b3050d9e31e9283d818f3640f3460ac8cfb5b'

* commit '9f1b3050d9e31e9283d818f3640f3460ac8cfb5b':
  rtmpdh: Check the output buffer size in the openssl version of dh_compute_key

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

Merge commit '127d813bcb5705202b7100cf1eccd1e26d72ba14'

* commit '127d813bcb5705202b7100cf1eccd1e26d72ba14':
  rtmpdh: Fix a local variable name in the nettle/gcrypt codepath

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

Merge commit '78efc69e7c990226f4b913721ef1b308ca5bfa04'

* commit '78efc69e7c990226f4b913721ef1b308ca5bfa04':
  rtmpdh: Make sure ret is initialized in the nettle version of bn_hex2bn

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

rtmpdh: Create sufficiently long private keys for gcrypt/nettle

There was a misunderstanding betewen bits and bytes for the parameter
value for generating random big numbers.

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

rtmpdh: Remove an unnecessary check in the gcrypt/nettle dh_compute_key

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

rtmpdh: Add fate test for the DH handshake routine

This helps if these functions need to be implemented using another
crypto API.

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

rtmpdh: Pass the actual buffer size of the output secret key

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

rtmpdh: Check the output buffer size in the openssl version of dh_compute_key

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

rtmpdh: Fix a local variable name in the nettle/gcrypt codepath

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

rtmpdh: Make sure ret is initialized in the nettle version of bn_hex2bn

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

dxva2_hevc: re-write reference frame handling

The old logic required and explicit clearing of the lists first and was
prone to overflow the DXVA2 struct in some circumstances.

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

dxva2_hevc: fix 32x32 scaling lists

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

avcodec/mpegvideo: Reset bitstream_buffer_size on allocation failure

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

avcodec/mpegvideo: Check pointer when allocation fail

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

avformat/avienc: Correct possible dereference of null

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

Merge commit 'a64a5773ea5f7337cd1d87cfdf511914d317fe81'

* commit 'a64a5773ea5f7337cd1d87cfdf511914d317fe81':
  pixfmt: remove misleading and broken documentation

Conflicts:
libavutil/pixfmt.h

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

Merge commit 'e4610300de6869bd6b3b00e76cfeabb6d7653dcd'

* commit 'e4610300de6869bd6b3b00e76cfeabb6d7653dcd':
  x86: cavs: Remove an unneeded scratch buffer

Conflicts:
libavcodec/x86/cavsdsp.c

See: d79f7bf0d63a81ee66026ee92a6946a7303d04bd
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '1b1bb2c4efc126d74d44d8c421860c85f932ecb1'

* commit '1b1bb2c4efc126d74d44d8c421860c85f932ecb1':
  rl: Add error checking to ff_rl_init().

Conflicts:
libavcodec/rl.h

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

Merge commit '324e50ee95929a9491b855c5e15451145bd5d1ec'

* commit '324e50ee95929a9491b855c5e15451145bd5d1ec':
  rl: Add a function for freeing dynamically allocated tables.

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

Merge commit '6f57375d707de40dcec28d3cef886c364e032c21'

* commit '6f57375d707de40dcec28d3cef886c364e032c21':
  rl: Rename ff_*_rl() to ff_rl_*()

Conflicts:
libavcodec/mpeg4videodec.c
libavcodec/rl.c
libavcodec/rl.h

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

Merge commit 'fa1923f18205410a3b0aa6c0e77cb31443ef340d'

* commit 'fa1923f18205410a3b0aa6c0e77cb31443ef340d':
  mpegvideo: Move ff_*_rl functions to a separate file

Conflicts:
libavcodec/mpegvideo.c

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

Merge commit '419e3404d07acaac019e8f363c281e17c3a3d622'

* commit '419e3404d07acaac019e8f363c281e17c3a3d622':
  mpegvideo: Drop exchange_uv() function and use its code directly

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

Merge commit 'd4d90504a687d2c0ef77ccf11d831f24dcff9cf1'

* commit 'd4d90504a687d2c0ef77ccf11d831f24dcff9cf1':
  tls_gnutls: Add missing includes for the gcrypt thread safety callbacks

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

Merge commit 'dd4d709be705edaec0bc35c426bf8434e942b7df'

* commit 'dd4d709be705edaec0bc35c426bf8434e942b7df':
  x86inc: Clear __SECT__

See: 204b228a1d884343dbd974e6326a06901736b3b4
Merged-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/mips/hevcdsp_msa: Restructure as per avutil/mips/generic_macros_msa.h

This patch modifies HEVC mc MIPS-SIMD optimized code according to improved version of generic macros.

Overall, this patch is just upgrading the code with styling changes and will bring it in sync with MIPS-SIMD optimized latest codebase at our end.

Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

pixfmt: remove misleading and broken documentation

This was probably broken some time ago. The breakage is now part of the
ABI. For example, we have:

    AV_PIX_FMT_XYZ12BE
    AV_PIX_FMT_NV16
    AV_PIX_FMT_NV20LE

AV_PIX_FMT_NV20LE is wrong. It has the value 113, but as little-endian
format it should be even. This must have been quite obvious when these
formats were added (because of the AV_PIX_FMT_XYZ12BE entry), but
nobody cared or knew about this.

The future libavutil major bump will also break this additionally,
because disabling FF_API_VDPAU will remove an odd number of entries from
the middle of the enum.

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

x86: cavs: Remove an unneeded scratch buffer

Simplifies the code and makes it build on certain compilers
running out of registers on x86.

CC: libav-stable@libav.org
Reported-By: mudler

avcodec/mips: Split uni mc optimizations to new file

This patch moves HEVC code of uni mc cases to new file hevc_mc_uni_msa.c.
(There are total 5 sub-modules of HEVC mc functions, if we add all these modules in one single file, its size would be huge (~750k) & difficult to maintain, so splitting it in multiple files)
This patch also adds new HEVC header file libavcodec/mips/hevc_macros_msa.h

Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

rl: Add error checking to ff_rl_init().

rl: Add a function for freeing dynamically allocated tables.

Such tables are not used anywhere currently, but that should change.

rl: Rename ff_*_rl() to ff_rl_*()

mpegvideo: Move ff_*_rl functions to a separate file

mpegvideo: Drop exchange_uv() function and use its code directly

Code is small enough that there is no advantage in a separate function.