Merge commit 'bde2bba45c2f2df27a8534028bda09a6e7f835e2'

* commit 'bde2bba45c2f2df27a8534028bda09a6e7f835e2':
  rtpenc: Restructure if statements in packetizers to simplify adding more conditions

Conflicts:
libavformat/rtpenc_xiph.c

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

Merge commit 'd4c7fc02f9f59e721e76debf4a595df529707545'

* commit 'd4c7fc02f9f59e721e76debf4a595df529707545':
  rtpenc: Skip redundant initialization

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

Merge commit 'f8c01257f93ceda3e03bc4e540a51022d1e2bff2'

* commit 'f8c01257f93ceda3e03bc4e540a51022d1e2bff2':
  rtpenc: Always do the default initialization regardless of codecs

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

Merge commit '11edeaea3293c41ecf577a330422eabba35f76a2'

* commit '11edeaea3293c41ecf577a330422eabba35f76a2':
  rtpenc_xiph: Don't exclude headers from max_payload_size

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

Merge commit '7c1e2e64667421f931ab48141517f19d309c7eea'

* commit '7c1e2e64667421f931ab48141517f19d309c7eea':
  rtpenc_xiph: Use AV_WB16 instead of manual bitshifts

Conflicts:
libavformat/rtpenc_xiph.c

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

Merge commit 'd16c8d28d4e2fca3af1054ffbf635c8cee755fc8'

* commit 'd16c8d28d4e2fca3af1054ffbf635c8cee755fc8':
  rtpenc_aac: Use AV_WB16 instead of manual bitshifts

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

Merge commit '9c9b0218e85fcd969308632f75af48a4ce229541'

* commit '9c9b0218e85fcd969308632f75af48a4ce229541':
  rtpenc_aac: Merge a definition with a declaration

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

Merge commit '1fc64e2e07787bbca82a72c146588e850e6d098a'

* commit '1fc64e2e07787bbca82a72c146588e850e6d098a':
  rtpenc: Write conditional statements on separate lines

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

Merge commit '0662440b991361fdb5e732712d997a73e4692e34'

* commit '0662440b991361fdb5e732712d997a73e4692e34':
  rtpenc_aac: Set a default value for max_frames_per_packet at init

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

Merge commit '12b3459979f5ea6481660cd2c99a0381e2b5ba37'

* commit '12b3459979f5ea6481660cd2c99a0381e2b5ba37':
  rtpenc_amr: Use s->num_frames instead of s->buf_ptr - s->buf

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

Merge commit '98563953442560dd83aab938f86de3e5a22a891f'

* commit '98563953442560dd83aab938f86de3e5a22a891f':
  rtpenc_aac: Fix sending fragmented frames

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

Merge commit '990e4a6639d0714198583c2812b1074b5751898f'

* commit '990e4a6639d0714198583c2812b1074b5751898f':
  Add a QSV decoding example.

Conflicts:
configure
doc/Makefile

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

Merge commit 'ea1d0b7ece1881c2f1360f8c1e2116791105af21'

* commit 'ea1d0b7ece1881c2f1360f8c1e2116791105af21':
  avcodec/utils: use correct printf specifier in ff_set_sar

See: 732c3ebffaff5005367d7f947fa903f3b6e92f68
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '802987f8c7033ec8b82b35438d3822cf7f761166'

* commit '802987f8c7033ec8b82b35438d3822cf7f761166':
  x11grab: Unbreak building

Conflicts:
libavdevice/x11grab.c

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

Merge commit '71f1ad37d858b810b71a4af1c25771beaa50b27b'

* commit '71f1ad37d858b810b71a4af1c25771beaa50b27b':
  lavc: do not compile fmtconvert unconditionally

Conflicts:
configure
libavcodec/ppc/Makefile
libavcodec/x86/Makefile

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

Merge commit 'd74a8cb7e42f703be5796eeb485f06af710ae8ca'

* commit 'd74a8cb7e42f703be5796eeb485f06af710ae8ca':
  fmtconvert: drop unused functions

Conflicts:
libavcodec/arm/fmtconvert_vfp_armv6.S
libavcodec/x86/fmtconvert.asm
libavcodec/x86/fmtconvert_init.c

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

Merge commit 'ee964145b5d229571e00bf6883a44189d02babe2'

* commit 'ee964145b5d229571e00bf6883a44189d02babe2':
  lavc: remove unused traces of fmtconvert usage

Conflicts:
libavcodec/aac.h
libavcodec/aacdec.c
libavcodec/atrac3.c
libavcodec/vorbisdec.c
libavcodec/wma.c
libavcodec/wma.h

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

Merge commit 'b9d2d6843a49f9df1d1ae1afe817d9b48c445919'

* commit 'b9d2d6843a49f9df1d1ae1afe817d9b48c445919':
  tls: Pass AVOptions dictionaries through to the chained protocol

Conflicts:
libavformat/tls.c

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

Merge commit 'e14f98c62fdf8744b07419314095d1b3248cce75'

* commit 'e14f98c62fdf8744b07419314095d1b3248cce75':
  tcp: Clarify the units for the timeout avoptions

Conflicts:
libavformat/tcp.c

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

Merge commit 'c86d8aed748adc3ba8df96fb1fe14aa45da77a31'

* commit 'c86d8aed748adc3ba8df96fb1fe14aa45da77a31':
  avio: Rename avclass symbols relating to avio

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

avformat/flvenc: check that the codec_tag fits in the available bits

flags is later written with avio_w8 and if it doesn't fit in one byte it
triggers an av_assert2.

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

avcodec/utils: use correct printf specifier in ff_set_sar

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

avutil/imgutils: correctly check for negative SAR components

These could trigger assert failures previously

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

rtpenc: Restructure if statements in packetizers to simplify adding more conditions

Factorize out the s->num_frames check at the start of the if statements,
simplifying adding more alternative causes for sending the buffered
frames.

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

rtpenc: Skip redundant initialization

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

rtpenc: Always do the default initialization regardless of codecs

This avoids having to jump to the defaultcase in the switch. Manually
override the stream time base back to 90 kHz for the few audio codecs
that don't use the sample rate as time base (mp2, mp3).

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

rtpenc_xiph: Don't exclude headers from max_payload_size

This makes things more consistent by using the variable in the same
way as in all other packetizers.

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

rtpenc_xiph: Use AV_WB16 instead of manual bitshifts

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

rtpenc_aac: Use AV_WB16 instead of manual bitshifts

This makes the code slightly more readable and understandable.

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

rtpenc_aac: Merge a definition with a declaration

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

rtpenc: Write conditional statements on separate lines

Intentionally keeping some conditional statements on single lines
in rtpenc_h263.c.

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

rtpenc_aac: Set a default value for max_frames_per_packet at init

This avoids having to conditionally set the default within the
packetizer function.

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

rtpenc_amr: Use s->num_frames instead of s->buf_ptr - s->buf

This doesn't fix any bug, but makes the code simpler for later
patches, and more straightforward to read as is.

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

rtpenc_aac: Fix sending fragmented frames

After sending a fragmented frame, len (s->buf_ptr - s->buf) isn't
zero, while s->num_frames is zero as intended. Using s->num_frames
makes it work as intended, and is less convoluted than keeping track
of (resetting) s->buf_ptr.

This avoids sending stray data after sending a fragmented aac packet.

CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

Add a QSV decoding example.

avcodec/utils: use correct printf specifier in ff_set_sar

Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>

x11grab: Unbreak building

The correct macro is DEC not D. Broken in
b31328d008985f87f0a7c83c700847cef1a4f08c

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

lavc: do not compile fmtconvert unconditionally

Only ac3dec and dcadec use it.

fmtconvert: drop unused functions

lavc: remove unused traces of fmtconvert usage

Those decoders have been switched to float output and so do not use
fmtconvert anymore.

tls: Pass AVOptions dictionaries through to the chained protocol

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

tcp: Clarify the units for the timeout avoptions

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

avio: Rename avclass symbols relating to avio

Don't prefix them ffio_url, which is misleading, sounding too
much like the urlprotocol layer (like ffurl_*).

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

avcodec/ppc/fdctdsp: fix const warning

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

avcodec/vc1_block: Optimize vc1_decode_i_block_adv()

9127 -> 8936 decicycles (fate-suite/vc1/SA10143.vc1)
13855 -> 10976 decicycles  (fate-suite/vc1/SA20021.vc1)
tests done by the author over this function but with the whole
patchset applied not just this commit

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

avcodec/vc1_block: Optimize vc1_decode_i_block()

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

avcodec/vc1_block: cosmetics

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

avcodec/vc1_block: move early exit code up in ff_vc1_pred_dc()

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

avcodec/vc1_block: Use av_clip_uintp2()

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

avcodec/vc1_block: Optimize vc1_decode_ac_coeff()

Avoid branches in sign handling
Move local variables to where they are used

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

avcodec/vc1_block: Simplify dcdiff initialization

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

avcodec/vc1_block: Simplify dir_ptr / pred initialization

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

avcodec/vc1: Simplify pred/mc loops

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

avcodec/vc1: Simplify dmv_y initialization

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

avcodec/vc1: Simplify code setting and using extend_x/y

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

avcodec/vc1: Make init_block_index() inline

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

avcodec/vc1: use uint8_t for size_table

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

avcodec/vc1: simplify offset table and access

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

ffmpeg: Fix -vstats when coded_frame is not set

There are several encoders which do not set coded_frame

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

lavc/pngenc: Support writing colorspace tags.

These are defined in http://www.w3.org/TR/PNG/#11addnlcolinfo.

avformat: Add prefix to get_codec_guid()

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

swscale: Add prefix to updateMMXDitherTables()

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

swresample: Add prefix to soxr_resampler

also move declaration to header

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

avcodec/dvdec: Make dv_iweight_720_c static

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

avcodec/v210dec: Add ff prefix to v210_x86_init()

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

avcodec/h264: Add ff_ prefix to the shared h264_init_dequant_tables() function

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

lavf/mov: correct printf specifier

The type is int64_t.

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

avcodec/snow: mark dwt init as av_cold

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

avcodec/diracdsp: Mark ff_diracdsp_init() as av_cold

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

avcodec/mjpegenc: Mark ff_mjpeg_encode_close() as av_cold

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

postproc: mark alloc/free context functions as av_cold

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

movenc: Write 'colr' box correctly for MP4

This also restricts it to MOV and MP4, since it is only
defined for those formats.

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

avfilter/palettegen: fix frame mem leak

avformat/ffmenc: Use ffio_free_dyn_buf()

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

avformat/rtpdec_vp9: Use ffio_free_dyn_buf()

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

avcodec/hqx: use init_get_bits8()

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

avcodec/hqx: fix typo

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

swscale/utils: clear formatConvBuffer on allocation

Fixes use of uninitialized memory
Fixes: asan_heap-oob_35ca682_1474_cov_3230122439_aletrek_tga_16bit.mov
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/bit: only accept the g729 codec and 1 channel

Other codecs/channel numbers are not supported by this muxer.

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

avformat/bit: check that pkt->size is 10 in write_packet

Ohter packet sizes are not supported by this muxer.

This avoids a null pointer dereference of pkt->data.

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

Merge remote-tracking branch 'cigaes/master'

* cigaes/master:
  ffmpeg: notify when the thread message queue blocks.
  ffmpeg: allow to set the thread message queue size.
  lavd/alsa: set frame_size field.

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

Fix buffer_size argument to init_put_bits() in multiple encoders.

Several encoders were multiplying the buffer size by 8, in order to get
a bit size. However, the buffer_size argument is for the byte size of
the buffer. We had experienced crashes encoding prores (Anatoliy) at
size 4096x4096.

avformat/mp3dec: Check for avcodec_alloc_context3() failure

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

avformat/mp3dec: properly allocate dummy AVCodecContext

Fixes (harmless) use of uninitialized variable

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

Merge commit '82b6e451161f19ab90bfb9565b710021caf5dcbd'

* commit '82b6e451161f19ab90bfb9565b710021caf5dcbd':
  configure: Move the cross_prefix setting after the toolchain one

Conflicts:
configure

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

colormatrix: Accept bt601 aliases as input

As far as matrix coefficients are concerned, BT470BG and SMPTE170M
match BT601. This is similar to the check performed in vf_scale.

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

ffmpeg: notify when the thread message queue blocks.

This can help finding the source of A-V desync with live input.

ffmpeg: allow to set the thread message queue size.

lavd/alsa: set frame_size field.

The value tells the typical size of a packet from the demuxer.

mips/acelp_filters: fix incorrect register constraint

Change register constraint on the v variable from = to +. This was causing GCC
to think that the v variable was never read and therefore not initialize it.

This fixes about 20 fate failures on mips64el.

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

mips/aacdec: refactor out duplicated assembly code

The float_copy and fmul_and_reverse functions are refactored out from the
multiple copies in this file.

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

mips/sbrdsp: remove sbr_neg_odd_64_mips

The optimized C version of this code actually runs faster than this
version, so remove it.

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

mips/fft: remove some useless assembly

Remove some assembly that the compiler can easily handle optimally on its own.
GCC produces almost identical assembly.

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

mips/aacpsdsp: fix definition of ps_decorrelate_mips

Q_fract should have be declared as 'const float*'.
Also fix the constness of some local variables affected by this.

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

mips/mathops: remove 64-bit code

GCC is perfectly happy generating optimized multiplication code on its own for
64-bit arches. GCC refuses to optimize the loongson code when in 32-bit mode,
so I've left that.

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

configure: Move the cross_prefix setting after the toolchain one

Makes passing to configure
--toolchain=gcc-asan --cross-prefix=armv7a-hardfloat-linux-gnueabi-
work as intended.

tests: add paletteuse test

tests: add palettegen test

avfilter/paletteuse: use AV_QSORT()

See previous commit for a rationale.

avfilter/palettegen: use AV_QSORT()

This makes the sorting of the colors along an axis (r, g or b)
predictible, and thus testable under FATE. The performance is not really
an issue here since the function is called only once at the end and will
need to sort very small number of entries, so an alternative would be to
make the sorting functions (see DECLARE_CMP_FUNC()) fallback on another
axis in case of equality. This approach was actually simpler.

I don't know if there is any advantage in using a multidimensional sort,
but it will affect the final palette one way or another.