movenc: fix tkhd height for imx

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit fffdee89cc0712ed3324bae1f86e271a36c02822)

avio: deprecate url_fskip

avio_seek should be used instead

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 0300db8ad778a194b4a8ec98f6da3de5b41c46ee)

lavf: replace all uses of url_fskip with avio_seek

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit e356fc57a2e9887370caec58d8aafeafd1f336dc)

avio: avio_ prefix for url_fseek

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 6b4aa5dac8f41aa452d0ce9a1bede9e59a303060)

huffyuv: Add multithreading support

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 76d8846c4e918749b045ea2ee7399a069af5e4a5)

pthreads: Fix bug introduced with thread_safe_callbacks

For intra codecs, ff_thread_finish_setup() is called before decoding starts
automatically. However, get_buffer can only be used before it's called, so
adding this requirement broke frame threading for them. Fixed by moving the
call until after get_buffer is finished.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit ad9791e12b6653a465803062e2543f25916300d3)

Fix duplicate & droped frame in Californication.S02E01.HDTV.XviD-NoTV.avi

Attribute used is correctly supported by the Intel C compiler since version 11.1.

Use macro __INTEL_COMPILER instead of __ICC.

__ICC is not supported on all platforms, this is an Intel documentation
bug that is supposed to be fixed in a future release.
(Intel issue 612289)

Use --sysroot flag for clang

Although not documented, clang does support the --sysroot flag, and it
does the right thing.  Use this flag intead of -isysroot which only
applies to header file searches, not the linker.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 00ba041cb396f88f68a1bf7907e7c98ded3760ab)

In mov muxer, fix adpcm mono muxing, patch by Alex Converse

In mpegts demuxer, do not output known sized packet if an unbounded packet
is already queued. Fix issue #2624.

ARM: VP8: fix build on systems with global symbol prefix

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 0b32da90f893b3e04ead18a546252e241a6a0988)

ARM: fix vp8 neon with pic enabled

The assembler emits literal pools too far from the load instructions,
so we must do it explicitly at a suitable location.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 8b454c352f49c2a61db37793d838b553db3da734)

configure: document FATE_SAMPLES env var in --help text

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 52b3cc6047e2b93e8a798734c65904243c1a7be3)

fate: get samples location from env var if not explicitly set

Use the FATE_SAMPLES environment variable if samples location
is not set with the --samples configure option or on the make
command line.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit aa3805a486831d9c9e40cc9e897c048d60b51c1a)

movenc: remove uses of deprecated API.

Replace put_tag() with ffio_wfourcc() and ByteIOContext with AVIOContext.
(cherry picked from commit 13ff92d19706d17f876ff676591dace0b3e34794)

store pasp atom for all types of quicktime movie

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
(cherry picked from commit d184c86cd309bbd61c5d1a1dfcf60d133da4c1e3)

reindent after tapt patch

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
(cherry picked from commit 77d207cbe64710b7b34865f347e608b9c0e15968)

use tapt atom for sample aspect ratio

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
(cherry picked from commit ea1afa124cb58c7eebd4fd3706434fafb58b00c0)

bink: prevent overflows within binkidct by using int-sized intermediate array

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit e211e255aa399d68239ffa42c9cc7a52eb3d55a9)

bink: decode audio track identifiers into AVStream.id

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 588a3ffd9687cbef60cc26d5b3a247b8ecdbf75b)

Bink version 'b' audio decoder

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit ccfcddb3f287545a20831f266c2a2f734d6a9f31)

binkaudio: simplify frame_len_bits and frame_len calculation

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 8a8c283edd0da5863eb729a31a94b70820400941)

binkaudio: remove unnecessary loop

decode_init sets bands[0] == 2, so this loop always sets the band table
index (k) to zero.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit a304def1dca50d63bf2a39651f84792980db3508)

binkaudio: perform band table scaling in decode_init

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 582ac86d19c61ad73c2e1a2d5ed54a5fe1e20bd1)

bink: set audio stream codec_tag such that binkaudio decoder can identify bitstream version

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit f0ca29eb5f476a281c8066214e3745244656d417)

bink: use LOCAL_ALIGNED for aligned stack data

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 8997bb8807336dd6d6c96364952bb71c161ac3b2)

dxva2: define required feature selection macros

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 04973f8082c5a822112d2e42d535b7f3f59dccc0)

OS/2: lxlite should use stdout

This causes lxlite to use stdout instead of vioXXX
functions. This improves fate and build logs readability.
Affects OS/2 only.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit cc4e9d2a24a9ca79d4e912df21053426411aa14b)

avio: deprecate put_tag

it's not used internally anymore and shouldn't be public.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 61840b43607679e8c4a1a63f005770f69dda12fa)

lavf: replace remaining uses of put_tag with avio_write

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit bbc413f943b53d96aa85fe272ec413c85794019f)

avienc: replace &tag[0] with tag.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 99f42c27abfe916fa19ac0c18489c8bbf8525406)

lavf: use a new ffio_wfourcc macro instead of put_tag() where possible

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 0abdb2931719d96dee725e555e9b46b2b2f8a6be)

dca: use EXT_AUDIO_ID field to determine core extensions

This avoids the core substream extensions scan when the EXT_AUDIO_ID
field indicates no extensions or only unsupported extensions. The scan
is done only if the value of EXT_AUDIO_ID is unknown or indicates a
present XCh extension which we can decode.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 7e06e0ede3b798f591634b277e8dfa6507b196de)

vmdaudio: output 8-bit audio as AV_SAMPLE_FMT_U8.

There is no need to expand to 16-bits. Just use memcpy() to copy the raw data.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 1108f8998c6536bb44d6ecbe5adfa18e0c1478e8)

vmdaudio: remove unnecessary fields from VmdAudioContext and use the corresponding AVCodecContext fields instead.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 2ec7862db8b782020cc220dae827c3438d108b3a)

vmdaudio: add out_bps to VmdAudioContext and use it to replace hard-coded sample size.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 1e86d685e0935077766c645e49b8533d41ca11cb)

vmdaudio: simplify vmdaudio_decode_frame() by handling block_type first, then making a single call to vmdaudio_loadsound().

This also adds output buffer size checks for AUDIO and SILENCE block types.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 1574eff3d23ad799d25454a449b01f94795495ea)

cosmetics: reindent after previous commit

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit ba9516cca845f8d3fb7ac08ef53a996c3ee0dbf5)

vmdaudio: move all silence chunk handling to vmdaudio_loadsound().

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 762b386e4aafc7bbdbec367bc652cf1199e81f51)

cosmetics: remove debugging cruft

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 8e9027d266ef39ab9f88b4bbad5cf9e425d0696c)

cosmetics: reindent after previous commit

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 868f2f4d90e07edd9a65bd9b917bb5940643ec75)

vmdaudio: simplify buffer pointer and header size handling.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 2d213695fce86e740800ddcf43d8d0864d2ecea5)

vmdaudio: set *data_size to zero when skipping small packets and add a warning log message.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 7a4fb3fd9357dfdd27431a0a8d7250dab54a9938)

vmdaudio: validate block type

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 22f893e1c9f9387f0a021f775757130fa48e0180)

vmdaudio: use macros and a local variable for block type.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit dd1af5136fe7767f2f18ac943efe994946864640)

vmdaudio: correct the silent chunk count in the first block.

This fixes A/V sync with several samples, notably:
http://samples.mplayerhq.hu/game-formats/sierra-vmd/swat_*.vmd

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 6989cb2dae85ea455ffcc8a36a763134fb311e29)

vmdaudio: output audio samples for standalone silent blocks.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 9b73f786005f31429d7c88092edfeef6696a5f69)

vmdaudio: remove duplicated code by merging mono and stereo decoding.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 1328d433137c055df06f99772b243ceec2bbf36e)

vmdaudio: fix raw_block_size calculation.

The size should depend on the output sample size, not the internal bit depth.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit a58bcb40b164b92957db73e702465808a9180126)

Mark MS-MPEG4v1 encoder as experimental.

The encoder has never produced files that could be decoded
with any software.

Add libmpcodecs to the list of libavfilter subdirectories.

lavf: update documentation of AVOutputFormat.flags

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

Make FF_DEBUG_DCT_COEFF output coeffs via av_log() instead of just via AVFrame.
This allows the values to be used without changing C code and is closer to how
the other DEBUG flags work.
If this causes a problem for any user of this flag, please tell me and
ill split the flag in 2.

lavf: postpone removal of deprecated avio functions

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

lavf: postpone removal of public metadata conversion API

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

avienc: fix AVI stream index for files with >10 streams

Fixes issue 2563.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 4acc94e97a9551d11ead29735e23283d71f1d4c2)

avio: rename url_fopen/fclose -> avio_open/close.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 22a3212e32b696028e21f00871f3cb48c044029d)

libavformat: Remove FF_NETERRNO()

Map EAGAIN and EINTR from ff_neterrno to the normal AVERROR()
error codes. Provide fallback definitions of other errno.h network
errors, mapping them to the corresponding winsock errors.

This eases catching these error codes in common code, without having
to distinguish between FF_NETERRNO(EAGAIN) and AVERROR(EAGAIN).

This fixes roundup issue 2614, unbreaking blocking network IO on
windows.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 28c4741a6617a4c1d2490cb13fc70ae4c9c472da)

Add more AVC Intra FOURCCs

Also change the comments a bit since the FOURCCs aren't specific to Flip4Mac
and different ones are used for 720 versus 1080 lines.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 8f935b9271052be8f97d655081b94b68b6c23bfb)

aacdec: Reduce the size of buf_mdct.

It was doubled in size for the LTP implementation. This brings it back
down to its original size.
(cherry picked from commit e22910b21a6c78b0159f98426b10c204f12bc15a)

Remove old VCSs from version.sh
(cherry picked from commit 203df50d1013ea59e558d63f3ea69ffae1249124)

mpegtsenc: use correct PES stream_id for AAC

This adds the AAC codec to the list of audio codecs that results
in a PES stream_id of 0xc0 (audio stream).

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 6c065f0acd5b50a9037b561c6fc370b316832543)

spdifenc.c: fix compile because of missing include avio_internal.h.
(cherry picked from commit 0f86fcabdf74f658323140111dfd4de83f7f2eee)

vp8: ppc: fix invalid reads in altivec epel mc

The 4-tap filters should only access one row/column before the
reference block.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit e0e46cae377347cbe1cd27c0d85568921b12c2ad)

ppc: fix vc1 inverse transform, unbreak build

GCC 4.3 and later are more particular about signedness matching
in vector operations.  The operations under if(rangered) were
missing assignments and thus had no effect.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 381efba0ecedd41575f99ba9e9bd3826551079f6)

avio: make put_nbyte internal.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 0ac8e2bf2bf3d636241bf2811018d9974687a63c)

avio: avio: avio_ prefixes for put_* functions

In the name of consistency:
put_byte           -> avio_w8
put_<type>         -> avio_w<type>
put_buffer         -> avio_write

put_nbyte will be made private
put_tag will be merged with avio_put_str

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 77eb5504d3b3e1047900382350e0bc5e0bfb16b5)

targa: prevent integer overflow in bufsize check.
(cherry picked from commit 78e2380a6d09e7a8b2a74d090abfb0a922e046f6)

avio: make get_partial_buffer internal.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit b3db9ceef13037f2b6963a84b7abf8871a1e0f8a)

avio: avio_ prefixes for get_* functions

In the name of consistency:
get_byte           -> avio_r8
get_<type>         -> avio_r<type>
get_buffer         -> avio_read

get_partial_buffer will be made private later

get_strz is left out becase I want to change it later to return
something useful.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit b7effd4e8338f6ed5bda630ad7ed0809bf458648)

VC1: merge idct8x8, coeff adjustments and put_pixels.

Merging these functions allows merging some loops, which makes the
results (particularly after SIMD optimizations) much faster.
(cherry picked from commit f8bed30d8b176fa030f6737765338bb4a2bcabc9)

avio: rename av_alloc_put_byte -> avio_alloc_context for consistency

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 8d9ac969cb4ac3e3e18f6425703af4d7aec6c513)

dsputil: make {add/put/put_signed}_pixels_clamped() non-static.
(cherry picked from commit 484a337cd7cd8bb180c4a1bd3321881f1c874a92)

VC1: inline vc1_put_block() in vc1_decode_i_blocks().

Advantage is that it allows us to combine several loops into a single
one, and these can eventually be merged into the IDCT itself. Also, it
allows us to remove vc1_put_block(), and makes CODEC_FLAG_GRAY faster.
(cherry picked from commit bbfd2e7ab4e2ae0b934657fe51afdbbbaead52b7)

Makefile: include deps from tools directory

This ensures the tools are rebuilt when necessary.  Specifically,
lavfi-showfiltfmts was sometimes not rebuilt causing spurious test
failures.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 08df7f866663f83ba42398cea9f3eeff4681ea9a)

amrnb: use correct size when copying lsf_r array

lsf_r is an array of int16_t, not float.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 1efa772e20be5869817b2370a557bb14e7ce2fff)

x86: use raw opcode for xgetbv instruction

This allows the CPU detection to work with assemblers not supporting
the xgetbv mnemonic.  These include clang and some BSD versions.

All AVX code will be written for yasm, where the main assembler
is not involved.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit ef6695387522fc783105ac187c84beaa284e7cff)

Add shadow support to the drawtext filter

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

factor draw_glyphs out of drawtext filter

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

Fix and optimize yuv blend in the drawtext filter

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

get rid of divisions in the inner loop of the drawtext filter

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

fix chroma alpha in the drawtext filter

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

Simplify chroma blend in the drawtext filter

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

Port drawtext filter by Hemanth from the libavfilter soc repo, with
the following additions:
* support to anti-aliased glyph rendering
* support to UTF-8 text and Unicode chars rendering
* support for RGB packed formats

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

lavfi: add fade filter

Port fade filter from libavfilter soc repo, with minor fixes by
Stefano.

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

lavfi: put color source in a dedicated file

Move the color source code from vf_pad.c to vsrc_color.c.

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

lavfi: add drawutils

Add drawutils.h and drawutils.c, and use them in the pad filter.
The new functions are going to be shared by other filters.

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

Update version and APIchanges.

Update libavformat/version.h and doc/APIChanges after renaming
init_put_byte() and ByteIOContext to ffio_init_context() (private)
and AVIOContext, (public), and deprecating the originals.
(cherry picked from commit d2bbf82e65e4a128ab1f0bf48df5d1e43c43c698)

avio: move init_put_byte() to a new private header and rename it

init_put_byte should never be used outside of lavf, since
sizeof(AVIOContext) isn't part of public ABI.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit e731b8d8729e75bfb69f5540e6446d6118dac549)

avio: rename ByteIOContext to AVIOContext.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit ae628ec1fd7f54c102bf9e667a3edd404b9b9128)

VC1: don't use vc1_put_block() in vc1_decode_i_blocks_adv().

Advanced profile never uses "range reduction", so vc1_put_block() quite
literally just calls put_pixels_clamped() from vc1_decode_i_blocks_adv().
By inlining the function, we can prevent calling IDCT8x8 if
CODEC_FLAG_GRAY is set, and we don't have to scale the coeffs in the
[0,256] range, but can instead use put_signed_pixels_clamped().
(cherry picked from commit 70aa916e4630bcec14439a2d703074b6d4c890a8)

x86: check for AVX support

This adds configure and runtime checks for AVX support on x86 CPUs.

Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 87f1355f9b4fc11414d0e6a91404203c2745f89f)

MMS: also discover streams in extended stream properties object

Allows playback of nonprimary audio streams in multiple bitrate sources,
such as mmsh://wmscr1.dr.dk/e02ch03m

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
(cherry picked from commit 74d6871d6244865b5863a01c3dab16a2f06a1706)

dsputil_mmx.c: remove ff_vector128.

Remove ff_vector128, it is identical to ff_pb_80.
(cherry picked from commit bf6fa732459399fac215bdfa44dd39a6fb1d1e01)

Fix invalid reads in VC1 decoder

Patch discussed and taken from https://roundup.ffmpeg.org/issue2584
(cherry picked from commit 2bbec1eda46d907605772a8b6e8263caa4bc4c82)

Fix VP3 edge emulation

With negative stride, the start of the edge_emu buffer should be pointing to
the last line, not the end of the buffer.
With positive stride, pointing to the end of the buffer was completely wrong.
(cherry picked from commit a89f4ca005efa5ee82089583c7b8bc00bcee816a)

aacdec: dsputilize the scalar multiplication in intensity stereo
(cherry picked from commit 9707f84fa73c23352937fc7e4e0a85eaf3135cbc)

VP3: fix decoding of videos with stride > 2048

Also remove qscale_table code; this didn't make sense anyways as VP3 doesn't
use an MPEG-like quantizer scale.
(cherry picked from commit 902685b8abf4e2089802528a32a50f3c68033fda)