anm: convert to bytestream2 API

Protects from overreads.

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

bytestream: add more unchecked variants for bytestream2 API

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

jvdec: unbreak video decoding

The safe bitstream reader broke it since the buffer size was specified
in bytes instead of bits.

Signed-off-by: Janne Grunau <janne-libav@jannau.net>
CC: libav-stable@libav.org

jv demux: set video stream duration

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Janne Grunau <janne-libav@jannau.net>

fate: add pam image regression test

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Janne Grunau <janne-libav@jannau.net>

doc/general: update supported devices table.

Add libcdio and fbdev, remove v4l1.

doc/general: add missing @tab to codecs table.

h264: Fix invalid interlaced/progressive MB combinations for direct mode prediction.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

avconv: reindent

CC: libav-stable@libav.org

avconv: link '-passlogfile' option to libx264 'stats' AVOption.

Fixes bug 204.

CC: libav-stable@libav.org

libx264: add 'stats' private option for setting 2pass stats filename.

x264 always opens the file itself with fopen, so we cannot use the
standard lavc stats mechanism.

CC: libav-stable@libav.org

libx264: fix help text for slice-max-size option.

CC: libav-stable@libav.org

http: Clear the auth state on redirects

Currently we only try continuing with the same auth mechanism
as the initial request.

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

http: Retry auth if it failed due to being stale

Allow up to 4 retries for normal requests, where both the
proxy and the target server might need to authenticate.

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

rtsp: Resend new keepalive commands if they used stale auth

These commands are sent asynchronously, not waiting for the reply.
This reply is parsed later by ff_rtsp_tcp_read_packet or
udp_read_packet. If the reply indicates that we used stale
authentication and need to use a new nonce, resend a new keepalive
command immediately.

This is the only request sent asynchronously, so currently there's
no other command that needs to be resent in the same way.

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

rtsp: Retry authentication if failed due to being stale

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

httpauth: Parse the stale field in digest auth

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

dxva2_vc1: pass the overlap flag to the decoder

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

dxva2_vc1: fix decoding of BI frames

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

FATE: add shorthand to wavpack test

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

dfa: convert to bytestream2 API

Protects from overreads.

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

anm decoder: move buffer allocation from decode_init() to decode_frame()

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

h264: improve parsing of broken AVC SPS

Parsing the entire NAL as SPS fixes decoding of some AVC bitstreams
with broken escaping. Since the size of the NAL unit is known and
checked against the buffer end we can parse it entirely without buffer
overreads.

Fixes playback of
http://streams.videolan.org/streams/mp4/Mr_MrsSmith-h264_aac.mp4

Signed-off-by: Janne Grunau <janne-libav@jannau.net>

pcm-mpeg: convert to bytestream2 API

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Janne Grunau <janne-libav@jannau.net>

Revert "h264: clear trailing bits in partially parsed NAL units"

This reverts commit 729ebb2f185244b0ff06d48edbbbbb02ceb4ed4e.

There was an off-by-one error in the bit mask calculation clearing
actually the last valid bit and causing
http://bugzilla.libav.org/show_bug.cgi?id=227

The broken sample (Mr_MrsSmith-h264_aac.mp4) the commit was fixing
does not work after correcting the off-by-one error.

CC: libav-stable@libav.org

remove iwmmxt optimizations

The were broken since August of 2010 without anyone noticing until
three weeks ago. Nobody cares about it anymore and hopefully Marvell
will support NEON like in the PXA978 from now on.

mimic: do not continue if swap_buf_size is 0

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

mimic: convert to bytestream2 API

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

frwu: use MKTAG to check marker instead of AV_RL32

Using intreadwrite.h for this is overkill.

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

txd: port to bytestream2 API

Protects against overreads.

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

c93: convert to bytestream2 API

Protects against overreads.

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Martin Storsjö <martin@martin.st>

iff: make .long_name more descriptive

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Martin Storsjö <martin@martin.st>

FATE: add test for cdxl demuxer

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Martin Storsjö <martin@martin.st>

rtsp: Fix a typo

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

dsicinvideo: validate buffer offset before copying pixels.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

cook: error out on quant_index values outside [-63, 63] range.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

mpc: pad mpc_CC/SCF[] tables to allow for negative indices.

MPC8 allows indices of mpc_CC up to -1, and mpc_SCF up to -6, thus pad
the tables by that much on the left end.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

cook: expand dither_tab[], and make sure indexes into it don't overflow.

Fixes overflows in accessing dither_tab[].

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

xxan: reindent xan_unpack_luma().

It used 3-space indent instead of 4-space indent.

xxan: protect against chroma LUT overreads.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

xxan: convert to bytestream2 API.

Protects against overreads.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

xxan: don't read before start of buffer in av_memcpy_backptr().

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

vp8: convert mbedge loopfilter x86 assembly to use named arguments.

vp8: convert inner loopfilter x86 assembly to use named arguments.

Fix a bunch of common typos.

build: Skip compiling xvmc.h under the correct condition.

configure: darwin: Change dylib install names to include major version.

This will cause linkers to link against the major lib names, instead of the
base names, allowing multiple major versions of the libraries to co-exist.

Signed-off-by: Diego Biurrun <diego@biurrun.de>

mpegts: Always honor a registration descriptor if present and there is no other codec information.

aacdec: Fix SCE parity check.

An unpaired SCE preceding a CPE only makes sense for front SCEs
preceding the first CPE.

Split from FFmpeg commit a8d67efa53dae1d14614e3a7bd4e77e4eab066ab

Signed-off-by: Alex Converse <alex.converse@gmail.com>

aacdec: Fix out of array writes (stack).

Set the element to channel vector (e2c_vec) size to be the maximum
number of aac channel elements. This makes it slightly larger than it
needs to be because CCEs are never mapped to output channel locations.

Also add a check that all input tags (legal or not) will fit.

Split from FFmpeg commit a8d67efa53dae1d14614e3a7bd4e77e4eab066ab

Signed-off-by: Alex Converse <alex.converse@gmail.com>

rtsp: Only set the ttl parameter if the server actually gave a value

Passing ttl=0 to the rtp/udp url contexts makes packets never
leave the host machine.

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

udp: Set ttl for read-write streams, too, not only for write-only ones

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

udp: Only bind to the multicast address if in read-only mode

This fixes sending back RTCP RR packets if receiving RTP over
multicast.

If the multicast stream is sent on demand (set up and signalled
via RTSP), the sender might depend on getting RTCP RR packets
knowing that there are listeners, otherwise the stream can be
closed after a certain timeout.

This fixes receiving RTSP streams over multicast on unix, from
certain Axis cameras.

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

udp: Clarify the comment about binding the multicast address

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

udp: Reorder comments

When this code was added in 36b532815cb83, the new code was added
between the existing comment and the existing line of code, making
the old comment seem to refer to the new code. This makes it read
correctly.

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

ttadec: unbreak playback of matroska files

Matroska demuxer needs to recreate tta header, so just display
crc error without aborting.

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

vorbisdec: avoid invalid memory access

This fixes some invalid memory access caused later in the function
by res_chan[] not being set for all channels. This happens when a
channel doesn't appear a submap. This change simply returns a
decoder error when this situation is detected.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

Fix uninitialized reads on malformed ogg files.

The ogg decoder wasn't padding the input buffer with the appropriate
FF_INPUT_BUFFER_PADDING_SIZE bytes. Which led to uninitialized reads in
various pieces of parsing code when they thought they had more data than
they actually did.

Signed-off-by: Dale Curtis <dalecurtis@chromium.org>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

huffyuv: add padding to classic (v1) huffman tables.

We slightly overread the input buffer, so we require
padding at the end of the buffer, as is documented in the
get_bits API. Without padding, we'll read uninitialized
data or beyond the end of the .rodata, which may crash.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

png: convert to bytestream2 API.

Protects against overreads in the input buffer.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

dca: include libavutil/mathematics.h for possibly missing M_SQRT1_2

avs: fix infinite loop on end-of-stream.

The codec would keep returning the last decoded frame if the stream
contains B-frames, since it wouldn't clear that frame from the list of
frames to be returned to the user.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

tiffdec: Prevent illegal memory access caused by recycled pointers.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

rtpenc: Fix the AVRational used for av_rescale_q_rnd

The current one has a zero denominator - this is what was
intended in 14aecc50fae6.

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

wma: fix off-by-one in array bounds check.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

sbrdsp.asm: convert all instructions to float/SSE ones.

Since the values are floats, using the float operations
makes sense, improves performance on some CPUs and
makes the code SSE compatible instead of needing SSE2.

Based on suggestion by Jason.

Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

dv: cosmetics.

dv: check buffer size before reading profile.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

Revert "AAC SBR: group some writes."

This reverts commit ba36f14e5d34bcca6ecf77ba1c7512d1b51a9740. It
broke decoding on x86-32 on some systems.

udp: Print an error message if bind fails

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

cook: extend channel uncoupling tables so the full bit range is covered.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

roqvideo: cosmetics.

roqvideo: convert to bytestream2 API.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

dca: don't use av_clip_uintp2().

The argument is not a literal, thus causing the ARM v6 or later
builds to break.

wmall: fix build with -DDEBUG enabled.

smc: port to bytestream2 API.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

AAC SBR: group some writes.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

dsputil: remove shift parameter from scalarproduct_int16

There is only one caller, which does not need the shifting. Other use cases
are situations where different roundings would be needed.

The x86 and neon versions are modified accordingly.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

SBR DSP: unroll sum_square

The length is even, so some unrolling can be performed. Timings are for x86:
- 32bits: 102c -> 82c
- 64bits:  82c -> 69c

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

rv34: remove dead code in intra availability check

This was an incorrect copy-and-paste to a code not needing the original code.
Spotted by Jason in a previous review but forgotten in the commit.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

rv34: clean a bit availability checks.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

v4l2: update documentation

Update the examples and remove stale information.
Video4linux support had been already removed.

tgq: convert to bytestream2 API.

This protects against input buffer overreads.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

parser: remove forward declaration of MpegEncContext

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

dca: prevent accessing static arrays with invalid indexes.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

raw: move buffer size check up.

This way, it protects against overreads for 4bpp/2bpp content also.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

lpcm: fix sample size calculation for 20bit LCPM.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

swscale: clip unscaled output intermediates.

Fixes bug 240, as well as several integer overflows (visible as glitches)
in other scaling output routines, e.g. YUV422.

smacker: error out if palette copy-with-offset overruns palette size.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

algmm: convert to bytestream2 API.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

wma: Refactor common code to fix standalone compilation of WMA lossless decoder.

vc1: Move init code shared between decoder and parser to common code file.

This fixes standalone compilation of the VC-1 parser.

x86: Remove duplicated AVG_3DNOW_OP / AVG_MMX2_OP macros from h264_qpel_mmx.c.

mov: Allow last chunk to have an arbitrary number of samples.

Fixes ticket #673.
(cherry picked from commit 8dcd2a41ecff8cc1e9b20cc267df54c59878ab3b)

Signed-off-by: Alex Converse <alex.converse@gmail.com>

SBR DSP: fix SSE code to not use SSE2 instructions.

movq from SSE register _to_ memory is an SSE2 instruction.
Use the SSE movlps function instead that does the same thing.

Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

cpu: initialize mask to -1, so that by default, optimizations are used.

error_resilience: initialize s->block_index[].

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

svq3: protect against negative quantizers.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

Don't use ff_cropTbl[] for IDCT.

Results of IDCT can by far outreach the range of ff_cropTbl[], leading
to overreads and potentially crashes.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org

swscale: make filterPos 32bit.

Fixes overflows for large image sizes.

Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org