avutil: Fix compilation with inline asm disabled on mingw

Because of -Werror=implicit-function-declaration the build will fail.

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

lavu: Add interleaved 4:2:2 8/10-bit formats

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

h264dec: Add .avc file name extension

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

h264dec: K&R formatting cosmetics

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

mem: Handle av_reallocp(..., 0) properly

Previously this did a double free (and returned an error).

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

h264: remove an unused static constant

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

msvc/icl: Use __declspec(noinline)

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

msvc/icl: Use __declspec(deprecated)

Prior to this on msvc/icl there was no handling of deprecated functions
and the deprecated warning was disabled.

After enabling there are a number of warnings relating to the CRT and
the use of the non-secure versions of several functions.  Defining
_CRT_SECURE_NO_WARNINGS silences these warnings.

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

rtmp: Unbreak get_packet

Commit 5626f994f273af80fb100d4743b963304de9e05c broke it.

dcadec: Validate the lfe parameter

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

rtmpproto: Add a comment explaining the logic in handle_notify

This explains why the cleanup in 5626f994f was wrong and why
ae0f316a was needed.

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

build: Support cparser

r3d: Add more input value validation

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

rl2: Avoid a division by zero

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

wtv: Add more sanity checks for a length read from the file

Also make sure the existing length check can't overflow.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

yop: Clear all references to the AVBuffer in the local AVPacket

This makes sure that it doesn't try to free an AVBuffer belonging
to an earlier packet when we free the local packet at the end.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

fraps: Make the input buffer size checks more strict

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

segafilm: Validate the number of audio channels

This avoids divisions by zero later.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

svq3: Avoid a division by zero

If the height is zero, the decompression will probably end up
failing due to not fitting into the allocated buffer later
anyway, so this doesn't need any more elaborate check.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

qpeg: Add checks for running out of rows in qpeg_decode_inter

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

mpegaudiodec: Validate that the number of channels fits at the given offset

This is similar to the fix in 35cbc98b.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

asvdec: Verify the amount of extradata

The init function reads one byte of extradata.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

movenc: Use null buffers for measuring the amount of data to be written

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

aviobuf: Add functions for null buffers

Null buffers are useful for simulating writing to a real buffer
for the sake of measuring how many bytes are written.

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

asfenc: mux chapters in ASF files using an ASF "marker" section

ASF markers only have a start time, so we lose the chapter end times,
but that is ASF for you

Signed-off-by: Vladimir Pantelic <vladoman@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>

asfenc: remember send time and offset of the index entries

Signed-off-by: Vladimir Pantelic <vladoman@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>

asfenc: add ASF_Reserved_4 as defined in section 10.10 of the ASF spec

Signed-off-by: Vladimir Pantelic <vladoman@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>

asfdec: replace magic constant with DATA_HEADER_SIZE

Signed-off-by: Vladimir Pantelic <vladoman@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>

asfdec: substract preroll time from marker presentation time

this was forgotten when we changed ASF to not output the preroll time

Signed-off-by: Vladimir Pantelic <vladoman@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>

rv34: Fix a memory leak on errors

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

idroqdec: Make sure a video stream has been allocated before returning packets

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

rv10: Validate the dimensions set from the container

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

rtmp: Unbreak handle_notify

Commit 5626f994f273af80fb100d4743b963304de9e05c broke it.

aacdec: Add support for LD (Low Delay) AAC

aacdec: Add support for Error Resilience syntax.

This does not add support for any error resilience tools.

Add a WebP decoder

Container and lossy decoding by Aneesh Dogra <aneesh@sugarlabs.org>
Lossless decoding by Justin Ruggles <justin.ruggles@gmail.com>

lavc: add ff_u8_to_s8() for safe type punning of uint8_t to int8_t

avformat: Use av_reallocp() where suitable

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

rmdec: Validate the fps value

Abort if it is invalid if strict error checking has been requested.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

movenc: Add an option for omitting the tfhd base offset

This makes the output fragments independent of their position in
the output stream, making the output work better when streamed.

QuickTime Player doesn't support fragmented mp4 without the base
data offset, though.

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

movenc: Write the moof atom in two passes

This is a bit more work, but avoids having to fill in
the data offset field afterwards instead of directly when
the rest of the trun atom is written.

This simplifies future cases where this field needs to be set to
something different.

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

twinvqdec: Check the ibps parameter separately

This is required, since invalid parameters actually could
pass the switch check below.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

rtmp: Support reading interleaved chunks.

A given packet won't always come in contiguously; sometimes
they may be broken up on chunk boundaries by packets of another
channel.

This support primarily involves tracking information about the
data that's been read, so the reader can pick up where it left
off for a given channel.

As a side effect, we no longer over-report the bytes read if
(toread = MIN(size, chunk_size)) == size

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

xmv: Add more sanity checks for parameters read from the bitstream

Since the number of channels is multiplied by 36 and assigned to
to a uint16_t, make sure this calculation didn't overflow. (In
certain cases the calculation could overflow leaving the
truncated block_align at 0, leading to divisions by zero later.)

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

rtmpproto: Extend a comment to explain the prev_pkt arrays roles

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

rtmpproto: Consistently use the right prev_pkt array

prev_pkt[0] is used for input packets, while prev_pkt[1] is
used for output.

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

rtmpproto: Check for the right return code

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

dca_parser: Handle changes in DCA frame size

A change in framesize caused a perpetual loss of synchronization.
So read (and use) the frame size from the frame header instead of
setting it only once.

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

jpeg2000dec: Add more checks when parsing headers

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

doc: Add missing hashes and dates to APIChanges

Also fix the typo in the last entry version.

ffv1: Make sure at least one slice context is initialized

This avoids crashes when initializing the range coder for
the first slice context.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

eacmv: Make sure a reference frame exists before referencing it

This is similar to an existing check for the second-last frame
from 062421e3.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

truemotion2: Use av_freep properly in an error path

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

mpeg4videodec: Check the width/height in mpeg4_decode_sprite_trajectory

This avoids a potential division by zero.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

ivi_common: Make sure color planes have been initialized

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

mov: Don't use a negative duration for setting other fields

Some files have the duration set to -1 in the mdhd atom, more
or less legitimately. (We produce such files ourselves, for the
initial duration in fragmented mp4 files.)

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

oggparseogm: Convert to use bytestream2

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

rv34: Check the return value from ff_rv34_decode_init

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

matroskadec: Verify realaudio codec parameters

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

mace: Make sure that the channel count is set to a valid value

Also return a proper error code.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

svq3: Check for any negative return value from ff_h264_check_intra_pred_mode

Also pass on any returned error code.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

rtmp: rename main_channel_id to stream_id.

This more closely corresponds to the usage of the field.
Its usage here is unrelated to the channel ID.

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

rtmp: Follow Flash player numbering for channels.

Channel 4 is typically used by the Flash player to transmit
audio, channel 6 for video, and various stream-specific invokes
get sent over channel 8, which is designated the source channel.

This more closely matches the behavior of the Flash player,
including the transmission of play requests over channel 8.

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

rtmp: Send video on a separate channel.

Sending non-monotonic packets (e.g. when the audio and video
streams are monotonic within themselves but not muxed
monotonically) will lead to negative values the RTMP timestamp
field (where timestamps are transmitted only as deltas for each
channel), and this delta can end up being incorrectly written as
a large unsigned number.

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

lavf: Make probe_codec return an error code

This allows handling errors from av_realloc properly.

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

x86: Add an xmm clobbering wrapper for avcodec_encode_video2

This is required since 187105ff8 when we started trying to
wrap this function as well.

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

lavf: Don't explicitly flush after each written packet in muxers

Since 596e5d4783, this is not necessary anymore. It also allows to
actually disable the flushing, improving write performance (but
possibly giving worse latency in real-time streaming).

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

lavf: Add a flag to enable/disable per-packet flushing

This is enabled by default and can be disabled with
"-fflags -flush_packets".

Inspired by a patch from Nicolas George <nicolas.george@normalesup.org>.

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

flv: Do not export datastream as metadata

It is used internally.

rtmp: Store all the notify messages

The onTextData is used to implement text data streams in flv.

rtmp: Do not send the first field twice within the handshake

rtmp: Drop an unneeded warning

Apparently a widely used streaming server requires that the second
field always presents a version during C1 phase.

rtmp: Support play method in listen mode

rtmp: Factor out publish specific code

Will be reused for supporting play.

mem: Introduce av_reallocp

Fix references to deleted avcodec_encode_video() function

avpacket: Fix error checking in packet_alloc

Previously the wrong buffer pointer was checked, when buf
instead of *buf was checked. But checking the return value
instead is even better.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

vp3: Check the framerate for validity

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

proresdec: Properly make sure an index doesn't run past the limit

If idx equaled num_coeffs - 1 on entry to the loop, the previous
check failed to break out of the loop.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

cavsdec: Make sure a sequence header has been decoded before decoding pictures

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

vocdec: Don't update codec parameters mid-stream

If we really want to support parameter changes, they need to be
signalled along with the AVPackets as parameter change side data,
not just changing the AVCodecContext parameters when a packet
is demuxed (since there may be other earlier packets yet undecoded).

Something similar was already done for the sample rate in 0883109b2,
but some parameters were left changeable.

This avoids having to recheck the channel count for validity for
each decoded frame in (ad)pcm decoders, unless the decoders
explicitly say that they accept parameter changes.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

sierravmd: Do sanity checking of frame sizes

Limit the size to INT_MAX/2 (for simplicity) to be sure that
size + BYTES_PER_FRAME_RECORD won't overflow.

Also factorize other existing error return paths.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

omadec: Properly check lengths before incrementing the position

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

mathops/x86: work around inline asm miscompilation with GCC 4.8.1

The volatile is not required here, and prevents a miscompilation with GCC
4.8.1 when building on x86 with --cpu=i686

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

mem: Document the non-compatibility of av_realloc() and av_malloc()

pcm-dvd: Minor leftovers

Drop a pointless branch in uninit and use the compact copyright.

pcm-dvd: Support channel configuration changes

The sample buffering logic does not take into account that the blocksize
could change. Reset the buffer if the channel configuration changes,
since if there are leftover samples, it is most likely a broken or
misconcatenated stream. This could lead to negative numbers for
missing_samples during decoding.

Thanks to Michael Niedermeyer for pointing these out.

matroskaenc: Fix stray pointers left over from av_reallocp_array refactoring

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

mp3: add .mpa extension

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

configure: Mention that icl does not build both static and shared libs

g2meet: Allocate cursor buffers large enough to fit the aligned width

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

aic: Validate values read from the bitstream

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

mpc8: Make sure the first stream exists before parsing the seek table

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

mpc8: Check the seek table size parsed from the bitstream

Limit the size to INT_MAX/2 (for simplicity) to be sure that
size + FF_INPUT_BUFFER_PADDING_SIZE won't overflow.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

zmbvdec: Check the buffer size for uncompressed data

Also don't pointlessly set the buffer size to 1 after copying
one packet.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

ape: Don't allow the seektable to be omitted

The seektable is required for filling in ape->frames[i].pos
further down.

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

shorten: Break out of loop looking for fmt chunk if none is found

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

shorten: Use a checked bytestream reader for the wave header

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

smacker: Free memory properly if the init function fails

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

smacker: Make sure we don't fill in huffman codes out of range

Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>