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

All decode_channel_map calls together can easily read
more data than the amount of padding available.
Thus below patch adds an input length check before reading them.
Fixes some invalid reads with sample from
http://bugzilla.mplayerhq.hu/show_bug.cgi?id=1138

Fixes warning message with ticket #68.

Fixes ticket #68.

Based on a patch by Reimar Döffinger.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

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

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

Now that pix_fmt, width, and height are directly embedded in the AVFrame,
there is no need to pass these values as separate arguments.

This is required specifically for setting frame->format to -1,
otherwise it will be set to 0 = PIX_FMT_YUV420P and code reading
the format from the output decoded frame will get misled.

In particular fix regressions occurring with the pending vsrc_buffer
patch.

The new name is more consistent with the employed naming scheme, and
more meaningful.

Compact code, improve readability.

The check and the corresponding message in do_video_out() are
unnecessary in the libavfilter path, as a similar check and message
are performed within the buffer source.

Implement lazy initialization of the image resample context, so that
the initialization code does not need to be duplicated.

Also manage the case in which resample/size change mid-stream. For
each packet the code computes if resampling is needed.

Previously the resample check was done only at the initialization, in
case size/format changed and no resample was detected at the init
stage ffmpeg was silently exiting.

avfilter_copy_frame_props() avoids code duplication and increases
robustness.

The added files libavfilter/avcodec.[ch] are used for containing
utilities useful for gluing togheter libavfilter and libavcodec.

The format is a per-frame property, having it in AVFrame simplify the
operation of extraction of that information, since avoids the need to
access the codec/stream context.

width and height are per-frame properties, setting these values in
AVFrame simplify the operation of extraction of that information,
since avoids the need to check the codec/stream context.

recreating the value.

Avoids a memleak, is less code and might be faster.

twice and add a missing "const".

* qatar/master:
  mov: fix composition timestamps on movie fragments.
  wmavoice: Use proper size in memeset().

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

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

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

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

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

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

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

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

The patch below provides exactly that to the MP3 muxer. A XING header
containing

   * the numer of frames,
   * the size, and
   * a TOC

is generated.

It's based on an idea by Anton Khirnov (restricted to the number of
frames) found at

   http://patches.ffmpeg.org/patch/1891/

The TOC is generated as found in lame's "VbrTag.c".

According to my tests the following reproduces the number of frames, the
size and the TOC in "c.mp3" from "b.mp3" (except a shift due to shorter
XING header generated by FFmpeg):

   lame -V2 a.wav b.mp3
   ffmpeg -i b.mp3 -acodec copy -y c.mp3

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

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

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

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

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

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

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