Also, templatize the functions for 16-bit and 32-bit sample range. This will
be used for 24-bit FLAC encoding.

Needed to avoid integer overflows for 24-bit encoding.

This is much faster than calculating the MD5 one sample at a time.

This fixes compilation with --disable-everything and components enabled.

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

Documentation includes only the externally visible API of the installed
headers.

Based on a patch by Anton Khirnov <anton@khirnov.net>.
Signed-off-by: Anton Khirnov <anton@khirnov.net>

A sentence was mistakenly inserted in the middle of another sentence.

The packet belongs to lavf, the caller must not free it or access it in
any other way.
No change in practice, since destruct is set to NULL anyway.

Useful for detection of subtitles displayed during foreign language
scenes.
Signed-off-by: Diego Biurrun <diego@biurrun.de>

"mmxext" is a more sensible name and more common in outside projects.

This allows overriding the value from outside of the file.

DV1394_NTSC has a lower value than DV1394_PAL.
Signed-off-by: Martin Storsjö <martin@martin.st>

This avoids warnings about unused variables.
Signed-off-by: Martin Storsjö <martin@martin.st>

right above there is if (pkt) {<do stuff>; pkt = NULL}, so pkt is just a
fancy name for NULL at this point.

It has been disabled since it was added two years ago.

This location contains prebuilt binaries as well.
Signed-off-by: Martin Storsjö <martin@martin.st>

Same fix and issue as in a25d912dSigned-off-by: Anton Khirnov <anton@khirnov.net>

channels, sample_rate, bit_rate, and block_align can be used directly from
the AVCodecContext

A large invalid value could cause undefined behavior when left-shifted
by 8 later in the function.