Both conversion and streaming output wizard are covered.
However, demuxdump is _not_ covered.

To avoid disrupting existing workflows, the dialog is only shown when
overwrite is disabled. If there is no UI, then it will be assumed that
the file is not overwritten.

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

This stops rejecting 0-byte packets, adds checks everywhere we
access the first byte of a packet, updates the th_decode_packetin()
return value checking so that duplicate frames aren't thrown away,
and updates video pts generation so the resulting frames are
displayed correctly,

Fixes bug #3416.
Signed-off-by: Rafaël Carré <funman@videolan.org>

This patch uses the information passed by the demuxer in block_t's
i_nb_samples and i_length to properly handle pre-skip, seeking
pre-roll, and end-trim. Multi-channel decoding should work. It also
adds .opus to the list of supported formats.
Signed-off-by: Rafaël Carré <funman@videolan.org>

This properly handles pre-skip, seeking pre-roll, and end-trim. It
uses the i_nb_samples field of block_t to signal to the decoder how
many samples should be returned after (possibly) skipping some at
the start of the block (for both pre-skip and pre-roll). In
addition, it abuses the i_length field of block_t to signal to the
decoder how many samples to trim from the end of the packet (for
end-trimming).

This patch does not compute correct timestamps for streams which
start at a non-zero offset (e.g., live streams joined in the
middle), nor does it correctly compute the length of a stream. But
neither of those things work for Vorbis, either. I'm leaving them
for a follow-up patch.
Signed-off-by: Rafaël Carré <funman@videolan.org>

Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>

Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>

Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>

If --no-sout-file-overwrite is specified, the access output will refuse
to overwrite an existing file. Overwrite is enabled by default for
backward compatibility with existing command lines.

While less precise, ticks count should be more stable than performance
counters.

We do not use (non-standard) alloca() anymore.

macosx: fixed conditionals for the different OS X release by adapting to the newly updated version defines

No functional changes.

WIN32 is defined also on WinCE anyway.

On OS X, fontconfig searches font only from /System/Library/Fonts, so this patch adds more search paths to its initialization.
Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>