If this stays too counter-intuitive, revert this commit, 672c3f35
and 7fe659bf.
IMHO, the user who enables the record access-filter knows what he does.

Fixes #1902
Signed-off-by: Derk-Jan Hartman <hartman@videolan.org>

Note that all NS*Integer types were introduced on 10.5 only. They can tollfree bridged with the respective plain ints. They are undocumented anyway...

Proposed by Georg Seifert.

Don't send failure messages when probing filter. (It unfortunately seems impossible to check if a filter is being probed or asked for explicitly)

Fix forced audio filter2 usage in transcode (when using afilter=). Downmixing now works with channels=1 (auto probed filter) or afilter=mono.

Before setting end_date, remember to initialize the divider for future
calculations.
Signed-off-by: David Flynn <davidf@rd.bbc.co.uk>
Signed-off-by: Antoine Cellerier <dionoea@videolan.org>

18:36 antoine@macbook ~/videolan/vlc-build% ./vlc ~/media/music/Greenday/American\ Idiot/01\ -\ American\ Idiot.mp3 --sout '#transcode{acodec=mpga,ab=32,channels=1}:std{access=file,mux=raw,dst=/tmp/a.mp3}' -vvv -I dummy --no-media-library

(with pre 9107e1e0 builds you need to add --sout-mono-downmix to the command line)

Use the same order for buttons in the fullscreen interface than in the normal interface (there's still an extra space between the next and fullscreen toggle buttons ... don't know how to remove it)

Change the fullscreen button in the fullscreen controller (since clicking it switches from fullscreen and not to fullscreen).

I will properly fix it.

This reverts commit 1e6d1086.

Keep p_input->i_state when adding a subtitle on the fly. This prevents i_state from being changed to BUFFERING_S (which would in turn prevent the input from being paused). This should close #1904. (Should be backported in 0.8.6 if we keep that branch)

Reported by Georg Seifert.

Most likely due to the late and incomplete filter rework