Not only is an extra console obnoxious, but freopen() crashes.

This is required for Wine to find libvlc.dll and libvlccore.dll.

Just like we already do for (Unix) vlc.

This time, it should work (famous last words).

1kHz is way too much for human perception.

This unfortunately inserts silence shortly after resuming, right after
the samples that were held in the buffer during pause. This can possibly
be avoided with prebuffering (not tested).

Fixes #3078

do not wait until a state change event

GNOME went away from freedesktop DBus interface some time ago (for the sake of being 'special' ?)
Unfortunately that means we have to support 2 different (but very similar) interfaces

(GL/gl.h ought to be present if GL/glext.h is)

AC_CHECK_HEADERS does not work the way this change assumes. This commit
did not fix the stated problem.

This reverts commit 4e33295f.

Also scale maximum down to 2 times the nominal volume (i.e. +18dB) to
stay in the recommended range of +10 to +20dB.

The formula should be changed in any case. WaveOut has a logarithmic
scale, which is not consistent with the other audio outputs.

aout_VolumeUp() already handles this case correctly internally.

1.0 is nominal, not maximum. This is consistent with both VLC and the
MPRIS 2.1 specification. Also, the setter ought to have the same scale
as the getter!

(Do not assume that AOUT_VOLUME_MAX is 4 times AOUT_VOLUME_DEFAULT.)

VLC (like PulseAudio) stores the software volume i.e. the position of
the volume UI slider, rather than the linear amplification multiplier.

So do not use PulseAudio linear to volume conversion. Otherwise the
VLC volume gets wrongly cube-rooted.

We do not need to care which functions are in the math library. We only
need to care whether libm actually exists and can be linked against.

And by the way, lets define @LIBM@ for makefiles.

macosx: implemented both skipping within an item and item changing for the forward/backward buttons (closes #4943)