This closes #1395

From the beginning we were using -march=pentium-m -mtune=prescott
However this was based on the Intel development machine of Apple. All actually shipped Intel machines by Apple are at least Core machines. As such, we might as well use -march=prescott -mtune=generic now. For 10.6 I currently think we should use -march=prescott -mtune=core2, since that will allow us to use much more advanced optimizations, and only few machines were shipped as Core, most were Core2 machines.

Instead of depending on a full xml parser --unavailable in our lua-- we
just assume one bit of information to occupy one line (jpsaman). In case
there are no newlines, fake them (jpd).

(cherry picked from commit e3cb87d8)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

It should fix lost some picture errors that were critical with dvd menus.
It also seems to fix #2803 but I don't really understand why.

The following diff for VLC 1.0.0rc3 corrects the detection of POSIX threads
on OpenBSD.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

The DV demuxer (rawdv) can't deal with truncated frames, so discard them.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

This makes Windows use the same mode with transparency than on Linux.
This removes the opacity trick to hide() the code.
This fix a few issues that were happening and difficult to fix because of this difficulty.

It will probably break stuffs for people on OLD windows version (2000,XP) with broken graphic cards... (Deactivate it by yourself) But I have spent enough time on this code.

"the short description of --sout-ts-pcr is still "PCR delay (ms)" when it should be "PCR interval (ms)"
as suggested by Dnumgis.

(cherry picked from commit a6110fb1)
(cherry picked from commit b24fc9a9)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This will hopefully help avoid some of the reports where people fail to notice that there is an expand triangle next to what they clicked on.

The repeat was done before buffered data were consumed.

It is about [a0d23587].
Without this patch, with files that have a shorter track, the demuxer
would try seeking the track every time the Demux() function was called.

It is a partial revert of [a8c9c25d]

It is more a rewrite than a conversion ;)

It have removed the RGB support (it was broken for 15/16 bits
and seemed highly suspicious for 24/32 bits)
It might have been overzealous about some simplifications, if so
you are welcome to send patches.