* Fixed support for SPU and LPCM streams over network.

1) Title / Chapter selection
2) Subtitle selection
3) Audiochannel/Language selection
4) Angle selection

This is a problem because the intf is currently destroyed
before the vout.

without starting it from the command line)

Cleaned up playlist handling in Intf_Vlc_Wrapper.

  * Modified the pf_close kludge in input.c so that we don't segfault when
    pf_init fails.

from title selection ones.

Readonly filesystems like NTFS where being incorrectly added.

- Fixed a polioterie from previous commit

* Some optimization in imdct (all data are now aligned)
* SSE downmix now works for windows
* SSE imdct is desactivated for windows (MINGW32 doesn't know how to aligned data)

Also improved more "BeOS" CD detection.
Todo new graphics for the navigation controls.

  * Synced dummy_dvdcss with Stef's changes to libdvdcss.
  * configure now compiles the VCD plugin if linux/cdrom.h is present.

the VCD input arrives now :)

I'd better go to bed !

The Linux specific stuff is located in linux_cdrom_tools files.
For the time being, it only has title (tracks). Chapters will come if I
ever can read .vcd files.
I've modified Gnome and GTK+, not the others.
I've noticed a big bug, it segfaults at the end of a title. I will
have a look on this tomorrow.
Please be indulgent, it's my first peace of code ;-)

  * Fixed NetBSD (and probably OpenBSD) libdvdcss compilation.

  * xvideo.so is now built in by default. Mixing PIC and non-PIC code was not
  very nice.
  * Fixed compilation errors in src/interface/main.c when the compiler did
  not support SSE or 3DNow! inline assembly.

BUT this is UGLY ! We HAVE to find a way to make multicast function
under Win32.

BTW, it does now definitely read TS Streams, locally or from the network.

was crashing the windows port in TS_input.

The audio still segfaults.

* Fixed CSS ioctl for Darwin/OSX - still might need tweaking of 'dataLength' fields
* Added ReadTitleKey ioctl for Darwin/OSX
* Minor cleanup

-libdvdcss error messages are reported in input_dvd.c.
-vlc verbosity also enables more error messages from libdvdcss.

It should fix libdvdcss compilation issue on non-linux platforms.

  * Fixed an error in the ts://server[:port][/broadcast] input parsing.

method in libdvdcss, the key is only checked at title start, not before each
chapter.

The behaviour when we have decrypted the disc key is to check the key before
each seek (although I'm still fighting against a bug).

1) Crack the title key (the method that was here before). The only change
here is that we search the key for the exact chapter we are seeking with
DVDSetArea (in case the key has changed within a title). It is maybe not a
good idea.

2) Crack the disc key, which allows us to decode instantly all title keys.
I've used an algorithm from Frank Stevenson ; it eats much memory (64MB),
and takes about 15 s at launch time.

3) Decode the disc key with player keys (libcss method). However, you need
licensed player keys at build time for that to work.

To choose between libdvdcss methods, a command line options is supplied:

        vlc --dvdcss <method> where method is one of title, disc, key.

Note that all these changes only work with linux now, since we have to add a
specific ioctl to read title key. I hope that I haven't broken too many things.