* modules/access/dshow/dshow.cpp: added an "adev" and "vdev" option for the MRL.
   If not specified, then the default devices are used.
   (eg. vlc dshow://:vdev="Labtec Webcam":adev="foo")

* modules/access/dshow/*: audio is now supported as well.

        preliminary support for SSA.

 tested with vlc/mplayer.

 for mmsh protocol.

http output as mmsh server (see next commits).

* modules/codec/faad/decoder.c: for some reason my libfaad the frame samplerate to 0. I guess I should update it but anyway a sanity check avoids a segfault in this case.

* NEWS: keeping track of the new features.

  see bug 1136

* configure.ac, modules/access/dshow/: brand new DirectShow input plugin.
   Much work still needs to be done, like audio support, adding plenty of configuration options, etc... But the video part is already working quite well here.

  - reordered several parts
  - made it as close to the gentoo ebuild as possible.
  - fixed several changes/additions since 0.6.1

    . cleanups, use of top_builddir where possible, just to feel safer
    . fixed VLC.app inclusion of AUTHORS, COPYING etc.
    . propagated previous changes to package-osx
    . checked both build methods

special goom tree source as I was unable to use standard goom library.
(I will provide it once mmx/ppc included)
 * modules/visualization/Modules.am: added goom
 * include/vlc_block.h src/misc/block.c: introduce a new data block
 api (not yet tested, ported from my local new input work).

  that don't have a . char in their name. Also fixed a problem where the
  code was trying to open a empty string subtitle after autodetect couldn't
  find a suitable filename.

  make it easier for users to find their charset now.
* modules/demux/util/sub.?: try to find filename.subext based on filename.avi
  Very basic now, but might be become more advanced in the feature. (that's
  why it's so much code )

                 $(top_builddir) instead of $(srcdir)

(for the user filter part).
 For now, when a user filter can't be inserted it is just ignored. (before
the whole audio chain failed).

  fix broken tree

  in use by the OS. It's pretty nifty. if available it uses nl_langinfo. Then
  it falls back to trying setlocale (exept on OSX, where this funtion is useless)
  and then falls back to LC_ALL, LC_TYPE and LANG environment variables.
  If only the LANG variable is available (like on osx) it tries a countrycode
  to charset mapping (making an educated guess ).
  On windows it retrieves the charset with GetACP().
  It also has an aliases system to transform incompatible charset description
  strings to libiconv compatible charset descriptions. If you want you can even
  retrieve the charset on OS/2.
* modules/codec/subsdec/subsdec.c: By default we try to open a subtitle file in
  the systems charset (or language mapped to charset). It's no more than a guess,
  and if you open Latin1 subs on your Japanese system you will still need to
  specify the encoding by hand, but it's better then nothing.
* src/playlist/playlist.c: spelling error in a comment.