(previous commit foired)

 . this is a coding style patch which removes all "foo(bar){" constructions
   and most of the tabulations.
 . also, fixed a bug in the default subtitle track.
 . and made a few error messages more explicit, ie. changed "error: %s" to
   "foo error: couldn't initialize bar (%s)"
 . and removed a warning in intf_ctrl.c which isn't used anyway.

 . this is a coding style patch which removes all "foo(bar){" constructions
   and most of the tabulations.
 . also, fixed a bug in the default subtitle track.
 . and made a few error messages more explicit, ie. changed "error: %s" to
   "foo error: couldn't initialize bar (%s)"

 . splitted modules.h into modules.h and modules_core.h to enable
    clean compilation of the Null Module.

 . added the Null Module, the module that does nothing but that is going
    to be of great help for debugging my module code.

 . Added files needed for the forthcoming module management.

 Notes:

   Plugins are now called modules, because we will be able to compile
   them either as a dynamic plugin, or within the program. And, more
   important, I can commit this without having to break the vlc :)

   I tried to be as clear as possible in my comments, please tell me
   if a few prototypes still have an unclear beahaviour.

   Current features:
    . none, files haven't even been added to the Makefile yet.

   Future features:
    . dynamic loading (done)
    . built-in modules (still to do)
    . automatic unloading of plugins (done)
    . dynamic configuration (still to do)
    . automatic choosing of the most appropriate plugin for a given
       task (still to do but well prepared)

  - fixed segfault, it was probably a memory corruption but I'm not sure :) ;
  - video_output.c : compute picture size even if b_need_render == 0;

grrrr, I'll kill someone) ;
* Put config.h in front of others in audio_decoder.c.

 . added a few missing break;s
 (oui je sais je ne fais que des patches de deux lignes � trois balles,
  mais par telnet c'est pas facile)

- Removed i_iovec_start and i_iovec_end. Used i_data instead
- Meuuh : some questions for you in the commentaries

Happy new year !

* Fixed a bug in vpar_headers.c synchro handling introduced before ;
* Enhanced synchro's handling of dates and frame structure changes.

NB : b_has_pts has disappeared for brevity reasons, use i_pts != 0 instead.

 . avoid calling SafeRead() with a null packet size.

* Changed behaviour of preparsing code, faster startup ;
* Fixed DecodePSM(), cannot test ;
* Cleaned up ParsePES(), fixed a bug which could reject very small but
  valid PES packets ;
* Fixed cosmetic bugs in vpar_synchro frame_rate display.

 . SPU decoder now uses Meuuh's GetChunk() code.
 . lots of cleaning in the SPU decoder.
 . message queue is flushed before we spawn input.
 . fixed so that it compiles with VDEC_SMP (no good sync yet though).

* Fixed a bug in -c and -s options.

 . fixed AC3 and SPU PES reading (ph34r da one-byte patch)
 . some 79 char wrap.

 . no need to add "\n" at the end of intf_*Msg() messages anymore.

* Fixed an segfault in input_programs.c ;
* Added -fomit-frame-pointer optimization.

  be the last one.

* Moved p_input->pp_es -> p_input->stream.pp_es and clean up of
  input_programs.c & co ;
* Fixed memory leaks.

* Temporarily removed b&w kludge.

* Cleaned up input_ts.c, added it into the Makefile.

All decoders should now exit cleanly (if it's not the case, it should be
referenced as a bug).

ES in the pp_selected_es array).

Added netlist support.

Warning : it has not been tested. It just make no error at build time.