GitLab

* fixed some bugs in program management : selected ESs must be attached
to the selected program.

  * skip "//" if we find "://" in a resource we have to play : this allows us to use the syntax udpstream://host.

  * ./plugins/chroma/i420_rgb8.c: plain C 8 bpp transformation.
  * ./plugins/chroma/i420_rgb16.c: plain C 24/32 bpp transformations.
  * ./plugins/ggi/ggi.c: support for 8 bpp displays.
  * ./plugins/sdl/vout_sdl.c: support for 8 bpp displays.
  * ./plugins/x11/xcommon.c: support for 8 bpp displays.
  * ./src/video_output/video_output.c: fixed a segfault.

  * ./Makefile: fixed OS X vlc.app compilation dependencies.
  * ./configure.in: fixed the endianness check under Solaris.

  * ./src/input/input.c: strings are now initialized to "" instead of NULL.

  * ./plugins/spudec/spu_decoder.c: we now properly support the subtitle
    palette information we get from the DVD. Subtitles rox !
  * ./src/video_output/vout_subpictures.c: removed all DVD-specific routines
    and put them into spu_decoder.c.

*p_es->p_demux_data is available in p_config->p_demux_data so that the input
can pass information directly to the decoder.

*First application: aspect ratio for DVD is read from the ifo (should fix
the buf with aspect ratio in anamorphic DVDs).

*The DVD plugin also copy the spu yuv palette in p_demux_data so that the
spudec can display subtitle with the right color (sam !).

*Removed duplicate code and useless lock in gtk.

*New configuration option to choose audio output format at runtime.
It allows to select spdif pass-through decoder with --ac3_adec and to
remove the ugly b_ac3 in main.h

As a consequence --spdif doesn't work anymore: you should use --ac3_adec pass

*Fixed a typo in main.c

*Return the option name in error message for unknown option intead of the
argument.

* New GTK configuration interface. This interface is generated dynamically
  at run-time from the configuration data contained in each plugin.
* config_PutPszVariable and config_GetPszVariable are now thread safe.
* Few modifications/optimizations to the config module.
* Got rid of all the _VAR macros in config.h

This is the first version of the GTK configuration interface. There are still
a few raw corners and a fews things left to do, but it basically does what
I wanted.
Constructive comments like it sucks or it smells rotten fish are of course
very welcome. And if there are GTK gurus out there, please don't hesitate to
add your touch.

I also started to add usefull comments to the configuration options but I'm not
really good at this kind of thing so if people want to help, they can finish
the job. For this look at all the *_LONGTEXT macros at the top of main.c

(PS: Translations will also be appreciated but not right now as the interface
will likely change again in the few coming days)

*Fixed EOF segfault at end of DVD

*Buffer manager is renintialised at area change.

*Removed an occurance of former angle item in gtk.

*Fixed a bug in ES management introduced lately in DVD/DvdRead.

*Beginning of reorganisation in DVD plugin source files, in order to try to
make it easier to understand. There is some work left though :p.

*Decreased the score for VCD plugin, as the VCD demuxer was used even
for DVD :p. It should make autodetection work.
Some comment:
-I think that VCDInit should make a test on the access plugin and be launched
only if the access plugin is VCD,
-VCDOpen shouldn't set p_input->b_error: if it fails we just try another
module,
-PSRead doesn't need to be duplicated anymore: input_ReadPS is now available
for plugins (cf Christophe's commit).

Please comment around that: I'd like to know if the behaviour I've described
is the one everyone expects.

Enjoy my new bugs :p

* Dvd angle is now handled through a program structure, so that the interface
needs nothing specific to change angle.

* Fixed a bug in clock management where ES without a program couldn't get
timestamped.

*Fixed program management in gnome/gtk. Program now also appears in popup.

*In interface, only display ES associated with the selected program, or with
no program at all.

* fixed typo in the win32 change to the input parser.

for BT ;
* --6 and --4 options to force IPv6 or IPv4 (FIXME: it should be -6 and
-4 but I don't know how to do it with the new configuration module) ;
* Channel manager now uses a socket from the network plug-in and is more
portable (and probably broken, too :) ;
* input_ReadPS and input_ReadTS are now in mpeg_system.c to be more
easily reused by plug-ins other than mpeg_ps and mpeg_ts.

That's all for tonight.

* this time I actually commited the win32 modif to the input parser ;-)

* added network to the list of the plugins directories in Makefile
* small bugfixes

 * Fixed DVD input under Win32.
 * Fixed input parser under win32, chich couldn't understand that
c:\windows\bureau is a valid file and doesn't need to be accessed with
the plugin 'C' :-)
 * BTW, I added myself in the AUTHORS file :)

 * Various Win32 compilation fixes.
 * DVD compiles, but doesn't work yet.
 * Fixed some typos in my previous Win32 multicast commit

   * ./plugins/dummy/input_dummy.c: fixed `vlc vlc:quit'.
   * ./src/input/input.c: initialized uninitialized variables.

*Initiliaze p_access_module & p_demux_modules to NULL when input thread
in created.

*dvd and dvdread support for input III.

Dvdread is known to segfault at title change. I'm working on this.

- Major rewrite of the buffer core functions
- Modularity of access plugins (file, udp, http)
- Modularity of network stack (ipv4, soon ipv6)
- Autodetection of the type of stream

And a new killing feature : HTTP streams are now seekable.

Please note that dvd, dvdread and vcd plug-ins are broken and thus disabled.

  * ./src/input/input.c: BeOS network support (no idea if it works when you
    don't have BONE, though).

Multicast now works under Win32, but this is a _bad_ kludge.

It still segfaults 4 secondes after the first frame, but, at least, I
can see two or three frames from MI2 :-)

  * ./include/threads.h: support for the State Threads Library, a userland
    threads library (http://state-threads.sourceforge.net/). Useless to most
    people, but can be handy to debug stuff. Activate with `--enable-st'.

in multi-program streams (such as ones provided by a DVB-S card).

This is the first part of the new configuration architecture for vlc.
Two other parts will follow soon:
- config file implementation
- gtk configuration interface.

This part won't change much for end-users as these changes are mostly
internal, you will however notice that the command line interface has
changed. It will maybe be less user-friendly as I almost removed the
short options but it shouldn't a big problem when the config file is
implemented.

What as changed is that it is really easy now to add configuration
options to vlc. Plugins can also implement configuration options. All
you have to do is to declare a MODULE_CONFIG section in your plugin.
If you want to know more about this look at configuration.h and for some
example: main.c, dsp.c or xvideo.c

It would be nice if every developper could now start using configuration
options in there plugins and also enhance old plugins.

That's it, now let's hope I didn't break too many things ;-)

   * Got rid of TRACE and intf_DbgMsg which were seldom used anyway.
   * All intf_*Msg functions now write to stderr.
   * Implemented a message queue interface plugins can subscribe to.
   * Wrote the logger interface plugin. Usage:
      vlc -I logger:filename.log
   * Lots of fixes in the SPU renderer.

  * ALL: got rid of *_Probe functions because most of them were duplicates
    of the real _Init or _Create functions. That's 1000 lines of code gained.
  * ./src/misc/modules.c: we try not to keep the global lock in module_Need,
    which should speed up output spawning.
  * ./plugins/dummy/input_dummy.c: fixed vlc:quit command.
  * ./plugins/gtk/gtk_playlist.c: compilation fix.
  * ./plugins/macosx/*: preliminary fixes for the vout4 port.

   Notes on the *_Probe eradication:
  o I probably broke the DirectX video output, because of the "must init and
    run in the same thread" issue. I'll fix this ASAP.
  o Subtitles are broken (or at least, pretty ugly) in RGB mode.
  o The input plugins still have *_Probe functions. This will disappear when
    meuuh is finished with the access plugins.
  o The decoder plugins still have *_Probe functions. This is probably not
    going to change, it looks OK to me.

* New AC3 decoder plugin using liba52 from Aaron Holtzman and
    Michel Lespinasse.
    This plugin isn't used by default, you have to specify:
    --ac3_adec liba52

* mad PTS should finally be accurate
* Some configure changes.

The gnome/gtk interface should no longer user p_input->stream before it
has been initialized.

to have it initialized when we call it in the interface.

* Fixed issues in the PTS handling of the mad plug-in (incomplete).

  * ./src/input/input.c: fixed a few segfaults caused by the dummy plugin.
  * ./plugins/mpeg_adec/mpeg_adec.c: fixed a segfault due to misuse of
    DecoderError.
  * ./plugins/x11/xcommon.c: replaced XFlush with XSync. This must be the
    fourth time someone has to commit this, so I added a reminder :-)

  * ./include/modules_inner.h: replaced _X with __VLC_SYMBOL because _X was
    already a system macro under MacOS X.
  * ./plugins/dummy/dummy.c: fixed vlc:loop, vlc:quit, etc. entries.
  * ./plugins/glide/glide.c: activated double buffering.
  * ./plugins/mga/xmga.c: started writing an xmga plugin; doesn't work yet.
  * ./src/input/input.c: fixed the input memory leak, and the insane thread
    spawning we got with vlc:loop.
  * ./src/misc/intf_eject.c: disc ejection routine courtesy of Julien Blache,
    currently Linux-only.

  * ./plugins/mga/mga.c: removed the mgammx module and put the code in
    the chroma_i420_ymga module instead. This introduces a slight
    performance issue as we now need to memcpy() the Y plane, but I have
    some vout4 enhancements on their way which make this point moot.
  * ./debian/changelog: imported official Debian changelog.
  * ./debian/control: activated the vlc-mad and vlc-arts packages; package
    builds should now work again.
  * ./plugins/dvd/dvd.c: solved a symbol resolution issue when vlc was
    compiled without libdvdcss which overflowed the return stack.
  * ./plugins/ggi/ggi.c: ported the GGI plugin to vout4.
  * ./plugins/glide/glide.c: ported the Glide plugin to vout4 (untested).
  * ./src/input/input.c: prepared everything for multi-input mode; since
    no interface is yet prepared to handle this, I changed them so that
    they just use the first available input for now. This may break other
    things, but I promise I'll eventually repair everything :-)

  * ./BUGS: added a list of known bugs. Please add your findings!
  * ./MODULES: added a short description of each module. Maybe we should
    add more details to make this file really useful.
  * ./doc/bugreport-howto.txt: wishful thinking...
  * ./Makefile: added a workaround for the fact that .c files don't depend
    on <videolan/vlc.h>.
  * ./plugins/memcpy/memcpy.c: all memcpy modules are now built from the same
    single file.
  * ./src/interface/main.c: added the --nommx, --no3dnow, --nommxext, --nosse
    and --noaltivec options to vlc.
  * ./configure.in: removed the --disable-mmx option, now useless because
    of the vlc --no* options.
  * ./src/interface/main.c: fixed a crash caused by the memcpy module
    being released too early.
  * ./include/video.h: more vout4 changes, including margin support, more
    FOURCC formats recognized, transparent support of identical or almost
    identical FOURCC formats.
  * ./include/video_output.h: added a vout_Render function to video output
    modules which is to be executed before vout_Display.
  * ./include/common.h: this header is back.
  * ./plugins/chroma/i420_rgb.c: planar YUV 4:2:0 to packed RGB conversion
    functions, mostly taken from vout3's yuv functions, but heavily modified
    to use FAST_MEMCPY and to handle image dimensions which are not multiples
    of 16. Not all functions have been written yet, though.
  * ./plugins/chroma/i420_yuy2.c: planar YUV 4:2:0 to packed YUV conversion
    functions, various targets such as 4:2:2 YUYV, YVYU, UYVY, or 2:1:1 YUYV,
    with accelerated MMX versions.
  * ./plugins/chroma/i422_yuy2.c: same as i420_yuy2.c but takes planar YUV
    4:2:2 pictures in input. With accelerated MMX versions as well. This plugin
    makes it possible to display 4:2:2 files on most video cards without the
    quality loss of 4:2:0 downsampling.
  * ./plugins/filter/*: thanks to the vout_Render function, all filters now
    output their data perfectly in sync.
  * ./plugins/dvdread/Makefile: fix to prevent unnecessary target rebuilds.
  * ./plugins/sdl/vout_sdl.c: support for YV12, UYVY, YUY2 and YVYU overlays.
  * ./plugins/x11/xcommon.c: support for I420, YV12, YUY2 and Y211 overlays.

* Temporary kludge for vlc_mpeg_adec=mad ./vlc ...

Some heavy changes today:
  * Removed duplicate function checks from configure.in.
  * Added extra magic to Makefile.modules so that the module Makefiles
    are now ridiculously simple. And I mean *simple*. Check it! This will
    make a possible switch to full autoconf/automake a lot easier.
  * Added the vlc version name to the plugin symbols, to be sure we only load
    plugins with the same version number. A nasty consequence is that you
    need to rebuild your tree after midnight if you are using a CVS tree :-)
  * Got rid of modules_export.h by #defining exported functions in the same
    header as their prototype.
  * Added modules_inner.h and other commonly used .h files to common.h so
    there are less and less files to include, and renamed common.h to
    <videolan/vlc.h>.
  * First modifications to the module handling system towards my ultimate
    goal to get rid of the *_Probe functions. Got rid of TestMethod and
    TestCPU, as well as src/misc/tests.c.
  * Wrote the chroma plugin handling functions. No YUV functions have been
    ported yet because it'ls a lot of work, but the core system works, I
    tried it with a naive yv12->rgb16 plugin (which will disappear when the
    real functions are ready).
  * Made a lot of functions in dvd_summary.c one-liners to avoid wasting
    too many output lines.
  * Fixed a segfault in input_dvd.c:DVDInit.
  * Added a fixfiles.sh script in plugins/gtk to be run after Glade has
    generated its C files.
  * Did some work on the KDE interface to make it suck a bit less. It still
    segfaults, but at least it runs and it looks less ugly.
  * RGB SDL rendering works again, though in 16bpp only.
  * Made plugins/vcd/linux_cdrom_tools.c independent of any vlc structure
    so that it'll be easily put in a library. Maybe libdvdcss?
  * Fixed VCD date display.
  * Merged vout_xvideo.c, vout_x11.c and vout_common.c into xcommon.c.
  * Wrote non-Shm XVideo output.
  * Made X11 output work again. Still pretty unstable, only works for 16bpp.
  * Additional french translation in po/fr.po. Any taker for the rest?
  * Fixed a segfault in video_output.c when the allocated pictures were
    not direct buffers.
  * If $DISPLAY isn't set, don't try to run the Gtk+ interface.
  * Replaced 48x48 .xpm images with 32x32 ones to conform to Debian policy
    (Closes Debian bug #126939).
  * Removed the automatic ./configure launch when running `make all' for
    the first time.

Stuff currently more broken than it ought to be:
  * The wall filter. Being fixed.
  * x11 and sdl plugins for depth != 16bpp.
  * Software YUV.
  * gvlc, gnome-vlc, kvlc shortcuts. Use --intf instead for the moment.

  * Forgot a file in my last commit.

  * Put ErrorThread() in input_ext-dec.h because it was the same function
    in all decoders. Called it DecoderError() instead.
  * All decoders enter DecoderError() in case of error, even if they
    haven't allocated all their data. This prevents the input from getting
    stuck if a decoder has failed its initialization.
  * Namespace fixes through the decoder files.