xine-lib/src/post/visualizations, branch 1.1.3

two fixes to viz glue code:

2006-12-02T22:35:18+00:00

- avoid overrunning the provided input audio buffer.
- generate a bad frame if time is due but we don't have enough data
for updating the viz plugin. this could happen in some rare situations
but the result was pretty catastrophic: xine froze with 100% cpu usage.

CVS patchset: 8400
CVS date: 2006/12/02 22:35:18

Implement visibility support, available on GCC 4.0 and later and on some 3.4 (through backports), to avoid exporting unneeded internal symbols, making plugins' loading faster and use of internal copies of libraries more solid. It should automatically fall back to the old way in GCCs that does not support -fvisibility=hidden, but has to be tested carefully. No issues were found in the months of testing in Gentoo, but this requires special attention anyway.

2006-07-10T22:08:12+00:00

CVS patchset: 8101
CVS date: 2006/07/10 22:08:12

Various static/const fixes from Gentoo.

2006-06-02T22:18:56+00:00

CVS patchset: 8009
CVS date: 2006/06/02 22:18:56

Make various arrays and structures (mainly xine_plugin_info) const.

2006-05-03T19:46:06+00:00

(Diego Pettenò)

CVS patchset: 7985
CVS date: 2006/05/03 19:46:06

Adapt the engine to the new list code.

2006-01-27T07:46:09+00:00

CVS patchset: 7848
CVS date: 2006/01/27 07:46:09

BUGFIX

2005-08-25T15:36:29+00:00

Windows ports fixes and improvements due to my current work on toxine:
- first experiments with external win32 pthreads,
  more portable code (pthread_t may be a struct)
- headers refactored
- moved dirent win32 replacement to lib/, hide it for frontends,
  used system version, if found, not used non-POSIX dirent->d_reclen
  (this item doesn't work in MinGW), fix memleak in dvb
- separated settings for postproc and avcodec when using external ffmpeg
- check for malloc.h in public xine.m4, used it conditionally in xine headers
- replaced random() by POSIX more common rand()
- prevent one segfault in directx vo plugin, if fails
- M$VC port update

CVS patchset: 7709
CVS date: 2005/08/25 15:36:29

fix segfault

2004-12-13T19:05:01+00:00

CVS patchset: 7246
CVS date: 2004/12/13 19:05:01

separate the two semantics of querying a port's status with a NULL stream;

2004-05-29T14:45:25+00:00

before, NULL could mean two things: no stream at all or a stream that does not
want to be addressed; now the latter is represented by the new XINE_ANON_STREAM
resulting changes:
* the status() functions now behave differently for NULL and XINE_ANON_STREAM
  (as the commentary always implied, but this was not the case, so post
  plugin rewiring went wrong, because it relies on the status() function)
* the NULL_STREAM defines in audio_out and video_out are obsolete
* update the function comments in the headers
* update the post plugin rewire functions to use the status() functions to
  check, if the old port was opened and handle the new one accordingly;
  this makes open_count obsolete
* change all post plugins accordingly (mostly using XINE_ANON_STREAM instead
  of NULL)
additional change:
* the status() function of audio port now returns the bits/rate/mode values
  of the input and not the output; this is more likely to be what a post plugin
  wants
* the reimplementation of status() in the upmix plugin is obsolete

CVS patchset: 6603
CVS date: 2004/05/29 14:45:25

track number of times the port has being opened.

2004-05-18T03:16:12+00:00

rewiring a closed port won't cause the new destination to be opened.

CVS patchset: 6564
CVS date: 2004/05/18 03:16:12

swap usage of "up" and "down" to the more common meaning:

2004-04-17T19:54:31+00:00

 "down" is the usual datapath from decoder to output, "up" is the opposite way
-> this means swapping two function names
-> this means increasing the post plugin iface version
external plugin developers: just swap usage of
_x_post_frame_copy_up() and _x_post_frame_copy_down() functions

CVS patchset: 6412
CVS date: 2004/04/17 19:54:31