* src/misc/mtime.c: implemented the heuristic described in http://www.cs.man.ac.uk/fellowsd-bin/TIP/7.html to test whether the win32 performance counter is reliable. That should fix the clock problems that a few users reported on win32.
do the actual execution of this command. this is a (very ugly) workaround
to the problem we were having with the deinterlace menu on osx.
It works now, but now i'm running into the black screen and continous
"late picture skipped (-218000)" again which is also plagueing us when you
enter fullscreen with a filter enabled. This ought to be fixed before 0.6.0
as well.
* modules/gui/wxwindows/*: fixed bug in sout dialog box.
* modules/gui/skins/*: factorised the wxwindows dialogs code in dialogs.cpp. wxWindows is also initialised earlier on to allow opening a skins file with a wxWin dialog box. (Note: I badly broke the GTK2 skins).
- Changed the autogenerated menu's in a system working similar to wxWindows.
- The deinterlace menu is autogenerated now. However it seems the problem
with on the fly filter changes still exists.
* modules/gui/skins/*:
- fixed the wxwindows bug, due to a huge bug with broadcast events
- beginning of fonts in X11 skins
so, you can test the X11 skins module: it should work now !
* skins/controls/generic.h: Some public fields are now protected
* skins/src/window.cpp:
- Added a few debug messages
- Invisible controls don't receive events anymore
Note: There are still a few bugs in the code handling the controls and
their different states, and I think they will be difficult to fix with
the current code structure. A rewrite of this part of code could be a
good idea (perhaps using states and transitions between states, like
Zinf does?).
quite more convenient than, e.g, calling input_SetStatus( STATUS_FASTER )
two times if you want to switch from 1:1 to 4:1 from the interface.
Old functions still work.
* modules/gui/beos/VlcWrapper.cpp : use it
that had done all the work.
Paul Forgey's notes:
--------------------
Unfortunatley, the v4l interface shows how much it sucks here. I can't
wait for v4l2 to become more common. Anyway, the mjpeg mechanism
captures completely differently than the uncompressed frame capture
mechanism. This means the code has to split off a bunch to do similar
things depending on mjpeg being set or not. For some reason, I can't
use non v4l2 calls to capture uncompressed frames from the lml33,
however I do have another bttv card on my system to test that I didn't
break the non-mjpeg stuff.
The Zoran encoders don't use square pixels. So the 4:3 (or 16:9)
picture is going to show up as 720x480. I don't know my way around the
code well enough to know how to specify a non-square pixel aspect
ratio. If there isn't a way to do it, it would be nice to be able to
specify as an option the aspect ratio of the sourc. Either 4:3 or 16:9
in the case of anamorphic S-Video, which does occur.
I stole the videodev_mjpeg.h header from the lavrec project. Despite
the linux kernels having support for the Zoran cards, there doesn't seem
to be a reliable way to pick up this header, which is probably why
lavrec did it this way. Since it only defines ioctl definitions, I
don't think using it is a legal problem.
I also modified the audio handling a bit. If the card supports audio,
great, we'll set the card's parameters as we did before. But if adev=
is specified, use it regardless. The reason to do this is to capture
audio from the sound card with video coming from, say, a web cam or in
my case, an encoder card which doesn't have any audio capabilities at
all.
- added "en" to the list of languages in the preferences.
- added DIV1, DIV2, i263, WMV1 to video encoding options in prefs.
FENRIR: is this correct this way?
mem leaks will kernel panic there machine or something ;)A
This will generate less bugreports, yet keep them visible to those who
should and will care.