Re: [AD] X11 unresponsiveness

[ Thread Index | Date Index | More lists.liballeg.org/allegro-developers Archives ]


On Tue, 2004-07-13 at 18:29 +0200, Evert Glebbeek wrote:
> On Tuesday 13 July 2004 15:25, Elias Pschernig wrote:
> > Well, the patch works perfectly here so far. I'm only concerned about
> > possible race conditions. A signal can interrupt at any place, even
> > inside a C instructions. And with that in mind, using a simple lock
> > variable like we do with the signals driver isn't 100% safe. (We would
> > need to disable signals, use some special software-mutex-construct, or
> > rely on asm support, to be on the safe side.)
> 
> Indeed it isn't. The demo program crashes for me with Xlib complaining 
> about an asynchronous call or somesuch...
> 

Oh, in case I didn't mention, there's still the problem with the color
conversion patch. You need to revert it before applying this patch. And
it's an updated version of Chris' patch (small change to not recursively
call _xwin_handle_input, and leaving out the vsync change for now) - the
idea is to call handle_input in XUNLOCK in case it was blocked. So we
never lose any signals at all.

We should probably first investigate the color conversion crash (signals
version with it applied always crashes with that X11 async call error).
This is completely unrelated to the responsiveness changes.

-- 
Elias Pschernig





Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/