Keith Packard wrote (in a message from Wednesday 15)
> Around 23 o'clock on Jan 15, Matthieu Herrb wrote:
>
> > The patch below fixes that, by ignoring the BadValue error, but it
> > still has a problem, since it can kill a random client started after
> > the 1st instance of beforelight was killed. Any ideas on how to avoid
> > that ?
>
> Looks like the screen saver convention isn't very safe. If we assume that
> the beforelight convention of using a pixmap is consistent across screen
> savers, then we could validate the kill_id with an XGetGeometry and
> make sure the kill_id referred to a 1x1 depth 1 pixmap. That would at
> least avoid killing most other applications by accident.
>
> We should probably respecify how XScreenSaverRegister/
> XScreenSaverGetRegistered work to avoid this problem.
>
It also looks like the current implementation is not 64 bit clean. It
aborts with a BadDrawable error in ScreenSaverQueryInfo on
OpenBSD/alpha and it apparently causes the server to crash on
OpenBSD/sparc64 (Big endian 64 bits).
Matthieu
_______________________________________________
Devel mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/devel
