On Sun, Mar 24, 2002 at 02:57:02PM +0100, Denis Oliver Kropp wrote:
> > I just checked out XDirectFB and DirectFB from CVS this morning, rebuilt
> > both, and fired up XDirectFB. I'm getting what looks like corrupted
> > video memory about 90% of the time. The other 10% of the time, I get an
> > X display, but the keyboard and mouse input soon become unresponsive.
> 
> I just committed a fix to the keyboard driver that makes XDirectFB work
> again from linux console (I know it should have been an issue in vt.c,
> this will be cleaned up soon).
>
> I don't know exactly why this happened with XDirectFB only.
> I could not really debug it, because it works when started in strace ;-P
> But it didn't in gdb, so it's not a ptrace issue, I think.
> 

Aha. That sounds like a fix, considering I never tried it outside of the
linux console.

/me does a cvs update, subscribes to directfb-cvs/directfb-users, and
starts compiling. :)


> > The machine is alive at the kernel level (I can ssh in, etc), but all of
> > the virtual consoles, keyboard, and mouse are locked up, even when
> > XDirectFB terminates on SIGTERM.
> 
> Did you try Ctrl-Alt-BackSpace?
> 
> Anyway, it should deinitialize correctly on a signal.

Yep. Ctrl-Alt-Backspace managed to deinitialize it correctly about 1/2 of
the time. The other 1/2 of the time (usually the times when there was nothing
but a black screen on initialization) it didn't seem to.


> > All of the DirectFB demos work great on this (uniprocessor) box, so I'm
> > pretty confident it's not a configuration issue. I've tried kernels
> > 2.4.17 - 2.4.19-ac3, both with and without rml's preempt patches. Also tried
> > memory tests and the X mga server; both of these work great also.
> > 
> > Unless I'm really doing something boneheaded, it looks to me like I'm
> > managing to trigger a race during the initialization of the framebuffer.
> > I'm at a loss as to what to try next, so I thought I'd see if anyone has
> > any debugging hints (or can reproduce this behavior).
> 
> I could reproduce it. I always started it from xterm before and it worked.

/That/ I didn't try. :) When I rebuild from CVS, I'll give it a try from
within X first. All of my previous attempts were from a matroxfb
console with the same resolution/depth as the first entry in fb.modes.


> > I'm more than willing to freeze up this machine a few more times, send
> > configuration information, strace output, etc.
> 
> You didn't try strace, did you? You would have been surprised like me ;)
> 
> Please try again and don't worry about the error messages,
> these will disappear after I understood all this VT/TTY voodoo ;)

I may have tried strace once or twice, redirecting to a file, but I
can't remember for sure.

Thanks for all the help. I'll let you know how it goes today.


Regards,
- Dave
  [EMAIL PROTECTED]

Attachment: msg00556/pgp00000.pgp
Description: PGP signature

Reply via email to