FC3 and ruby woes

[James v. Z. <ja...@dv...>]

Just not stable. (for me)

Applying the same process that I've applied to rh9, fc1 and fc2, I have
not yet succeeded in producing a stable fc3 replica of an fc2 ruby
configuration.

I have applied the ruby-2.6.9-20041021.diff to the latest fc3 kernel
tree. It applied and compiled cleanly.

I have patched Xorg-6.8.1 most recent for fc3 with the mandrake
isolate_device patch.

I have compiled the kernel and X both with cpu-specific (athlon)
optomizations, something I have done with every other ruby config that
is working here.

I use nVidia cards exclusively, no TNT2 in service, all MX440, MX4000,
and FX5200. For this machine 1x FX5200, 2x MX440.

Symptoms are as follows, under 6111 or 6629 nV drivers, the system will
boot but experience a kernel oops/panic not long after X has started. It
is not consistent. I might crash while starting an Xinimera session with
rhgb graphical boot, or it might complete boot, give me three login
prompts, and crash soon after. I can login on any console if it doesn't
crash first.

If I don't start X, everything is stable. If I boot runlevel3 and start
X as Xinimera session, seems stable. I haven't done a long test of this,
but it doesn't lockup within seconds as it sometimes does with rhgb and
almost always does at runlevel 5 as gdm starts the consoles.

I also applied the following correction hunk to ruby, just in case,
although discussion suggested x86-64 problem that doesn't impact 32-bit
units. Doesn't solve any problems for me.

------------------------------------------------------------------
diff -Nurp --exclude-from=exclude linux-2.6.9/fs/compat_ioctl.c 
r269/fs/compat_ioctl.c
--- linux-2.6.9/fs/compat_ioctl.c       2004-10-19 00:55:27.000000000
+0300
+++ r269/fs/compat_ioctl.c      2004-10-20 19:31:58.000000000 +0300
@@ -1609,8 +1611,10 @@ static int do_fontx_ioctl(unsigned int f
                     get_user(data, &user_cfd->chardata))
                         return -EFAULT;
                 op.data = compat_ptr(data);
-               return con_font_op(fg_console, &op);
+               return con_font_op(vc, &op);
         case GIO_FONTX:
+               if (!user_cfd->chardata)
+                       return 0;
                 op.op = KD_FONT_OP_GET;
                 op.flags = 0;
                 op.width = 8;
------------------------------------------------------------------

I am currently suspecting Xorg to be unstable in this configuration, I
am wondering what success others have had with similar 32 bit FC3
configurations.

This same configuration was 100% stable under FC2 on the same machine.

Should I double check my kernel config ?-- it has changed from fc2 to
fc3, now having less built in and more modular and based on the fc3
kernel distribution. Should I avoid athlon optimizations?

Are there any known issues that I might not be aware of that would
produce such behaviour?

J

http://members.westnet.com.au/vanzeeland