> I have gotten a few reports of "strace" completely freezing a few
> machines. So far these reports have had two things in common: a Pentium
> processor and gcc-2.7.0. It may be that the Pentium 4MB page table code
> does strange things on some machines, or that gcc-2.7.0 compiles the
> code to something I never meant it to be.. I'd like to hear if the
> people that can make this happen could try an older compiler to check.
I believe that on the net someone specified (was it Thomas Koenig?) they
had used 2.6.3 to do it and it was still there. Also, unmounting /proc
(with the resulting failed open on /proc/[pid]/mem) was armor against the
bug.
Also note that, with me and as I remember with everyone else, the bug is
not a freeze bug, it is a reboot bug. Sudden, instant reboot, to the
point that
strace touch nuke.my.machine > strace.log
on a synchronous-mounted ext2fs fs doesn't leave any files. This
resembles very strongly the reboot bug I run into while compress-tarring
large directories.
> Anything else that has problems with this kernel?
I'm getting cross-VT corruption, where the text on one VT appears on
another (which is never active, and it doesn't happen to that VT after I
switch back to it, for a while). This corruption can be from a couple
lines of obvious garbage (random characters of random colors) to text from
other VTs (even VTs that aren't currently scrolling, but just sitting
there with for example a login prompt) to entire screenfuls of output
from another VT with a kernel compile going on in it.
This isn't a common occurrence, but it's certainly not rare, either. And
apparently totally random.
-- this mind intentionally left blank | Linux - anything else Ryan Smith-Roberts -- labrat@u.washington.edu | would be uncivilized!"In /this/ house, we obey the laws of THERMODYNAMICS!" -- Homer Simpson