Hi
I have made some tests with the Redhat 7.1 gcc-2.96-81.
If I remember correctly Steve mailed that he has no problemes with the gcc
and I always had problems. So I think about the difference about Steves and
my compilations. One thing is, that it could be for a different processor
type. I test it with a interesting result:
All test was using a CVS kernel from 2001-05-24. This source is stable with
kgcc. I have a k6-2 500 System. I normaly compilied the kernel for k6 of
course.
The first test with gcc-2.96-81 was using "pentium classic" as processor
family in make xconfig (-march=i586).
This kernel made no problems. I have used it for several days and medium
stressed it.
The second test was using k6 as processor family (-march=k6).
The 3rd reboot (after a unclean shutdown) crashed while processing the boot
scripts. It trapped in kdb (sorry no serial console this time).
I rebooted the system and it hangs in the boot scripts (no kdb). One error
message told me that /var/run/utmp is a directory.
I booted the XFS install CD and run xfs_check and xfs_repair. The usual
thing: pid files from /var/run were in /lost+found. If you want I can mail
you the output from xfs_check and xfs_repair.
I thing the code generation for k6 processors are buggy in gcc-2.96-81.
ps: Good luck for the future Russell and many thanks.
utz
|