Gcc 3.2.2 vs gcc 3.2.3

[Holger Kipp]

Justin Smith (jsmith at drexel.edu) wrote: 
>I'm running a P4 2.4 Ghz processor with 512meg of ram.
>
>I've had the problem of getting signal 11 and signal 4 when I make
>buildworld and buildkernel.
>
>I tested my system with memtest86 (for several hours) and it says
>everything is OK.

We had similar problems once. It turned out to be a 'lucky' combination of good
mainboard and good memory that were slightly incompatible.

memtest86 might not detect this, but make buildworld will. If you don't get the
errors at the same location every time, it might be the same issue. We found
out after contacting both the mainboard and the memory chip vendors.

imho memtest seems to be suitable for this-is-real-broken-memory(tm) only.

It might be that due to differnent optimisations (gcc 3.2.2 vs gcc 3.2.3) 
you might escape the bitpatterns that would otherwise trigger the sig 11.

This is only one possibility, but might be worth looking at.

Regards,
Holger Kipp
(Sysadmin alogis AG, Berlin)