I have tried my "make -j" stress test on several kernels, and the
results are as follows:
1.3.58 is OK, compiles until it runs out of swap space,
1.3.59 gives lots of sig11's while there is still plenty of free swap.
Things like setting lower CPU speed, slowest possible RAM timings,
disabling caches, even trying this on another known good old machine
(over 2 years old 486sx33, 4MB RAM, cheap slow ISA IDE interface)
doesn't change anything. It seems more likely to happen on machines
with less RAM (more swapping) - on this second slow machine I start
getting sig11's much faster than on the faster one with 8MB RAM.
Also, I think have seen someone else reporting similar problem on
this list - it's not just me.
So, it looks like a legitimate kernel bug, and almost certainly not
like a hardware problem. There are some changes in swapping code
between 1.3.58 and .59 - maybe this is the culprit?
The latest kernel at the time I tried it was 1.3.83 and it still
had the same problems. Haven't tried newer kernels yet...
I get sig11's only when using "make -j" on any larger source package
(tested with the kernel and the shadow suite, using gcc 2.5.8, 2.6.3
and 2.7.2). Normal kernel compiles worked fine several times.
I hope this will help tracking down the bug further. Thanks!
Marek