> ok, if you can reproduce it easily, could you try the following debug
Unfortunately, I cannot reproduce the errors easily. I got the signal 7
only once on every machine until now. I have applied your patches and am
running concurrent endless "make -j" on two machines right now.
> patch, it puts the reason for the SIGBUS (signal 7) into the syslog. [well
> there are two more possible reasons for SIGBUS that this patch doesnt
> handle, but they are _very_ unlikely in the case of GCC]
>
> are all those machines set up by you? Maybe you use the same swapsize or
> similar other stuff, which tickles the same type of memory shortage. But
> anyways, it would be interesting to see what the exact reason for your
> SIGBUS's is.
I don't know whether the sig 7 problem and the "Couldn't get a free
page.." have the same cause. I get the "Couldn't get..." on many machines.
Swap space never was exhausted. In fact, swap is ony rarely used. I never
had these problems before. It started with 2.0.30.
Kernel is plain 2.0.30 with Krzysztof Strasburger's memory leak patch
applied.
Hubert mantel@suse.de