|
From: Tom H. <th...@cy...> - 2004-09-12 12:30:17
|
In message <200...@ac...>
Julian Seward <js...@ac...> wrote:
> Strangely, Tom's nightly builds appear OK, except "standard" (RH 7.2);
> not sure if that failure is related or not.
That's been going on for a while, but it always works fine when
run by hand so I think it's something odd going on when it runs
from cron overnight.
> Another one which might somehow mess with memory layout is Tom's
> fix to #73818 (stack rlimit stuff).
>
> I have a vague memory from writing the leak checker, that SuSE
> kernels behave differently, in that they will just extend the
> stack downwards indefinitely (or something) if you endlessly prod
> just below %esp, so perhaps that's related.
All kernels work like that, and so does valgrind. The stack segment
for the main thread had SG_GROWSDOWN set and when valgrind see a
segmentation fault below the current base of that segment it extends
it and maps some more memory.
All I did was make valgrind stop doing that once the stack limit is
hit, which is exactly what the kernel will do.
Tom
--
Tom Hughes (th...@cy...)
Software Engineer, Cyberscience Corporation
http://www.cyberscience.com/
|