Re: Problems in newest netbsd-5 systems?

[Frank Wille <frank%phoenix.owl.de@localhost>]

Frank Wille wrote:

> Now I can (nearly) reproduce the problem. It reaches compilation of
> pcrecpp_unittest.cpp, where your segfault happened, but it seems to take
> forever! After more than 3 hours I gave up waiting for the segfault (I
> have to run it overnight this weekend).
> 
> What I observed: In the beginning the cc1plus process runs 90% of the time
> in user-mode and 10% in supervisor-mode. After 8 minutes it has allocated
> 56MB memory and starts running 100% in supervisor-mode, for the following
> hours. No more additional memory will be allocated after that.

Sorry no segfault here. Just an endloss loop.

After running it for nearly 10 hours today, I used gdb to find out in which
function it hangs. It looks to me like malloc_init_hard(), or some of the
neighbour functions, from stdlib/jemalloc.c.

Will try to compile a libc with the old phk-malloc, to see if it makes any
difference.

How long does it take until you see the segfault? Do you get a core file
which you can debug with gdb?

-- 
Frank Wille