------- Comment #16 from bangerth at dealii dot org 2006-06-15 03:39 -------
(In reply to comment #13)
> and the test case never exceeded 1 GB vm with r111111. I would certainly hope
> that gcc would politely report vm exhaustion as out-of-memory or some such
> rather than segfaulting.
That, unfortunately, is out of the hands of gcc: most operating systems allow
processes to allocate more memory than there physically is, on the assumption
that typically processes allocate more memory than they actually need, or that
the don't need it right away. Only when the process actually writes to a page
is it physically allocated. If at that time the operating system cannot honor
its commitment of providing a physical page for an allocated page, the
process segfaults. Since the OS typically doesn't provide a way to figure
out why the segfault happens, application programs don't usually have a way
to state the actual cause of the segfault.
All that said: does the problem persist? Could you repeat your tests with
the latest snapshots to see whether this still happens?
Thanks
Wolfgang
--
bangerth at dealii dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |bangerth at dealii dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24605
