On Fri, Sep 17, 2010 at 08:35:16AM +0300, Nikodemus Siivola wrote:
> On 17 September 2010 00:54, Josef Wolf <jw@...> wrote:
Thanks for your quick reply, Nikodemus!
> > I am trying to compile sbcl on opensuse-11.2 running on a dual-core amd64
> > box with 8GB of RAM and 20GB of swap.
>
> > ::: Running (:UNSYNCHRONIZED-HASH-TABLE)
> > Heap exhausted during allocation: 1016590336 bytes available, 1073741840 requested.
>
> This is very curious. Is this repeatable? That is, if you run the test
> suite ten times, how many times does this error occur?
It _was_ reproducible with the same build. Unfortunately, instead of keeping
this build, I deleted it and ran my build-script again. This time the tests
are OK. Since I always make me a small shell-script to build my stuff, the
commands that were executed to do the builds were identical in both cases.
When the outcome is different, some sort of race condition has to be suspected
or something.
I might try to run the build over and over again until it happens again. One
build lasts about 1 hour on this machine, so it might take a while until it
happens again (if at all).
> In any case, if after adding #+nil in front of
>
> (with-test (:name (:unsynchronized-hash-table))
>
> in tests/threads.impure.lisp nothing else fails. I would not worry:
> the test in question is quite an evil one.
Too late now.. Sorry for not having thought twice before deleting the
build.
> That said, heap-exhaustion from it is definitely not expected, and I
> would be very interested to hear how reproducible this is, and if
> anyone else sees this happening.
I'll go trying to rebuild until it happens again.
|