at r17619, i'm unable to run make test, due to failures in p5rx tests.
actually, none of them are failing, but my system becomes
non-responsive. after approximately test 200, tests (and there are
over 900) take a few seconds each to complete, and each one uses more
and more memory. at over 125MB, and 100% cpu, i can do little but kill
the process. this means that the harness dies, and i can't complete
'make test'.
i suspect a string gc problem. in the meantime, i'm turning on -G
locally so i can run the rest of the tests. although it uses 170+MB
even with -G, the test completes, and does so in 12s. if there's no
other solution pre-release, i'll commit this change.
~jerry
> at r17619, i'm unable to run make test, due to failures in p5rx tests.
> actually, none of them are failing, but my system becomes
> non-responsive. after approximately test 200, tests (and there are
> over 900) take a few seconds each to complete, and each one uses more
> and more memory. at over 125MB, and 100% cpu, i can do little but kill
> the process. this means that the harness dies, and i can't complete
> 'make test'.
>
> i suspect a string gc problem.
A GC problem might explain the STM test failure too. Do you see that? If so,
is it possible to do a binary search to figure out which checkin caused (or
revealed) these failures?
-- c