The Linux build has been failing with a segfault. It looks to me as though this started with 0e022e56b130ab9d277965b794e70d8d3fb29533: Turn EvTerm (almost) into CoreExpr.
Hi, Am Freitag, den 26.01.2018, 21:49 -0500 schrieb David Feuer:
The Linux build has been failing with a segfault. It looks to me as though this started with 0e022e56b130ab9d277965b794e70d8d3fb29533: Turn EvTerm (almost) into CoreExpr.
I have seen that, but it seemed to be intermittent. I saw a segfault in the differential revision for that patch, but restarting the build made it go away. Given that this patch is purely a refactoring in the type checker, I believe the problem is somewhere else. Did anyone observe the segfaults locally? Joachim -- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/
Hi, Am Freitag, den 26.01.2018, 22:34 -0500 schrieb Joachim Breitner:
Did anyone observe the segfaults locally?
JFTR, neither of perf.haskell.org, Travis CI and CircleCI observe the segfaults. Joachim -- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/
David Feuer
The Linux build has been failing with a segfault. It looks to me as though this started with 0e022e56b130ab9d277965b794e70d8d3fb29533: Turn EvTerm (almost) into CoreExpr.
Yes, I know. I believe this is just due to a long-standing crash when the RTS is unable to allocate heap. Unfortunately I had trouble reproducing it earlier in the week. I'll try to have another look today. Cheers, - Ben
participants (3)
-
Ben Gamari -
David Feuer -
Joachim Breitner