github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
AlexDaniel squashable6: status 17:53
squashable6 AlexDaniel, ⚠🍕 Next SQUASHathon in ≈10 hours (2019-12-07 UTC-12⌁UTC+20). See github.com/rakudo/rakudo/wiki/Mont...Squash-Day
nine That's a curious segfault: a sp_guardconc op in a frame without spesh cands and more importantly without effective_spesh_slots 18:56
tellable6 2019-12-05T10:09:07Z #raku-dev <tbrowder> nine thanks! 18:57
nine At least it's apparently not a regression by my NativeCall work. It's reproducible on 2019.07 19:47
And 2018.12 19:52
But not on 2017.12, so let the bisect begin 20:00
nine Darn...when bisecting I run into a range of commits that causes bogus errors like "Attribute $!name not declared in class Inline::Perl5::ClassHOW" 21:05
MasterDuke nine: not easy to use committable/bisectable? 21:21
nine with Inline::Perl5? 21:30
MasterDuke oh, bisecting commits in I:P5? 21:35
that you couldn't do. but you can install modules if you're bisecting rakudo
nine I am bisecting rakudo. Just curious: how does bisectable handle such broken commits? 21:49
Manual selection of commits got me out of the breakage :) 21:57
MasterDuke it doesn't really, just says the output is suspicious. but by manually setting where to start/end you could probably get something useful. or use committable to run over a whole range of commits to see all the different outputs 22:04
nine And the winner is: github.com/MoarVM/MoarVM/commit/38...99cbe588d6 22:14
MasterDuke nice. have a fix? 22:16
nine No, will have a look tomorrow. Now that I know what to look at 22:18
Actually I think that commit is innocent. It just unlocks some optimization that breaks things. But the commit should help me find this optimization. 22:27
As I can compare spesh dumps with and without it