Welcome to the main channel on the development of MoarVM, a virtual machine for NQP and Rakudo (moarvm.org). This channel is being logged for historical purposes.
Set by lizmat on 24 May 2021.
Nicholas *, * good 06:52
[Coke] *, * * 15:43
[Coke] gist.github.com/coke/ae63d0a6efee6...f449632780 18:15
getting a "weird" failure mode with recent rakudos (last month or so, bisect led me to skipped commits) where my script just exits mid-processing. gist.github.com/coke/ae63d0a6efee6...f449632780 - grabbed the backtrace from the VS debugger.
(running with JIT_DISABLE or SPESH_DISABLE = 1 has no effect, program still exits early in processing) 18:18
I did a half hearted attempt to golf which didn't go far. Can try that path again if this BT isn't helpful
Looks like the GC is trying to free something and exploding? 18:20
MasterDuke [Coke]: can you show the script that causes the invalid free? and/or try with a MoarVM built with `--no-mimalloc`? 19:41
ah ha. seems a `git clean -xdf` in 3rdparty/gmp has fixed the gcd problem 21:09
[Coke] MasterDuke: no, it's $dayjob stuff with a bit going on. 21:14
I can try the --no-mimalloc, though.
(it's a diferent machine than i IRC from, so expect delays, thank you. :) 21:15
MasterDuke ok, so there are a bunch of spectest fails, but they seem likely to not be related to gmp at all. rebuilding with master to give that a whirl and see if i get the same fails 21:43
if so, it might just be that warning and the abort handling left (assuming it is not currently 100% correct) and the gmp branch is ready 21:44
MasterDuke the fails seen on master are pretty much identical to those seen on gmp. so yeah, just clobbered by longjmp warning (i suspect caused by the signal handling added to neutralize the abort), said signal handling, and the mingw build to finish up and then i think it's good to go (plus some rebasing to clean up the commits) 22:14