github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
Guest23744 not much activity here today :) 12:13
nine Yeah....if there was more activity I could take the lack of a "NNNOOOOO don't do that!" as a sign that turning guard sets into MVMCollectable was sane... 17:02
dogbert11 perhaps it is sane 17:03
jnthn Ah, I meant to look at that 17:08
Hmm...not sure 17:15
I mean, it marks the things within the plugin guard set but I don't see how it prevents the memory related to them being freed 17:17
Hm, what's allocating during guard resolution anyway... 17:18
Oh, the collected objects
Since spesh plugin resolution needn't be reentrant, I wonder wonder if we can simply allocate such an array once and hang it off the thread context 17:19
otoh, that sadly won't help with MVM_repr_get_attr_o
nine jnthn: if the guard set were an MVMCollectable we wouldn't need the MVM_fixed_size_free_at_safepoint and the MVMROOT in resolve_using_guards would keep it alive. 20:35
Or maybe the whole plugin state 20:36
jnthn nine: Yes, that may be the cleanest option 21:54
At least, I didn't come up with a better idea
I didn't convince myself that the current solution is correct. I didn't prove it ain't either, but my gut feeling is something's not right 21:55
patrickb Could someone have a look at PR #1182? That's enabling paths with unicode symbols on Windows. I'd like to have that merged before the next release, as I see it as a prerequisite for a binary release on Windows. (github.com/MoarVM/MoarVM/pull/1182) 23:13
tellable6 2019-09-26T16:23:38Z #perl6-dev <vrurg> patrickb can we have perl6-m in rakudo source not run moar but be linked against moar's dynlib, same as inst-perl6-m?