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.
nine dogbert17: I wonder how you came up with that commit and diff? Neither reverting that part, nor reverting the whole commit fixes things for me. In fact I have to revert almost everything including "Allow inlining of dispatches with resume init args" to get it working. 15:22
tellable6 2021-09-12T20:26:30Z #raku-dev <vrurg> nine Could you, please, merge github.com/niner/Inline-Perl5/pull/168 because R#4495 depends on it.
dogbert17 nine: welcome back 15:41
dogbert17 tries to figure out what nine is referring to ...
if it's the callframe diff you're referring to then it seems that some comments, from me, should be taken with a grain of salt :( 15:43
nine Ah, yes, talking about callframe 15:44
dogbert17 the diff should be ignored but if I didn't mess up completely going back to commit github.com/MoarVM/MoarVM/commit/e3...c059a8ce88 seems to fix the problem 15:45
dogbert17 famous last words
nine Sorry, can't confirm 15:46
dogbert17 let's assume that I was speaking nonsense then :)
dogbert17 retries 15:47
hah, I messed up again my comment about 3c6d39dcd4 fixing the problem is incorrect, that's the commit which seems to introduce the problem 16:08
on the previous commit , i.e. a4853302b9ea8 I'm unable to repro the callfram problem. Perhaps I'm not trying hard enough. 16:09
s/3c6d39dcd4/e3c6d39dcd4/ 16:10
nine: did you have a nice vacation? 16:12
nine dogbert17: oh yes, very much so, thanks :) 17:11
dogbert17: oh, indeed! a4853302b9ea83a63e941a935055b9a9225ce233 does look fine 17:18
The major difference is that in the broken version EXISTS-KEY gets inlined. 17:19
dogbert17 nine: sounds like you're already zeroing in on the bug 17:36
I have, what I believe to be, a much simpler bug which might perhaps be of interest at a later time 18:02
nine dogbert17: I maz have missed that a4853302b9 is still working because from e3c6d39dc on it segfaults until a later commit 18:24
Huh...the broken version deopts in 'a' while the working version never does 18:30
That deopt happens from the sp_guardconc right before the return_o (idx 20 -> pc 240) 18:50

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!