@WG-prioritization/alerts issue #76387 has been requested for prioritization.
I'd say P-high? Looks pretty bad but they haven't said it's a regression
yup, confirmed not a regression
Let’s ping LLVM group for this?
I'm not sure it's LLVM
It sounded like a MIR opt gone bad
Oh wait they said
Dead Argument Elimination
So yeah, LLVM
Would be nice to have a reproduction in C then
Issue #76387's prioritization request has been removed.
@LeSeulArtichaut @Joshua Nelson I've just seen that you have agreed on
P-high but it seems
P-critical to me
sounds fine to me, @nagisa also suggested P-critical
I understand that this is an LLVM issue and maybe that make us lower a bit the priority, on the other hand I think I'd rather tag this kind of issues as
P-critical as in the end they affect Rust very badly and there's something we can do, like backing out the LLVM 11 upgrade if needed
anyway, I wanted to discuss this :point_up: briefly to see what others do and think, my opinion is that I wouldn't lower the priority of an issue because it's LLVM or a Rust 3rd party dependency because they still affect Rust and almost always there's something we can do :)
my reasoning originally is that no one had come across it for 5 years, but if it's unsound it should probably higher yeah
but in general, what do you think about my thoughts?
regardless of this particular issue
makes sense :)
anyway, back to this particular case, I'm kind of having second thoughts about the priority of it :)
was checking and yeah this issue exists since a TON
in general I tend to think that soundness issues are always
P-critical but ... unsure ...
cc @pnkfelix who might want to also weigh in
Can we assign the best label based on our estimate, then tomorrow during the meeting it will be discussed anyway and possibly the priorty adjusted?
yes, that would be ok
but just in case, I've already labelled as
just mentioned Felix in case they wanted to give an opinion
@Santiago Pastorino According to last week's meetings notes, the issue is being actively worked on, can we remove the I-prio label?
I-prioritize label in this issue
unsure what did you mean exactly
@Aaron Hill @pnkfelix (if you want to discuss here for LLVM cherry-pick)
to backport I guess the commit should be cherry-picked to https://github.com/rust-lang/llvm-project and then on the rust repo just update the llvm-project submodule
cc @Aaron Hill
I'm writing a regression test, and I'll open a rustc pr bumping llvm soon