Stream: t-compiler/wg-prioritization/alerts

Topic: I-prioritize #76432 Incorrect MIR optimizations at level 3:…


triagebot (Sep 18 2020 at 10:06, on Zulip):

@WG-prioritization/alerts issue #76432 has been requested for prioritization.

Procedure

James Gill (Sep 18 2020 at 16:21, on Zulip):

Hi, was looking into this, trying to get my head around the way this works. Due to it being an unsoundness bug we are in P-High/P-Crit territory?

Joshua Nelson (Sep 18 2020 at 16:23, on Zulip):

that sounds right

Joshua Nelson (Sep 18 2020 at 16:23, on Zulip):

although this one is a little odd because miri caught it, so it might not be affecting code in the wild

Joshua Nelson (Sep 18 2020 at 16:23, on Zulip):

I would say P-high maybe?

James Gill (Sep 18 2020 at 16:24, on Zulip):

Yeah, ngl first experience with Miri today haha, and yeah, isn't a crit in my eyes

lcnr (Sep 18 2020 at 16:26, on Zulip):

P-high makes sense I think. We recently changed mir-opt-level=3 to mean that these opts have a compile time cost but are not unsound

lcnr (Sep 18 2020 at 16:26, on Zulip):

iirc

lcnr (Sep 18 2020 at 16:27, on Zulip):

before that there were quite a few known unsound opts at mir-opt-level=3

Joshua Nelson (Sep 18 2020 at 16:27, on Zulip):

@lcnr do you happen to know the default for the compiler that's released?

lcnr (Sep 18 2020 at 16:28, on Zulip):

(mir-opt-level can only be set on nightly, so this is definitely not P-critical imo)

lcnr (Sep 18 2020 at 16:28, on Zulip):

mir-opt-level=1 I think

Joshua Nelson (Sep 18 2020 at 16:28, on Zulip):

thanks!

lcnr (Sep 18 2020 at 16:29, on Zulip):

See https://github.com/rust-lang/compiler-team/issues/319 for the current design (at least I think that this is the current one :laughing: )

Wesley Wiser (Sep 18 2020 at 16:54, on Zulip):

Yeah, I'm working on implementing part of that MCP.

Wesley Wiser (Sep 18 2020 at 16:55, on Zulip):

IMO this is P-medium since you have to use -Zmir-opt-level=3 and you can only do so on nightly. There's no way a stable or beta user could be affected. -Zmir-opt-level=3 has had unsoundness for a long time anyway so this not a recent issue by any means.

lcnr (Sep 18 2020 at 16:55, on Zulip):

ah, I thought that the MCP was already implemented. In this case P-medium also seems fine to me.

DPC (Sep 18 2020 at 17:20, on Zulip):

agreed with p-med

Camelid (Sep 19 2020 at 02:26, on Zulip):

It looks like no one applied the P-medium label. I tried to apply it but I guess it won't work until the team PR gets merged :)

triagebot (Sep 19 2020 at 02:58, on Zulip):

Issue #76432's prioritization request has been removed.

Joshua Nelson (Sep 19 2020 at 03:34, on Zulip):

applied P-medium

Last update: Apr 11 2021 at 19:00UTC