Stream: t-compiler/wg-prioritization/alerts

Topic: I-prioritize #79890 Failed to recover key for type_of with …


triagebot (Dec 10 2020 at 14:07, on Zulip):

@WG-prioritization/alerts issue #79890 has been requested for prioritization.

Procedure

Joshua Nelson (Dec 10 2020 at 14:10, on Zulip):

P-high maybe, someone ran into it almost immediately

apiraino (Dec 10 2020 at 14:13, on Zulip):

out of curiosity, what is the issue here? Looks like that rustc is using some cached stuff when it should not?

Joshua Nelson (Dec 10 2020 at 14:14, on Zulip):

looks like an issue with metadata serialization, maybe? I think there might be a hash collision

apiraino (Dec 10 2020 at 14:17, on Zulip):

right, I thought maybe a hash of an external module that it's not there anymore becuase removed from the main.rs

apiraino (Dec 10 2020 at 14:17, on Zulip):

anyway, not important just geek curiosity :smile:

Joshua Nelson (Dec 10 2020 at 14:17, on Zulip):

I don't know, sorry

triagebot (Dec 10 2020 at 14:19, on Zulip):

Issue #79890's prioritization request has been removed.

Camelid (Dec 10 2020 at 19:59, on Zulip):

It really feels like we've been getting a lot of incr-comp issues recently...

Joshua Nelson (Dec 10 2020 at 20:00, on Zulip):

they're all related

Joshua Nelson (Dec 10 2020 at 20:00, on Zulip):

this is a regression from https://github.com/rust-lang/rust/pull/79721/, which fixed https://github.com/rust-lang/rust/issues/79661

Camelid (Dec 10 2020 at 20:00, on Zulip):

Ah, I see

Joshua Nelson (Dec 10 2020 at 20:00, on Zulip):

which was in turn caused by https://github.com/rust-lang/rust/pull/74967, a pretty significant refactor of metadata loading

Aaron Hill (Dec 10 2020 at 20:29, on Zulip):

yeah, sorry about that

Aaron Hill (Dec 10 2020 at 20:29, on Zulip):

it's hard to test incremental changes in the same way that you test other changes

Aaron Hill (Dec 10 2020 at 20:30, on Zulip):

since doing a Crater run (or manually testing a bunch of crates) doesn't really exercise much code

Santiago Pastorino (Dec 10 2020 at 20:31, on Zulip):

we could run crater on each crate up to HEAD~1, and then incrementally build up to HEAD :smiley:

Santiago Pastorino (Dec 10 2020 at 20:32, on Zulip):

or well to previous version and then up to latest

Aaron Hill (Dec 10 2020 at 20:32, on Zulip):

previous versions would probably be better

Santiago Pastorino (Dec 10 2020 at 20:32, on Zulip):

right

Aaron Hill (Dec 10 2020 at 20:32, on Zulip):

HEAD~1 might just give you things like typo fixes/version bumps in a bunch of cases

Aaron Hill (Dec 10 2020 at 20:33, on Zulip):

which doesn't really exercise any incr comp code

Santiago Pastorino (Dec 10 2020 at 20:33, on Zulip):

yep

Aaron Hill (Dec 10 2020 at 20:33, on Zulip):

anyway, I have a fix for this issue, an I'm currently adding the test case to src/test/incremental

Santiago Pastorino (Dec 10 2020 at 20:33, on Zulip):

I'm not sure if doing something like that is reasonable and how much time would that take but something that may worth trying I guess

Santiago Pastorino (Dec 10 2020 at 20:34, on Zulip):

cc @simulacrum

Aaron Hill (Dec 10 2020 at 20:34, on Zulip):

we'd also need to see how difficult it would be to add that to Crater

Santiago Pastorino (Dec 10 2020 at 20:34, on Zulip):

yep

Aaron Hill (Dec 10 2020 at 20:34, on Zulip):

currently, I think Crater always re-tests the same crate version

simulacrum (Dec 10 2020 at 20:35, on Zulip):

I have no insight how hard this would be

simulacrum (Dec 10 2020 at 20:35, on Zulip):

File an issue though

Santiago Pastorino (Dec 10 2020 at 20:35, on Zulip):

:+1:

Santiago Pastorino (Dec 10 2020 at 20:38, on Zulip):

https://github.com/rust-lang/crater/issues/556

Last update: Apr 11 2021 at 18:15UTC