Stream: t-compiler

Topic: weekly meeting 2020-03-19 #54818


Santiago Pastorino (Mar 18 2020 at 19:00, on Zulip):

Hi @T-compiler/meeting ; the triage meeting will be starting in 19 hours

Santiago Pastorino (Mar 18 2020 at 19:04, on Zulip):

I will be doing pre-triage in #t-compiler/wg-prioritization > pre-meeting triage 2020-03-19 #54818

Santiago Pastorino (Mar 18 2020 at 19:06, on Zulip):

btw, as you may notice that's a new stream. As per #t-compiler > design meeting 2020-02-28 about "focused and efficient triage" we are just starting to form a working group.

Santiago Pastorino (Mar 18 2020 at 19:16, on Zulip):

this week's checkins are PGO which is retired and Pipelining Working Group which seems to be retired too cc @Alex Crichton @Nicholas Nethercote

Nicholas Nethercote (Mar 18 2020 at 21:55, on Zulip):

Yes, I think the Pipelining WG can be disbanded. Pipelining shipped some time ago, it works well, I have no plans to work more on it, and I suspect Alex also has no further plans.

Santiago Pastorino (Mar 19 2020 at 13:37, on Zulip):

hi @T-compiler/meeting, triage meeting will be starting in 23 minutes

Santiago Pastorino (Mar 19 2020 at 13:39, on Zulip):

you can check out the meeting agenda here

pnkfelix (Mar 19 2020 at 14:02, on Zulip):

hi @T-compiler/meeting , we will start now. please add a :wave: to show you are here.

pnkfelix (Mar 19 2020 at 14:02, on Zulip):

we will start off with 5 minutes for ...

pnkfelix (Mar 19 2020 at 14:02, on Zulip):

Announcements

nikomatsakis (Mar 19 2020 at 14:03, on Zulip):
centril (Mar 19 2020 at 14:03, on Zulip):

@nikomatsakis Happy birthday in advance!

eddyb (Mar 19 2020 at 14:04, on Zulip):
eddyb (Mar 19 2020 at 14:04, on Zulip):

feel free to throw things my way (but not too many)

Santiago Pastorino (Mar 19 2020 at 14:04, on Zulip):
centril (Mar 19 2020 at 14:04, on Zulip):

/me is writing their first clippy lint... 830-ish LOC in... dayum.

Santiago Pastorino (Mar 19 2020 at 14:04, on Zulip):

that's as part of rustc-dev-guide

Santiago Pastorino (Mar 19 2020 at 14:05, on Zulip):

if someone have a nice PR that serves the purposes of code reading and understanding something would be great if you share

pnkfelix (Mar 19 2020 at 14:05, on Zulip):

Oh yeah let me cut-and-paste the announcements from the agenda too

pnkfelix (Mar 19 2020 at 14:05, on Zulip):
pnkfelix (Mar 19 2020 at 14:06, on Zulip):

also, an emoji straw-poll: Should we remove I-nominated on toolstate breakage issues? Put :thumbs_up: for yes and :thumbs_down: for no.

Wesley Wiser (Mar 19 2020 at 14:07, on Zulip):

What happens if a tool breaks right before we cut a release? Is there another mechanism where we'll notice that it's broken or will it just be missing from the release?

nikomatsakis (Mar 19 2020 at 14:07, on Zulip):

seems related to wg-prioritization formation :)

pnkfelix (Mar 19 2020 at 14:08, on Zulip):

(for context: the main reason that they are nominated is to ensure that they get attention and (maybe) that they get a priority label (P-high etc)

nikomatsakis (Mar 19 2020 at 14:08, on Zulip):

Wesley Wiser said:

What happens if a tool breaks right before we cut a release? Is there another mechanism where we'll notice that it's broken or will it just be missing from the release?

this was my question too

centril (Mar 19 2020 at 14:08, on Zulip):

I think it will be noticed :slight_smile:

nikomatsakis (Mar 19 2020 at 14:08, on Zulip):

I think I would say :+1: if we hvae a backup mechanism to monitor status and raise alarm if it seems like a problem is persisting

Wesley Wiser (Mar 19 2020 at 14:08, on Zulip):

If marking it I-nominated was supposed to be that mechanism, I'm not sure we should remove it

nikomatsakis (Mar 19 2020 at 14:08, on Zulip):

e.g. we could just poll/announce status as part of the triage meeting

pnkfelix (Mar 19 2020 at 14:08, on Zulip):

one option would be to start nominating them only during the period before a release?

centril (Mar 19 2020 at 14:08, on Zulip):

it's kinda hard to have no one notice it basically, given how many eyes there are on the repo

simulacrum (Mar 19 2020 at 14:09, on Zulip):

mostly I think our experience has been that tool developers are pretty eager to fix their tool :)

nikomatsakis (Mar 19 2020 at 14:09, on Zulip):

(wg-prioritization could copy/paste the info into the meeting briefing)

pnkfelix (Mar 19 2020 at 14:09, on Zulip):

i.e. change the bot to have time-dependent behavior

nikomatsakis (Mar 19 2020 at 14:09, on Zulip):

I'm fine w/ removing I-nominated but I also think we should make "look at tool status" part of our procedure

nikomatsakis (Mar 19 2020 at 14:09, on Zulip):

it seems pretty low overhead

pnkfelix (Mar 19 2020 at 14:10, on Zulip):

which procedure? Release, or weekly triage?

Santiago Pastorino (Mar 19 2020 at 14:10, on Zulip):

yep, maybe we should add that

Wesley Wiser (Mar 19 2020 at 14:10, on Zulip):

Ok, removing the nomination seems fine to me then. Thanks!

nikomatsakis (Mar 19 2020 at 14:10, on Zulip):

I think wg-prioritization should look, as part of pre-triage, at tool status -- if a tool is broken for 2 weeks in a row or something, they can look at why

nikomatsakis (Mar 19 2020 at 14:10, on Zulip):

since people are normally fixing quickly, that should be unusual

pnkfelix (Mar 19 2020 at 14:11, on Zulip):

okay, maybe we can let the wg-prioritization group fine tune the process here then.

Santiago Pastorino (Mar 19 2020 at 14:11, on Zulip):

agreed

pnkfelix (Mar 19 2020 at 14:11, on Zulip):

great.

centril (Mar 19 2020 at 14:11, on Zulip):

(typically there's a fix PR up the same day... it's so fast that it sorta gets in the way of rollups... ^^)

pnkfelix (Mar 19 2020 at 14:12, on Zulip):

seems like everyone is on-board for the proposal to stop auto-nominating tool-state breakage, at least in the common case.

pnkfelix (Mar 19 2020 at 14:12, on Zulip):

So, next up: beta-nominations

pnkfelix (Mar 19 2020 at 14:13, on Zulip):
pnkfelix (Mar 19 2020 at 14:15, on Zulip):

I'm trying to remember if this was a deliberate decision

pnkfelix (Mar 19 2020 at 14:15, on Zulip):

back in the day

centril (Mar 19 2020 at 14:15, on Zulip):

to accept negation?

pnkfelix (Mar 19 2020 at 14:15, on Zulip):

to treat it as part of the literal rather than an operation

centril (Mar 19 2020 at 14:16, on Zulip):

$lit:literal does intentionally accept a negated literal if that's what you're asking

pnkfelix (Mar 19 2020 at 14:16, on Zulip):

oh oh this was a regression, never mind

centril (Mar 19 2020 at 14:17, on Zulip):

yea; I broke it, I fix it :slight_smile:

pnkfelix (Mar 19 2020 at 14:17, on Zulip):

I didn't look carefully enough at the underlying issue description

pnkfelix (Mar 19 2020 at 14:17, on Zulip):

Okay sounds great, beta-accepted.

pnkfelix (Mar 19 2020 at 14:19, on Zulip):

we have no stable-nominations

pnkfelix (Mar 19 2020 at 14:19, on Zulip):

(which is what I would hope, since the release was only last week...)

pnkfelix (Mar 19 2020 at 14:19, on Zulip):

and there's no PR's marked S-waiting-on-team

pnkfelix (Mar 19 2020 at 14:20, on Zulip):

As noted in the agenda there are 54 P-high issues, 32 of which are unassigned.

pnkfelix (Mar 19 2020 at 14:21, on Zulip):

one P-high stable-to-beta regression, 5 P-high stable-to-nightly regressions. (and some P-medium ones in both buckets; see the agenda for more details.)

pnkfelix (Mar 19 2020 at 14:21, on Zulip):

actually one detail that isn't spelled out in the agenda is if any of the P-high regressions are unassigned. We probably should call such cases out specifically if we find them.

pnkfelix (Mar 19 2020 at 14:22, on Zulip):

3 of the P-high stable-to-nightly regressions are indeed unassigned.

Santiago Pastorino (Mar 19 2020 at 14:22, on Zulip):

this one https://github.com/rust-lang/rust/issues/70041 is not assigned

pnkfelix (Mar 19 2020 at 14:22, on Zulip):

"ICE when building docopt with rustc nightly on Travis CI" #70041

pnkfelix (Mar 19 2020 at 14:22, on Zulip):

(i'm just going to list them all right now, explicitly here)

pnkfelix (Mar 19 2020 at 14:23, on Zulip):

"Compiler error while compiling Winrt" #66402

Santiago Pastorino (Mar 19 2020 at 14:23, on Zulip):

"is_x86_feature_detected!("avx512f") fails to build on beta and nightly" #68905

eddyb (Mar 19 2020 at 14:23, on Zulip):

#70041 has fix in #70051, you can assign @Zoxc

nikomatsakis (Mar 19 2020 at 14:23, on Zulip):

Santiago Pastorino said:

this one https://github.com/rust-lang/rust/issues/70041 is not assigned

pending fix https://github.com/rust-lang/rust/pull/70051 ?

pnkfelix (Mar 19 2020 at 14:23, on Zulip):

and @Wesley Wiser did look into #66402

centril (Mar 19 2020 at 14:24, on Zulip):

odd; I'm trying to assign @Zoxc but it doesn't work

Wesley Wiser (Mar 19 2020 at 14:24, on Zulip):

Yeah it seems to have already been fixed quite a while ago

pnkfelix (Mar 19 2020 at 14:24, on Zulip):

okay, closing #66402 as fixed then.

Wesley Wiser (Mar 19 2020 at 14:25, on Zulip):

Should we write a test for this?

pnkfelix (Mar 19 2020 at 14:25, on Zulip):

oh. darn.

pnkfelix (Mar 19 2020 at 14:25, on Zulip):

reopening, and reprioritizing as P-medium, E-needstest.

Wesley Wiser (Mar 19 2020 at 14:25, on Zulip):

(sorry lol)

nikomatsakis (Mar 19 2020 at 14:25, on Zulip):

Santiago Pastorino said:

"is_x86_feature_detected!("avx512f") fails to build on beta and nightly" #68905

so we have a plan of action here but we don't I guess have a PR? @Amanieu did you plan on making a PR here?

nikomatsakis (Mar 19 2020 at 14:26, on Zulip):

probably pretty easy one to mentor

Amanieu (Mar 19 2020 at 14:26, on Zulip):

#68905 is waiting for the FCP to finish (still 1 week to go), but I'm tempted to skip that and merge the fix now. We canthen backport to beta.

nikomatsakis (Mar 19 2020 at 14:26, on Zulip):

and/or just do :)

Amanieu (Mar 19 2020 at 14:26, on Zulip):

The PR is here: https://github.com/rust-lang/stdarch/pull/842

nikomatsakis (Mar 19 2020 at 14:26, on Zulip):

er, wait, I think it's impl'd in https://github.com/rust-lang/stdarch/pull/842

nikomatsakis (Mar 19 2020 at 14:26, on Zulip):

Amanieu said:

#68905 is waiting for the FCP to finish (still 1 week to go), but I'm tempted to skip that and merge the fix now. We canthen backport to beta.

yes, it's fine

nikomatsakis (Mar 19 2020 at 14:27, on Zulip):

we often skip waiting out the full FCP period for "reversible, unlikely to be controversial" decisions like this

pnkfelix (Mar 19 2020 at 14:27, on Zulip):

okay, so now we just need the backport?

centril (Mar 19 2020 at 14:27, on Zulip):

@Amanieu that only stabilizes feature detection right?

pnkfelix (Mar 19 2020 at 14:28, on Zulip):

and #68905 should stay open until the backport lands. Still, seems like its under control.

pnkfelix (Mar 19 2020 at 14:28, on Zulip):

Should I assign the bug to @Amanieu ? Who is point-person on this?

Amanieu (Mar 19 2020 at 14:28, on Zulip):

Alright I'll merge it and submit a PR to bump the submodule.

Amanieu (Mar 19 2020 at 14:28, on Zulip):

Yes

Amanieu (Mar 19 2020 at 14:29, on Zulip):

Normally @gnzlbg is the point person for stdarch, but he's been away from Rust-related stuff lately. I can take over maintenance of stdarch for now.

nikomatsakis (Mar 19 2020 at 14:30, on Zulip):

Amanieu said:

Normally gnzlbg is the point person for stdarch, but he's been away from Rust-related stuff lately. I can take over maintenance of stdarch for now.

it seems obvious that we should be writing this sort of thing down somewhere

nikomatsakis (Mar 19 2020 at 14:30, on Zulip):

just noting for later but I feel like "who to ping about X?" has been coming up more and more often

pnkfelix (Mar 19 2020 at 14:30, on Zulip):

okay, now all the unassigned regressions are only P-medium priority.

Santiago Pastorino (Mar 19 2020 at 14:31, on Zulip):

I was thinking that experts map may need to be called something like who maintains what or who works on what

nikomatsakis (Mar 19 2020 at 14:31, on Zulip):

yeah, I was thinking the experts map was a possible place

nikomatsakis (Mar 19 2020 at 14:31, on Zulip):

I think we already divided up the entries a bit into "maintainers" and "folks who know things" and stuff like that, right?

Santiago Pastorino (Mar 19 2020 at 14:31, on Zulip):

we can discuss this later but I fear that the name experts is not great for a lot of reasons :)

simulacrum (Mar 19 2020 at 14:31, on Zulip):

(In some sense, perhaps the right thing is a github team for each repository, so you don't even need to go hunt down a map, just cc rust-lang/repo

pnkfelix (Mar 19 2020 at 14:33, on Zulip):

okay so there are two nominated issues

centril (Mar 19 2020 at 14:33, on Zulip):

@simulacrum that's a cool idea!

pnkfelix (Mar 19 2020 at 14:34, on Zulip):
nikomatsakis (Mar 19 2020 at 14:34, on Zulip):

simulacrum said:

(In some sense, perhaps the right thing is a github team for each repository, so you don't even need to go hunt down a map, just cc rust-lang/repo

Ideally this would be reflected in the team repo: I've been thinking about making them be "ice-breaker" style groups. The other case where this comes up a lot is "architectures".

nikomatsakis (Mar 19 2020 at 14:34, on Zulip):

pnkfelix said:

cleanup crew?

centril (Mar 19 2020 at 14:34, on Zulip):

It's a rather big codebase, will paste a link later on.

sadly #70117 has this ^--

centril (Mar 19 2020 at 14:35, on Zulip):

so no point in cleanup crew ping yet

pnkfelix (Mar 19 2020 at 14:35, on Zulip):

ah it got a link just 32 minutes ago

nikomatsakis (Mar 19 2020 at 14:35, on Zulip):

I guess there's literally no repo right now?

pnkfelix (Mar 19 2020 at 14:35, on Zulip):

it was unactionable until then

pnkfelix (Mar 19 2020 at 14:35, on Zulip):

but now we can

Santiago Pastorino (Mar 19 2020 at 14:35, on Zulip):

nikomatsakis said:

pnkfelix said:

  • nom: [ ] "rustc segfault" #70117

cleanup crew?

yes, 30 minutes ago the reporter shared "code"

nikomatsakis (Mar 19 2020 at 14:35, on Zulip):

oh, no there is now

centril (Mar 19 2020 at 14:35, on Zulip):

oh; it seems OP was not updated

centril (Mar 19 2020 at 14:36, on Zulip):

@pnkfelix aha; @rustbot cleanup-crew seems like another alias to add :D

pnkfelix (Mar 19 2020 at 14:36, on Zulip):

@centril yeah definitely

Santiago Pastorino (Mar 19 2020 at 14:36, on Zulip):

repo https://github.com/haraldh/enarx_sev_kvm_demo

pnkfelix (Mar 19 2020 at 14:36, on Zulip):

for those of us who do find that memorable. :smile:

pnkfelix (Mar 19 2020 at 14:37, on Zulip):

It also is likely to be an LLVM bug

nikomatsakis (Mar 19 2020 at 14:37, on Zulip):

plausible

pnkfelix (Mar 19 2020 at 14:37, on Zulip):

but maybe we can wait until cleanup crew dives in before we ping the LLVM icebreakers ?

pnkfelix (Mar 19 2020 at 14:37, on Zulip):

(i.e. they can work on reducing to mcve)

Santiago Pastorino (Mar 19 2020 at 14:38, on Zulip):

note the while cargo build -vvv --release ; do cargo clean; done; part

pnkfelix (Mar 19 2020 at 14:38, on Zulip):

okay anyway my guess is that we should call this P-high for now, just for the initial investigation. Okay?

nikomatsakis (Mar 19 2020 at 14:38, on Zulip):

I guess it depends how confident we are this is an LLVM bug

pnkfelix (Mar 19 2020 at 14:38, on Zulip):

(it seems unlikely to be P-critical without more information.)

nikomatsakis (Mar 19 2020 at 14:38, on Zulip):

Santiago Pastorino said:

note the while cargo build -vvv --release ; do cargo clean; done; part

heh .. oh dear

bjorn3 (Mar 19 2020 at 14:39, on Zulip):

Santiago Pastorino said:

repo https://github.com/haraldh/enarx_sev_kvm_demo

enarx is the name of an organization he is member of. SEV are hardware encrypted VMs. KVM is the linux hardware VM system.

pnkfelix (Mar 19 2020 at 14:39, on Zulip):

might need rr expertise for this one, then.

centril (Mar 19 2020 at 14:39, on Zulip):

@pnkfelix ask and ye shall receive, https://github.com/rust-lang/rust/pull/70150

pnkfelix (Mar 19 2020 at 14:40, on Zulip):

pnkfelix said:

might need rr expertise for this one, then.

(due to the non-deterministic nature, based on the while shell loop to drive the test)

pnkfelix (Mar 19 2020 at 14:40, on Zulip):

anyway I'll tag as P-high.

Amanieu (Mar 19 2020 at 14:41, on Zulip):

PR for updating stdarch: #70151

pnkfelix (Mar 19 2020 at 14:42, on Zulip):

nom: "ICE using associated type from higher ranked trait" #70120

pnkfelix (Mar 19 2020 at 14:42, on Zulip):

this does sound like a duplicate

centril (Mar 19 2020 at 14:42, on Zulip):

I suspect #70120 is a dupe of some form

eddyb (Mar 19 2020 at 14:42, on Zulip):

every single time you say "nom" and I actually see the word I think of the nom crate
(not relevant, sorry, just a recurring thing. for half a second I wonder "what have parser combinators done now?")

centril (Mar 19 2020 at 14:43, on Zulip):

(and probably hard to fix?)

pnkfelix (Mar 19 2020 at 14:43, on Zulip):

namely of this bucket issue, right? "ICE field: higher-rank trait bound (HRTB) for<'a> ... hits OutputTypeParameterMismatch in librustc/traits/codegen" #62529

nikomatsakis (Mar 19 2020 at 14:44, on Zulip):

yes, looks like the standard sort of case

nikomatsakis (Mar 19 2020 at 14:44, on Zulip):

not sure what's the best bucket issue

pnkfelix (Mar 19 2020 at 14:44, on Zulip):

@eddyb but would "I-nom" help in any way?

centril (Mar 19 2020 at 14:44, on Zulip):

@pnkfelix ah nice, you got it on hand; yeah that's the one :P

pnkfelix (Mar 19 2020 at 14:44, on Zulip):

(i suspect that would just make things worse)

pnkfelix (Mar 19 2020 at 14:44, on Zulip):

I could find a greek letter. Or use :exclamation:

pnkfelix (Mar 19 2020 at 14:45, on Zulip):

anyway, my suggestion is that we add #70120 to the list on #62529.

eddyb (Mar 19 2020 at 14:45, on Zulip):

don't worry about it :P

Santiago Pastorino (Mar 19 2020 at 14:46, on Zulip):

we could also spell everything out :)

pnkfelix (Mar 19 2020 at 14:46, on Zulip):

yeah I happened to remember I called it an "ICE field"

Santiago Pastorino (Mar 19 2020 at 14:46, on Zulip):

I-nominated 1/x:

pnkfelix (Mar 19 2020 at 14:46, on Zulip):

and also I'm going to down grade #70120 to P-medium.

pnkfelix (Mar 19 2020 at 14:48, on Zulip):

the problem with spelling out "I-nominated" is that it takes up screen space

pnkfelix (Mar 19 2020 at 14:48, on Zulip):

that would be better occupied by the title.

pnkfelix (Mar 19 2020 at 14:48, on Zulip):

I think I'm going to stick with "nom" (or use some emoji if I can find the perfect one...)

pnkfelix (Mar 19 2020 at 14:49, on Zulip):

(I do need to consistently use 1/n though. Its not so bad when there are just two issues, but in general its important, I think.)

pnkfelix (Mar 19 2020 at 14:50, on Zulip):

Okay so that's all the nominations!

pnkfelix (Mar 19 2020 at 14:50, on Zulip):

and both of the scheduled WG checkins are with groups that should be disbanded.

nikomatsakis (Mar 19 2020 at 14:50, on Zulip):

I propose using :point_right: instead of "nom"

pnkfelix (Mar 19 2020 at 14:50, on Zulip):

So I think we can call the meeting to a close a little early. (Which is good, because "circle time" is starting now for my kids, in our attempt to recreate their school schedule.)

Santiago Pastorino (Mar 19 2020 at 14:50, on Zulip):

I have noted down to go over the rotation list and fix it

nikomatsakis (Mar 19 2020 at 14:51, on Zulip):

/me makes sure we stay focused on the important questions

Santiago Pastorino (Mar 19 2020 at 14:51, on Zulip):

unsure where that lives but gonna find out :)

pnkfelix (Mar 19 2020 at 14:51, on Zulip):

Bye everyone, thanks for attending! And stay safe; practice social distancing! :dancer: ...... :dancing:

pnkfelix (Mar 19 2020 at 14:51, on Zulip):

Bye @T-compiler/meeting

Last update: May 29 2020 at 17:25UTC