Stream: t-compiler

Topic: pre-meeting triage 2019-05-16 #54818


pnkfelix (May 16 2019 at 10:40, on Zulip):

Hello @T-compiler/meeting ; the triage meeting is in 3 hours (+ 20 minutes). I'll be doing pre-triage in this topic.

pnkfelix (May 16 2019 at 10:42, on Zulip):

unfortunately the WG-checkin calendar does not extend long enough to include today's date.

pnkfelix (May 16 2019 at 10:43, on Zulip):

looking at the previous scheduled checkin, it seems like we have not had recent checkins from WG-traits, WG-paralllel-rustc, and WG-pgo.

pnkfelix (May 16 2019 at 10:45, on Zulip):

the WG-traits page currently does not list a main contact.

pnkfelix (May 16 2019 at 10:45, on Zulip):

@WG-traits who is the main contact for WG-traits? I assume @nikomatsakis would be one option, but I'd kind of like to spread the responsibility around. @scalexm , are you a potential lead for the WG?

pnkfelix (May 16 2019 at 10:46, on Zulip):

Regarding WG-parallel-rustc: @Zoxc , would you be available to provide a summary of what's happened since the last checkin?

pnkfelix (May 16 2019 at 10:47, on Zulip):

as for WG-pgo: perhaps if we don't hear from one of the two groups above, perhaps @mw would be willing to provide a quick checkin from WG-pgo?

pnkfelix (May 16 2019 at 10:47, on Zulip):

having said that, lets continue with pre-triage.

pnkfelix (May 16 2019 at 10:48, on Zulip):

first pre-pass: unprioritized nominated T-compiler issues

pnkfelix (May 16 2019 at 10:50, on Zulip):

first up: " crater run to estimate impact of full NLL transition" #60680

pnkfelix (May 16 2019 at 10:50, on Zulip):

I'm going to mark this as P-medium. I'm not sure why Centril nominated it.

centril (May 16 2019 at 10:52, on Zulip):

I'm not sure why Centril nominated it.

To get it off the ground

centril (May 16 2019 at 10:52, on Zulip):

and hopefully get someone to do it soon

pnkfelix (May 16 2019 at 10:53, on Zulip):

okay

pnkfelix (May 16 2019 at 10:53, on Zulip):

we have got to resolve that future-compat lint thing.

pnkfelix (May 16 2019 at 10:54, on Zulip):

At least, it seems like I keep seeing a variety of inquiries that end up connecting to our question of how we're going to address that problem.

centril (May 16 2019 at 10:54, on Zulip):

Well it's on t-lang now; and I don't want to think about it atm

pnkfelix (May 16 2019 at 10:54, on Zulip):

(the problem of "future compatibilty)

pnkfelix (May 16 2019 at 10:54, on Zulip):

Right, its possible that someone else should just take up the mantle there.

centril (May 16 2019 at 10:54, on Zulip):

Well we need to discuss what to do

pnkfelix (May 16 2019 at 10:54, on Zulip):

(maybe I should. I sort of have other things on my plate, but that task at least seems ... like "a simple matter of programming", once we figure out what want to do.)

centril (May 16 2019 at 10:54, on Zulip):

It's not about execution at this point, but about policy

pnkfelix (May 16 2019 at 10:55, on Zulip):

right. That's the real problem: Deciding what we want to do.

pnkfelix (May 16 2019 at 10:55, on Zulip):

Which given niko's concerns, seems to me like something that should be floated to the core team.

centril (May 16 2019 at 10:56, on Zulip):

I strongly disagree with that; Lints are explicitly a matter for T-Lang and this impacts us the most.

pnkfelix (May 16 2019 at 10:57, on Zulip):

What?

pnkfelix (May 16 2019 at 10:57, on Zulip):

you honestly think its solely the domain of T-lang?

pnkfelix (May 16 2019 at 10:57, on Zulip):

If you think that is the consequence of it being a lint

pnkfelix (May 16 2019 at 10:58, on Zulip):

...

pnkfelix (May 16 2019 at 10:58, on Zulip):

well, I don't want to argue about it here.

pnkfelix (May 16 2019 at 11:00, on Zulip):

(that is, we can talk about it, but lets make a separate topic for it if we do.)

pnkfelix (May 16 2019 at 11:01, on Zulip):

okay, anyway, regarding #60680: marked as P-medium and left I-nominated. Hopefully that will lead to it getting assigned, and then worked on. (Given that I'm on the hook in the end for NLL's progression, I will assign it to myself if no one else volunteers at the meeting.)

pnkfelix (May 16 2019 at 11:01, on Zulip):

next: "ICE impl_trait_in_bindings with elided lifetime" #60473

centril (May 16 2019 at 11:02, on Zulip):

(that is, we can talk about it, but lets make a separate topic for it if we do.)

I think this is something we should discuss at the t-lang meeting.

pnkfelix (May 16 2019 at 11:02, on Zulip):

#60473 triage: P-medium, removing nomination tag.

pnkfelix (May 16 2019 at 11:04, on Zulip):

next: "error: internal compiler error: src/librustc/ty/query/plumbing.rs:1195: Cannot force dep node: coherent_trait(core[30a9]::ops[0]::drop[0]::Drop[0])" #59731

pnkfelix (May 16 2019 at 11:06, on Zulip):

incremental compilation bug. Does seem to have a nicely described example for reproduction.

pnkfelix (May 16 2019 at 11:08, on Zulip):

going to prioritize as P-high for now, under the theory that a well-documented incr-comp bug will be easier to squash and thus have very high value.

pnkfelix (May 16 2019 at 11:09, on Zulip):

next: "rls no longer builds after rust-lang/rust#60679" #60768

pnkfelix (May 16 2019 at 11:11, on Zulip):

posted Q about whether this should be closed as fixed

pnkfelix (May 16 2019 at 11:11, on Zulip):

in terms of prioritization, i'm pretty sure its P-medium.

pnkfelix (May 16 2019 at 11:14, on Zulip):

next: "ICE: thread 'main' panicked at 'index out of bounds: the len is 0 but the index is 1', /usr/src/rustc-1.32.0/src/libcore/slice/mod.rs:2463:10" #60821

pnkfelix (May 16 2019 at 11:16, on Zulip):

the dialogue there leads me to think that this is hard to reproduce, and was only witnessed on an relatively old version of the compiler. :sad:

pnkfelix (May 16 2019 at 11:17, on Zulip):

I'm going to leave this nominated and try to discuss it at the meeting today, in terms of getting feedback on whether we have any real hope of trying to resolve bugs like this.

pnkfelix (May 16 2019 at 11:21, on Zulip):

next: "linker error "undefined symbol: __aeabi_uidiv" for target thumbv6m-none-eabi" #60782

pnkfelix (May 16 2019 at 11:22, on Zulip):

do we know if PR rust-lang-nursery/compiler-builtins#290 fixed this?

pnkfelix (May 16 2019 at 11:23, on Zulip):

there is related bug "Cannot create rv64gc binary anymore since nightly-2019-05-10" #60747

scalexm (May 16 2019 at 11:23, on Zulip):

@pnkfelix (re: wg-traits lead, I dont think so, I thought @Aaron Turon was supposed to be co-leading with @nikomatsakis though)

pnkfelix (May 16 2019 at 11:23, on Zulip):

@scalexm okay, I'll look into updating that WG-traits web page. Thanks!

pnkfelix (May 16 2019 at 11:25, on Zulip):

I'll mark #60782 and #60747 as P-high. I don't know for sure that this target is actually a sufficiently high tier platform to warrant that prioritization, but lets start high and deprioritize later if it turns out that the compiler-builtins PR didn't fix it.

centril (May 16 2019 at 11:25, on Zulip):

Well Aaron is on leave, so that can't be true anymore

mw (May 16 2019 at 11:25, on Zulip):

I can do a quick update on WG-pgo, yes

pnkfelix (May 16 2019 at 11:28, on Zulip):

@mw great, thanks!

pnkfelix (May 16 2019 at 11:34, on Zulip):

last: "forego caching cycles leads to a severe perf regression" #60846

pnkfelix (May 16 2019 at 11:35, on Zulip):

a 16x performance loss sounds like its pretty important to investigate. P-high.

pnkfelix (May 16 2019 at 11:36, on Zulip):

okay, that was all the nominated unpriortized T-compiler issues.

pnkfelix (May 16 2019 at 11:37, on Zulip):

there are no unprioritized beta-regressions for us.

pnkfelix (May 16 2019 at 11:37, on Zulip):

and there are no unprioritized nightly regressions for us.

pnkfelix (May 16 2019 at 11:38, on Zulip):

we have 28 open P-high T-compiler issues.

pnkfelix (May 16 2019 at 11:38, on Zulip):

I think I am going to revise the agenda slightly

pnkfelix (May 16 2019 at 11:39, on Zulip):

and actually separate assigned from unassigned issues, if I can. let me see.

pnkfelix (May 16 2019 at 11:41, on Zulip):

there's an easy way to search for all unassigned issues, but I don't see a way to view solely the assigned ones.

pnkfelix (May 16 2019 at 11:42, on Zulip):

okay, so, never mind that idea.

pnkfelix (May 16 2019 at 11:42, on Zulip):

I'll just add an (optional) link for unassigned to the agenda

pnkfelix (May 16 2019 at 11:44, on Zulip):

hmm. We have the triage link set up to sort by updated-desc. but I might think updated-asc could make more sense, in that we would want to encourage visiting issues that have otherwise been overlooked... (Update: actually, I don't think the link specifies any sort. I don't know whether its some setting of my own that's making github do that, or a site default...)

pnkfelix (May 16 2019 at 11:44, on Zulip):

(this is probably something to discuss at the Steering meeting on our Issue Triage strategy...)

pnkfelix (May 16 2019 at 11:51, on Zulip):

anyway, just to mix things up, I'm going to try to iterate through the P-high T-compiler issues now by sorting by updated-asc

pnkfelix (May 16 2019 at 11:53, on Zulip):

first up: "Compiler panic when using a slice pattern" #59016

pnkfelix (May 16 2019 at 11:54, on Zulip):

this is waiting on PR #59369, which is marked as S-waiting-on-author. Last update from @oli was 23 days ago.

pnkfelix (May 16 2019 at 11:54, on Zulip):

@oli , if you get a chance, maybe write a note about your plans regarding PR #59369 on that PR?

pnkfelix (May 16 2019 at 11:55, on Zulip):

I also am having trouble believing that the issue #59016 can really be P-high, given how relatively low priority its associated PR has gottne.

pnkfelix (May 16 2019 at 11:56, on Zulip):

I'm going to lower its priority to P-medium.

pnkfelix (May 16 2019 at 11:58, on Zulip):

next: "ICE in librustc_codegen_llvm when building kernel" #59548

centril (May 16 2019 at 11:59, on Zulip):

btw, https://github.com/rust-lang/rust/issues/60479#issuecomment-493029604 is no longer nominated for t-compiler but with the new details it should be?

pnkfelix (May 16 2019 at 11:59, on Zulip):

I continue to believe that #59548 is a (known) limitation of #[linkage]

pnkfelix (May 16 2019 at 12:00, on Zulip):

I don't know if we're going to readily find volunteers to implement the feature requested in issue #60479

pnkfelix (May 16 2019 at 12:00, on Zulip):

but I can believe we should at least float it for discussion

pnkfelix (May 16 2019 at 12:00, on Zulip):

I'll tag and nominate.

centril (May 16 2019 at 12:00, on Zulip):

Would be nice to have it implemented so we don't repeat these mistakes :slight_smile:

pnkfelix (May 16 2019 at 12:01, on Zulip):

sure, but the team that would benefit the most explicitly did not volunteer.

centril (May 16 2019 at 12:01, on Zulip):

@pnkfelix maybe an E-easy issue?

centril (May 16 2019 at 12:01, on Zulip):

(T-release would also benefit I think)

pnkfelix (May 16 2019 at 12:02, on Zulip):

maybe. Maybe a mentor issue also.

pnkfelix (May 16 2019 at 12:04, on Zulip):

I'll assign isue #59548 to myself, with the attention of figuring out if this is indeed a known limitation of #[linkage] and if so, attempting to replace the ICE with a more sensible error message.

pnkfelix (May 16 2019 at 12:10, on Zulip):

next: "Compiler panic with generic-typed nested closures" #59494

pnkfelix (May 16 2019 at 12:12, on Zulip):

this really would be nice to get assigned. I'll post a request for a volunteer in the main meeting topic

pnkfelix (May 16 2019 at 12:17, on Zulip):

next: "assertion failure with src/test/ui/pattern/const-pat-ice.rs" #59996

pnkfelix (May 16 2019 at 12:17, on Zulip):

I triaged this ancient bug as P-high 28 days ago, with the caveat that I wanted someone to talk me out of downgrading it. No one has. I'm going to downgrade to P-medium and also assign to myself so I do not lose track of it.

pnkfelix (May 16 2019 at 12:18, on Zulip):

next: "ICE running clippy on embedded crate in release mode" #59898

pnkfelix (May 16 2019 at 12:21, on Zulip):

we did discuss 28 days ago whether to backport PR #58605 to stable and declined to do so. There was some pushback regarding that decision but I didn't see that comment myself and I don't know if anyone else saw it either.

pnkfelix (May 16 2019 at 12:22, on Zulip):

in any case that same comment also suggested we expand the test suite for thumb-* ...

pnkfelix (May 16 2019 at 12:22, on Zulip):

and there was also a question there about whether cargo-check and cargo-clippy are "included in the general Rust stability guarantee"

pnkfelix (May 16 2019 at 12:24, on Zulip):

(i don't really know how best to respond to that question. Like, software has bugs, and developers have to measure tradeoffs. The author of the patch thought the trade-off wasn't worth it. Not sure what else there is to say, apart from whether other teams like the T-release team should potentially have the authority to override T-compiler when it comes to backports?)

centril (May 16 2019 at 12:29, on Zulip):

T-release team should potentially have the authority to override T-compiler when it comes to backports?

Speaking as a T-release member, but not for the team, no please :slight_smile:

pnkfelix (May 16 2019 at 12:30, on Zulip):

I will however note, on the subject of the hypothesized backport of PR #58605 ...

pnkfelix (May 16 2019 at 12:31, on Zulip):

... that there were comments to the effect of "this PR wouldn't justify a point release. but maybe it should be included if there is a point release."

pnkfelix (May 16 2019 at 12:31, on Zulip):

and then it looks like we've had two point releases...

simulacrum (May 16 2019 at 12:32, on Zulip):

the second fwiw was intended as a security release and we didn't consider including other things in it

simulacrum (May 16 2019 at 12:32, on Zulip):

primarily to make sure it went well from my side of things

pnkfelix (May 16 2019 at 12:32, on Zulip):

ah okay thanks @simulacrum

pnkfelix (May 16 2019 at 12:32, on Zulip):

that's a very good point, I had forgotten about the motivation for 1.34.2, but now I remember it (the Error::type_id thing)

centril (May 16 2019 at 12:32, on Zulip):

yeah, it was particularly "we need to get this security release out the door immediately"

Pietro Albini (May 16 2019 at 12:33, on Zulip):

what I personally think for stable backports is the teams should nominate and approve things they think are safe for backport, and the release team decide whether to make the release or not based on the approved stuff

pnkfelix (May 16 2019 at 12:33, on Zulip):

but I think the stable-nomination of PR #58605 even predated 1.34.1 ... is that right?

Pietro Albini (May 16 2019 at 12:33, on Zulip):

yep

Pietro Albini (May 16 2019 at 12:33, on Zulip):

1.34.1 was a bit weird, but shrugs

pnkfelix (May 16 2019 at 12:33, on Zulip):

what I personally think for stable backports is the teams should nominate and approve things they think are safe for backport, and the release team decide whether to make the release or not based on the approved stuff

I actually personally agree with this. The whole "lets backport if there is a point release" does not really make sense to me.

pnkfelix (May 16 2019 at 12:34, on Zulip):

maybe we just need to settle on the actual policy there and codify it.

pnkfelix (May 16 2019 at 12:34, on Zulip):

I'll make a note in the meeting topic.

Pietro Albini (May 16 2019 at 12:34, on Zulip):

(please ping me if you end up discussing it at the meeting)

pnkfelix (May 16 2019 at 12:36, on Zulip):

anyway, it seems to me like we are not going to backport anything for #59898 at this point.

Pietro Albini (May 16 2019 at 12:36, on Zulip):

it's hard to make a 1.34.3 literally one week before 1.35.0 :D

pnkfelix (May 16 2019 at 12:37, on Zulip):

next: " ICE in src/librustc/ty/query/plumbing.rs when compiling incrementally" #59716

pnkfelix (May 16 2019 at 12:38, on Zulip):

under the assumption that #59716 is fixed, I'm going to downgrade it to P-medium. For now I'll continue to leave it open (as E-needstest) and assigned to self to investigate said test, as previously commented on that issue.

pnkfelix (May 16 2019 at 12:42, on Zulip):

next: "Incorporate RLS bug tracking into compiler team triage" #58858

pnkfelix (May 16 2019 at 12:42, on Zulip):

there's been lip service paid here, in that links have been added to the agenda page for this meeting

pnkfelix (May 16 2019 at 12:42, on Zulip):

but its at the bottom of the agenda

pnkfelix (May 16 2019 at 12:42, on Zulip):

which, for now, means it is very unlikely to actually get visited.

pnkfelix (May 16 2019 at 12:43, on Zulip):

the whole process here is something that we will probably discuss at the steering meeting regarding issue triage.

pnkfelix (May 16 2019 at 12:43, on Zulip):

next: "ICE on higher-trait bounds" #60218

pnkfelix (May 16 2019 at 12:44, on Zulip):

@nikomatsakis if you have anything to report on #60218, it would be great if you could leave a note in the issue. If you think it is not realistic for you to look into this in the next two weeks, let us know and we can attempt to reassign.

pnkfelix (May 16 2019 at 12:44, on Zulip):

also, if #60218 should not be P-high, that would be good to know too.

pnkfelix (May 16 2019 at 12:45, on Zulip):

next: "reached the type-length limit while instantiating" #58952

pnkfelix (May 16 2019 at 12:45, on Zulip):

no one is assigned to issue #58952. We have a suspicion that this is a T-libs issue rather than a T-compiler one, but have not confirmed this.

pnkfelix (May 16 2019 at 12:50, on Zulip):

I'm going to assume that #58952 is a duplicate of #54540. In turn, #54540 ("Exponential compile-time and type_length_limit blowup when nesting closure wrappers") is, for better or for worse, P-medium (but also I-nominated).

pnkfelix (May 16 2019 at 12:51, on Zulip):

so I'm going to remove the nominated tag from #58952 and downgrade it to P-medium.

pnkfelix (May 16 2019 at 12:52, on Zulip):

however, I'm worried that #54540 is going to get lost in the shuffle of P-medium bugs

pnkfelix (May 16 2019 at 12:52, on Zulip):

(it is I-nominated at least...)

pnkfelix (May 16 2019 at 12:54, on Zulip):

@eddyb 's comment here on #54540 is really good.

pnkfelix (May 16 2019 at 12:54, on Zulip):

I'm going to post a request for volunteer to investigate to the main meeting area.

eddyb (May 16 2019 at 12:56, on Zulip):

(apologies for not doing anything on that. wg-grammar got me working on parser generator stuff)

pnkfelix (May 16 2019 at 12:56, on Zulip):

its okay, it doesn't need to be just your problem.

pnkfelix (May 16 2019 at 12:56, on Zulip):

and it looks like an interesting one to resolve.

eddyb (May 16 2019 at 12:56, on Zulip):

I think it came up in meeting last week and I said something but I forget what

eddyb (May 16 2019 at 12:58, on Zulip):

or maybe it was some random topic on here. either way, I think we should be smarter about computing this value

eddyb (May 16 2019 at 12:59, on Zulip):

even if we compute the same value as today, we should do it faster. and I might've said I will try to implement that

pnkfelix (May 16 2019 at 12:59, on Zulip):

hey @Pietro Albini , regarding the continuing thread on #58605, ... we don't want to have continued discussion on that PR, right? I was about to post a comment recommending to jamesmunns that they open a fresh issue regarding the Tier status of Thumb, but then I realized I wasn't sure where would be best to have that discussion.

pnkfelix (May 16 2019 at 13:00, on Zulip):

next: "llvm lint: "Undefined behavior: Call argument type mismatches callee parameter type" with mixing debug and release" #48310

pnkfelix (May 16 2019 at 13:00, on Zulip):

needs someone to followup. Will post request for help.

centril (May 16 2019 at 13:01, on Zulip):

but then I realized I wasn't sure where would be best to have that discussion.

Or who to have it with? Seems to me like tier bumping is a shared concern between libs, infra, compiler, and lang

pnkfelix (May 16 2019 at 13:02, on Zulip):

but then I realized I wasn't sure where would be best to have that discussion.

Or who to have it with? Seems to me like tier bumping is a shared concern between libs, infra, compiler, and lang

all the more reason to try to migrate the conversation away from a T-compiler PR ?

centril (May 16 2019 at 13:02, on Zulip):

@pnkfelix sure

pnkfelix (May 16 2019 at 13:02, on Zulip):

but yes, the point remains: Do I advise them to open an issue?

centril (May 16 2019 at 13:03, on Zulip):

@pnkfelix sgtm

pnkfelix (May 16 2019 at 13:03, on Zulip):

or is that just adding noise to our system and is unlikely to actually get the problem resolved?

centril (May 16 2019 at 13:04, on Zulip):

@pnkfelix issues come cheap... they grow steadily every day with no prospect of ever shrinking in total so I wouldn't lose any sleep over adding extra noise

pnkfelix (May 16 2019 at 13:07, on Zulip):

filed "Should Thumb bugs get increased visibility, e.g. Tier 1 status?" #60880

pnkfelix (May 16 2019 at 13:08, on Zulip):

next: "ICE when running kcov with proptest as dev-dependency" #60372

pnkfelix (May 16 2019 at 13:09, on Zulip):

@nikomatsakis would you be willing to take on the task of verifying (or at least increasing confidence) that #60372 is a duplicate of #58375 (where the latter you self-assigned last week) ?

pnkfelix (May 16 2019 at 13:10, on Zulip):

next: "ICE with unsized associated type" #60431

pnkfelix (May 16 2019 at 13:10, on Zulip):

I'll post a request in general meeting area

pnkfelix (May 16 2019 at 13:12, on Zulip):

next: "ICE with existential_type feature #60407

pnkfelix (May 16 2019 at 13:25, on Zulip):

downgrading to P-medium (now that I've explicitly noted that this ICE should block stabilization of #34511 (impl Trait)

centril (May 16 2019 at 13:26, on Zulip):

@pnkfelix I can auto-P-medium all "unstable buggy features that don't cause an ICE without gate" for you if you like

pnkfelix (May 16 2019 at 13:27, on Zulip):

its tough because my gut I know some of these ICE's are cases of real bugs that can be witnessed without the feature

pnkfelix (May 16 2019 at 13:27, on Zulip):

I don't know how many of them fall into that category though

centril (May 16 2019 at 13:28, on Zulip):

(As for existential_type, we'll need a concerted effort to fix it and that will likely have to wait until after async/await is stabilized on nightly)

centril (May 16 2019 at 13:29, on Zulip):

(and ofc we need the syntax RFC to get merged)

pnkfelix (May 16 2019 at 13:29, on Zulip):

anyway, for now, its probably best to not auto-P-medium anything

centril (May 16 2019 at 13:30, on Zulip):

ok

pnkfelix (May 16 2019 at 13:30, on Zulip):

I mean, I guess I myself am auto P-medium'ing things during this pre-triage process

Pietro Albini (May 16 2019 at 13:41, on Zulip):

we don't want to have continued discussion on that PR, right? I was about to post a comment recommending to jamesmunns that they open a fresh issue regarding the Tier status of Thumb, but then I realized I wasn't sure where would be best to have that discussion.

@pnkfelix eh, the issue is there is no process for promoting a target to tier 1

Pietro Albini (May 16 2019 at 13:42, on Zulip):

like what are the requirements for it? who needs to approve the tier change?

pnkfelix (May 16 2019 at 14:00, on Zulip):

feel free to generalize the issue accordingly. :wink: Or open a new one.

Last update: Nov 16 2019 at 01:40UTC