Stream: t-compiler

Topic: weekly meeting 2019-05-16 #54818


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

Hi @T-compiler/meeting ; the triage meeting will be started in this topic in about an 1 hour and 45 minutes. In the meantime, I'll be throwing small notes here that come from the pre-triage topic

pnkfelix (May 16 2019 at 12:14, on Zulip):

today we will be having WG-checkin from WG-pgo via @mw

pnkfelix (May 16 2019 at 12:15, on Zulip):

I don't know yet we'll have a second WG checkin (you can see my callout here).

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

:construction_worker: request for investigation: "Compiler panic with generic-typed nested closures" #59494 needs an assignee.

pnkfelix (May 16 2019 at 12:35, on Zulip):

:info: we probably need to codify our rules about what should (and should not) drive the backport decisions. See in particular this

pnkfelix (May 16 2019 at 12:55, on Zulip):

:construction_worker: request for investigate: "Exponential compile-time and type_length_limit blowup when nesting closure wrappers" #54540 needs an assignee. @eddyb has excellent notes here on the fundamental problem (type duplication during a walk performed by type_length_limit)

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

:construction_worker: request for investigation: "llvm lint: "Undefined behavior: Call argument type mismatches callee parameter type" with mixing debug and release" #48310; specifically, we need to resolve whether this is in fact an LLVM bug.

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

(maybe I can find someone in @WG-llvm to do this?)

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

:construction_worker: request for investigation: "ICE with unsized associated type" #60431. ICE itself is internal compiler error: src/librustc_codegen_llvm/context.rs:867: failed to get layout for &Ref<u8, Obstack>: the type <Obstack as Arena>::Dyn has an unknown layout

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

hello @T-compiler/meeting , meeting starts now!

pnkfelix (May 16 2019 at 14:01, on Zulip):

Following our agenda (#54818), we'll devote the first five to ten minutes to ad-hoc announcements.

pnkfelix (May 16 2019 at 14:02, on Zulip):

(you can see the ones I've been adding up above)

pnkfelix (May 16 2019 at 14:05, on Zulip):

okay it seems like no one wants to jump in

pnkfelix (May 16 2019 at 14:06, on Zulip):

just as a heads up: I listed some "requests to invesigate" up above, but in total we have twelve P-high unassigned issues

pnkfelix (May 16 2019 at 14:07, on Zulip):

please do review them and see if there's any you might want to take point on

mw (May 16 2019 at 14:07, on Zulip):

https://github.com/rust-lang/rust/issues/59731 is likely already fixed

pnkfelix (May 16 2019 at 14:08, on Zulip):

well then that's a great candidate for anyone who wants to just verify that the bug is fixed

mw (May 16 2019 at 14:08, on Zulip):

I assigned myself

pnkfelix (May 16 2019 at 14:08, on Zulip):

(I marked that one as P-high in part because it has a very well written out steps-to-reproduce, which is always welcome when it comes to A-incr-comp bugs...)

nikomatsakis (May 16 2019 at 14:09, on Zulip):

:info: we probably need to codify our rules about what should (and should not) drive the backport decisions. See in particular this

I don't quite follow this .. oh, I see

pnkfelix (May 16 2019 at 14:09, on Zulip):

there are a couple of bugs that I do think I'll assign to myself if they aren't otherwise taken, but I don't want to do that until I've maximized the amount of time/guilt I've tried to give to others to take the issues. :wink:

nikomatsakis (May 16 2019 at 14:09, on Zulip):

I thought the "this" was part of the :info: emoji :)

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

lets move along (if there's time at end maybe I'll circle back to the unassigned issues Q)

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

beta-nominations

pnkfelix (May 16 2019 at 14:12, on Zulip):

just one: "save-analysis: Pull associated type definition using qpath_def" #59894

centril (May 16 2019 at 14:12, on Zulip):

I thought the "this" was part of the :info: emoji :slight_smile:

Me as well :P

varkor (May 16 2019 at 14:12, on Zulip):

I think there's another that maybe wasn't marked T-compiler

varkor (May 16 2019 at 14:12, on Zulip):

let me find it quickly

pnkfelix (May 16 2019 at 14:12, on Zulip):

I thought the "this" was part of the :info: emoji :)

Me as well :P

(one of you can perhaps enlighten me as to what you mean by this in a privmsg)

pnkfelix (May 16 2019 at 14:13, on Zulip):

regarding PR #59894

pnkfelix (May 16 2019 at 14:13, on Zulip):

the beta is being cut, like, next week

pnkfelix (May 16 2019 at 14:13, on Zulip):

right?

Pietro Albini (May 16 2019 at 14:13, on Zulip):

yep

pnkfelix (May 16 2019 at 14:13, on Zulip):

like, maybe we should have considered this for beta-nom back in April

varkor (May 16 2019 at 14:13, on Zulip):

https://github.com/rust-lang/rust/pull/60710

Pietro Albini (May 16 2019 at 14:13, on Zulip):

we still have time for backports approved today

centril (May 16 2019 at 14:13, on Zulip):

@pnkfelix tuesday to be exact

pnkfelix (May 16 2019 at 14:13, on Zulip):

but if we approve PR #59894 for beta-backport today

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

we're essentially really backporting to stable, in the sense that we won't have time to undo it

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

right?

centril (May 16 2019 at 14:14, on Zulip):

pretty much

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

having said that

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

this seems low risk...

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

oh I forgot to add the voting emojis, let me add them now

Pietro Albini (May 16 2019 at 14:14, on Zulip):

we could undo it

Pietro Albini (May 16 2019 at 14:15, on Zulip):

if it lands between today and tomorrow and we notice everything is broken we can revert the backport in the beta->stable promotion

centril (May 16 2019 at 14:15, on Zulip):

@Pietro Albini huh?

pnkfelix (May 16 2019 at 14:15, on Zulip):

well, via a point release you mean?

nikomatsakis (May 16 2019 at 14:15, on Zulip):

well it's been on nightly for a while

nikomatsakis (May 16 2019 at 14:15, on Zulip):

so presumably it's been tested

pnkfelix (May 16 2019 at 14:15, on Zulip):

oh, you mean we could undo it in the super-short term

nikomatsakis (May 16 2019 at 14:15, on Zulip):

I trust @Igor Matuszewski on this one

Pietro Albini (May 16 2019 at 14:15, on Zulip):

yep

pnkfelix (May 16 2019 at 14:15, on Zulip):

I meant more "we won't get the usual six week cycle to undo."

Pietro Albini (May 16 2019 at 14:15, on Zulip):

oh yeah, not that

centril (May 16 2019 at 14:16, on Zulip):

(Well to be fair it's usually never the full 6 weeks unless it is like directly the day after master=>beta promotion)

pnkfelix (May 16 2019 at 14:17, on Zulip):

okay. I'm willing to go along with 1. its been on nightly for a while, and (perhaps more importantly) 2. its super-low risk since its its "just" in the save_analysis crate, which is not the most critical bit of infrastructure in the compiler, right?

pnkfelix (May 16 2019 at 14:18, on Zulip):

@centril sure, but there's some argument that you usually get X days on nightly before the backport itself occurs, and then Y days on beta.

pnkfelix (May 16 2019 at 14:18, on Zulip):

(this is assuming that the semantics of the original patch actually matches that of the backport)

centril (May 16 2019 at 14:18, on Zulip):

Sure; just saying that depending on when backport happens you get varying amounts of time

pnkfelix (May 16 2019 at 14:18, on Zulip):

okay anyway cool lets go for it

centril (May 16 2019 at 14:19, on Zulip):

(which I suppose is an obvious boring statement)

nikomatsakis (May 16 2019 at 14:19, on Zulip):

Yesterday, nrc was saying that there is some sort of clippy problem that is wreaking havoc around the release

nikomatsakis (May 16 2019 at 14:19, on Zulip):

That is, a change to rustc that broke clippy, but only in the context of RLS, or something like that?

pnkfelix (May 16 2019 at 14:19, on Zulip):

hold on

pnkfelix (May 16 2019 at 14:19, on Zulip):

um

nikomatsakis (May 16 2019 at 14:20, on Zulip):

Do we have a PR en route to fixing this? Is that related to this backport?

pnkfelix (May 16 2019 at 14:20, on Zulip):

is this a shift in topic?

pnkfelix (May 16 2019 at 14:20, on Zulip):

or related to these backports?

nikomatsakis (May 16 2019 at 14:20, on Zulip):

Does anyone have any idea what i'm talking about? :)

nikomatsakis (May 16 2019 at 14:20, on Zulip):

That's what I'm trying to establish

centril (May 16 2019 at 14:20, on Zulip):

@nikomatsakis is this related to https://github.com/rust-lang/rust/pull/60861#issuecomment-492789936 ?

nikomatsakis (May 16 2019 at 14:20, on Zulip):

quite possibly

nikomatsakis (May 16 2019 at 14:20, on Zulip):

anyway, ok, I'll try to get more data, carry on

pnkfelix (May 16 2019 at 14:21, on Zulip):

okay

pnkfelix (May 16 2019 at 14:21, on Zulip):

so a second beta-nom was JIT'ed

pnkfelix (May 16 2019 at 14:21, on Zulip):

Use delay_span_bug for error cases when checking AnonConst parent #60710

varkor (May 16 2019 at 14:21, on Zulip):

this is https://github.com/rust-lang/rust/issues/60848, I think

varkor (May 16 2019 at 14:21, on Zulip):

which is linked to in Centril's link

varkor (May 16 2019 at 14:22, on Zulip):

#60710 just downgrades some bug!s to delay_span_bugs

centril (May 16 2019 at 14:22, on Zulip):

Use delay_span_bug for error cases when checking AnonConst parent #60710

Seems risk free? But also idk if it is important

varkor (May 16 2019 at 14:23, on Zulip):

it causes ICEs in code with easy typos to make

centril (May 16 2019 at 14:23, on Zulip):

Fair enough

varkor (May 16 2019 at 14:23, on Zulip):

there's an example in the original issue

pnkfelix (May 16 2019 at 14:23, on Zulip):

yeah I have no problem with backporting this

pnkfelix (May 16 2019 at 14:23, on Zulip):

The "<32> should be <u32>" is notable.

pnkfelix (May 16 2019 at 14:24, on Zulip):

(as in, that's the easy typo)

pnkfelix (May 16 2019 at 14:24, on Zulip):

okay great lets do it

pnkfelix (May 16 2019 at 14:25, on Zulip):

okay what else what else

pnkfelix (May 16 2019 at 14:25, on Zulip):

there aren't anything PR's marked S-waiting-on-team for us

pnkfelix (May 16 2019 at 14:25, on Zulip):

there are ten I-nominated issues for us.

pnkfelix (May 16 2019 at 14:26, on Zulip):

lets try to run through these

pnkfelix (May 16 2019 at 14:26, on Zulip):

first: "Panics in destructors can cause the return value to be leaked" #47949

pnkfelix (May 16 2019 at 14:27, on Zulip):

I guess you all discussed this last week as well?

pnkfelix (May 16 2019 at 14:27, on Zulip):

oh oh, it was in lang team meeting

pnkfelix (May 16 2019 at 14:27, on Zulip):

is this solely nominated for lang team, @nikomatsakis ?

centril (May 16 2019 at 14:27, on Zulip):

Probably shouldn't be nominated for t-compiler (or t-lang for that matter since we already discussed it)?

pnkfelix (May 16 2019 at 14:27, on Zulip):

(@nikomatsakis had comment explicitly saying they were leaving nominated for followup meeting)

pnkfelix (May 16 2019 at 14:28, on Zulip):

but

pnkfelix (May 16 2019 at 14:28, on Zulip):

I don't think this is for us in this meeting

pnkfelix (May 16 2019 at 14:28, on Zulip):

so lets move along

centril (May 16 2019 at 14:28, on Zulip):

oh he did, guess I forgot :P

pnkfelix (May 16 2019 at 14:28, on Zulip):

next: "Exponential compile-time and type_length_limit blowup when nesting closure wrappers" #54540

nikomatsakis (May 16 2019 at 14:28, on Zulip):

(I don't really recall)

pnkfelix (May 16 2019 at 14:28, on Zulip):

see in particular @eddyb 's comment here

pnkfelix (May 16 2019 at 14:29, on Zulip):

I suspect we can do much better here

pnkfelix (May 16 2019 at 14:29, on Zulip):

@eddyb mentioned that you all may have discussed this last week, or something related to it

pnkfelix (May 16 2019 at 14:29, on Zulip):

but I'd like to see if someone wants to try to take this on

pnkfelix (May 16 2019 at 14:29, on Zulip):

@eddyb also commented that they had ideas on how to tackle it

pnkfelix (May 16 2019 at 14:29, on Zulip):

but I think @eddyb already has too much on their plate as it is

pnkfelix (May 16 2019 at 14:30, on Zulip):

so I'd love if someone wanted to try to work with @eddyb to resolve this.

pnkfelix (May 16 2019 at 14:30, on Zulip):

I guess I already posted this ticket with a :construction_worker: above

pnkfelix (May 16 2019 at 14:30, on Zulip):

so ... its already gotten twice as much attention as any other bug

nikomatsakis (May 16 2019 at 14:31, on Zulip):

=)

pnkfelix (May 16 2019 at 14:31, on Zulip):

if no one wants to speak up about it, then I'll assign it to myself and unnominate it.

pnkfelix (May 16 2019 at 14:31, on Zulip):

next nomination: "Decouple nightly RLS from Clippy" #59761

eddyb (May 16 2019 at 14:31, on Zulip):

(random idea: can we make the Ty::walk algorithm use a TypeVisitor that pushes to a queue without recursing, in visit_ty? instead of having a copy of some of the visiting logic)

reason: it would make the solution to #54540 much simpler, since we could copy Ty::walk then and add a cache to it. then again, doing the cache with breadth-first searching seems hard, ugh

nikomatsakis (May 16 2019 at 14:33, on Zulip):

next nomination: "Decouple nightly RLS from Clippy" #59761

I think I meant to un-nominate that -- though I think this might be an interesting topic for a meeting.

pnkfelix (May 16 2019 at 14:33, on Zulip):

next nomination: "Decouple nightly RLS from Clippy" #59761

Its not clear to me whether we have action items here for ourselves

oli (May 16 2019 at 14:33, on Zulip):

I think we should unnominate #59761 there's not much to discuss there... oh...

nikomatsakis (May 16 2019 at 14:33, on Zulip):

But it's kind of a "cross-team" thing

pnkfelix (May 16 2019 at 14:34, on Zulip):

yeah I nominated it to try to ensure that we did discuss it

pnkfelix (May 16 2019 at 14:34, on Zulip):

but at this point, I think we need more direction before we can/would do anything

pnkfelix (May 16 2019 at 14:35, on Zulip):

next: "Proc macro errors can lead to rustc panics on non-Linux" #59998

pnkfelix (May 16 2019 at 14:35, on Zulip):

OMG

pnkfelix (May 16 2019 at 14:36, on Zulip):

I love macros, but bugs like this remind me of what a nightmare they can be

pnkfelix (May 16 2019 at 14:37, on Zulip):

@eddyb the actual action item here is ... revise our handling of panics on server-side?

eddyb (May 16 2019 at 14:37, on Zulip):

less "revise" and more "throw another copy of the suppressor hack, on the server too"

pnkfelix (May 16 2019 at 14:38, on Zulip):

okay.

pnkfelix (May 16 2019 at 14:38, on Zulip):

Is anyone here interested in taking point on this?

Zoxc (May 16 2019 at 14:38, on Zulip):

I do get these errors when running tests locally. I wonder why appveyor doesn't spot them.

eddyb (May 16 2019 at 14:38, on Zulip):

(the hack involves registering a panic handler that just doesn't print the panic, in certain situations)

eddyb (May 16 2019 at 14:39, on Zulip):

I wish we had a better way to catch panics "silently" than this hack, tbh

pnkfelix (May 16 2019 at 14:39, on Zulip):

does appveyor not run these tests itself?

eddyb (May 16 2019 at 14:39, on Zulip):

such as a version of catch_panic that specifically prevents output from being emitted

eddyb (May 16 2019 at 14:39, on Zulip):

silently_catch_panic :P

pnkfelix (May 16 2019 at 14:39, on Zulip):

anyway

pnkfelix (May 16 2019 at 14:39, on Zulip):

this bug actually has tickled my fancy

pnkfelix (May 16 2019 at 14:39, on Zulip):

I'm going to assign it to myself

pnkfelix (May 16 2019 at 14:40, on Zulip):

even though I'm a little worried that its going to fall down the P-medium hole

pnkfelix (May 16 2019 at 14:40, on Zulip):

(should it be P-high ??)

eddyb (May 16 2019 at 14:40, on Zulip):

(feel free to bother me about it, I have had a tab open about it since it was open, lol)

eddyb (May 16 2019 at 14:40, on Zulip):

@pnkfelix I think it just results in more output than desired, in cases when the proc macro panicked (so compilation would've failed anyway)

eddyb (May 16 2019 at 14:40, on Zulip):

so P-high doesn't seem necessary

pnkfelix (May 16 2019 at 14:41, on Zulip):

okay.

pnkfelix (May 16 2019 at 14:41, on Zulip):

next: "Rust 1.34 generates significantly less debug information for libstd functions vs. Rust 1.33" #60020

pnkfelix (May 16 2019 at 14:41, on Zulip):

the nomination here is quite old

mw (May 16 2019 at 14:41, on Zulip):

yeah, I should look into that

pnkfelix (May 16 2019 at 14:42, on Zulip):

@mw maybe just post an update on this bug. But I don't think it warrants discussing at this meeting right now.

mw (May 16 2019 at 14:42, on Zulip):

but I don't think it should still be nominated

pnkfelix (May 16 2019 at 14:42, on Zulip):

yeah sorry; this is something I should deal with at pre-triage

mw (May 16 2019 at 14:43, on Zulip):

np :)

pnkfelix (May 16 2019 at 14:43, on Zulip):

next: "Cargo build ICE when run with closed STDIN" #60447

pnkfelix (May 16 2019 at 14:43, on Zulip):

I had nominated this to try to keep its visibility up

pnkfelix (May 16 2019 at 14:43, on Zulip):

but I cannot yet bring myself to re-prioritize as P-high

pnkfelix (May 16 2019 at 14:43, on Zulip):

anyone want to argue in favor of P-high for this?

pnkfelix (May 16 2019 at 14:44, on Zulip):

(or better still, just take it as a bug?)

mw (May 16 2019 at 14:44, on Zulip):

I might have run into a related issue where make wasn't able to handle the output fast enough (?)

pnkfelix (May 16 2019 at 14:44, on Zulip):

that sounds to me like a volunteer!

pnkfelix (May 16 2019 at 14:44, on Zulip):

:wink:

mw (May 16 2019 at 14:44, on Zulip):

no way :)

pnkfelix (May 16 2019 at 14:44, on Zulip):

heh

pnkfelix (May 16 2019 at 14:44, on Zulip):

okay I'm going to un-nominate

pnkfelix (May 16 2019 at 14:45, on Zulip):

it just doesn't warrant further attention

pnkfelix (May 16 2019 at 14:45, on Zulip):

next: "Compiler-internal lint for public libcore items not reexported in libstd" #60479

pnkfelix (May 16 2019 at 14:46, on Zulip):

this is a feature request from the libs team

centril (May 16 2019 at 14:46, on Zulip):

@oli JINX! :P

pnkfelix (May 16 2019 at 14:47, on Zulip):

at the very least, it needs prioritization

pnkfelix (May 16 2019 at 14:47, on Zulip):

I can't imagine its P-high.

nikomatsakis (May 16 2019 at 14:47, on Zulip):

clearly not P-high

centril (May 16 2019 at 14:47, on Zulip):

next: "Compiler-internal lint for public libcore items not reexported in libstd" #60479

From a T-release POV this would be nice to have; I prefer not to have to look through the stuff when making the blog post and such

pnkfelix (May 16 2019 at 14:47, on Zulip):

is anyone willing to provide mentorship for it?

mw (May 16 2019 at 14:47, on Zulip):

well, it _is_ another team, so maybe p-high after all?

centril (May 16 2019 at 14:47, on Zulip):

@simulacrum perhaps?

centril (May 16 2019 at 14:48, on Zulip):

I only have marginal tidy experience myself

pnkfelix (May 16 2019 at 14:48, on Zulip):

@mw I dunno, the other team itself explicitly said they weren't willing to volunteer to do it

simulacrum (May 16 2019 at 14:48, on Zulip):

hm?

simulacrum (May 16 2019 at 14:48, on Zulip):

I don't think this is p-high, maybe med

mw (May 16 2019 at 14:48, on Zulip):

/me was joking

pnkfelix (May 16 2019 at 14:48, on Zulip):

and these include people with compiler hacking experience, I believe. So how important can it

pnkfelix (May 16 2019 at 14:48, on Zulip):

oh

pnkfelix (May 16 2019 at 14:48, on Zulip):

okay "just a joke"

mw (May 16 2019 at 14:48, on Zulip):

sorry :)

nikomatsakis (May 16 2019 at 14:48, on Zulip):

I thnk it's P-medium :)

pnkfelix (May 16 2019 at 14:48, on Zulip):

okay lets mark it P-medium and un-nominate.

centril (May 16 2019 at 14:48, on Zulip):

@simulacrum maybe you can provide mentoring instructions for https://github.com/rust-lang/rust/issues/60479 ?

pnkfelix (May 16 2019 at 14:50, on Zulip):

(I'm letting our Nomination enumeration go long today because I only got WG-checkin confirm from a single person.)

pnkfelix (May 16 2019 at 14:50, on Zulip):

((we need to extend the WG-checkin calendar))

pnkfelix (May 16 2019 at 14:50, on Zulip):

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

simulacrum (May 16 2019 at 14:50, on Zulip):

@centril probably not because I don't have lint experience

pnkfelix (May 16 2019 at 14:50, on Zulip):

after seeing how eager everyone has been today to take on issues

pnkfelix (May 16 2019 at 14:50, on Zulip):

I'll just assign #60680 to myself

pnkfelix (May 16 2019 at 14:51, on Zulip):

since I'm in prinicple the one on the hook for NLL deployment

nikomatsakis (May 16 2019 at 14:51, on Zulip):

that one I guess is fairly easy

Pietro Albini (May 16 2019 at 14:51, on Zulip):

remember that crater is broken right now

pnkfelix (May 16 2019 at 14:51, 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

Pietro Albini (May 16 2019 at 14:51, on Zulip):

until https://github.com/rust-lang/rust/pull/60874 is merged

centril (May 16 2019 at 14:51, on Zulip):

Well just make the PR and we can crater when it is available

pnkfelix (May 16 2019 at 14:52, on Zulip):

I left #60821 nominated and unprioritized because at the time of pre-triage, it wasn't clear to me whether there was anything actionable there.

Pietro Albini (May 16 2019 at 14:52, on Zulip):

yeah, just remember to do the try builds after that PR is merged, otherwise you won't get the cargo that fixes crater

pnkfelix (May 16 2019 at 14:52, on Zulip):

but it sounds like the current status on #60821 is that we do not have a ready way to reproduce the bug as described.

mw (May 16 2019 at 14:53, on Zulip):

looks like a bug in incremental compilation

pnkfelix (May 16 2019 at 14:53, on Zulip):

right, but I'm not clear on whether we have enough info to reproduce it

pnkfelix (May 16 2019 at 14:53, on Zulip):

(which has been a big problem for A-incr-comp in the past)

mw (May 16 2019 at 14:54, on Zulip):

yes, I know :(

pnkfelix (May 16 2019 at 14:54, on Zulip):

I wonder if there's something we could do, infrastructurally, to try to address that

pnkfelix (May 16 2019 at 14:54, on Zulip):

e.g., somehow summarize the dep-graph

pnkfelix (May 16 2019 at 14:54, on Zulip):

and snapshot it over time?

mw (May 16 2019 at 14:54, on Zulip):

maybe

pnkfelix (May 16 2019 at 14:54, on Zulip):

with the intentino being that when people report a bug, they'd actually include the before and after

mw (May 16 2019 at 14:54, on Zulip):

sounds non-trivial though

nikomatsakis (May 16 2019 at 14:54, on Zulip):

re: https://github.com/rust-lang/rust/issues/60821, I'm inclined to close

pnkfelix (May 16 2019 at 14:55, on Zulip):

definitely non-trivial

nikomatsakis (May 16 2019 at 14:55, on Zulip):

sooner or later a bug report will come that is actionable

pnkfelix (May 16 2019 at 14:55, on Zulip):

but its also very non-trivial to reproduce these bugs

pnkfelix (May 16 2019 at 14:55, on Zulip):

yeah okay I'll close #60821

nikomatsakis (May 16 2019 at 14:56, on Zulip):

I guess there might be some value in trying to .. nah. I was wondering if there is something to "keep track of" in terms of the general shape of the error message, but who cares really?

nikomatsakis (May 16 2019 at 14:57, on Zulip):

Sorry, that was misinterpreted :)

nikomatsakis (May 16 2019 at 14:57, on Zulip):

What I mean is: would it help us to know "there were 22 ICEs related to incremental that involved out-of-bounds accesses on arrays"

nikomatsakis (May 16 2019 at 14:57, on Zulip):

i.e., if we tracked them in some form of meta bug

nikomatsakis (May 16 2019 at 14:57, on Zulip):

it's not like we have test cases we can go back and re-test once we have a PR

pnkfelix (May 16 2019 at 14:57, on Zulip):

so

pnkfelix (May 16 2019 at 14:58, on Zulip):

@Pietro Albini has pointed out some beta-noms that need our attetion

Esteban Küber (May 16 2019 at 14:58, on Zulip):

We can do so by properly tagging before closing and then reporting on them after the fact

pnkfelix (May 16 2019 at 14:58, on Zulip):

(that we missed due to a lack of T-compiler label on them)

Esteban Küber (May 16 2019 at 14:58, on Zulip):

Add a new tag needsrepro

pnkfelix (May 16 2019 at 14:58, on Zulip):

beta-nom: Use delay_span_bug for "Failed to unify obligation" #60644

pnkfelix (May 16 2019 at 14:59, on Zulip):

beta-nom: "Instead of ICEing on incorrect pattern, use delay_span_bug" #60641

pnkfelix (May 16 2019 at 14:59, on Zulip):

beta-nom: "conditionally modify darwin targets to macosx targets with versions" #60378

pnkfelix (May 16 2019 at 15:00, on Zulip):

(sorry to spam, I know the conversation goes better when we do them piece-meal, but we are also out of time.)

pnkfelix (May 16 2019 at 15:00, on Zulip):

((sorry also to @mw for failing to get around to the WG-pgo checkin)

mw (May 16 2019 at 15:01, on Zulip):

no worries

pnkfelix (May 16 2019 at 15:01, on Zulip):

was PR #60788 also supposed to be beta-nom'ed, @nikomatsakis ?

nikomatsakis (May 16 2019 at 15:02, on Zulip):

yes sort of

nikomatsakis (May 16 2019 at 15:02, on Zulip):

I think we would have to do #60378 and #60788 together or not at all

pnkfelix (May 16 2019 at 15:02, on Zulip):

That is, @nikomatsakis said we would need to backport PR #60788 along with PR #60378. But then @Björn Steinbrink seems to have reported issues with #60788 ... but maybe that was a mistake?

pnkfelix (May 16 2019 at 15:03, on Zulip):

okay then, one more in that case

pnkfelix (May 16 2019 at 15:03, on Zulip):

beta-nom: "default to $ARCH-apple-macosx10.7.0 LLVM triple for darwin targets" #60788

nikomatsakis (May 16 2019 at 15:03, on Zulip):

I'm not sure if I think backporting #603678/#60788 is worth it but it sounds like bjorn3's comment is not a reason not to

nikomatsakis (May 16 2019 at 15:04, on Zulip):

( i.e., that change will come )

pnkfelix (May 16 2019 at 15:04, on Zulip):

so many negatives

pnkfelix (May 16 2019 at 15:04, on Zulip):

/me overlooked the first "not" in "is not a reason not to" when he first read that

nikomatsakis (May 16 2019 at 15:04, on Zulip):

Heh. What I mean is:

nikomatsakis (May 16 2019 at 15:05, on Zulip):

The FF team asked me to nominate #603678 and #60788 because it's blocking them from adopting ThinLTO. I think however that they can use beta -- which is branchign very soon -- so long as they don't use any feature-gates. So maybe it's not blocking them that badly, though they'd prefer to use stable. Also, ThinLTO is an advertised feature that others might want, and you need this fix to use it.

nikomatsakis (May 16 2019 at 15:05, on Zulip):

On the other hand, bjorn3 reported a problem, but it sounds like that problem is not something we plan to fix or react to

centril (May 16 2019 at 15:06, on Zulip):

(Doesn't FF use RUSTC_BOOTSTRAP :big_frown: anyways?)

nikomatsakis (May 16 2019 at 15:06, on Zulip):

The fix itself seems limited in scope so I guess backport seems ok to me

nikomatsakis (May 16 2019 at 15:06, on Zulip):

(Doesn't FF use RUSTC_BOOTSTRAP :big_frown: anyways?)

I think in some very specific cases.

nikomatsakis (May 16 2019 at 15:06, on Zulip):

but their general policy is "no feature gates, beta/stable", but I'm not an expert.

Esteban Küber (May 16 2019 at 15:07, on Zulip):

I feel letting it bake and maybe backporting after next release would give us time to deal with potential fallout in the ecosystem

Wesley Wiser (May 16 2019 at 15:07, on Zulip):

(#603678 is 404-ing for me EDIT: is that supposed to be #60378?)

pnkfelix (May 16 2019 at 15:07, on Zulip):

I don't understand the combo of "if they don't use any feature gates" when their goal is already to use stable.

pnkfelix (May 16 2019 at 15:07, on Zulip):

i.e., aren't they already meeting that constraint?

pnkfelix (May 16 2019 at 15:07, on Zulip):

oh, I see, if they're using RUSTC_BOOTSTRAP then its all out the window.

nikomatsakis (May 16 2019 at 15:07, on Zulip):

yes, sorry, point is, they could use beta

mw (May 16 2019 at 15:07, on Zulip):

I think their policy is to use stable except for one platform that can use beta

nikomatsakis (May 16 2019 at 15:09, on Zulip):

I guess the question is whether we want to ship a "working" ThinLTO but risk some further, unknown fallout?

pnkfelix (May 16 2019 at 15:09, on Zulip):

I'd personally prefer to let PR #60788 and #60378 bake

nikomatsakis (May 16 2019 at 15:09, on Zulip):

(er, I mean specifically cross-language ThinLTO I guess)

nikomatsakis (May 16 2019 at 15:09, on Zulip):

yeah, I think that might be the right call here too

mw (May 16 2019 at 15:10, on Zulip):

I agree

pnkfelix (May 16 2019 at 15:10, on Zulip):

#60788 landed only two days ago; its another instance of the "effectively backporting to stable" phenomenon I described at outset

pnkfelix (May 16 2019 at 15:10, on Zulip):

Okay

pnkfelix (May 16 2019 at 15:10, on Zulip):

lets go ahead and beta-accept PR #60644 and #60641

nikomatsakis (May 16 2019 at 15:10, on Zulip):

right, my concerns are (a) that #60788 hasn't had time on nightly and (b) that the other PR already required 1 fix, so clearly this isn't "obviously right"

pnkfelix (May 16 2019 at 15:11, on Zulip):

and decline to backport #60378 and #60788

nikomatsakis (May 16 2019 at 15:11, on Zulip):

plus (c) there was at least some unexpected breakage for people, allowed or not

nikomatsakis (May 16 2019 at 15:12, on Zulip):

(gotta run, bbs)

Björn Steinbrink (May 16 2019 at 15:12, on Zulip):

That is, nikomatsakis said we would need to backport PR #60788 along with PR #60378. But then Björn Steinbrink seems to have reported issues with #60788 ... but maybe that was a mistake?

JFYI, that's a different Björn. I'm dotdash on github

mw (May 16 2019 at 15:13, on Zulip):

thanks for driving, @pnkfelix !

centril (May 16 2019 at 15:15, on Zulip):

(Aside: Something should be done bout RUSTC_BOOSTRAP and FF's use of it...)

pnkfelix (May 16 2019 at 15:15, on Zulip):

JFYI, that's a different Björn. I'm dotdash on github

ugh sorry, I knew that. :oops:

Philipp Hansch (May 16 2019 at 15:15, on Zulip):

For completeness sake, the Clippy/RLS issue that was mentioned here (does that link work?), is tracked here: https://github.com/rust-lang/rust/issues/60848 and oli said he's working on it

pnkfelix (May 16 2019 at 15:17, on Zulip):

thanks for the info @Philipp Hansch !

pnkfelix (May 16 2019 at 15:17, on Zulip):

and bye to everyone @T-compiler/meeting ! Thanks for attending!

Last update: Nov 22 2019 at 05:00UTC