Stream: t-compiler

Topic: weekly meeting 2019-07-25 #54818


nikomatsakis (Jul 25 2019 at 13:43, on Zulip):

Hi @T-compiler/meeting -- meeting starting in 15 minutes -- pre-triage ongoing in this topic

nikomatsakis (Jul 25 2019 at 13:44, on Zulip):

Hey @davidtwco or @T-compiler/WG-meta -- do we have a "working group checkin calendar" on the new compiler-team website?

davidtwco (Jul 25 2019 at 13:44, on Zulip):

Not yet.

nikomatsakis (Jul 25 2019 at 13:45, on Zulip):

Also, in general, people should feel free to post announcements in this topic =)

davidtwco (Jul 25 2019 at 13:45, on Zulip):

I'll aim to add that this week after things settle down with getting everything deployed.

nikomatsakis (Jul 25 2019 at 13:46, on Zulip):

One announcement (not sure if already announced) is that https://rust-lang.github.io/compiler-team/ exists =)

varkor (Jul 25 2019 at 13:50, on Zulip):

the links to experts map and procedures directory 404s

Francesco Dainese (Jul 25 2019 at 13:55, on Zulip):

the links to experts map and procedures directory 404s

experts map doesn't 404 me, but following the here link prompts a download. Procedure directory isn't really accessible unless you type the url manually, but the links under it work fine

nikomatsakis (Jul 25 2019 at 14:01, on Zulip):

(be right there, 2 nominated issues left, gonna try to get through those in pre-triage)

nikomatsakis (Jul 25 2019 at 14:05, on Zulip):

OK, hello @T-compiler/meeting, weekly meeting time :)

nikomatsakis (Jul 25 2019 at 14:06, on Zulip):

I didn't get a chance to see which working groups had a check-in last time

nikomatsakis (Jul 25 2019 at 14:06, on Zulip):

Maybe none

nagisa (Jul 25 2019 at 14:06, on Zulip):

We did not checkin last time, I couldn’t find a calendar.

nikomatsakis (Jul 25 2019 at 14:06, on Zulip):

Yeah

nikomatsakis (Jul 25 2019 at 14:06, on Zulip):

Has to get fixed

nikomatsakis (Jul 25 2019 at 14:07, on Zulip):

OK, let's start with highlights from pre-triage

nagisa (Jul 25 2019 at 14:07, on Zulip):

Instead we spent most of the time voting on backports.

nikomatsakis (Jul 25 2019 at 14:08, on Zulip):

Few things

nikomatsakis (Jul 25 2019 at 14:09, on Zulip):

#60297 -- we are looking for someone with access to ARM/musl to help debug this

nikomatsakis (Jul 25 2019 at 14:09, on Zulip):

#62896, #62932 -- there are some embedded failures, quite possibly related to LLVM, that could use bisection

nikomatsakis (Jul 25 2019 at 14:09, on Zulip):

similarly #62781

nagisa (Jul 25 2019 at 14:10, on Zulip):

For #60297, I’m fairly confident that we’ll need to virtualize this either way, so everyone has equally much of an access to this platform

nikomatsakis (Jul 25 2019 at 14:10, on Zulip):

I was thinking something similar

nikomatsakis (Jul 25 2019 at 14:10, on Zulip):

that would be an even better option

nikomatsakis (Jul 25 2019 at 14:10, on Zulip):

not sure 100% what that means :) maybe somebody getting a working setup and writing up instructions?

nikomatsakis (Jul 25 2019 at 14:12, on Zulip):

Maybe we can open an issue around that? (Creating some kind of virtualization setup to help debugging esoteric platforms and documenting how to do it?)

nikomatsakis (Jul 25 2019 at 14:12, on Zulip):

It seems like the sort of thing we could put out a "call for help" around and get some people who are familiar with that

nikomatsakis (Jul 25 2019 at 14:12, on Zulip):

I'll take a note to do it, seems like it would go well in the rustc-guide

nikomatsakis (Jul 25 2019 at 14:14, on Zulip):

OK, those were the major things I noted from the I-nominated issues; one question that came up for me was how to prioritize things like #62708 -- a fairly basic ICE but not (apparently) a regression etc.

nikomatsakis (Jul 25 2019 at 14:14, on Zulip):

One could make a case for P-high but I was trying to hold the line to some extent

nikomatsakis (Jul 25 2019 at 14:14, on Zulip):

In any case I tagged it as E-needs-mentor for now, and I might try to investigate that (and some other long-standing things assigned to me) later today -- though if others want to poke at it, feel free.

nikomatsakis (Jul 25 2019 at 14:15, on Zulip):

I guess next up is beta backports

nikomatsakis (Jul 25 2019 at 14:15, on Zulip):

there's only one though

nikomatsakis (Jul 25 2019 at 14:15, on Zulip):

Make the parser TokenStream more resilient after mismatched delimiter recovery #62887

nagisa (Jul 25 2019 at 14:16, on Zulip):

Yeah, we cleared that list last week.

nikomatsakis (Jul 25 2019 at 14:17, on Zulip):

OK, looks like there is general approval to backport

nikomatsakis (Jul 25 2019 at 14:18, on Zulip):

Nothing stable-nominated

nikomatsakis (Jul 25 2019 at 14:18, on Zulip):

Looking through the list of regressions though I found a few that were not nominated and don't have a priority

nikomatsakis (Jul 25 2019 at 14:18, on Zulip):

Undefined symbol _fltused when compiling to x86_64-unknown-uefi #62785

nikomatsakis (Jul 25 2019 at 14:19, on Zulip):

perhaps related to compiler-builtins crate -- not sure who was making changes in that area -- maybe we'll start with a bisection?

nikomatsakis (Jul 25 2019 at 14:20, on Zulip):

ICE "no ImplicitCtxt stored in tls" in Servo on Windows and macOS #62717

nikomatsakis (Jul 25 2019 at 14:21, on Zulip):

seems like there is a pending fix

nikomatsakis (Jul 25 2019 at 14:22, on Zulip):

indirect_structural_match lint misfiring on enum variant unused in const item #62614

centril (Jul 25 2019 at 14:22, on Zulip):

a) I'm not sure why we label this as P-high, b) At what point do we give up on the plugin interface?

nikomatsakis (Jul 25 2019 at 14:22, on Zulip):

I'm marking this as P-medium since @pnkfelix kind of resolved it

nikomatsakis (Jul 25 2019 at 14:23, on Zulip):

a) I'm not sure why we label this as P-high, b) At what point do we give up on the plugin interface?

regression, was my motivation but (b) is a good question

centril (Jul 25 2019 at 14:24, on Zulip):

We should probably lift https://github.com/rust-lang/rust/issues/62614 to a lang team discussion at some point also

nikomatsakis (Jul 25 2019 at 14:24, on Zulip):

but all the PR does is to add a "deprecated" placeholder? seems good to give people an "off ramp"

nikomatsakis (Jul 25 2019 at 14:25, on Zulip):

it'd be good to be tracking it though and have some plan for when to remove entirely

nikomatsakis (Jul 25 2019 at 14:25, on Zulip):

We should probably lift https://github.com/rust-lang/rust/issues/62614 to a lang team discussion at some point also

(yes, I agree, I think it'd be good to resolve as part of a "const generics" effort, personally)

centril (Jul 25 2019 at 14:25, on Zulip):

(yes, I agree, I think it'd be good to resolve as part of a "const generics" effort, personally)

Excellent idea :+1:

nagisa (Jul 25 2019 at 14:26, on Zulip):

At what point do we give up on the plugin interface?

Whenever there is an alternative for servo.

nikomatsakis (Jul 25 2019 at 14:27, on Zulip):

We should certainly try to figure out what they are doing and how to create alternatives

nagisa (Jul 25 2019 at 14:27, on Zulip):

I believe their primary and most critical use-case is for their root-checking plugin

nagisa (Jul 25 2019 at 14:27, on Zulip):

there is an issue somewhere about that

nikomatsakis (Jul 25 2019 at 14:27, on Zulip):

yeah, makes sense

centril (Jul 25 2019 at 14:27, on Zulip):

Most recent issue is https://github.com/rust-lang/rust/issues/62868

nagisa (Jul 25 2019 at 14:28, on Zulip):

https://github.com/rust-lang/rust/issues/62868 is that, yeah

centril (Jul 25 2019 at 14:29, on Zulip):

However, my understanding is that:

Access to full type information. For example in let x = foo.bar(); the lint wants to find out whether the struct definition for the concrete type of x has one of the previously mentioned custom attributes, even if bar is a trait method that returns an associated type.

is not realistic from a stability and forward compat POV and that this will largely mean that we forever support only servo on the plugin interface

centril (Jul 25 2019 at 14:29, on Zulip):

anyways... let's not dig too deep into this now

nikomatsakis (Jul 25 2019 at 14:29, on Zulip):

Agreed, I was going to say, sounds like a good topic for a design meeting ...

nikomatsakis (Jul 25 2019 at 14:29, on Zulip):

... and has lang-team interactions too

nikomatsakis (Jul 25 2019 at 14:30, on Zulip):

or at least part of the fix might be trying to create alternatives at other levels for what they're actually trying to do

nikomatsakis (Jul 25 2019 at 14:30, on Zulip):

OK, so, S-waiting-on-team is one issue:

Turn HIR indexing into a query #59064

This PR I think wants some amount of big picture design discussion. As @**mw** noted here, they've got some concerns, and I think we should have a "plan". This was what the design meeting was intended to try and address but it's not really gotten off the ground. =)

nikomatsakis (Jul 25 2019 at 14:30, on Zulip):

Maybe in August, not sure. I definitely want us to make progress in general around the end-to-end query setup.

nikomatsakis (Jul 25 2019 at 14:31, on Zulip):

Welp, I think we reached the end of our "triage agenda"

nikomatsakis (Jul 25 2019 at 14:31, on Zulip):

Did I miss something? :)

oli (Jul 25 2019 at 14:32, on Zulip):

I didn't see https://github.com/rust-lang/rust/issues/62189 but I'm on it

oli (Jul 25 2019 at 14:32, on Zulip):

not sure in what triage part it should have shown up

nikomatsakis (Jul 25 2019 at 14:32, on Zulip):

oh yeah I didn't mention it because it seemed like you were on it

nagisa (Jul 25 2019 at 14:32, on Zulip):

I have a topic

nikomatsakis (Jul 25 2019 at 14:32, on Zulip):

Ideally we'd do working group check-in but I'm not sure which groups, we have to fix the calendar... one thing I personally would like to get though is an update regarding async-await status.

nikomatsakis (Jul 25 2019 at 14:33, on Zulip):

Oh, @nagisa, go for it

nagisa (Jul 25 2019 at 14:33, on Zulip):

Alex is looking to transfer the stacker crate, either to T-compiler overall or to me, as they don’t have the time needed to maintain the crate anymore. Do we want to maintain it? I can get it transferred to myself and am willing to maintain it too.

nikomatsakis (Jul 25 2019 at 14:34, on Zulip):

Well, if the compiler relies on it, I think we might as well move it to rust-lang

nagisa (Jul 25 2019 at 14:34, on Zulip):

It does not currently, but I think the hope to make it so is not dead yet.

nikomatsakis (Jul 25 2019 at 14:34, on Zulip):

see also https://rust-lang.github.io/compiler-team/docs/procedures/crates/

nikomatsakis (Jul 25 2019 at 14:34, on Zulip):

ah, ok -- I think we basically have to rely on it to handle various scenarios, so it seems good to me

nagisa (Jul 25 2019 at 14:35, on Zulip):

I’ll ask alex to transfer it under rust-lang then

nagisa (Jul 25 2019 at 14:35, on Zulip):

next topic.

centril (Jul 25 2019 at 14:36, on Zulip):

I haven't nominated it, but I would love to see some progress on fixing https://github.com/rust-lang/rust/issues/28728.

nikomatsakis (Jul 25 2019 at 14:37, on Zulip):

Yeah, I remember there being some discussion about that in #t-compiler/wg-llvm at some point

centril (Jul 25 2019 at 14:37, on Zulip):

Has a PR open to fix it afaik, https://github.com/rust-lang/rust/pull/59546

nikomatsakis (Jul 25 2019 at 14:38, on Zulip):

I see. And current status is basically: run lolbench

nagisa (Jul 25 2019 at 14:38, on Zulip):

All we really need to do there is to gather some benchmarks to see that generated code quality does not suffer significantly

nikomatsakis (Jul 25 2019 at 14:38, on Zulip):

that seems like a good e-mentor sort of issue too

nikomatsakis (Jul 25 2019 at 14:39, on Zulip):

I wonder if @Adam Perry could do it easily

nikomatsakis (Jul 25 2019 at 14:39, on Zulip):

does it need rebasing?

nikomatsakis (Jul 25 2019 at 14:39, on Zulip):

looks like no

centril (Jul 25 2019 at 14:39, on Zulip):

to see that generated code quality does not suffer significantly

One thing I'd like to ask is whether we know that, assuming it does suffer significantly, whether that is possible to actually prevent...?
i.e. there could be some fundamental limitations here and such...

centril (Jul 25 2019 at 14:39, on Zulip):

Well, we should first see whether it does "suffer significantly" tho...

nagisa (Jul 25 2019 at 14:40, on Zulip):

if my memory serves me right Adam said they were not in a position to run the benchmarks themselves.

nikomatsakis (Jul 25 2019 at 14:40, on Zulip):

I see there are some instructions there

nagisa (Jul 25 2019 at 14:40, on Zulip):

but they were willing to provide help necessary.

nikomatsakis (Jul 25 2019 at 14:41, on Zulip):

OK, I'll write some quick mentoring instructions and try to call some attention to it.

nikomatsakis (Jul 25 2019 at 14:43, on Zulip):

Done, though I may actually try to get that going in the background

nikomatsakis (Jul 25 2019 at 14:43, on Zulip):

In the meantime, I propose we end the meeting early this week.

nikomatsakis (Jul 25 2019 at 14:43, on Zulip):

I'm going to spend the rest of the day trying to catch up on status of things, feel free to ping me -- I'll be looking also for PRs that need reviewing etc.

centril (Jul 25 2019 at 14:44, on Zulip):

@nikomatsakis btw, idk how many folks will show up to the T-lang mtg but I think we are a go?

centril (Jul 25 2019 at 14:44, on Zulip):

will you be there?

nikomatsakis (Jul 25 2019 at 14:44, on Zulip):

Yeah, I was planning to attend today :)

centril (Jul 25 2019 at 14:44, on Zulip):

Excellent

nikomatsakis (Jul 25 2019 at 14:44, on Zulip):

Thanks all for attending! :heart:

eddyb (Jul 25 2019 at 14:56, on Zulip):

ooops, I had the Zulip window killed because I was doing some silly wg-grammar things that used a lot of RAM :(

Last update: Nov 16 2019 at 02:15UTC