Stream: t-compiler

Topic: weekly meeting 2020-03-12 #54818


Santiago Pastorino (Mar 11 2020 at 16:09, on Zulip):

Hi @T-compiler/meeting ; the triage meeting will be starting in 21 hours 50 minutes

Santiago Pastorino (Mar 11 2020 at 16:09, on Zulip):

We are going over this process with @pnkfelix right now

Santiago Pastorino (Mar 11 2020 at 16:14, on Zulip):

We will be doing pre-triage in a parallel topic

Santiago Pastorino (Mar 11 2020 at 16:20, on Zulip):

We'll be having a WG checkin from WG-nll tomorrow

Santiago Pastorino (Mar 11 2020 at 16:20, on Zulip):

hey @Matthew Jasper if there's anything to share, let us know :)

Santiago Pastorino (Mar 11 2020 at 16:23, on Zulip):

Likewise we will be having WG-parallel-rustc, cc @simulacrum to share about this

simulacrum (Mar 11 2020 at 16:24, on Zulip):

no updates from parallel unfortunately

simulacrum (Mar 11 2020 at 16:24, on Zulip):

we've not made any changes/progress I think

Santiago Pastorino (Mar 11 2020 at 16:33, on Zulip):

We have also started filling the Agenda for this meeting

Matthew Jasper (Mar 11 2020 at 16:48, on Zulip):

Santiago Pastorino said:

hey Matthew Jasper if there's anything to share, let us know :)

There's not. Why hasn't wg-nll been removed from the rotation?

Santiago Pastorino (Mar 12 2020 at 13:06, on Zulip):

@Matthew Jasper yeah we were talking about that with @pnkfelix but asked just in case there were something else to share

Santiago Pastorino (Mar 12 2020 at 13:06, on Zulip):

we need to remove nll from rotation then

Santiago Pastorino (Mar 12 2020 at 13:07, on Zulip):

btw, hi @T-compiler/meeting, triage meeting will be starting in 53 minutes

pnkfelix (Mar 12 2020 at 13:58, on Zulip):

@Matthew Jasper I had thought there was still one last thing for the migration to NLL; removing region inference from typeck or some such?

pnkfelix (Mar 12 2020 at 14:00, on Zulip):

Hi @T-compiler/meeting! Add a :wave: emoji to show you're here :)

Matthew Jasper (Mar 12 2020 at 14:01, on Zulip):

There's #69189 open

pnkfelix (Mar 12 2020 at 14:01, on Zulip):

@Matthew Jasper (ah, okay, and that should tie things off for NLL? Sounds good. We'll catch up on that later then.)

simulacrum (Mar 12 2020 at 14:01, on Zulip):

https://github.com/rust-lang/blog.rust-lang.org/pull/534 is a post asking for side effect feedback, would appreciate some eyes (particularly from LLVM folk, I guess).

pnkfelix (Mar 12 2020 at 14:02, on Zulip):

Heh I think we have thus started off with the few minutes for ...

pnkfelix (Mar 12 2020 at 14:02, on Zulip):

Announcements

nikomatsakis (Mar 12 2020 at 14:02, on Zulip):
pnkfelix (Mar 12 2020 at 14:03, on Zulip):

The agenda already has some announcements listed, namely:

pnkfelix (Mar 12 2020 at 14:03, on Zulip):
pnkfelix (Mar 12 2020 at 14:03, on Zulip):
Santiago Pastorino (Mar 12 2020 at 14:04, on Zulip):
Santiago Pastorino (Mar 12 2020 at 14:04, on Zulip):

and now lives on https://rustc-dev-guide.rust-lang.org

pnkfelix (Mar 12 2020 at 14:04, on Zulip):

hmm we should add a link from that forge page to the actual proposals list

pnkfelix (Mar 12 2020 at 14:05, on Zulip):

people may not realize how much work is necessary to alpha rename the rustc-guide

pnkfelix (Mar 12 2020 at 14:05, on Zulip):

so I think its safe to say we probably won't rename it again. :)

Santiago Pastorino (Mar 12 2020 at 14:06, on Zulip):

right, it may have not been the best decision :P

pnkfelix (Mar 12 2020 at 14:06, on Zulip):

(or at least, I didn't realize how many moving parts were involved until @Santiago Pastorino explained it to me yesterday.)

Wesley Wiser (Mar 12 2020 at 14:06, on Zulip):

I'm not sure this is exactly what announcements are typically for, but I would like to remove support for consuming optimization fuel (-Zfuel) from the layout code. Currently, when layout runs out of fuel it results in an ICE which makes using fuel difficult for debugging other optimizations which use it (#55524). Removing this would close that issue and make debugging other optimization issues a bit easier.

If anyone has any objections, feel free to :thumbs_down: this comment otherwise I'll open a PR, tag the compiler team and let it go through the normal PR process.

Wesley Wiser (Mar 12 2020 at 14:07, on Zulip):

(I figured this might be the lightest weight way to get a sense of the rest of the teams opinion on this)

Santiago Pastorino (Mar 12 2020 at 14:07, on Zulip):

pnkfelix said:

(or at least, I didn't realize how many moving parts were involved until Santiago Pastorino explained it to me yesterday.)

the working group learned the hard way that for these things that may look trivial, it's anyway a good idea to ping infra

pnkfelix (Mar 12 2020 at 14:07, on Zulip):

@Wesley Wiser does it need to be removed from all of the layout code? Is there a way to isolate the optimization parts of layout from other things?

Wesley Wiser (Mar 12 2020 at 14:07, on Zulip):

@pnkfelix There's just one line of code that calls consider_optimizing()

pnkfelix (Mar 12 2020 at 14:08, on Zulip):

hmm okay

nikomatsakis (Mar 12 2020 at 14:08, on Zulip):

@Wesley Wiser Hmm. Are you saying the ICE is hard to fix in any other way? It seems like it'd still be a useful tool for folks who are getting segfaults due to layout ..

nikomatsakis (Mar 12 2020 at 14:08, on Zulip):

...but no strong objection from me I guess.

pnkfelix (Mar 12 2020 at 14:08, on Zulip):

It certainly seems safe to remove it now

pnkfelix (Mar 12 2020 at 14:08, on Zulip):

but maybe open an issue to investigate re-adding it?

Wesley Wiser (Mar 12 2020 at 14:09, on Zulip):

I'm not sure. I don't know very much about the layout code so it would be difficult for me to fix personally. If others are using it, I don't want to break their workflows but I've never seen anyone mention using it to debug a layout issue.

pnkfelix (Mar 12 2020 at 14:09, on Zulip):

(since I cannot imagine a fundamental reason we could not support -Zfuel on layout optimizations on things that have compiler-defined representation.)

nikomatsakis (Mar 12 2020 at 14:10, on Zulip):

@Wesley Wiser I doubt anyone is using it "actively"

pnkfelix (Mar 12 2020 at 14:10, on Zulip):

okay thanks

nikomatsakis (Mar 12 2020 at 14:10, on Zulip):

which is partly why I'm ok removing it

pnkfelix (Mar 12 2020 at 14:10, on Zulip):

Okay lets move along to the agenda items...

nagisa (Mar 12 2020 at 14:10, on Zulip):

o/

pnkfelix (Mar 12 2020 at 14:10, on Zulip):

we have one beta nomination

pnkfelix (Mar 12 2020 at 14:10, on Zulip):

beta-nom: Use TypeRelating for instantiating query responses #69591

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

at this point

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

is this implicitly a stable nomination?

nagisa (Mar 12 2020 at 14:12, on Zulip):

I think stable has been already prep'd?

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

@simulacrum when is the the nightly-to-beta promotion?

Wesley Wiser (Mar 12 2020 at 14:12, on Zulip):

Based on pietro's comment I think not

nagisa (Mar 12 2020 at 14:12, on Zulip):

today :slight_smile:

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

I guess what I mean is, I'm not sure there's any point in backporting

simulacrum (Mar 12 2020 at 14:13, on Zulip):

I think nightly to beta already happened

pnkfelix (Mar 12 2020 at 14:13, on Zulip):

ah okay

pnkfelix (Mar 12 2020 at 14:13, on Zulip):

sounds great then

simulacrum (Mar 12 2020 at 14:13, on Zulip):

we've not switched master to bootstrap off of (new) beta yet though

pnkfelix (Mar 12 2020 at 14:13, on Zulip):

I see, this bug, I think, is young enough that this wasn't on the imminent release?

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

anyway, it all sounds great, beta-accepted.

pnkfelix (Mar 12 2020 at 14:15, on Zulip):

there are no stable-nominations

pnkfelix (Mar 12 2020 at 14:16, on Zulip):

Last time we (@Santiago Pastorino and I) checked, we have 52 P-high issues, 32 of which are unassigned.

pnkfelix (Mar 12 2020 at 14:16, on Zulip):

there are 2 P-medium issues that are stable-to-beta regressions

pnkfelix (Mar 12 2020 at 14:17, on Zulip):

namely this one: "internal compiler error: "byte index 4 is not a char boundary"" #69130, which is assigned to @Esteban Küber

pnkfelix (Mar 12 2020 at 14:17, on Zulip):

and this one: "Worsened debug build codegen in beta" #68855, which is assigned to @Wesley Wiser

pnkfelix (Mar 12 2020 at 14:18, on Zulip):

but since they are P-medium, we have basically said "its okay to ship the release with these bugs"

nagisa (Mar 12 2020 at 14:18, on Zulip):

yeah, the first one defintely feels that way.

nikomatsakis (Mar 12 2020 at 14:18, on Zulip):

pnkfelix said:

namely this one: "internal compiler error: "byte index 4 is not a char boundary"" #69130, which is assigned to Esteban Küber

I'm not entirely sure why I put this as p-medium (though I see that I did)

pnkfelix (Mar 12 2020 at 14:18, on Zulip):

there are also four issues that are stable-to-nightly regressions

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

though I do think we can "afford" to ship with it, but seems like it's probably a relatively easy fix..?

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

one of them is P-medium, the other three are P-high

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

@nikomatsakis I was assuming you assigned that P-medium because it was result of fuzzing

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

and ... unlikely to occur in the wild?

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

ah, perhaps so

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

I missed that it resulted from fuzzing

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

though I don't know; maybe my bias towards ASCII is showing

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

I'm not sure though how unlikely it is. Certainly somewhat

nikomatsakis (Mar 12 2020 at 14:20, on Zulip):

I have stray copy-and-pastes inserting weird characters in my code from time to time, at least :)

nikomatsakis (Mar 12 2020 at 14:20, on Zulip):

(In any case I don't feel inclined to worry too much about this specific bug at this specific time, but it's an interesting policy question)

pnkfelix (Mar 12 2020 at 14:21, on Zulip):

I'm going to explicitly walk through the P-high stable-to-nightly regressions

pnkfelix (Mar 12 2020 at 14:21, on Zulip):

just to raise awareness

pnkfelix (Mar 12 2020 at 14:22, on Zulip):

(and because our agenda wasn't overloaded today)

pnkfelix (Mar 12 2020 at 14:22, on Zulip):

P-high regr : "is_x86_feature_detected!("avx512f") fails to build on nightly" #68905

pnkfelix (Mar 12 2020 at 14:22, on Zulip):

this is on the path to being resolved, I think.

simulacrum (Mar 12 2020 at 14:23, on Zulip):

correct, yes, and may need a t-libs backport or so but I'm not (yet) worried

pnkfelix (Mar 12 2020 at 14:23, on Zulip):

(you can read the comment thread; re-stabilizing the feature in question had to be approved by the lang and libs teams...)

pnkfelix (Mar 12 2020 at 14:23, on Zulip):

P-high regr: "cargo check --message-format json regression of error span data on empty main.rs file" #68808

pnkfelix (Mar 12 2020 at 14:23, on Zulip):

oh yikes I self-assigned this over a month ago...

pnkfelix (Mar 12 2020 at 14:24, on Zulip):

um, sorry.

pnkfelix (Mar 12 2020 at 14:24, on Zulip):

anyway I have nothing to report, it obviously fell off my radar

nikomatsakis (Mar 12 2020 at 14:24, on Zulip):

maybe we can try to find a new assignment?

pnkfelix (Mar 12 2020 at 14:24, on Zulip):

(the P-high categorization is interesting. )

pnkfelix (Mar 12 2020 at 14:25, on Zulip):

(I think its right to make this P-high. Incorrect/misleading spans can be very annoying...)

pnkfelix (Mar 12 2020 at 14:26, on Zulip):

I can allocate time for it

nikomatsakis (Mar 12 2020 at 14:26, on Zulip):

I wonder if some of the diagnostics folks might take a look. I guess we don't yet have a convenient way to ping that group

pnkfelix (Mar 12 2020 at 14:26, on Zulip):

maybe I'll ask @centril for help since they seemed to be involved with injecting it.

pnkfelix (Mar 12 2020 at 14:26, on Zulip):

we'll see.

pnkfelix (Mar 12 2020 at 14:26, on Zulip):

P-high regr: "Compiler error while compiling Winrt" #66402

eddyb (Mar 12 2020 at 14:27, on Zulip):

what is a daylight saving

pnkfelix (Mar 12 2020 at 14:27, on Zulip):

hey @Wesley Wiser , you have a window box, right? Could you maybe look at this?

eddyb (Mar 12 2020 at 14:27, on Zulip):

(sorry, I'm here now and reading backlog)

pnkfelix (Mar 12 2020 at 14:27, on Zulip):

eddyb said:

what is a daylight saving

https://en.wikipedia.org/wiki/Daylight_saving_time

Wesley Wiser (Mar 12 2020 at 14:27, on Zulip):

Sure!

pnkfelix (Mar 12 2020 at 14:28, on Zulip):

It is the result of the world being driven by farmers

nagisa (Mar 12 2020 at 14:28, on Zulip):

pnkfelix said:

It is the result of the world being driven by farmers

hey, europe is turning it off this year AFAIK

pnkfelix (Mar 12 2020 at 14:28, on Zulip):

(actually that seems to be completely false. :laughing: )

nikomatsakis (Mar 12 2020 at 14:28, on Zulip):

https://www.youtube.com/watch?v=_XFojiOqB3o

Wesley Wiser (Mar 12 2020 at 14:28, on Zulip):

Do you just want an mcve that's not Windows specific?

pnkfelix (Mar 12 2020 at 14:29, on Zulip):

nagisa said:

pnkfelix said:

It is the result of the world being driven by farmers

hey, europe is turning it off this year AFAIK

oh maybe that's what @eddyb was implicitly referring to

eddyb (Mar 12 2020 at 14:29, on Zulip):

nagisa said:

pnkfelix said:

It is the result of the world being driven by farmers

hey, europe is turning it off this year AFAIK

great now the US will be wrong half the year instead of just a few weeks

pnkfelix (Mar 12 2020 at 14:29, on Zulip):

Wesley Wiser said:

Do you just want an mcve that's not Windows specific?

an MCVE wolud be a good start, yeah.

pnkfelix (Mar 12 2020 at 14:29, on Zulip):

if you get us that far, then we can go from there more generally next week.

eddyb (Mar 12 2020 at 14:30, on Zulip):

@pnkfelix I just keep forgetting every yeah that the US doesn't switch to summer time around the spring equinox like the rest of the world

pnkfelix (Mar 12 2020 at 14:30, on Zulip):

boy I wish we had "Dynamically Sized Types" on the agenda today

pnkfelix (Mar 12 2020 at 14:30, on Zulip):

so that I could say "lets stop talking about DST so that we can start talking about DST"

pnkfelix (Mar 12 2020 at 14:31, on Zulip):

eddyb said:

pnkfelix I just keep forgetting every yeah that the US doesn't switch to summer time around the spring equinox like the rest of the world

yeah that was the result of an attempt from the USA govt to boost its economy

pnkfelix (Mar 12 2020 at 14:31, on Zulip):

which is bonkers

centril (Mar 12 2020 at 14:31, on Zulip):

Damn it... missed that meeting was 1h earlier, :wave:

pnkfelix (Mar 12 2020 at 14:31, on Zulip):

See @Santiago Pastorino , I told you we wouldn't have a problem filling up the meeting time

eddyb (Mar 12 2020 at 14:31, on Zulip):

(in my defense I didn't mean to get anyone started on this just grumbling out loud that I almost completely missed it)

pnkfelix (Mar 12 2020 at 14:32, on Zulip):

So: @Santiago Pastorino has selected five nominated issues for us to discuss

pnkfelix (Mar 12 2020 at 14:32, on Zulip):

nom 1/5: "Compiler incorrectly assumes int will never be one" #69841

centril (Mar 12 2020 at 14:32, on Zulip):

pnkfelix said:

P-high regr: "cargo check --message-format json regression of error span data on empty main.rs file" #68808

Yea sorry, this one slipped off my radar, although @Esteban Küber is more familiar with this code even though I injected it :P

pnkfelix (Mar 12 2020 at 14:32, on Zulip):

this apparently is an LLVM injection

Santiago Pastorino (Mar 12 2020 at 14:33, on Zulip):

pnkfelix said:

See Santiago Pastorino , I told you we wouldn't have a problem filling up the meeting time

I realized earlier of TZ changes, something to remember for the future, so we can make everyone know about it :)

pnkfelix (Mar 12 2020 at 14:33, on Zulip):

I realized earlier of TZ changes, something to remember for the future, so we can make everyone know about it :)

Well this is in part why I always use relative addressing when I announce the meeting times

simulacrum (Mar 12 2020 at 14:33, on Zulip):

Q: Are people finding these by fuzzing or something? That input looks quite odd

centril (Mar 12 2020 at 14:33, on Zulip):

pnkfelix said:

nom 1/5: "Compiler incorrectly assumes int will never be one" #69841

Seems like a possible point release candidate if we can fix it

nikomatsakis (Mar 12 2020 at 14:34, on Zulip):

we could ping the LLVM ICE-breakers, though it seems like we've kind of already bisected the llvm side so maybe there's no point

pnkfelix (Mar 12 2020 at 14:34, on Zulip):

simulacrum said:

Q: Are people finding these by fuzzing or something? That input looks quite odd

there is a user who is actively using fuzzing to file tickets, but I don't think it is this user.

nikomatsakis (Mar 12 2020 at 14:34, on Zulip):

simulacrum said:

Q: Are people finding these by fuzzing or something? That input looks quite odd

that looks to me like a reduced bit of code

nagisa (Mar 12 2020 at 14:34, on Zulip):

simulacrum said:

Q: Are people finding these by fuzzing or something? That input looks quite odd

Looks minimised from an actual codebase to me.

simulacrum (Mar 12 2020 at 14:34, on Zulip):

yeah I think we'd at this point want an upstream fix presumably

nagisa (Mar 12 2020 at 14:34, on Zulip):

Not really feasible to fuzz an entire compiler.

nikomatsakis (Mar 12 2020 at 14:35, on Zulip):

I guess one question is whether "edit: I bisected the regression specifically to llvm/llvm-project@0290a77" has a LLVM bug attachd to it?

pnkfelix (Mar 12 2020 at 14:35, on Zulip):

nikomatsakis said:

we could ping the LLVM ICE-breakers, though it seems like we've kind of already bisected the llvm side so maybe there's no point

yeah, I guess my only question is whether we have an LLVM upgrade scheduled any time soon? Or perhaps we should try to cherry-pick a fix here?

simulacrum (Mar 12 2020 at 14:35, on Zulip):

we do have an upgrade "in progress" though I think it's stalled out a bit

simulacrum (Mar 12 2020 at 14:35, on Zulip):

not sure it fixes this though

nikomatsakis (Mar 12 2020 at 14:35, on Zulip):

is it fixed upstream?

nikomatsakis (Mar 12 2020 at 14:35, on Zulip):

ah, ok

pnkfelix (Mar 12 2020 at 14:36, on Zulip):

well anyway

nikomatsakis (Mar 12 2020 at 14:36, on Zulip):

I'd be in favor of trying to get an LLVM ICE-breaker person to look into (a) is this fixed upstream? (b) is there an attached bug? and (c) can we backport a narrow fix

nikomatsakis (Mar 12 2020 at 14:36, on Zulip):

and/or push the upgrade through I guess

nagisa (Mar 12 2020 at 14:36, on Zulip):

It is fixed on LLVM master as per comment thread, I’d say chances are good that’s in LLVM 10, but I didn’t git log the commit

pnkfelix (Mar 12 2020 at 14:36, on Zulip):

yes, lets ping LLVM ICE-breakers for that

pnkfelix (Mar 12 2020 at 14:36, on Zulip):

(for your first suggestion, that is)

pnkfelix (Mar 12 2020 at 14:37, on Zulip):

(or both of them, the "and/or" that is...)

pnkfelix (Mar 12 2020 at 14:37, on Zulip):

I'm not clarifying my position, am I

pnkfelix (Mar 12 2020 at 14:37, on Zulip):

next nomination

pnkfelix (Mar 12 2020 at 14:37, on Zulip):

nom 2/5: “ICE in collect_and_partition_mono_items” #69785

pnkfelix (Mar 12 2020 at 14:38, on Zulip):

this bug surprised me

pnkfelix (Mar 12 2020 at 14:38, on Zulip):

in that its something where replacing a pub(crate) with pub fixes the problem.

centril (Mar 12 2020 at 14:38, on Zulip):

is that reproducible on stable?

Santiago Pastorino (Mar 12 2020 at 14:38, on Zulip):

@jonhoo says the (crate) part is key

pnkfelix (Mar 12 2020 at 14:39, on Zulip):

@centril the bug as written uses a feature flag

Santiago Pastorino (Mar 12 2020 at 14:39, on Zulip):

interesting to see their workaround on their app https://github.com/mit-pdos/noria/commit/d1f78f6c4299fe7c724d16b50b6a7011180ada05

centril (Mar 12 2020 at 14:39, on Zulip):

if not reproducible on stable we should add requires-nightly & downgrade to P-medium

nagisa (Mar 12 2020 at 14:39, on Zulip):

collector will act differently based on visibility

pnkfelix (Mar 12 2020 at 14:39, on Zulip):

so one of the open question is "is there a variant that reproduces the issue without the feature flag"

nikomatsakis (Mar 12 2020 at 14:39, on Zulip):

pnkfelix said:

in that its something where replacing a pub(crate) with pub fixes the problem.

that can certainly happen as a result of some of the "visibility" logic

eddyb (Mar 12 2020 at 14:39, on Zulip):

Wesley Wiser said:

I'm not sure this is exactly what announcements are typically for, but I would like to remove support for consuming optimization fuel (-Zfuel) from the layout code. Currently, when layout runs out of fuel it results in an ICE which makes using fuel difficult for debugging other optimizations which use it (#55524). Removing this would close that issue and make debugging other optimization issues a bit easier.

If anyone has any objections, feel free to :thumbs_down: this comment otherwise I'll open a PR, tag the compiler team and let it go through the normal PR process.

the irony here is that we added -Zfuel to debug layout (and specifically randomized field order IIRC, idk what happened to that, I don't think there is a flag to enable randomization), and no it shouldn't ICE, it should just not optimize after fuel runs out

I guess we should have is a per-category fuel so one would use -Zfuel=layout=123,mir-opt=567

nagisa (Mar 12 2020 at 14:39, on Zulip):

pub(crate) is an obvious signal to just codegen function code into the rlib

centril (Mar 12 2020 at 14:39, on Zulip):

@pnkfelix I would suggest we assume there's not until otherwise shown

mw (Mar 12 2020 at 14:40, on Zulip):

it would be good to get a backtrace with line numbers for the ICE

Santiago Pastorino (Mar 12 2020 at 14:40, on Zulip):

fwiw, my guess is that the feature flag is not necessary, and that the relevant type = impl Trait can be replaced by an async fn.

but that's not proven as @centril noted

pnkfelix (Mar 12 2020 at 14:41, on Zulip):

@centril yes, well, the filer has at least theorized that the type = impl Trait can be replaced with an async fn

centril (Mar 12 2020 at 14:41, on Zulip):

we can always upgrade again when that's demonstrated :slight_smile:

pnkfelix (Mar 12 2020 at 14:42, on Zulip):

hmm. well I'm just vaguely worried that it will fall off our radar if we downgrade...

nagisa (Mar 12 2020 at 14:42, on Zulip):

this looks fairly trivial to "prove otherwise" to me, but I don’t have the bandwidth for doing so

pnkfelix (Mar 12 2020 at 14:43, on Zulip):

is there anyone here who has the ability and bandwidth to attempt the replacement with async fn that is suggested? Or do we just push back on the issue filer to do so?

nagisa (Mar 12 2020 at 14:44, on Zulip):

passive thought: what if we have allowed-to-fail tests for ICEs that need unstable features that start failing when feature is being stabilised? So that we don’t miss those.

pnkfelix (Mar 12 2020 at 14:44, on Zulip):

I guess I'm okay with reclassifying as P-medium and telling the issue filer that we will upgrade priority after seeing demo

centril (Mar 12 2020 at 14:44, on Zulip):

@nagisa ooh! I had similar thoughts recently

nikomatsakis (Mar 12 2020 at 14:45, on Zulip):

I think it's probably ok to mark it as "nightly only", but a lot of the code that deals with impl Trait is fairly common to different ways of using it, and I wouldn't be surprised if there is some way to trigger on stable.

centril (Mar 12 2020 at 14:45, on Zulip):

like, what if we just add ICEs as tests, and then we notice when they don't ICE anymore

nikomatsakis (Mar 12 2020 at 14:45, on Zulip):

I'm in favor of that -- we used to do that on a compiler I worked on and it was very useful

pnkfelix (Mar 12 2020 at 14:45, on Zulip):

yeah a lot of other projects do this

centril (Mar 12 2020 at 14:45, on Zulip):

@nagisa feels like a good design meeting candidate

pnkfelix (Mar 12 2020 at 14:45, on Zulip):

though I don't know if I'd block CI on it?

pnkfelix (Mar 12 2020 at 14:46, on Zulip):

anyway yeah lets maybe propose for design meeting

nagisa (Mar 12 2020 at 14:46, on Zulip):

pnkfelix said:

though I don't know if I'd block CI on it?

Allowed-to-fail is the key.

pnkfelix (Mar 12 2020 at 14:46, on Zulip):

obligatory link: https://github.com/rust-lang/compiler-team/issues?q=is%3Aissue+is%3Aopen+label%3Ameeting-proposal

centril (Mar 12 2020 at 14:46, on Zulip):

Switching the label to P-medium then

pnkfelix (Mar 12 2020 at 14:47, on Zulip):

oh right, this meeting

pnkfelix (Mar 12 2020 at 14:47, on Zulip):

nom 3/5: “ICE: src/librustc/middle/region.rs:1037: Encountered greater count 28” #69307

pnkfelix (Mar 12 2020 at 14:48, on Zulip):

@nikomatsakis what was AsyncAwait-OnDeck denote again?

pnkfelix (Mar 12 2020 at 14:48, on Zulip):

ah, "next focus items"

nikomatsakis (Mar 12 2020 at 14:48, on Zulip):

we actually were moving away, I think, from those labels

centril (Mar 12 2020 at 14:48, on Zulip):

I would say this one is not P-high for t-compiler as it's not a regression

nikomatsakis (Mar 12 2020 at 14:48, on Zulip):

towards priority labels

pnkfelix (Mar 12 2020 at 14:48, on Zulip):

heh

centril (Mar 12 2020 at 14:48, on Zulip):

but would be nice to fix; and has a small reproducer

eddyb (Mar 12 2020 at 14:48, on Zulip):

for issues like this we should get a backtrace from a debuginfo-level=1 build

pnkfelix (Mar 12 2020 at 14:48, on Zulip):

because we make such great use of those (priority labels, that is)...

eddyb (Mar 12 2020 at 14:48, on Zulip):

maybe one day the nightly will have that debuginfo :(

eddyb (Mar 12 2020 at 14:49, on Zulip):

although this one is surprisingly readable, but maybe that's just an illusion

pnkfelix (Mar 12 2020 at 14:49, on Zulip):

@nikomatsakis anyway I'm trying to infer whether this will be handled by the AsyncAwait WG

nikomatsakis (Mar 12 2020 at 14:50, on Zulip):

in theory, if anyone picks it up

pnkfelix (Mar 12 2020 at 14:50, on Zulip):

and also, I'm definitely unhappy with the I-nominated system at this point

nikomatsakis (Mar 12 2020 at 14:50, on Zulip):

it's in the pool of "bugs we'd like to handle"

pnkfelix (Mar 12 2020 at 14:50, on Zulip):

in terms of the ambiguity of what team is targetted by a nomination, and whether they have discharged it or not

nikomatsakis (Mar 12 2020 at 14:50, on Zulip):

"sooner rather than later", I guess

centril (Mar 12 2020 at 14:50, on Zulip):

sounds like something up @Matthew Jasper or @Aaron Hill's alleys

nikomatsakis (Mar 12 2020 at 14:51, on Zulip):

I'm not really clear on why this is nominated at all :) but async-wait team doesn't use that label

pnkfelix (Mar 12 2020 at 14:51, on Zulip):

@nikomatsakis oh really? For some reason I had thought that that team did.

pnkfelix (Mar 12 2020 at 14:51, on Zulip):

Okay then, that makes things easier.

pnkfelix (Mar 12 2020 at 14:52, on Zulip):

so, in terms of the prioritization

pnkfelix (Mar 12 2020 at 14:52, on Zulip):

even though this is not a regression, AsyncAwait is a high-profile, heavily used feature, right?

pnkfelix (Mar 12 2020 at 14:53, on Zulip):

I had thought that was a reason that we would continue classifying bugs like this as P-high.

nikomatsakis (Mar 12 2020 at 14:54, on Zulip):

it's on the 'short list' of bugs that async await wants to handle

nikomatsakis (Mar 12 2020 at 14:54, on Zulip):

so I guess that is "p-high", I don't know, I think it's kind of unclear as you said what those labels should mean :)

nikomatsakis (Mar 12 2020 at 14:54, on Zulip):

(I think @tmandry moved over to using GH projects for the async-await WG internal tracking, so you can see it's in that "To Do" column, along with about 5 or 6 other bugs)

pnkfelix (Mar 12 2020 at 14:54, on Zulip):

my suggestion is that we leave as P-high and remove the I-nominated tag. That way, T-compiler won't revisit it every week

nikomatsakis (Mar 12 2020 at 14:54, on Zulip):

probably the most immediate thing is to do mentoring

nikomatsakis (Mar 12 2020 at 14:54, on Zulip):

which I might try to do

pnkfelix (Mar 12 2020 at 14:55, on Zulip):

okay next

pnkfelix (Mar 12 2020 at 14:55, on Zulip):

nom 4/5: “Replace our fragile safety scheme around erroneous constants” #67191

nagisa (Mar 12 2020 at 14:55, on Zulip):

should this be a design meeting?

pnkfelix (Mar 12 2020 at 14:55, on Zulip):

this was nominated ages ago ... lets see

oli (Mar 12 2020 at 14:56, on Zulip):

I think we discussed all the points in there, I just need to impl it?

nikomatsakis (Mar 12 2020 at 14:56, on Zulip):

I suspect the nomination is basically "stale"

pnkfelix (Mar 12 2020 at 14:56, on Zulip):

oh in fact, I think it was nominated for a different team

pnkfelix (Mar 12 2020 at 14:56, on Zulip):

and then just got moved to us?

pnkfelix (Mar 12 2020 at 14:56, on Zulip):

okay yeah

centril (Mar 12 2020 at 14:56, on Zulip):

it's in t-compilers court to impl now

pnkfelix (Mar 12 2020 at 14:56, on Zulip):

so this is awaiting an implementor

pnkfelix (Mar 12 2020 at 14:57, on Zulip):

but we don't need to visit it every week

pnkfelix (Mar 12 2020 at 14:57, on Zulip):

my suggestion is that we remove the nomination flag

oli (Mar 12 2020 at 14:57, on Zulip):

nope

oli (Mar 12 2020 at 14:57, on Zulip):

(nope to visiting every week)

pnkfelix (Mar 12 2020 at 14:57, on Zulip):

I don't even know if it is P-high ...

centril (Mar 12 2020 at 14:58, on Zulip):

It should be P-high; it's a stability hole, essentially

centril (Mar 12 2020 at 14:58, on Zulip):

the longer it is open the harder it gets to fix

pnkfelix (Mar 12 2020 at 14:58, on Zulip):

I made it P-high in terms of resolving the open questions

pnkfelix (Mar 12 2020 at 14:58, on Zulip):

those questions are now resolved

pnkfelix (Mar 12 2020 at 14:58, on Zulip):

centril said:

the longer it is open the harder it gets to fix

there's a lot of things in that bucket...

nikomatsakis (Mar 12 2020 at 14:59, on Zulip):

any idea how hard it is to fix, @oli? has there been any motion on that?

pnkfelix (Mar 12 2020 at 14:59, on Zulip):

anyway it can stay as P-high

oli (Mar 12 2020 at 14:59, on Zulip):

there has been zero work towards it, but I think it's pretty straightforward to do

pnkfelix (Mar 12 2020 at 14:59, on Zulip):

so okay, keep P-high, removing I-nominated.

centril (Mar 12 2020 at 14:59, on Zulip):

@oli think you can fix / leave mentoring notes?

Santiago Pastorino (Mar 12 2020 at 14:59, on Zulip):

oli said:

there has been zero work towards it, but I think it's pretty straightforward to do

do you want me to help?

pnkfelix (Mar 12 2020 at 14:59, on Zulip):

lets move along, we're almost done with the selected nominations for this week

pnkfelix (Mar 12 2020 at 15:00, on Zulip):

nom 5/5: “ICE field: higher-rank trait bound (HRTB) for<'a> ... hits OutputTypeParameterMismatch in librustc/traits/codegen” #62529

pnkfelix (Mar 12 2020 at 15:00, on Zulip):

any chance of us addressing lazy-normalization this year?

pnkfelix (Mar 12 2020 at 15:01, on Zulip):

(that question was why I nominated it0

centril (Mar 12 2020 at 15:01, on Zulip):

I suspect we'll know if we fix it this year :P

pnkfelix (Mar 12 2020 at 15:01, on Zulip):

@nikomatsakis thoughts?

centril (Mar 12 2020 at 15:01, on Zulip):

We've seen some movement re. lazy norm & const generics

pnkfelix (Mar 12 2020 at 15:02, on Zulip):

mmmm. Okay well our hour time slot for the meeting is up. I guess I'll remove the nominated tag from this too. We don't need to revisit it every week.

pnkfelix (Mar 12 2020 at 15:03, on Zulip):

As for WG checkins: @Santiago Pastorino and I established ahead of time that neither WG-NLL nor WG-parallel-rustc have anything to report. So that's all!

nikomatsakis (Mar 12 2020 at 15:03, on Zulip):

(Sorry, missed that ping)

nikomatsakis (Mar 12 2020 at 15:03, on Zulip):

I think we're making progress, I don't know

pnkfelix (Mar 12 2020 at 15:03, on Zulip):

Thanks to everyone in @T-compiler/meeting for attending, including those who were flummoxed by the USA time change!

nagisa (Mar 12 2020 at 15:04, on Zulip):

Make sure you express your concerns wrt landing https://github.com/rust-lang/rust/pull/55617 if you have any

centril (Mar 12 2020 at 15:04, on Zulip):

nagisa said:

Make sure you express your concerns wrt landing https://github.com/rust-lang/rust/pull/55617 if you have any

That seems rather bitrotted; might want to attempt a new perf run?

nagisa (Mar 12 2020 at 15:04, on Zulip):

I’ll rebase and look into landing it… right about now

nagisa (Mar 12 2020 at 15:05, on Zulip):

centril said:

nagisa said:

Make sure you express your concerns wrt landing https://github.com/rust-lang/rust/pull/55617 if you have any

That seems rather bitrotted; might want to attempt a new perf run?

It was done half a month ago and results didn’t change since the one done > year ago. But sure.

pnkfelix (Mar 12 2020 at 15:06, on Zulip):

(18 days doesn't sound so badly bitrotted to me.)

centril (Mar 12 2020 at 15:07, on Zulip):

my mistake, PR seemed older :slight_smile:

centril (Mar 12 2020 at 15:08, on Zulip):

@nagisa the diff in lowering seems fairly arbitrary and not well documented

centril (Mar 12 2020 at 15:08, on Zulip):

other than that I have no thoughts

nagisa (Mar 12 2020 at 15:08, on Zulip):

centril said:

nagisa the diff in lowering seems fairly arbitrary and not well documented

Concerns on the PR please.

Last update: May 29 2020 at 18:00UTC