Stream: t-compiler

Topic: pre-meeting triage 2019-04-18 #54818


pnkfelix (Apr 18 2019 at 11:31, on Zulip):

This week's pre-triage will be tracked here

eddyb (Apr 18 2019 at 11:33, on Zulip):

quick question: I was wondering whether https://github.com/rust-lang/rfcs/pull/2603#issuecomment-483838495 could be brought up in the meeting, or if FCP proposals are triaged regularly (I frankly can't remember)

pnkfelix (Apr 18 2019 at 11:34, on Zulip):

there's a couple different ways to try to ensure a PR / issue is discussed

eddyb (Apr 18 2019 at 11:34, on Zulip):

(to discuss the fact that we have an implementation, and we want to merge it and the RFC at the same time, and how we might want to get more eyes on either/both, and maybe where we might plan the messaging for getting people to test this on nightly)

pnkfelix (Apr 18 2019 at 11:35, on Zulip):

the current agenda structure doesn't have any link for FCP stuff, but we probably should add one.

eddyb (Apr 18 2019 at 11:36, on Zulip):

it's not urgent, but it could be significant to some people, so I would want to at least direct some attention towards it

pnkfelix (Apr 18 2019 at 11:36, on Zulip):

(the links we do have are for I-nominated T-compiler and S-waiting-on-team T-compiler

pnkfelix (Apr 18 2019 at 11:36, on Zulip):

I'll nominate it now so that it hopefully won't get overlooked.

eddyb (Apr 18 2019 at 11:37, on Zulip):

thanks!

pnkfelix (Apr 18 2019 at 11:38, on Zulip):

as for pre-triage itself...

pnkfelix (Apr 18 2019 at 11:39, on Zulip):

first prepass: unprioritized T-compiler nominated issues

pnkfelix (Apr 18 2019 at 11:39, on Zulip):

holy cow there's 11 of them

pnkfelix (Apr 18 2019 at 11:39, on Zulip):

first: "Can define and use async fn without feature gate on nightly" #60069

eddyb (Apr 18 2019 at 11:42, on Zulip):

were we relying on futures being unstable or something?

pnkfelix (Apr 18 2019 at 11:43, on Zulip):

maybe.

pnkfelix (Apr 18 2019 at 11:43, on Zulip):

I posted a question in the issue

eddyb (Apr 18 2019 at 11:43, on Zulip):

this is bad because we are stabilizing futures soon

pnkfelix (Apr 18 2019 at 11:43, on Zulip):

I guess I can make it P-high in that case

eddyb (Apr 18 2019 at 11:43, on Zulip):

(at first I thought we stabilized futures already and that's why you need 0 feature-gates, so I was close but wrong temporal frame of reference)

pnkfelix (Apr 18 2019 at 11:45, on Zulip):

triage: P-high. Leaving I-nominated tag due to questions posed in ticket.

pnkfelix (Apr 18 2019 at 11:45, on Zulip):

next: "Futures involving uninhabited variables are incorrectly considered uninhabited." #59972

pnkfelix (Apr 18 2019 at 11:47, on Zulip):

#59972 is already assigned to @tmandry and it sounds like they have a fix in development on WIP PR #59897. Marking P-high rather than debate about prioritization.

pnkfelix (Apr 18 2019 at 11:48, on Zulip):

next: " miri no longer builds after rust-lang/rust#60030" #60040

pnkfelix (Apr 18 2019 at 11:49, on Zulip):

based on previous notes from @RalfJ regarding prioritization of miri, marking P-medium.

pnkfelix (Apr 18 2019 at 11:50, on Zulip):

next: "assert! allows invalid trailing tokens in expression" #60024

pnkfelix (Apr 18 2019 at 11:51, on Zulip):

apparently this leaked into stable. 1.25.0 rejected such input; 1.26.0 accepted it.

pnkfelix (Apr 18 2019 at 11:52, on Zulip):

PR #60039 is meant to fix it

pnkfelix (Apr 18 2019 at 11:52, on Zulip):

there's a crater run being applied to #60039.

pnkfelix (Apr 18 2019 at 11:53, on Zulip):

I think we should categorize this as "just a bug" (rather than worry about future compat warning type stuff), unless the crater run turns up cases where this is being used in the wild

eddyb (Apr 18 2019 at 11:54, on Zulip):

"it's an implementation accident, not a design accident"

pnkfelix (Apr 18 2019 at 11:55, on Zulip):

... have we not applied future-compat warnings to past implementation accidents?

pnkfelix (Apr 18 2019 at 11:56, on Zulip):

next: "regression in deterministic generation of binary from 1.32 to 1.33" #59542

eddyb (Apr 18 2019 at 11:57, on Zulip):

so this was tricky but I think its status now is "what changes in the input is source directory path, and we have a regression in how much that factors into the output"

eddyb (Apr 18 2019 at 11:58, on Zulip):

or maybe it was a bug fix, I have no idea, I wish they provided a simpler way to reproduce

pnkfelix (Apr 18 2019 at 11:59, on Zulip):

did they actually provide any way to reproduce?

eddyb (Apr 18 2019 at 11:59, on Zulip):

I don't even know what's supposed to work with --remap-path-prefix, we'd need someone who knows that feature (@mw?) to figure out if their usecase is even supported

eddyb (Apr 18 2019 at 12:00, on Zulip):

@pnkfelix I think it's "build the same source with the same compiler in two separate directories but using --remap-path-prefix to remap those directories to the same thing, and compare the output (hashes in symbol names, for example)

eddyb (Apr 18 2019 at 12:01, on Zulip):

but I don't know how to use --remap-path-prefix "as intended", or what other factors might exist from their whole build infra they have

centril (Apr 18 2019 at 12:14, on Zulip):

@pnkfelix I nominated some more I-ICEs

pnkfelix (Apr 18 2019 at 12:28, on Zulip):

In any case, re #59542, it seems like they were not able to reproduce atop simple programs, and do not have permission to share the actual case where its occurring, right?

pnkfelix (Apr 18 2019 at 12:30, on Zulip):

I'm going to leave #59542 nominated and unprioritized. Hopefully we will discuss in the meeting. (I'd like to know what priority we want to assign to build-reproducibility in general; I knew there were efforts to make builds reproducible, but I didn't know we had clients currently relying on it.)

pnkfelix (Apr 18 2019 at 12:30, on Zulip):

next: "Compiler panic: 'internal error: entered unreachable code' in src/libsyntax/parse/parser.rs" #60075

pnkfelix (Apr 18 2019 at 12:30, on Zulip):

ICE, marking P-high and unnominating.

pnkfelix (Apr 18 2019 at 12:31, on Zulip):

next: "ICE: OutputTypeParameterMismatch when combining closures and higher-rank lifetime bounds" #60070

pnkfelix (Apr 18 2019 at 12:31, on Zulip):

ICE, marking P-high and unnominating.

pnkfelix (Apr 18 2019 at 12:32, on Zulip):

next: "ICE: can't type-check body of DefId" #60049

pnkfelix (Apr 18 2019 at 12:34, on Zulip):

ICE, marking P-high and unnominating.

pnkfelix (Apr 18 2019 at 12:36, on Zulip):

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

pnkfelix (Apr 18 2019 at 12:38, on Zulip):

it seems like the paths for methods in the debug info do not always have their prefixes? Seems interesting.

pnkfelix (Apr 18 2019 at 12:39, on Zulip):

the description actually has even more info, its not quite as simple as I made it out to be above

pnkfelix (Apr 18 2019 at 12:42, on Zulip):

leaving nominated and unprioritized for now; moving along

pnkfelix (Apr 18 2019 at 12:46, on Zulip):

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

pnkfelix (Apr 18 2019 at 12:47, on Zulip):

assigning P-medium (it appears to be a long known problem with open questions as to how we should resolve it), but leaving nomination tag.

pnkfelix (Apr 18 2019 at 12:47, on Zulip):

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

pnkfelix (Apr 18 2019 at 12:48, on Zulip):

oy

pnkfelix (Apr 18 2019 at 12:49, on Zulip):

I don't understand, the issue makes it sound like this was a pre-existing test ?

pnkfelix (Apr 18 2019 at 12:49, on Zulip):

/me looks

pnkfelix (Apr 18 2019 at 12:50, on Zulip):

oh I see, we are literally testing that we continue to get an ICE here

pnkfelix (Apr 18 2019 at 12:50, on Zulip):

const-pat-ice.stderr

pnkfelix (Apr 18 2019 at 12:51, on Zulip):

okay well lets call it P-high for now; though given its age (@centril says it dates back to 1.26.0 at least) it might be P-medium.

pnkfelix (Apr 18 2019 at 12:53, on Zulip):

next: "Change powerpc64 base CPU" #59932

pnkfelix (Apr 18 2019 at 12:54, on Zulip):

I don't know much about our official PPC support

pnkfelix (Apr 18 2019 at 12:54, on Zulip):

leaving unprioritized and nominated.

pnkfelix (Apr 18 2019 at 12:54, on Zulip):

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

pnkfelix (Apr 18 2019 at 12:55, on Zulip):

apparently this is fixed on nightly but is a stable-to-stable regression.

pnkfelix (Apr 18 2019 at 12:55, on Zulip):

the fix is PR #58605

pnkfelix (Apr 18 2019 at 12:56, on Zulip):

I guess we'll discuss the stable-nomination of that PR when we get to the stable-nominations

pnkfelix (Apr 18 2019 at 12:57, on Zulip):

in terms of prioritization, I'll call #59898 P-high, mark E-needstest, and move along

pnkfelix (Apr 18 2019 at 12:58, on Zulip):

okay. another prepass is nominated issues with "no team"

pnkfelix (Apr 18 2019 at 12:59, on Zulip):

the pattern I match for that actually currently includes issues tagged T-dev-tools

pnkfelix (Apr 18 2019 at 12:59, on Zulip):

namely this list

pnkfelix (Apr 18 2019 at 12:59, on Zulip):

which seem like either clippy or RLS bugs

pnkfelix (Apr 18 2019 at 12:59, on Zulip):

clippy is not our domain, but RLS might be

pnkfelix (Apr 18 2019 at 12:59, on Zulip):

do we need to discuss "Decouple nightly RLS from Clippy" #59761 ?

pnkfelix (Apr 18 2019 at 13:00, on Zulip):

(ugh I need to get to another meeting and I haven't finished even the prepasses...)

pnkfelix (Apr 18 2019 at 13:50, on Zulip):

I'm going to tag #59761 with T-compiler to try to make sure it stays on my/our radar

pnkfelix (Apr 18 2019 at 13:52, on Zulip):

next pre-pass: beta-regressions without P-label

pnkfelix (Apr 18 2019 at 13:52, on Zulip):

just one: "Tuple indexing regression" #59553

pnkfelix (Apr 18 2019 at 13:52, on Zulip):

hmm, I would have thought we looked at this last week

pnkfelix (Apr 18 2019 at 13:53, on Zulip):

in terms of it being marked I-needs-decision

pnkfelix (Apr 18 2019 at 13:53, on Zulip):

I guess we are in any case waiting for the beta crater run to see how much breakage this causes

pnkfelix (Apr 18 2019 at 13:54, on Zulip):

marking P-high

pnkfelix (Apr 18 2019 at 13:55, on Zulip):

next pre-pass: nightly regressions without P-label

pnkfelix (Apr 18 2019 at 13:55, on Zulip):

just one: "ICE while bootstrapping (./x.py build)" #59945

pnkfelix (Apr 18 2019 at 13:56, on Zulip):

okay, four minutes left and only the pre-passes done

pnkfelix (Apr 18 2019 at 13:57, on Zulip):

next P-high T-compiler issues

pnkfelix (Apr 18 2019 at 13:57, on Zulip):

eleven of above are unassigned

pnkfelix (Apr 18 2019 at 14:01, on Zulip):

I'll assign "ICE in src/librustc/ty/query/plumbing.rs when compiling incrementally" #59716 to myself to try to make a regression test.

Last update: Nov 16 2019 at 02:20UTC