Stream: t-compiler

Topic: pre-meeting triage 2019-07-11 #54818


pnkfelix (Jul 11 2019 at 12:37, on Zulip):

I will be doing pre-triage in this channel.

pnkfelix (Jul 11 2019 at 12:40, on Zulip):

There are three unprioritized nominated T-compiler issues

pnkfelix (Jul 11 2019 at 12:41, on Zulip):

first unprioritized nominated: "serde no longer builds with latest rustc (master)" #62562

pnkfelix (Jul 11 2019 at 12:41, on Zulip):

Looks like this was recently fixed by @Vadim Petrochenkov . Thanks!

pnkfelix (Jul 11 2019 at 12:43, on Zulip):

(marking as P-high and removing nomination and assigning to petrochenkov accordingly)

pnkfelix (Jul 11 2019 at 12:44, on Zulip):

next unprioritized nominated: "ICE: thread 'rustc' panicked at 'capacity overflow'" #62554

pnkfelix (Jul 11 2019 at 12:44, on Zulip):

it an issue with our parser recovery

pnkfelix (Jul 11 2019 at 12:45, on Zulip):

We do emit a nice message prior to ICE'ing due to the overflow. Personally I do not see resolving this as a P-high matter. (I imagine it may or may not be easy to fix, but my only point is that since we do emit a reasonable message, it seems likely to be P-medium.)

pnkfelix (Jul 11 2019 at 12:46, on Zulip):

The only reason I could see to make this P-high is if somehow our (mis)handling of this case has implications for IDE experience. @Igor Matuszewski , do you have any thoughts on that aspect of this?

pnkfelix (Jul 11 2019 at 12:46, on Zulip):

Anyway I'm going to mark as P-medium but leave it nominated so that we double-check that priority assignment at the meeting.

pnkfelix (Jul 11 2019 at 12:48, on Zulip):

last unprioritized nominated: "compiler panic: 'byte index 10 is not a char boundary'" #62524

pnkfelix (Jul 11 2019 at 12:50, on Zulip):

again, it seems here we provide a reasonable error message before the ICE occurs.

pnkfelix (Jul 11 2019 at 12:51, on Zulip):

I'm going to mark this one P-medium. The previous case (#62554) I saw potential issue due to the potential impact on memory usage and UX. But this case just seems like one we do not have to prioritize.

pnkfelix (Jul 11 2019 at 12:53, on Zulip):

there are no nominations with no team assigned

pnkfelix (Jul 11 2019 at 12:53, on Zulip):

there is one unprioritized beta-regression

pnkfelix (Jul 11 2019 at 12:54, on Zulip):

namely: "internal compiler error: Error constructed but not emitted" #62546

Pietro Albini (Jul 11 2019 at 12:56, on Zulip):

(note about beta regressions, the crater run should finish in 1/2 days)

pnkfelix (Jul 11 2019 at 13:02, on Zulip):

for #62546, I am again seeing a reasonable error message before the ICE. But the nature of this ICE is a little bit worriesome; I don't like it when our errors are swallowed up.

pnkfelix (Jul 11 2019 at 13:02, on Zulip):

so I'm going to traige #62546 as P-high, just in terms of looking in the problem to make sure something deeper isn't hiding here. (It may be something simple.)

pnkfelix (Jul 11 2019 at 13:03, on Zulip):

there are no nightly-regressions without a P-label.

pnkfelix (Jul 11 2019 at 13:04, on Zulip):

we have three P-high unassigned bugs

pnkfelix (Jul 11 2019 at 13:06, on Zulip):

We'll go over those at the meeting itself.

pnkfelix (Jul 11 2019 at 13:07, on Zulip):

hmm. there are three stable-to-beta regressions that i did not visit earlier, because they are all tagged with both T-compiler and T-rustdoc

pnkfelix (Jul 11 2019 at 13:07, on Zulip):

lets look at those quickly.

pnkfelix (Jul 11 2019 at 13:08, on Zulip):

first doubly-tagged beta-regression: "regression: doctest failure display-as-0.4.4" #61562

pnkfelix (Jul 11 2019 at 13:09, on Zulip):

going to ask whether this needs T-compiler label.

simulacrum (Jul 11 2019 at 13:09, on Zulip):

this is now stable/stable

simulacrum (Jul 11 2019 at 13:09, on Zulip):

(based on the date, at least)

simulacrum (Jul 11 2019 at 13:10, on Zulip):

retagged

pnkfelix (Jul 11 2019 at 13:10, on Zulip):

next doubly-tagged regression (I'll leave out the question of where it regressed): "regression: rustdoc resolution changes" #61560

pnkfelix (Jul 11 2019 at 13:11, on Zulip):

Hmm I'd like someone to double-check severity of this, see if name resolution for the crate itself, even if only w.r.t. tests or doc tests, changed somehow

pnkfelix (Jul 11 2019 at 13:11, on Zulip):

triage: P-high

simulacrum (Jul 11 2019 at 13:12, on Zulip):

this is also stable/stable but I agree with p-high probably for initial investigation

pnkfelix (Jul 11 2019 at 13:12, on Zulip):

last doubly-tagged regression: "rustdoc ICE (2/3) in markdown - unexpected node in tree (Header)" #61478

simulacrum (Jul 11 2019 at 13:12, on Zulip):

this needs to be confirmed for a fix

pnkfelix (Jul 11 2019 at 13:13, on Zulip):

yep, okay

simulacrum (Jul 11 2019 at 13:13, on Zulip):

It _should_ be fixed afaik

pnkfelix (Jul 11 2019 at 13:13, on Zulip):

is it worth adding regression tests for the case(s) in question?

pnkfelix (Jul 11 2019 at 13:13, on Zulip):

(is that something the T-rustdoc team does when encountering problems like this?)

simulacrum (Jul 11 2019 at 13:14, on Zulip):

not sure, I think t-rustdoc is... rather slow, if anything, about doing so

simulacrum (Jul 11 2019 at 13:14, on Zulip):

I guess I'll mark E-needstest

pnkfelix (Jul 11 2019 at 13:14, on Zulip):

great

pnkfelix (Jul 11 2019 at 13:14, on Zulip):

Skimming over the nightly regressions now

pnkfelix (Jul 11 2019 at 13:14, on Zulip):

they all have P-labels, so that's good.

simulacrum (Jul 11 2019 at 13:15, on Zulip):

most may now be -to-beta regressions, fwiw, I don't think anybody bumped tags this cycle yet

pnkfelix (Jul 11 2019 at 13:15, on Zulip):

two are unassigned; they were ones I just covered earlier in triage.

pnkfelix (Jul 11 2019 at 13:15, on Zulip):

@simulacrum okay, noted.

pnkfelix (Jul 11 2019 at 13:15, on Zulip):

has nightly been bumped to beta yet?

simulacrum (Jul 11 2019 at 13:15, on Zulip):

yeah

pnkfelix (Jul 11 2019 at 13:15, on Zulip):

okay

simulacrum (Jul 11 2019 at 13:16, on Zulip):

it happened on schedule (for once) -- friday after release (so a week ago? two?)

simulacrum (Jul 11 2019 at 13:16, on Zulip):

a week, I think

pnkfelix (Jul 11 2019 at 13:17, on Zulip):

we have two PR's marked waiting-on-team and three nominated T-compiler issues+PR's

pnkfelix (Jul 11 2019 at 13:17, on Zulip):

(and the two sets actually overlap on #60966)

simulacrum (Jul 11 2019 at 13:17, on Zulip):

I wasn't aware of regular visits of waiting-on-team, nomination is nothing more than that

simulacrum (Jul 11 2019 at 13:18, on Zulip):

(on #60966)

pnkfelix (Jul 11 2019 at 13:18, on Zulip):

I try to visit it as part of the weekly triage

simulacrum (Jul 11 2019 at 13:18, on Zulip):

that's good to hear/know

pnkfelix (Jul 11 2019 at 13:18, on Zulip):

it has its own bullet separate from nominated (in the standing agenda, #54818)

pnkfelix (Jul 11 2019 at 13:18, on Zulip):

but I think nominated is, strictly speaking, considered higher priority

simulacrum (Jul 11 2019 at 13:19, on Zulip):

yep, agreed -- I had the impression we didn't visit anything but nominated, regressions, and P-{high,med}

pnkfelix (Jul 11 2019 at 13:20, on Zulip):

okay so that's all the main pre-triage items, apart from hypothetically visiting all the P-high bugs

simulacrum (Jul 11 2019 at 13:20, on Zulip):

(I may update that waiting-on-team search to drop is:pr so we can label issues as such as well, hm)

pnkfelix (Jul 11 2019 at 13:21, on Zulip):

I'm mostly in shock right now about having 39 minutes left before the meeting starts

pnkfelix (Jul 11 2019 at 13:21, on Zulip):

so lets see, we have 37 open P-high T-compiler issues

pnkfelix (Jul 11 2019 at 13:22, on Zulip):

I guess I'll try to run through all the assigned ones and see if there are anything that could use a ping

pnkfelix (Jul 11 2019 at 13:22, on Zulip):

We already discussed "serde no longer builds with latest rustc (master)" #62562

pnkfelix (Jul 11 2019 at 13:23, on Zulip):

next, "ICE on self-referential typedef" #62364: has a PR, looks well in hand. Thanks @Alexander Regueiro !

pnkfelix (Jul 11 2019 at 13:23, on Zulip):

ah that even landed; it just didn't auto-close due to how the message was formatted.

pnkfelix (Jul 11 2019 at 13:24, on Zulip):

hmm, sad: Playground is only on nightly 2019-07-09? We've had nightlies since then, right?

simulacrum (Jul 11 2019 at 13:25, on Zulip):

https://github.com/rust-lang/rust/issues/62562 I believe is the cause

simulacrum (Jul 11 2019 at 13:25, on Zulip):

should be fixed tonight

pnkfelix (Jul 11 2019 at 13:25, on Zulip):

oh, okay, right.

pnkfelix (Jul 11 2019 at 13:28, on Zulip):

same story goes for "Nightly compiler panic on using Self with associated type in type alias" #62305; fixed by PR #62417

pnkfelix (Jul 11 2019 at 13:29, on Zulip):

next: "ICE on HRTB" #62203; this one is mine. It has a (trivial?) PR (#62523) which needs review.

pnkfelix (Jul 11 2019 at 13:30, on Zulip):

next: "Static cyclic references not compiled in current nightly rust version" #62189. Assigned to @oli a week ago; @RalfJ notes that @oli has been on vacation so they won't get to this for at least another week.

pnkfelix (Jul 11 2019 at 13:32, on Zulip):

next: " Incorrect span / broken rustfix: help: use dyn: dyn #[dom_struct]" #61963; currently assigned. @davidtwco , do you think you'll have a chance to look at this ? If not, maybe I should reassign to @centril or someone else from @WG-diagnostics ?

pnkfelix (Jul 11 2019 at 13:32, on Zulip):

At the very leas the issue looks like it could use a minimal test case.

pnkfelix (Jul 11 2019 at 13:33, on Zulip):

tagged E-needs-mcve

eddyb (Jul 11 2019 at 13:33, on Zulip):

(nit: I keep reading MCVE as MSVC)

pnkfelix (Jul 11 2019 at 13:33, on Zulip):

(I assume it is something with proc attribute macors)

pnkfelix (Jul 11 2019 at 13:33, on Zulip):

@eddyb and I read it as CVE

pnkfelix (Jul 11 2019 at 13:34, on Zulip):

next: "Forgone caching in cycles caused much overflow in trait solving" #61960

centril (Jul 11 2019 at 13:34, on Zulip):

If not, maybe I should reassign to @centril or someone else from @WG-diagnostics ?

How do I keep becoming parts of teams I never joined? :joy:

pnkfelix (Jul 11 2019 at 13:34, on Zulip):

@centril oh I'm sorry, for some reasonI thought you were on that team.

centril (Jul 11 2019 at 13:35, on Zulip):

(hehe; np -- someone thought I was on T-infra a few days ago ^^)

pnkfelix (Jul 11 2019 at 13:35, on Zulip):

regarding #61960, I'm going to unassign myself given that I'm going on leave for 8 weeks at the end of this week.

davidtwco (Jul 11 2019 at 13:36, on Zulip):

next: " Incorrect span / broken rustfix: help: use dyn: dyn #[dom_struct]" #61963; currently assigned. davidtwco , do you think you'll have a chance to look at this ? If not, maybe I should reassign to centril or someone else from @WG-diagnostics ?

I've been looking into it, but haven't had much time. Got a build of servo that reproduces and have been trying to craft a minimal test case that will but not having much luck yet.

pnkfelix (Jul 11 2019 at 13:36, on Zulip):

next: "Forgone caching in cycles caused much overflow in trait solving" #61960

pnkfelix (Jul 11 2019 at 13:36, on Zulip):

#61960 is me.I have a PR (#62503) which I think @eddyb just reviewed today.

pnkfelix (Jul 11 2019 at 13:37, on Zulip):

@davidtwco okay, thanks for the update

pnkfelix (Jul 11 2019 at 13:38, on Zulip):

next: nonsense #NNN "Soundness hole in pattern matching on enums with an uninhabited variant" #61696. @eddyb has this under control (i reviewed the PR for it earlier today).

pnkfelix (Jul 11 2019 at 13:38, on Zulip):

next: "Self as default type isnt typechecked" #61631

eddyb (Jul 11 2019 at 13:39, on Zulip):

looks like you copy-pasted the same thing twice?

pnkfelix (Jul 11 2019 at 13:40, on Zulip):

Oh let me fix that

pnkfelix (Jul 11 2019 at 13:41, on Zulip):

regarding #61631, it seems like there's been discussion but it has sort of stalled since 10 days ago

pnkfelix (Jul 11 2019 at 13:42, on Zulip):

it seems like @Alexander Regueiro is thinking of potential rules that we could try to enforce to rule out the bad case here

pnkfelix (Jul 11 2019 at 13:43, on Zulip):

but I"m not sure if this will turn into a T-lang issue, depending on the nature of such rules...

Alexander Regueiro (Jul 11 2019 at 13:43, on Zulip):

@pnkfelix Yeah, I think @eddyb's suggestion was a reasonable one too.

pnkfelix (Jul 11 2019 at 13:43, on Zulip):

next: "ICE using a type alias to a const generic struct across crates" #61624

pnkfelix (Jul 11 2019 at 13:44, on Zulip):

I had threatened to downgrade #61624 from P-high to P-medium

Alexander Regueiro (Jul 11 2019 at 13:44, on Zulip):

looks like you copy-pasted the same thing twice?

Do you fancy tackling #61361? :-)

centril (Jul 11 2019 at 13:45, on Zulip):

I had threatened to downgrade #61624 from P-high to P-medium

Seems fair; I think you should act on the threat.

pnkfelix (Jul 11 2019 at 13:45, on Zulip):

#61624: there was some commentary on ways the bug could arise, but it seems to me like the point still stands that if you use stable Rust, it will be used on all of your crates.

pnkfelix (Jul 11 2019 at 13:46, on Zulip):

(don't get me wrong; I've commented elsewhere, a rust-call bug in particular, where I do worry about impact on downstream crates from upstream crates using certain features that have bugs)

pnkfelix (Jul 11 2019 at 13:46, on Zulip):

but yeah, lets reclassify #61624 as P-medium

pnkfelix (Jul 11 2019 at 13:49, on Zulip):

next: "1.30 -> 1.31 dylib late-binding regression with GNU binutils 2.28 or older." #61539

pnkfelix (Jul 11 2019 at 13:50, on Zulip):

hmm.

pnkfelix (Jul 11 2019 at 13:50, on Zulip):

@eddyb @nagisa do you two have further thoughts on whether we should bother to dynamically check the linker version?

pnkfelix (Jul 11 2019 at 13:51, on Zulip):

Personally it strikes me as a bit too much "spooky action"

nagisa (Jul 11 2019 at 13:52, on Zulip):

I… have no idea. My kneejerk reaction is to tell people to update their binutils or specify -Cplt=yes or whatever.

pnkfelix (Jul 11 2019 at 13:52, on Zulip):

but also, what was the intended effect after doing such a check? Just turning on -C plt=yes silently?

pnkfelix (Jul 11 2019 at 13:52, on Zulip):

because AFAIK there was not a more specific change to codegen identfied, right?

pnkfelix (Jul 11 2019 at 13:54, on Zulip):

anyway, I honestly don't know what we should do here. I do not even have context to determine whether this should remain P-high or not. is binutils 2.28 reletively old in the Linux world? (does it depend on distribution?)

pnkfelix (Jul 11 2019 at 13:55, on Zulip):

but I'm also fine with saying "tell people to update their binutils." that might be a way I would justify dynamically checking the Linker version

nagisa (Jul 11 2019 at 13:55, on Zulip):

2.28 is relatively recent. Recent enough for the current line of LTS distributions still have that or an earlier version.

pnkfelix (Jul 11 2019 at 13:56, on Zulip):

but I'm also fine with saying "tell people to update their binutils." that might be a way I would justify dynamically checking the Linker version

as in, we use such detection just to drive the emission of a diagnostic warning

pnkfelix (Jul 11 2019 at 13:56, on Zulip):

but we do not otherwise change our codegen.

pnkfelix (Jul 11 2019 at 13:56, on Zulip):

You know what, I'll nominate this ticket to discuss that approach.

pnkfelix (Jul 11 2019 at 13:59, on Zulip):

next: "Creating a recursive type with infinite size leads to internal compiler error" #61323

pnkfelix (Jul 11 2019 at 14:00, on Zulip):

well the plan there (of having @nikomatsakis find someone to whom to delegate this) has failed.

pnkfelix (Jul 11 2019 at 14:01, on Zulip):

okay time for meeting

Igor Matuszewski (Jul 11 2019 at 14:03, on Zulip):

@pnkfelix we dont have a good recovery from compiler ICEs I’m afraid

Igor Matuszewski (Jul 11 2019 at 14:03, on Zulip):

if it’s a dependency we mostly emit „compiling dependent crate failed” or similar

Igor Matuszewski (Jul 11 2019 at 14:04, on Zulip):

And we forward any rustc errors that are emitted to the IDE but we don’t forward the information that it’s an ICE

Igor Matuszewski (Jul 11 2019 at 14:04, on Zulip):

if that helps

pnkfelix (Jul 11 2019 at 14:04, on Zulip):

@Igor Matuszewski and if its the local crate?

pnkfelix (Jul 11 2019 at 14:04, on Zulip):

I'm mostly asking: Is it going to cause your IDE to crash or become non-responsive

pnkfelix (Jul 11 2019 at 14:04, on Zulip):

and I'm guessing the answer is "no that would be really bad"

pnkfelix (Jul 11 2019 at 14:04, on Zulip):

but I could be incorrect

Igor Matuszewski (Jul 11 2019 at 14:15, on Zulip):

Nope

Igor Matuszewski (Jul 11 2019 at 14:15, on Zulip):

It’s not gonna destabilize the IDE in any way

Igor Matuszewski (Jul 11 2019 at 14:15, on Zulip):

@pnkfelix ^

Last update: Nov 20 2019 at 01:10UTC