Stream: t-compiler

Topic: #54818 weekly meeting 2018-11-29


nagisa (Nov 29 2018 at 12:41, on Zulip):

@pnkfelix are you gonna drive or should I do it again?

pnkfelix (Nov 29 2018 at 12:41, on Zulip):

I'm back from PTO and I was planning to drive

nagisa (Nov 29 2018 at 12:43, on Zulip):

okay

pnkfelix (Nov 29 2018 at 13:05, on Zulip):

in the meantime, I am going to do some preparation for the meeting now (i.e. triage) and post the decisions I make in this channel, so that people can review them later at their leisure. Obviously I will defer any case that I don't feel comfortable deciding on my own.

pnkfelix (Nov 29 2018 at 13:21, on Zulip):

for "prohibit "two-phase borrows" with existing borrows?" #56254, I left a note summarizing the outcome from the NLL meeting and I don't think we should dedicate much time to this in the T-compiler meeting. It remains tagged as I-nominated because T-lang may or may not want to discuss.

pnkfelix (Nov 29 2018 at 13:22, on Zulip):

"NLL error on closure, but not on equivalent function" #55526: no change since last week, but that's OK. I plan to dedicate some time to this soon (and have re-assigned it to self)

pnkfelix (Nov 29 2018 at 13:25, on Zulip):

"regression: stack overflow on macosx with xcode 6.4" #55471 : left note. progress continues on PR #55617. @Oli may want to chime in with responses to my Q's there (either during the meeting, or on the issue itself).

pnkfelix (Nov 29 2018 at 13:26, on Zulip):

those were the P-high T-compiler issues. Next, 2018 Release milestone ones.

pnkfelix (Nov 29 2018 at 13:26, on Zulip):

first up: "strange absolute paths warning if ed2018 crate is build with warn( rust_2018_compatibility)" #56311

pnkfelix (Nov 29 2018 at 13:27, on Zulip):

tagged as P-high and left a comment (it appears to have been addressed and the fix is en route to landing).

pnkfelix (Nov 29 2018 at 13:29, on Zulip):

(and I already noted #56254 above, so that's all the Release milestone items)

pnkfelix (Nov 29 2018 at 13:29, on Zulip):

next up: beta-nominations

pnkfelix (Nov 29 2018 at 13:30, on Zulip):

in principle we really are supposed to discuss these live, so I'll defer all of them (there are four) to the meeting.

pnkfelix (Nov 29 2018 at 13:31, on Zulip):

oh but let me at least make sure they're tagged with T-compiler if need be

pnkfelix (Nov 29 2018 at 13:33, on Zulip):

(the stable-nominations list is currently empty, and in any case it would all be deferred to meeting same way that beta-nominations are anyway.)

pnkfelix (Nov 29 2018 at 13:36, on Zulip):

There's a couple of I-nominated things that were intended solely for the NLL meeting, not T-compiler.

pnkfelix (Nov 29 2018 at 13:36, on Zulip):

(from around three weeks ago when I went through and was trying to stir up some discussion there.)

pnkfelix (Nov 29 2018 at 13:37, on Zulip):

I'm going to un-nominate them now; they're tagged as NLL-deferred and the WG-nll will deal with them as a natural outcome of triaging the deferred issues.

pnkfelix (Nov 29 2018 at 13:44, on Zulip):

also, "Add a forever unstable opt-out of const qualification checks" #56123 had a nomination tag but I think that was just left over from last week. I left a comment and removed the tag.

pnkfelix (Nov 29 2018 at 13:47, on Zulip):

"rustc 1.30 beta 12 incremental compilation hangs on win 7" #54874 was only nominated four days ago, so I'll leave the tag. It should be given a priority tag though. I'll make it P-high for now to ensure we at least talk about it first, but I'm not sure if we may want to downgrade to P-medium. Anyway we'll discuss.

pnkfelix (Nov 29 2018 at 13:49, on Zulip):

(it would be nice to somehow determine if the previous bug is indeed a duplicate of issue #54627, which we already decided we would not attempt to fix ourselves...)

pnkfelix (Nov 29 2018 at 14:04, on Zulip):

okay I've basically gone through the main agenda that I always try to cover

pnkfelix (Nov 29 2018 at 14:06, on Zulip):

since I've got some time before the meeting, I'm going to take the addition step of going through the stage-to-stable regressions that are missing P-tags and add them (and make notes here about the ones I add). My main intent will be to add P-high to the cases where I want the issue to be addressed at the start of today's meeting. If I don't want discussion but still think issue is high priority, I'll say so in a comment on the issue itself (and here).

pnkfelix (Nov 29 2018 at 14:09, on Zulip):

"Internal Compiler Error in 1.30 and 1.31" #55846: its fixed and is E-needstest. left comment, tagged as P-medium.

pnkfelix (Nov 29 2018 at 14:09, on Zulip):

"Cargo doc explodes with ICE (rest of the tests seem to work?)" #55690: fixed and backported. tagged as P-high and closed as fixed.

pnkfelix (Nov 29 2018 at 14:11, on Zulip):

"1.30.0 fails to build for target powerpc-unknown-netbsd" #55465: I didn't know what P-tag to assign three weeks ago, and I still don't. I-nominating to determine P-value.

pnkfelix (Nov 29 2018 at 14:14, on Zulip):

"rustc 1.30.0's linker flavor inference is a non-backwards compat change to -Clinker" #55396 - I tagged this as T-compiler three weeks ago. Tagging as P-high.

pnkfelix (Nov 29 2018 at 14:16, on Zulip):

"Concrete type only used in impl Trait is unused" #55124 ; marking as P-medium and leaving a comment. But its also not tagged as T-compiler; should it be? We don't have a separate WG in charge of lints, right?

pnkfelix (Nov 29 2018 at 14:17, on Zulip):

"Breaking macro name resolution change in 1.29" #54783 - I have no idea what priority to assign for this. Marking as T-compiler and I-nominating.

pnkfelix (Nov 29 2018 at 14:21, on Zulip):

"35% performance regression in generated code since 1.24" #53833 . Hmm. "Old" performance bug. I don't really think P-high is warranted, and yet I think it'd be good to investigate...

pnkfelix (Nov 29 2018 at 14:21, on Zulip):

I'm going to mark #53833 as P-medium but also self assign it for some preliminary investigation.

pnkfelix (Nov 29 2018 at 14:24, on Zulip):

"Binary search performance regressed in 1.25" #53823 - marking as P-medium via same reasoning as #53833

pnkfelix (Nov 29 2018 at 14:26, on Zulip):

"'solaris' runtime stack guard conflicts with OS stack-clash mitigation" #52577 - I don't know what priority to assign here. But I also don't really want to add it today's agenda. So... I'm just going to silently (w.r.t. github comments) pretend that I didn't notice it on the list.

pnkfelix (Nov 29 2018 at 14:27, on Zulip):

remaining two non-prioritized stable-to-stable regressions (#52267 and #47321) are both T-libs, so skipping. (I'll let them figure out prioritization, if they want.)

pnkfelix (Nov 29 2018 at 14:28, on Zulip):

okay that was fun but I'm going to take a break now.

pnkfelix (Nov 29 2018 at 14:29, on Zulip):

Here is just one last advance warning to @T-compiler that our weekly meeting will actually start here in about 30 minutes.

nikomatsakis (Nov 29 2018 at 14:53, on Zulip):

"rustc 1.30.0's linker flavor inference is a non-backwards compat change to -Clinker" #55396 - I tagged this as T-compiler three weeks ago. Tagging as P-high.

I'm not sure I agree with the stability level of this one -- unused code is a pretty useful lint overall. Particularly as this is a regression, I think we should make an effort to fix it.

nikomatsakis (Nov 29 2018 at 14:54, on Zulip):

(In general, I think regressions in "warn by default" lints should be taken seriously)

nagisa (Nov 29 2018 at 15:01, on Zulip):

o/

pnkfelix (Nov 29 2018 at 15:01, on Zulip):

sorry little late to arrival, hi @T-compiler

pnkfelix (Nov 29 2018 at 15:02, on Zulip):

@nikomatsakis sorry, trying to parse your note on #55396: You don't agree about the regression-from-stable-to-stable regression? Or the P-high ?

nikomatsakis (Nov 29 2018 at 15:03, on Zulip):

wait

pnkfelix (Nov 29 2018 at 15:03, on Zulip):

or maybe you meant to refer to a different issue

nikomatsakis (Nov 29 2018 at 15:03, on Zulip):

I quoted the wrong one :)

nikomatsakis (Nov 29 2018 at 15:03, on Zulip):

"Concrete type only used in impl Trait is unused" #55124 ; marking as P-medium and leaving a comment. But its also not tagged as T-compiler; should it be? We don't have a separate WG in charge of lints, right?

this one

nikomatsakis (Nov 29 2018 at 15:03, on Zulip):

I think should maybe be P-high

pnkfelix (Nov 29 2018 at 15:03, on Zulip):

okay lets just do that first

pnkfelix (Nov 29 2018 at 15:04, on Zulip):

and then we'll start the meeting proper

pnkfelix (Nov 29 2018 at 15:04, on Zulip):

and can rediscuss when it naturally arises

nikomatsakis (Nov 29 2018 at 15:04, on Zulip):

ok

pnkfelix (Nov 29 2018 at 15:04, on Zulip):

so, first up, P-high

pnkfelix (Nov 29 2018 at 15:05, on Zulip):

the first item is "prohibit "two-phase borrows" with existing borrows?" #56254

pnkfelix (Nov 29 2018 at 15:05, on Zulip):

but as I noted above, I don't think we have to talk about this here.

pnkfelix (Nov 29 2018 at 15:05, on Zulip):

(my summary comment has the heart of my current feelings on the matter)

pnkfelix (Nov 29 2018 at 15:06, on Zulip):

so under assumption that no one objects to that, I'll move along to ....

pnkfelix (Nov 29 2018 at 15:06, on Zulip):

"NLL error on closure, but not on equivalent function" #55526

pnkfelix (Nov 29 2018 at 15:06, on Zulip):

as noted above, there's been no change on this. But I've self-assigned it and hope to look at it... tomorrow...?

pnkfelix (Nov 29 2018 at 15:07, on Zulip):

next: "regression: stack overflow on macosx with xcode 6.4" #55471

pnkfelix (Nov 29 2018 at 15:07, on Zulip):

@Oli continues to toil away on this

pnkfelix (Nov 29 2018 at 15:07, on Zulip):

and says they hope to have a fix up this week

pnkfelix (Nov 29 2018 at 15:08, on Zulip):

next: "rustc 1.30.0's linker flavor inference is a non-backwards compat change to -Clinker" #55396

pnkfelix (Nov 29 2018 at 15:08, on Zulip):

hey look at that. :wink:

nikomatsakis (Nov 29 2018 at 15:08, on Zulip):

as noted above, there's been no change on this. But I've self-assigned it and hope to look at it... tomorrow...?

I think I know what the problem is here @pnkfelix, or at least I have some suspicions. I'll leave a few notes.

pnkfelix (Nov 29 2018 at 15:08, on Zulip):

so there's been no movement on #55396 since the end of october

pnkfelix (Nov 29 2018 at 15:09, on Zulip):

@Alex Crichton has proposed a path forward here

nagisa (Nov 29 2018 at 15:09, on Zulip):

I’m not sure why the detection is necessary at all.

pnkfelix (Nov 29 2018 at 15:09, on Zulip):

@nagisa well alex said " it fixed a number of issues and empowered a number of embedded use cases to become stable" ?

nagisa (Nov 29 2018 at 15:09, on Zulip):

-Clinker-flavour exists

pnkfelix (Nov 29 2018 at 15:09, on Zulip):

-C, or -Z ?

nikomatsakis (Nov 29 2018 at 15:10, on Zulip):

seems like the standing plan of action is probably still good,

nikomatsakis (Nov 29 2018 at 15:10, on Zulip):

although the window to change inference rules is ... narrow?

nikomatsakis (Nov 29 2018 at 15:10, on Zulip):

i.e., we risk breaking other things now

nagisa (Nov 29 2018 at 15:10, on Zulip):

might be -Z, but I don’t see any reason why we couldn’t keep the old default for that and stabilise this flag

nikomatsakis (Nov 29 2018 at 15:10, on Zulip):

(was there an RFC on this?)

nikomatsakis (Nov 29 2018 at 15:10, on Zulip):

I don't remember anything about it

pnkfelix (Nov 29 2018 at 15:11, on Zulip):

I doubt it, and I wonder if we need one (an RFC)

pnkfelix (Nov 29 2018 at 15:11, on Zulip):

in any case

pnkfelix (Nov 29 2018 at 15:11, on Zulip):

I don't actually want to try to resolve these Q's in this meeting

nikomatsakis (Nov 29 2018 at 15:11, on Zulip):

+1

pnkfelix (Nov 29 2018 at 15:11, on Zulip):

instead, I want to know if anyone wants to take point on this

pnkfelix (Nov 29 2018 at 15:11, on Zulip):

i.e. does anyone volunteer to get assigned to this ticket?

pnkfelix (Nov 29 2018 at 15:11, on Zulip):

I can take it as a fallback if necessary, but to be frank its not a topic area I'm terribly ... interested in...

nikomatsakis (Nov 29 2018 at 15:12, on Zulip):

heh, I was going to say the same :)

nagisa (Nov 29 2018 at 15:12, on Zulip):

I might be interested working on this, and so could be anybody else from embedded.

nagisa (Nov 29 2018 at 15:12, on Zulip):

but I have a ton of things that are more interesting as well :slight_smile:

pnkfelix (Nov 29 2018 at 15:12, on Zulip):

@nagisa maybe you can take it for now and if youc an find someone else to take it, you delegate accordingly?

nagisa (Nov 29 2018 at 15:12, on Zulip):

Sure

davidtwco (Nov 29 2018 at 15:13, on Zulip):

I could take a look, I'm not familiar with that area at all though, I could coordinate with @nagisa to get me started on it?

Split off to new topic.

nikomatsakis (Nov 29 2018 at 15:13, on Zulip):

@nagisa ps https://github.com/rust-lang/rust/pull/52101 has some interesting side-notes; my suspicion is that it enabled embedded use cases in part because -Clinker-flavor was not available on stable, though

pnkfelix (Nov 29 2018 at 15:13, on Zulip):

sounds good. feel free to fork off a side topic

nagisa (Nov 29 2018 at 15:13, on Zulip):

@davidtwco I can definitely mentor :slight_smile:

nikomatsakis (Nov 29 2018 at 15:13, on Zulip):

personally, I feel like this whole thing deserved an RFC

pnkfelix (Nov 29 2018 at 15:13, on Zulip):

okay last P-high matter: "rustc 1.30 beta 12 incremental compilation hangs on win 7" #54874

pnkfelix (Nov 29 2018 at 15:14, on Zulip):

do we have any Windows 7 users present here in this meeting today?

mw (Nov 29 2018 at 15:14, on Zulip):

I've been working on Windows 10 a bit lately and never encountered the problem there.

nikomatsakis (Nov 29 2018 at 15:14, on Zulip):

I wonder if this has to do with our jobserver support?

nikomatsakis (Nov 29 2018 at 15:14, on Zulip):

not sure why else we'd be waiting on a condition variable

pnkfelix (Nov 29 2018 at 15:14, on Zulip):

it used to be my Windows variant of choice when working in a VM...

pnkfelix (Nov 29 2018 at 15:15, on Zulip):

well, in this case I again don't want to try to resolve Q's in meeting. Instead, I want to either 1. downgrade to P-medium, or 2. find someone to take point on it.

pnkfelix (Nov 29 2018 at 15:15, on Zulip):

i.e. if no one volunteers, I'm going to downgrade.

pnkfelix (Nov 29 2018 at 15:16, on Zulip):

(which admittedly seems like a poor policy)

nagisa (Nov 29 2018 at 15:16, on Zulip):

W7 EOL’d mainstream support 3 years ago

pnkfelix (Nov 29 2018 at 15:16, on Zulip):

I was just about to mention the EOL schedule, yeah

nagisa (Nov 29 2018 at 15:16, on Zulip):

though extended support is still good till 2020

mw (Nov 29 2018 at 15:16, on Zulip):

does not sound like tier 1 then

pnkfelix (Nov 29 2018 at 15:16, on Zulip):

okay lets downgrade

nikomatsakis (Nov 29 2018 at 15:17, on Zulip):

still https://github.com/rust-lang/rust/issues/54627 looks quite similar

nikomatsakis (Nov 29 2018 at 15:17, on Zulip):

I guess that's also "arm"

nikomatsakis (Nov 29 2018 at 15:17, on Zulip):

but still a more important arch-OS combo

pnkfelix (Nov 29 2018 at 15:17, on Zulip):

(which I also tagged as P-medium, for different reasons)

nikomatsakis (Nov 29 2018 at 15:18, on Zulip):

ah, hmm, so that one is related to LLVM specifically

nikomatsakis (Nov 29 2018 at 15:18, on Zulip):

might be worth at least asking the #54874 poster to attempt those workarounds

nikomatsakis (Nov 29 2018 at 15:18, on Zulip):

I don't know if they mentioned trying that

pnkfelix (Nov 29 2018 at 15:18, on Zulip):

they were aware of it

pnkfelix (Nov 29 2018 at 15:18, on Zulip):

the issue, taht is

pnkfelix (Nov 29 2018 at 15:18, on Zulip):

since they linked to it

pnkfelix (Nov 29 2018 at 15:19, on Zulip):

(or someone added a link to the description, at least)

pnkfelix (Nov 29 2018 at 15:19, on Zulip):

okay well that's all the P-high issues.

pnkfelix (Nov 29 2018 at 15:19, on Zulip):

next, Release milestone issues

pnkfelix (Nov 29 2018 at 15:20, on Zulip):

I already went through these above

pnkfelix (Nov 29 2018 at 15:20, on Zulip):

nothing we need to talk about in the meeting

nikomatsakis (Nov 29 2018 at 15:20, on Zulip):

ah, hmm, so that one is related to LLVM specifically

actually reading in more detail I don't think they are related

pnkfelix (Nov 29 2018 at 15:20, on Zulip):

next: beta-nominations

nagisa (Nov 29 2018 at 15:21, on Zulip):

ah, hmm, so that one is related to LLVM specifically

actually reading in more detail I don't think they are related

That is entirely a bug in LLVM optimisation.

pnkfelix (Nov 29 2018 at 15:21, on Zulip):

(i know for some of these @nikomatsakis specifically untagged their beta-accepted labels to ensure they'd be discussed. So, listen up soldiers!) :wink:

pnkfelix (Nov 29 2018 at 15:21, on Zulip):

first up: "rustc_codegen_llvm: don't overalign loads of pair operands." #56329

Ariel Ben-Yehuda (Nov 29 2018 at 15:22, on Zulip):

is it OK to have Windows 7 as T2?

Ariel Ben-Yehuda (Nov 29 2018 at 15:22, on Zulip):

how popular is it?

nagisa (Nov 29 2018 at 15:22, on Zulip):

how popular is it?

fairly.

Ariel Ben-Yehuda (Nov 29 2018 at 15:22, on Zulip):

among Rust installs?

nikomatsakis (Nov 29 2018 at 15:22, on Zulip):

is it OK to have Windows 7 as T2?

I was pondering this question myself

pnkfelix (Nov 29 2018 at 15:22, on Zulip):

among Rust installs?

hmm we might even have data about that, no?

Ariel Ben-Yehuda (Nov 29 2018 at 15:23, on Zulip):

yea

Ariel Ben-Yehuda (Nov 29 2018 at 15:23, on Zulip):

maybe someone on the release team knows?

pnkfelix (Nov 29 2018 at 15:23, on Zulip):

Shall I open an actual bug to have the infra team get an answer to that Q?

nikomatsakis (Nov 29 2018 at 15:23, on Zulip):

I was just checking the survey

nikomatsakis (Nov 29 2018 at 15:23, on Zulip):

but it isn't broken down by windows version

nagisa (Nov 29 2018 at 15:23, on Zulip):

I want to accept both overalign fixes

pnkfelix (Nov 29 2018 at 15:23, on Zulip):

I'll open a bug

Ariel Ben-Yehuda (Nov 29 2018 at 15:23, on Zulip):

@pnkfelix If we want to downgrade it to T2, looks like a good idea

nagisa (Nov 29 2018 at 15:23, on Zulip):

I see no reason to not do so and they fix genuine UB.

nikomatsakis (Nov 29 2018 at 15:23, on Zulip):

given that they affect num-cpus, which I think is a dependencvy of rayon and no doubt other popular crates, seems good to me

Ariel Ben-Yehuda (Nov 29 2018 at 15:24, on Zulip):

yea they look fairly non-risky and they fix I-wrong that is risky

Ariel Ben-Yehuda (Nov 29 2018 at 15:25, on Zulip):

e.g. I'm quite sure they can cause crashes on Sparc

nagisa (Nov 29 2018 at 15:25, on Zulip):

There is a POC that crashes on x86_64

nagisa (Nov 29 2018 at 15:25, on Zulip):

sigsegvs.

nagisa (Nov 29 2018 at 15:25, on Zulip):

on one of the issues.

nikomatsakis (Nov 29 2018 at 15:26, on Zulip):

seems like we should approve

Ariel Ben-Yehuda (Nov 29 2018 at 15:26, on Zulip):

couldn't find it, but yea UB can cause that

pnkfelix (Nov 29 2018 at 15:26, on Zulip):

okay, so the two overalign fixes are #56329 and #56300

pnkfelix (Nov 29 2018 at 15:26, on Zulip):

Sounds like we'll accept both.

Ariel Ben-Yehuda (Nov 29 2018 at 15:26, on Zulip):

non-theoretical UB = stable point release for me

nagisa (Nov 29 2018 at 15:27, on Zulip):

https://github.com/rust-lang/rust/pull/56329#issuecomment-442830094 has the POc

pnkfelix (Nov 29 2018 at 15:28, on Zulip):

okay, there are two remaining beta-nominated issues

pnkfelix (Nov 29 2018 at 15:28, on Zulip):

(I've marked #56329 and #56300 as beta-accepted)

pnkfelix (Nov 29 2018 at 15:28, on Zulip):

next is: "resolve: Fix false-positives from lint absolute_paths_not_starting_with_crate" #56322

pnkfelix (Nov 29 2018 at 15:28, on Zulip):

seems low risk

pnkfelix (Nov 29 2018 at 15:29, on Zulip):

any objections?

nikomatsakis (Nov 29 2018 at 15:29, on Zulip):

I think we should backport it

nikomatsakis (Nov 29 2018 at 15:29, on Zulip):

(low-risk, very confusing)

pnkfelix (Nov 29 2018 at 15:30, on Zulip):

okay, #56322 is marked beta-accepted

pnkfelix (Nov 29 2018 at 15:30, on Zulip):

last beta-nom: "resolve: Extern prelude is for type namespace only" #56264

Pietro Albini (Nov 29 2018 at 15:31, on Zulip):

[I'll prepare a rollup later today]

pnkfelix (Nov 29 2018 at 15:31, on Zulip):

seems like @nikomatsakis has already presented an argument for beta-accept of #56264

pnkfelix (Nov 29 2018 at 15:32, on Zulip):

this seems even lower risk than the last one

nikomatsakis (Nov 29 2018 at 15:32, on Zulip):

similarly, I think we should backport this; it came up in the internals thread too

nikomatsakis (Nov 29 2018 at 15:32, on Zulip):

seems like a shame to ask people for feedback

nikomatsakis (Nov 29 2018 at 15:32, on Zulip):

and then fail to fix the bugs they find :)

pnkfelix (Nov 29 2018 at 15:32, on Zulip):

marking as beta-accepted

pnkfelix (Nov 29 2018 at 15:33, on Zulip):

okay. that's all the beta-noms. there are no stable-noms.

pnkfelix (Nov 29 2018 at 15:33, on Zulip):

next up: stable-to-beta regressions

pnkfelix (Nov 29 2018 at 15:34, on Zulip):

we already discussed #56311

pnkfelix (Nov 29 2018 at 15:34, on Zulip):

next: "ICE when variable shadows ::-qualified use'd crate" #56263

pnkfelix (Nov 29 2018 at 15:34, on Zulip):

seems P-high, no?

pnkfelix (Nov 29 2018 at 15:34, on Zulip):

oh never mind

pnkfelix (Nov 29 2018 at 15:34, on Zulip):

this is fixed and I just reopened it

pnkfelix (Nov 29 2018 at 15:35, on Zulip):

:facepalm:

nikomatsakis (Nov 29 2018 at 15:35, on Zulip):

yep and we accepted fix for backport :)

pnkfelix (Nov 29 2018 at 15:35, on Zulip):

okay then we've covered everything relevant to us on this list already

pnkfelix (Nov 29 2018 at 15:35, on Zulip):

next up, stable-to-nightly regressions

pnkfelix (Nov 29 2018 at 15:35, on Zulip):

first: "ICE: Encountered a freshend type with id 0 but our counter is only at 0" #56288

pnkfelix (Nov 29 2018 at 15:36, on Zulip):

I was just recently wrestling with our 0 values in freshened types...

pnkfelix (Nov 29 2018 at 15:36, on Zulip):

I'm willing to take this is no one else want it

pnkfelix (Nov 29 2018 at 15:37, on Zulip):

and I think it should be P-high.

Ariel Ben-Yehuda (Nov 29 2018 at 15:37, on Zulip):

assign to me?

Ariel Ben-Yehuda (Nov 29 2018 at 15:37, on Zulip):

I'll have some time

nikomatsakis (Nov 29 2018 at 15:37, on Zulip):

I can also look, maybe related to some of the recent traits stuff changes

nikomatsakis (Nov 29 2018 at 15:38, on Zulip):

but I'm happy for @Ariel Ben-Yehuda to look

nikomatsakis (Nov 29 2018 at 15:38, on Zulip):

(particularly as next week is moz all hands)

pnkfelix (Nov 29 2018 at 15:38, on Zulip):

next up: "rayon_1_0_0::sort perf regression" #56283

pnkfelix (Nov 29 2018 at 15:38, on Zulip):

so ... as a policy issue, I'm curious how we should assign P- labels to performance regressions

pnkfelix (Nov 29 2018 at 15:39, on Zulip):

(this was something I encountered a couple times during my triage.)

nikomatsakis (Nov 29 2018 at 15:39, on Zulip):

I was just thinking that we would ideally have a group focused on perf

pnkfelix (Nov 29 2018 at 15:39, on Zulip):

I'm inclined to tag them initially as P-high until someone does the work to identify whether the actual impact is limited.

nikomatsakis (Nov 29 2018 at 15:39, on Zulip):

ps, @Alex Crichton tells me we have no figures on Windows 7 specifically

pnkfelix (Nov 29 2018 at 15:39, on Zulip):

(I suppose another option would indeed be to delegate the prioritization to another subteam/WG)

nikomatsakis (Nov 29 2018 at 15:39, on Zulip):

I'm inclined to tag them initially as P-high until someone does the work to identify whether the actual impact is limited.

seems reasonable for now

pnkfelix (Nov 29 2018 at 15:40, on Zulip):

anyway I think for now I will tag #56283 as P-high and self-assign

pnkfelix (Nov 29 2018 at 15:40, on Zulip):

I'll have time since I'm not going to mozlando

nagisa (Nov 29 2018 at 15:40, on Zulip):

WG-codegen is the perf-oriented workgroup (formed last all-hands)

pnkfelix (Nov 29 2018 at 15:40, on Zulip):

and it seems like it might fit my zombie state.

nikomatsakis (Nov 29 2018 at 15:40, on Zulip):

yes, it's just not (afaik) esp active

nikomatsakis (Nov 29 2018 at 15:40, on Zulip):

( we should perhaps discuss separately )

nikomatsakis (Nov 29 2018 at 15:41, on Zulip):

(sort of ties into the things we've been discussing at steering meeting...)

pnkfelix (Nov 29 2018 at 15:41, on Zulip):

sometimes performance is not about codegen, too

pnkfelix (Nov 29 2018 at 15:41, on Zulip):

e.g. stdlib code regresses...

pnkfelix (Nov 29 2018 at 15:41, on Zulip):

anyway...

pnkfelix (Nov 29 2018 at 15:41, on Zulip):

that's all the stable-to-nightly regressions

pnkfelix (Nov 29 2018 at 15:42, on Zulip):

there's nothing tagged as Waiting-for-team for us

pnkfelix (Nov 29 2018 at 15:42, on Zulip):

so next up is I-nominated T-compiler

pnkfelix (Nov 29 2018 at 15:42, on Zulip):

I already said we're skipping "prohibit "two-phase borrows" with existing borrows?" #56254 (its nominated for T-lang, at most)

pnkfelix (Nov 29 2018 at 15:43, on Zulip):

next, I nominated "1.30.0 fails to build for target powerpc-unknown-netbsd" #55465

pnkfelix (Nov 29 2018 at 15:43, on Zulip):

because I don't know what P- label to give it

pnkfelix (Nov 29 2018 at 15:43, on Zulip):

does anyone object to P-high?

pnkfelix (Nov 29 2018 at 15:44, on Zulip):

(and if not, does anyone want to take point on it...?)

nagisa (Nov 29 2018 at 15:44, on Zulip):

I… recall seeing similar issue recently that hadn’t as many comments...

nikomatsakis (Nov 29 2018 at 15:45, on Zulip):

I don't really understand what's going on here

pnkfelix (Nov 29 2018 at 15:45, on Zulip):

from skimming the comments ... is this actually a libs issue...?

pnkfelix (Nov 29 2018 at 15:45, on Zulip):

I'm the one who added T-compiler based on very little info a month ago

nikomatsakis (Nov 29 2018 at 15:45, on Zulip):

there's a PR linked from the issue -- does this PR fix the issue? or cause it?

nikomatsakis (Nov 29 2018 at 15:46, on Zulip):

it seems like it's not compiler per se

mw (Nov 29 2018 at 15:46, on Zulip):

(@mw needs to run... :wave:)

nikomatsakis (Nov 29 2018 at 15:46, on Zulip):

but kind of in that grey area of "Low level lib"

pnkfelix (Nov 29 2018 at 15:46, on Zulip):

I think the issue might be that the compiler isn't pulling in the up-to-date crates.io libc ...?

pnkfelix (Nov 29 2018 at 15:46, on Zulip):

but i misunderstand the thread

pnkfelix (Nov 29 2018 at 15:47, on Zulip):

okay well at this point, I am going to tag as P-high and assign to self to dissect what's has gone on here.

nikomatsakis (Nov 29 2018 at 15:47, on Zulip):

seems like @eddyb might know :)

pnkfelix (Nov 29 2018 at 15:47, on Zulip):

(Though I'm guessing @eddyb has already been involved on thread and might be able to better illuminate matters)

nagisa (Nov 29 2018 at 15:48, on Zulip):

Well the error appears to make it a libs issue for sure :slight_smile:

nikomatsakis (Nov 29 2018 at 15:48, on Zulip):

or maybe @memoryruins actually -- are they around? :)

pnkfelix (Nov 29 2018 at 15:48, on Zulip):

okay, last nominated issue for today: "Breaking macro name resolution change in 1.29" #54783

nagisa (Nov 29 2018 at 15:48, on Zulip):

(unless we are willing to play with the idea that our typeck/infer are broken despite all evidence going against it)

pnkfelix (Nov 29 2018 at 15:49, on Zulip):

I again didn't and don't know what P-label to assign here (to #54783)

pnkfelix (Nov 29 2018 at 15:49, on Zulip):

I'm guessing P-high

eddyb (Nov 29 2018 at 15:49, on Zulip):

I don't actually know

pnkfelix (Nov 29 2018 at 15:49, on Zulip):

its already assigned to @Vadim Petrochenkov ; I assume they forgot to follow up on it.

pnkfelix (Nov 29 2018 at 15:50, on Zulip):

So unless someone objects, I'm going to tag #54783 as P-high and then hope that @Vadim Petrochenkov posts an update about it in the near term.

eddyb (Nov 29 2018 at 15:50, on Zulip):

(the libc thing looks like a rebase + libc update mixup and @Alex Crichton might be able to help, otherwise, idk, libs+release team more than compiler team)

nikomatsakis (Nov 29 2018 at 15:50, on Zulip):

seems good (to tag #54783 as P-high)

pnkfelix (Nov 29 2018 at 15:51, on Zulip):

okay, next up

eddyb (Nov 29 2018 at 15:51, on Zulip):

(in related news does anyone know where to get a hold of the libs team these days?

nagisa (Nov 29 2018 at 15:51, on Zulip):

(in related news does anyone know where to get a hold of the libs team these days?

… discord?

pnkfelix (Nov 29 2018 at 15:51, on Zulip):

stable-to-stable regressions not already labelled with P-*

pnkfelix (Nov 29 2018 at 15:52, on Zulip):

((eventually all will discover the glory of Zulip topics.....))

pnkfelix (Nov 29 2018 at 15:52, on Zulip):

(((and also I don't know about you all but I have not been able to get my head around the search UX for discord)))

nikomatsakis (Nov 29 2018 at 15:52, on Zulip):

(actually there doesn't seem to be a discord section for libs?)

varkor (Nov 29 2018 at 15:52, on Zulip):

(I think T-libs is on IRC)

nikomatsakis (Nov 29 2018 at 15:53, on Zulip):

oh that reminds me that I wanted to discuss de-voicing #rustc -- many other teams (e.g., infra, release) have done the same

pnkfelix (Nov 29 2018 at 15:53, on Zulip):

so there are five stable-to-stable regressions that are not prioritized

nikomatsakis (Nov 29 2018 at 15:53, on Zulip):

but I can resurrect that e-mail chain instead

nikomatsakis (Nov 29 2018 at 15:53, on Zulip):

so there are five stable-to-stable regressions that are not prioritized

most seem not to be T-compiler

pnkfelix (Nov 29 2018 at 15:53, on Zulip):

right

pnkfelix (Nov 29 2018 at 15:53, on Zulip):

the only two might be

pnkfelix (Nov 29 2018 at 15:53, on Zulip):

"Binary search performance regressed in 1.25" #53823

pnkfelix (Nov 29 2018 at 15:54, on Zulip):

and "'solaris' runtime stack guard conflicts with OS stack-clash mitigation" #52577

pnkfelix (Nov 29 2018 at 15:54, on Zulip):

(#52577 is tagged as T-libs, but I have the impression that we could address it if we reverted PR #48575...)

nagisa (Nov 29 2018 at 15:54, on Zulip):

#52577 is another of those PRs that should have landed with more look from T-compiler

pnkfelix (Nov 29 2018 at 15:54, on Zulip):

so, first up, #53823

pnkfelix (Nov 29 2018 at 15:55, on Zulip):

following the same reasoning that I used before for performance regressions, I'm going to tag #53823 as P-high for the initial investigation of the scope of the regression

pnkfelix (Nov 29 2018 at 15:56, on Zulip):

that leaves "'solaris' runtime stack guard conflicts with OS stack-clash mitigation" #52577

pnkfelix (Nov 29 2018 at 15:57, on Zulip):

it sounded like ishitatsuyuki was amenable to us going back to having an extra thread by default

pnkfelix (Nov 29 2018 at 15:57, on Zulip):

(if you look at the conversation on PR #48575)

nagisa (Nov 29 2018 at 15:57, on Zulip):

There is (technically) a simple patch to T-libs…

pnkfelix (Nov 29 2018 at 15:57, on Zulip):

ah and that is why its tagged with that label?

nagisa (Nov 29 2018 at 15:58, on Zulip):

the patch involves thread setup that’s implemented in libstd

nagisa (Nov 29 2018 at 15:58, on Zulip):

makes me wonder whether whatever we do in rustc is likely to occur in natural manner elsewhere.

pnkfelix (Nov 29 2018 at 15:59, on Zulip):

@nagisa how about if I assign both myself and you to this

pnkfelix (Nov 29 2018 at 15:59, on Zulip):

I'm interested in exploring reverting PR #48575, but it sounds like we might want to also look at the other mitigation mentioned in the comments on #52577

nagisa (Nov 29 2018 at 15:59, on Zulip):

I don’t really want to deal with this… I tried to bring attention to this when it actually mattered, but now I cannot really care about it anymore

pnkfelix (Nov 29 2018 at 15:59, on Zulip):

and it might be useful for you and I to collaborate further on it.

pnkfelix (Nov 29 2018 at 16:00, on Zulip):

hmm okay that is a fair response.

nikomatsakis (Nov 29 2018 at 16:00, on Zulip):

I can't really tell if we have an idea what the fix should be

pnkfelix (Nov 29 2018 at 16:00, on Zulip):

Though "when it actually mattered" seems like not the right viewpoint. Just because we broke Solaris doesn't mean we shouldn't fix it...

nagisa (Nov 29 2018 at 16:01, on Zulip):

FWIW I don’t think this is high priority

pnkfelix (Nov 29 2018 at 16:01, on Zulip):

that's is the other importnat Q

nagisa (Nov 29 2018 at 16:01, on Zulip):

as upstream has applied some sort of patch to their package and it works for them

pnkfelix (Nov 29 2018 at 16:01, on Zulip):

what Prioirty to assign

nagisa (Nov 29 2018 at 16:02, on Zulip):

this is their patch

pnkfelix (Nov 29 2018 at 16:02, on Zulip):

ooookay well we are now out of time

pnkfelix (Nov 29 2018 at 16:02, on Zulip):

I'm going to self-assign

pnkfelix (Nov 29 2018 at 16:04, on Zulip):

that's everything I think in our standing agenda

pnkfelix (Nov 29 2018 at 16:04, on Zulip):

and like I said, we are overtime (but not as much as usual!!)

pnkfelix (Nov 29 2018 at 16:04, on Zulip):

oh one more thing

pnkfelix (Nov 29 2018 at 16:04, on Zulip):

I want to say a big thank you to @nagisa for running the meeting while I was out on leave

pnkfelix (Nov 29 2018 at 16:05, on Zulip):

hmm

pnkfelix (Nov 29 2018 at 16:05, on Zulip):

We didn't assign someone to #55124 "Concrete type only used in impl Trait is unused"

pnkfelix (Nov 29 2018 at 16:06, on Zulip):

because I forgot to tag it as T-compiler

pnkfelix (Nov 29 2018 at 16:06, on Zulip):

whoops

nikomatsakis (Nov 29 2018 at 16:06, on Zulip):

I guess I'll take it ;)

nikomatsakis (Nov 29 2018 at 16:06, on Zulip):

since I complained

nikomatsakis (Nov 29 2018 at 16:06, on Zulip):

at least I'll try to figure out what it's about

pnkfelix (Nov 29 2018 at 16:07, on Zulip):

okay bye everyone!!!

pnkfelix (Nov 29 2018 at 16:07, on Zulip):

(and thanks for attending!)

Last update: Nov 22 2019 at 05:55UTC