Stream: t-compiler

Topic: weekly meeting 2019-06-13 #54818


pnkfelix (Jun 13 2019 at 12:33, on Zulip):

Hi @T-compiler/meeting ; the triage meeting will be starting in 1 hour 27 minutes

pnkfelix (Jun 13 2019 at 12:35, on Zulip):

I will be doing pre-triage in a parallel topic

pnkfelix (Jun 13 2019 at 12:39, on Zulip):

@nagisa are you around today? Just wondering if @WG-llvm has anything to report.

pnkfelix (Jun 13 2019 at 12:40, on Zulip):

Likewise, @Taylor Cramer or @nikomatsakis : is there anything to report at triage meeting regarding @WG-async-foundations ?

nikomatsakis (Jun 13 2019 at 12:46, on Zulip):

@pnkfelix yep, I can do that

Nikita Popov (Jun 13 2019 at 12:56, on Zulip):

On the LLVM side, I think the only thing that happened is that @varkor switched float min/max to use LLVM intrinsics, after I fixed a small codegen issue on x86 (#18384). I'm working to get CVP nowrap inference reenabled in LLVM, but fell into the rabbit hole that is LFTR (D60935, D62939). I'm now very depressed about poison and undef semantics in LLVM.

pnkfelix (Jun 13 2019 at 14:02, on Zulip):

Okay hi @T-compiler/meeting

pnkfelix (Jun 13 2019 at 14:03, on Zulip):

(sorry for not using the right format for notifying everyone in the message up top; that was a cut-and-paste mistake that I'll need to fix in my template.)

pnkfelix (Jun 13 2019 at 14:04, on Zulip):

so, first of all: Any announcements?

pnkfelix (Jun 13 2019 at 14:08, on Zulip):

Here's my "announcement": today I want to make sure we traverse the I-nominated list. We forgot to do it last week, and I do not know the last time that we did do a full traversal of it.

pnkfelix (Jun 13 2019 at 14:08, on Zulip):

(I didn't quite finish all the pre-triage; there were still six nominated unprioritized bugs that I didn't traverse; traversing the I-nominated list will just have to include those...)

pnkfelix (Jun 13 2019 at 14:09, on Zulip):

we have 37 open P-high issues, 15 of which are unassigned

pnkfelix (Jun 13 2019 at 14:10, on Zulip):

there are also 7 open stable-to-beta regressions for our team with no prioritization. There may or may not be overlap between those and the P-high issues.

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

But lets get the really easy thing out the way: beta-nomination(s)

Zoxc (Jun 13 2019 at 14:11, on Zulip):

TyCtxt has one less lifetime, that's probably quite noticeable though =P

pnkfelix (Jun 13 2019 at 14:12, on Zulip):

beta-nomination: "Handle index out of bound errors during const eval without panic" #61598

pnkfelix (Jun 13 2019 at 14:13, on Zulip):

(... its possible there was an implicit stable-nomination intended for #61598, given this comment)

pnkfelix (Jun 13 2019 at 14:14, on Zulip):

(but let's resolve the beta backport question first)

pnkfelix (Jun 13 2019 at 14:15, on Zulip):

the information on the bug being fixed (#61595) nor the PR itself (#61598) does not indicate when this regression was injected.

pnkfelix (Jun 13 2019 at 14:16, on Zulip):

The main benefit of a backport here is that you get a nicer diagnostic for your erroneous code, right?

pnkfelix (Jun 13 2019 at 14:16, on Zulip):

It doesn't seem like anyone present is interested in arguing for a backport.

pnkfelix (Jun 13 2019 at 14:16, on Zulip):

So I'm just going to decline it in that case

pnkfelix (Jun 13 2019 at 14:16, on Zulip):

hup I just saw a vote for backport from @Esteban Küber

pnkfelix (Jun 13 2019 at 14:17, on Zulip):

@Esteban Küber : You want to say anything to sway @Zoxc (or me) ?

Esteban Küber (Jun 13 2019 at 14:17, on Zulip):

It's an ice, the nicer diagnostic is just a consequence of not iceing

Zoxc (Jun 13 2019 at 14:17, on Zulip):

It's just seems like a nice to have thing. We seem to emit a relevant error already?

Esteban Küber (Jun 13 2019 at 14:17, on Zulip):

Ices are bad for rls users

nikomatsakis (Jun 13 2019 at 14:18, on Zulip):

I lean towards fixing it. It's a regression and the patch is small and seems "obviously ok"

Esteban Küber (Jun 13 2019 at 14:18, on Zulip):

Doesn't affect people using rustc or cargo

pnkfelix (Jun 13 2019 at 14:18, on Zulip):

Ices are bad for rls users

(maybe this is something we should consider addressing, unless we can seriously up our game when it comes to fixing ICE's)

nikomatsakis (Jun 13 2019 at 14:18, on Zulip):

(it is stable to stable, which I think we've traditionally considered lesser priority, but the point about RLS I found persuasive)

Zoxc (Jun 13 2019 at 14:19, on Zulip):

Doesn't RLS recover from ICEs?

pnkfelix (Jun 13 2019 at 14:20, on Zulip):

@Esteban Küber : were you saying ICE's are bad for RLS users because you miss out on errors further in compilation that you'd normally get from compilre internally recovering?

Esteban Küber (Jun 13 2019 at 14:20, on Zulip):

Not gracefully in large codebases

pnkfelix (Jun 13 2019 at 14:20, on Zulip):

or is there something deeper / more severe that goes wrong?

Esteban Küber (Jun 13 2019 at 14:20, on Zulip):

That argument would also go for cli users

pnkfelix (Jun 13 2019 at 14:20, on Zulip):

(true)

centril (Jun 13 2019 at 14:20, on Zulip):

Imo if something is risk free then backporting is OK when it does have value to do so

centril (Jun 13 2019 at 14:21, on Zulip):

and it does have value to do so

pnkfelix (Jun 13 2019 at 14:21, on Zulip):

Okay well this PR does seem pretty low-risk to me.

centril (Jun 13 2019 at 14:21, on Zulip):

(imo let's backport)

pnkfelix (Jun 13 2019 at 14:21, on Zulip):

we probably should firm up our guidelines/policies when it comes to backports

pnkfelix (Jun 13 2019 at 14:21, on Zulip):

which would be good to incorporate into the upcoming planning meeting regarding triage/prioritization/etc

Esteban Küber (Jun 13 2019 at 14:22, on Zulip):

I've seen iced make the rls crash and start reindexing from scratch which block my editor for a significant amount of time on a big project im in

pnkfelix (Jun 13 2019 at 14:22, on Zulip):

Okay, approved for backport. Lets move on

pnkfelix (Jun 13 2019 at 14:23, on Zulip):

there are 3 PR's waiting on our team

pnkfelix (Jun 13 2019 at 14:23, on Zulip):

"Stabilize support for Profile-guided Optimization" #61268 -- may just be an artifact of FCP process?

pnkfelix (Jun 13 2019 at 14:24, on Zulip):

no, that was false

nikomatsakis (Jun 13 2019 at 14:24, on Zulip):

afaik that's ready to go

centril (Jun 13 2019 at 14:24, on Zulip):

@pnkfelix yes, things in FCP should be -waiting-on-team

pnkfelix (Jun 13 2019 at 14:24, on Zulip):

but we approved it? 9 days ago?

centril (Jun 13 2019 at 14:24, on Zulip):

once it finishes I'll just r=alex

centril (Jun 13 2019 at 14:25, on Zulip):

@pnkfelix FCP has not finished

pnkfelix (Jun 13 2019 at 14:25, on Zulip):

okay maybe we need a S-waiting-on-universe

pnkfelix (Jun 13 2019 at 14:25, on Zulip):

next

pnkfelix (Jun 13 2019 at 14:25, on Zulip):

"rustdoc: make #[doc(include)] relative to the containing file" #60938

centril (Jun 13 2019 at 14:25, on Zulip):

(waiting on fcp rather)

pnkfelix (Jun 13 2019 at 14:26, on Zulip):

seems like this might need an owner within T-compiler responsible for review/feedback

pnkfelix (Jun 13 2019 at 14:27, on Zulip):

but it also doesn't seem worth blocking this meeting on finding that "volunteer"

pnkfelix (Jun 13 2019 at 14:27, on Zulip):

last: "Turn HIR indexing into a query" #59064

pnkfelix (Jun 13 2019 at 14:28, on Zulip):

I don't see any record here of progress on an end-to-end plan here

pnkfelix (Jun 13 2019 at 14:28, on Zulip):

do we need to dedicate a planning meeting to this topic?

pnkfelix (Jun 13 2019 at 14:28, on Zulip):

I'll suggest that on the PR itself.

nikomatsakis (Jun 13 2019 at 14:28, on Zulip):

I think so, I was just saying that to @eddyb yesterday

nikomatsakis (Jun 13 2019 at 14:28, on Zulip):

(in an offhand conversation...)

nikomatsakis (Jun 13 2019 at 14:28, on Zulip):

if I recall @mw had some specific concerns

nikomatsakis (Jun 13 2019 at 14:28, on Zulip):

(also)

pnkfelix (Jun 13 2019 at 14:28, on Zulip):

okay, now the fun stuff

nikomatsakis (Jun 13 2019 at 14:29, on Zulip):

(or maybe those were on some other PR)

pnkfelix (Jun 13 2019 at 14:29, on Zulip):

we have 18 I-nominated T-compiler issues

pnkfelix (Jun 13 2019 at 14:29, on Zulip):

and I really want to try to get through all of them, if only superficially

Zoxc (Jun 13 2019 at 14:29, on Zulip):

My opinion is to merge the first PR in my series and discuss the other ones.

pnkfelix (Jun 13 2019 at 14:29, on Zulip):

lets go older first

pnkfelix (Jun 13 2019 at 14:30, on Zulip):

My opinion is to merge the first PR in my series and discuss the other ones.

( it might be good to state that opinion on the PR; the dialogue there makes it seem like we're blocked on overall planning for the series of PR's)

centril (Jun 13 2019 at 14:30, on Zulip):

(@pnkfelix for future meetings I would suggest segmenting and doing the p-high ones first)

pnkfelix (Jun 13 2019 at 14:30, on Zulip):

first nominated issue: "single_use_lifetimes warns when deriving a struct" #53738

pnkfelix (Jun 13 2019 at 14:31, on Zulip):

I bascially wanted to check about prioritization here

pnkfelix (Jun 13 2019 at 14:31, on Zulip):

I applied P-medium

pnkfelix (Jun 13 2019 at 14:31, on Zulip):

because I cannot imagine it matters to most of our current users

pnkfelix (Jun 13 2019 at 14:32, on Zulip):

but it seemed like @eddyb was hoping to make progress on applying this lint to rustc source itself.

nikomatsakis (Jun 13 2019 at 14:32, on Zulip):

I bascially wanted to check about prioritization here

this is tricky

pnkfelix (Jun 13 2019 at 14:32, on Zulip):

so anyone who wants to fight for P-high (or just fix the bug), feel free to chime in on the issue.

nikomatsakis (Jun 13 2019 at 14:32, on Zulip):

I feel like Rust 2018 is "not done" when it comes to this stuff

nikomatsakis (Jun 13 2019 at 14:32, on Zulip):

and in theory it'd be good to prioritize

nikomatsakis (Jun 13 2019 at 14:32, on Zulip):

but I also feel like we're oversubscribed as is

pnkfelix (Jun 13 2019 at 14:33, on Zulip):

Given that we have 18 nominated issues to get through, I'm just going to say "I advertised this"

pnkfelix (Jun 13 2019 at 14:33, on Zulip):

and move along

nikomatsakis (Jun 13 2019 at 14:33, on Zulip):

have we pinged @Zack M. Davis -- not sure if they're busy or what -- but they've done great work there

pnkfelix (Jun 13 2019 at 14:33, on Zulip):

next: "Coherence can be bypassed by an indirect impl for a trait object" #57893

pnkfelix (Jun 13 2019 at 14:34, on Zulip):

I am sad that this was filed in January but I dont' recall seeing it until now

pnkfelix (Jun 13 2019 at 14:35, on Zulip):

we do have some people who have taken a stab at trying to fix it

centril (Jun 13 2019 at 14:35, on Zulip):

On this one I guess the question is what the right fix is:
1. make the assumption correct (ban some blanket impls)
2. verify the assumption (what Aaron1011 suggested)
And independently, I think @RalfJ's point:

Once again, exploiting assumptions implicitly instead of inferring them to an explicit form proves to be an issue. This reminds me that we still implicitly exploit WF everywhere instead of turning that into explicit assumptions...

is spot on.

nikomatsakis (Jun 13 2019 at 14:35, on Zulip):

Hmm, I hadn't seen that, but I did have that same scenario filed in the back of my mind for "I bet this is broken"

nikomatsakis (Jun 13 2019 at 14:35, on Zulip):

(Seems like we should try to address that via #wg-traits work)

pnkfelix (Jun 13 2019 at 14:35, on Zulip):

@nikomatsakis : who is good person to own this?

pnkfelix (Jun 13 2019 at 14:36, on Zulip):

(that was not me attempting to volunteer you as owner)

pnkfelix (Jun 13 2019 at 14:36, on Zulip):

its already tagged as A-traits, are there other tags to use to get attention of WG-traits?

centril (Jun 13 2019 at 14:37, on Zulip):

I'd volunteer @nikomatsakis :slight_smile:

nikomatsakis (Jun 13 2019 at 14:37, on Zulip):

No, it's more a matter of bandwidth than anything.

pnkfelix (Jun 13 2019 at 14:37, on Zulip):

@nikomatsakis can I assign you to delegate to someone on WG-traits?

nikomatsakis (Jun 13 2019 at 14:38, on Zulip):

Yes, let me think about this. There is another related issue as well I think (realted in a general sense)

pnkfelix (Jun 13 2019 at 14:39, on Zulip):

okay assigning to @nikomatsakis with expectation that they will delegate

nikomatsakis (Jun 13 2019 at 14:39, on Zulip):

/me makes a note

pnkfelix (Jun 13 2019 at 14:40, on Zulip):

next: "Limit dylib symbols" #59752

nikomatsakis (Jun 13 2019 at 14:40, on Zulip):

(I plan next week at Moz All Hands to spend a big chunk of time concentrating on planning for traits WG, so I will prob wind up trying to factor the need to handle issues like this then)

pnkfelix (Jun 13 2019 at 14:40, on Zulip):

next: "Limit dylib symbols" #59752

oh , that's nominated for T-release; never mind.

pnkfelix (Jun 13 2019 at 14:41, on Zulip):

(but it hasn't merged? okay well I don't know, its fine)

pnkfelix (Jun 13 2019 at 14:41, on Zulip):

likewise "crater run to estimate impact of full NLL transition" #60680 was nominated largely for lang team

centril (Jun 13 2019 at 14:41, on Zulip):

@pnkfelix I removed t-compiler from that PR so you don't get bothered by it again

pnkfelix (Jun 13 2019 at 14:42, on Zulip):

and likewise I think "Use const generics for array impls" #61415 is really only nominated for T-libs

pnkfelix (Jun 13 2019 at 14:42, on Zulip):

(but here's peoples chance to chime in arguing that we do need to talk about #61415)

centril (Jun 13 2019 at 14:42, on Zulip):

No; it is really nominated for t-lang

pnkfelix (Jun 13 2019 at 14:42, on Zulip):

okay, next one is definitely us

pnkfelix (Jun 13 2019 at 14:43, on Zulip):

next: " 1.30 -> 1.31 dylib late-binding regression with less recent Linux distro toolchains." #61539

pnkfelix (Jun 13 2019 at 14:43, on Zulip):

@eddyb are you around?

pnkfelix (Jun 13 2019 at 14:43, on Zulip):

They believe this is a linker bug

pnkfelix (Jun 13 2019 at 14:43, on Zulip):

but there are other ways to work around it from our end

pnkfelix (Jun 13 2019 at 14:43, on Zulip):

e.g. -Z plt=yes

pnkfelix (Jun 13 2019 at 14:44, on Zulip):

which leads us to ask: How much work should we do to try to make the linker(s) happy?

pnkfelix (Jun 13 2019 at 14:44, on Zulip):

or do we first need to take the effort to confirm that it truly is a linker bug?

pnkfelix (Jun 13 2019 at 14:45, on Zulip):

There is this note from @nagisa "To me it seems that exactly what happens with PLT here is prevention and the core issue still remains regardless of whether PLT is enabled or not." which implies that we should have someone looking at this

pnkfelix (Jun 13 2019 at 14:45, on Zulip):

the good news is, as the title says, the problem is exposed by "less recent" (aka "old(-ish?)") toolchains

pnkfelix (Jun 13 2019 at 14:46, on Zulip):

i.e binutils 2.28.1 vs binutils 2.30

pnkfelix (Jun 13 2019 at 14:47, on Zulip):

Anyway I know @eddyb wanted it discussed

pnkfelix (Jun 13 2019 at 14:47, on Zulip):

and I didn't know what priority to assign to it

pnkfelix (Jun 13 2019 at 14:47, on Zulip):

because I cannot immediately tell which Linux distributions (or versions of distributions) are likely to be affected by this

pnkfelix (Jun 13 2019 at 14:47, on Zulip):

I could just assign it P-high but then I'd really want to make sure it has an owner.

pnkfelix (Jun 13 2019 at 14:48, on Zulip):

or maybe we revert PR #54592 to "resolve" it? I cannot imagine that being a great solution.

nikomatsakis (Jun 13 2019 at 14:48, on Zulip):

hmm

pnkfelix (Jun 13 2019 at 14:49, on Zulip):

I also don't want to punish @nagisa for speaking up in the comment thread of #61539, but they do seem like a natural person to assign it to

pnkfelix (Jun 13 2019 at 14:49, on Zulip):

maybe I go P-high and assign it to @nagisa and myself?

pnkfelix (Jun 13 2019 at 14:50, on Zulip):

I guess given the lack of commentary here, that's the easiest path forward in this meeting.

nikomatsakis (Jun 13 2019 at 14:50, on Zulip):

too bad @eddyb's not here to provide more context

pnkfelix (Jun 13 2019 at 14:51, on Zulip):

next: "error updating smallvec in rustc" #61549

pnkfelix (Jun 13 2019 at 14:51, on Zulip):

I should have probably cleared nomination here; it seems like its been resolved (and its assigned to me)

pnkfelix (Jun 13 2019 at 14:52, on Zulip):

((this is a conseuqence of not getting through all the supposed pre-passes of pre-triage))

pnkfelix (Jun 13 2019 at 14:52, on Zulip):

the next one, "TypeId differs between builds" #61553, is really a T-lang thing

pnkfelix (Jun 13 2019 at 14:52, on Zulip):

but if people want to speak up here about it, feel free.

pnkfelix (Jun 13 2019 at 14:53, on Zulip):

(You can see the last comments on the comment thread to see my take on it, and the issue filer's ... objection (?) to my take.)

pnkfelix (Jun 13 2019 at 14:53, on Zulip):

next: " ICE on nightly: "Forcing query with already existing DepNode."" #61594

pnkfelix (Jun 13 2019 at 14:54, on Zulip):

I left this nominated in the hopes that there would be people present who could discuss relationship with #61530

pnkfelix (Jun 13 2019 at 14:54, on Zulip):

I'm going to leave it nominated for now.

pnkfelix (Jun 13 2019 at 14:54, on Zulip):

okay lets do one more and then do a WG-checkin

pnkfelix (Jun 13 2019 at 14:55, on Zulip):

last for today: "Self as default type isnt typechecked" #61631

centril (Jun 13 2019 at 14:55, on Zulip):

not a high prio item; should discuss @ t-lang first

centril (Jun 13 2019 at 14:55, on Zulip):

to answer "is this a bug?"

pnkfelix (Jun 13 2019 at 14:55, on Zulip):

I assume all known cases where this could cause issues cause an ICE to occur?

centril (Jun 13 2019 at 14:56, on Zulip):

well it's checked on use

centril (Jun 13 2019 at 14:56, on Zulip):

I haven't seen any ICEs

pnkfelix (Jun 13 2019 at 14:57, on Zulip):

okay. Maybe I'll downgrade to P-medium next week

pnkfelix (Jun 13 2019 at 14:57, on Zulip):

I'll leave nomination tag on it for T-lang meeting.

pnkfelix (Jun 13 2019 at 14:58, on Zulip):

so with 3 minutes left

pnkfelix (Jun 13 2019 at 14:58, on Zulip):

@nikomatsakis anything to say about @WG-async-foundations ?

nikomatsakis (Jun 13 2019 at 14:58, on Zulip):

last for today: "Self as default type isnt typechecked" #61631

we did spend a lot of time defining the rules here --

nikomatsakis (Jun 13 2019 at 14:58, on Zulip):

nikomatsakis anything to say about @WG-async-foundations

ok -- yes, very briefly:

nikomatsakis (Jun 13 2019 at 14:58, on Zulip):

I hope to post a kind of general update blog post in any case

nikomatsakis (Jun 13 2019 at 14:58, on Zulip):

we're still working down a list of blocking issues that we currently think would prevent stabilization

nikomatsakis (Jun 13 2019 at 14:59, on Zulip):

there's actually coming to be a number of smaller issues, so it may be a good time to do a push for volunteers

nikomatsakis (Jun 13 2019 at 15:00, on Zulip):

that said, I think the two most major issues are starting to get closed:

nikomatsakis (Jun 13 2019 at 15:00, on Zulip):

we are working on the "async book", that will explain how async I/O works and the async-await feature in praticular

nikomatsakis (Jun 13 2019 at 15:00, on Zulip):

and finally one big area where we need to do more work is around testing

nikomatsakis (Jun 13 2019 at 15:00, on Zulip):

trying to make a comprehensive set of test cases, @centril and I have talked some but nobody has been able to put significant time into that

nikomatsakis (Jun 13 2019 at 15:00, on Zulip):

seems like a good case where we might be able to outline the work needed and volunteer it, but maybe not

nikomatsakis (Jun 13 2019 at 15:01, on Zulip):

ok that's my 3 minute summary :) ping in #wg-async-foundations if you'd like more details I geuss :)

pnkfelix (Jun 13 2019 at 15:02, on Zulip):

okay, thanks to everyone in @T-compiler/meeting for attending!

nikomatsakis (Jun 13 2019 at 15:02, on Zulip):

we're still working down a list of blocking issues that we currently think would prevent stabilization

just a note: this is intended to cover major diagnostic or other usability issues as well as (clearly) backwards incompatible semantic changes

Last update: Nov 16 2019 at 01:55UTC