Stream: t-compiler

Topic: #54818 weekly meeting 2018-12-13


pnkfelix (Dec 13 2018 at 13:48, on Zulip):

Hi @T-compiler , we'll be starting the weekly meeting in about 72 minutes. I'm going to try to do some pre-triage for the next 30 minutes or so

pnkfelix (Dec 13 2018 at 13:57, on Zulip):

P-high: "ICE: cannot relate region: LUB(ReErased, ReEmpty)" #56350, has a PR that landed and seems to have fixed it. Issue remains open until Q of whether to beta-accept that PR (#56282) is resolved.

pnkfelix (Dec 13 2018 at 13:58, on Zulip):

P-high: "problems with trait objects with a trait projection value that contains Self" #56288, still waiting for T-lang to resolve Q. (T-lang meeting was cancelled last week so its not surprising there's been no movement here.)

pnkfelix (Dec 13 2018 at 13:59, on Zulip):

P-high: "prohibit "two-phase borrows" with existing borrows?" #56254, I am planning to look more carefully at this this week. Haven't started yet, but its assigned to me as of last night.

pnkfelix (Dec 13 2018 at 14:04, on Zulip):

P-high: "regression: stack overflow on macosx with xcode 6.4" #55471. The short-term fix to this was to bump stack size to 32 MB, which landed on nightly via PR #56467 ; that nightly then was promoted to beta, and subsequently the bump was backported to stable via PR #56518

pnkfelix (Dec 13 2018 at 14:06, on Zulip):

the long term fix is to use stacker; work continues on PR #55617 for this. I think the latest benchmarking finished literally minutes ago.

Nikita Popov (Dec 13 2018 at 14:12, on Zulip):

IIRC the backported stack size bump didn't fix this, because of https://github.com/rust-lang/rust/pull/48575, which hasn't been reverted yet.

pnkfelix (Dec 13 2018 at 14:13, on Zulip):

P-high: "1.30.0 fails to build for target powerpc-unknown-netbsd" #55465. I have posted PR #56562 to "fix" this. I put "fix" in quotes because our Cargo.lock was independently updated to include the version update anyway (but my fix should be beta-backported IMO). There was some independent discussion of (mis)use of git submodules in our source tree; there is related work to that in PR #56092

pnkfelix (Dec 13 2018 at 14:14, on Zulip):

@Nikita Popov I had thought briefly the two might be related, but to be honest I'm not even clear on how one mechanism is successfully changing the stack size while the other does not.

pnkfelix (Dec 13 2018 at 14:14, on Zulip):

@Nikita Popov my own internal confusion on that point led me to decide to not even mention PR #48575 on my note on the issue.

pnkfelix (Dec 13 2018 at 14:15, on Zulip):

Having said that, I'm becoming more interested now in reverting that PR.

pnkfelix (Dec 13 2018 at 14:15, on Zulip):

(something I was already interested in trying to do as it was)

pnkfelix (Dec 13 2018 at 14:17, on Zulip):

P-high: "rustc 1.30.0's linker flavor inference is a non-backwards compat change to -Clinker" #55396, has one work-item PR (#56349) that landed, and then the final step is in PR #56351, which has gotten approval via rfcbot so I think all that remains is to bors r+ it?

pnkfelix (Dec 13 2018 at 14:19, on Zulip):

I'm not sure the docs on PR #56351 are 100% complete, but I also don't see that as a reason to block landing the PR.

pnkfelix (Dec 13 2018 at 14:20, on Zulip):

(when I say the docs are not 100% complete, I mean, for example, that I don't think any examples are provided of what the valid values are for the linker-flavor option...)

pnkfelix (Dec 13 2018 at 14:23, on Zulip):

okay that's all the P-highs. I don't think there's any fires we need to put out; they all have someone assigned, and they are almost all moving along. I'll try to make sure the lang team looks at PR #56288 tonight. (There is a slight chance I'll miss the T-lang meeting, in which case I guess I better delegate that aforementioned responsibility to @nikomatsakis )

pnkfelix (Dec 13 2018 at 14:27, on Zulip):

beta-nominations: just this PR "Fix #56237: normalize type before deferred sizedness checking." #56282. This is believed to address #56350 as well. We'll all talk about it together. Maybe @nikomatsakis can explain the bit about using fresh vars instead of placeholders.

pnkfelix (Dec 13 2018 at 14:28, on Zulip):

no stable-nominations

pnkfelix (Dec 13 2018 at 14:28, on Zulip):

nothing waiting on our team

pnkfelix (Dec 13 2018 at 14:28, on Zulip):

I already touched on the one nominated issue (#56288)

pnkfelix (Dec 13 2018 at 14:29, on Zulip):

(and AFAICT all of the other nominated issues not tagged with T-compiler have pre-existing T-other tags that look appropriate.)

pnkfelix (Dec 13 2018 at 14:30, on Zulip):

I will say, however, this issue looks pretty scary for our collective workflow (e.g. for bisection): "No nightly since 2018-12-09" #56667

pnkfelix (Dec 13 2018 at 14:31, on Zulip):

next, stable-to-stable regressions

pnkfelix (Dec 13 2018 at 14:31, on Zulip):

Almost all are from October or earlier, but there is one that is tagged with T-compiler with no P-tag

pnkfelix (Dec 13 2018 at 14:31, on Zulip):

this: "Compiler panics unexpectedly in 1.31 when constructing struct with wrong syntax" #56611

pnkfelix (Dec 13 2018 at 14:31, on Zulip):

I'll mark that P-high

simulacrum (Dec 13 2018 at 14:32, on Zulip):

@pnkfelix FYI: We're doing a point release next week; if the compiler team would like to backport anything we'll need patches by Monday (we're operating on such a quick turnaround to avoid pushing deep into the holidays).

pnkfelix (Dec 13 2018 at 14:33, on Zulip):

Okay thanks for the heads up @simulacrum !!

simulacrum (Dec 13 2018 at 14:33, on Zulip):

I won't be able to make the meeting myself but if anything needs a backport please stable-accept and ping release team/me on it

pnkfelix (Dec 13 2018 at 14:36, on Zulip):

The fact that #56611 had no P-tag (and thus was readily overlooked during triage) makes me realize that we need to regularly look at the T-compiler issues with no P-label. I'll add a link for that to our agenda in #54818

pnkfelix (Dec 13 2018 at 14:38, on Zulip):

Here's my current Unprioritized T-compiler link. There are 1,021 (!) open issues

pnkfelix (Dec 13 2018 at 14:39, on Zulip):

let me just ... try to filter that down a bit more...

pnkfelix (Dec 13 2018 at 14:41, on Zulip):

maybe a new plan: make one link that just has the ICEs

pnkfelix (Dec 13 2018 at 14:41, on Zulip):

and another that doesn't.

pnkfelix (Dec 13 2018 at 14:43, on Zulip):

(and also I'll try to filter out the issues tagged with other teams as well, just to focus energy)

pnkfelix (Dec 13 2018 at 14:43, on Zulip):

that gets us down to 73 ICEs with no P-label, and 866 non-ICEs.

pnkfelix (Dec 13 2018 at 14:44, on Zulip):

Hmm there is one tagged regression-stable-to-beta but not regression-stable-to-nightly (#56618) ?

pnkfelix (Dec 13 2018 at 14:46, on Zulip):

argh and I hate our emoji in I-unsound :boom:

pnkfelix (Dec 13 2018 at 14:46, on Zulip):

Okay now I have to actually hop on a metro to take the meeting from hoem...

pnkfelix (Dec 13 2018 at 14:47, on Zulip):

hey @nagisa I realize now that I am leaving too late and so I may miss the start of the meeting. If you can run the start of it if I am a few minutes late I would appreciate that.

nagisa (Dec 13 2018 at 14:47, on Zulip):

o/

nagisa (Dec 13 2018 at 14:47, on Zulip):

sure

nagisa (Dec 13 2018 at 14:49, on Zulip):

/me is surprised that @pnkfelix lives just 13 minutes away from their office.

Wesley Wiser (Dec 13 2018 at 14:56, on Zulip):

/me is traveling but seems to have a stable internet connection :fingers_crossed:

nagisa (Dec 13 2018 at 14:59, on Zulip):

Okay, lets start I guess (now that a great part of this sound has started…)

nagisa (Dec 13 2018 at 15:00, on Zulip):

:music:

nagisa (Dec 13 2018 at 15:00, on Zulip):

Hi all @T-compiler!

nagisa (Dec 13 2018 at 15:00, on Zulip):

First up, P-high stuff, most of which has been already reviewed by @pnkfelix, so I’ll just quote the separate issues and we can discuss

nagisa (Dec 13 2018 at 15:01, on Zulip):

P-high: "ICE: cannot relate region: LUB(ReErased, ReEmpty)" #56350, has a PR that landed and seems to have fixed it. Issue remains open until Q of whether to beta-accept that PR (#56282) is resolved.

nagisa (Dec 13 2018 at 15:01, on Zulip):

P-high: "problems with trait objects with a trait projection value that contains Self" #56288, still waiting for T-lang to resolve Q. (T-lang meeting was cancelled last week so its not surprising there's been no movement here.)

nagisa (Dec 13 2018 at 15:01, on Zulip):

These two seem to require no discussion

nagisa (Dec 13 2018 at 15:02, on Zulip):

P-high: "prohibit "two-phase borrows" with existing borrows?" #56254, I am planning to look more carefully at this this week. Haven't started yet, but its assigned to me as of last night.

nagisa (Dec 13 2018 at 15:02, on Zulip):

@RalfJ seems to have written a long comment on this

nagisa (Dec 13 2018 at 15:03, on Zulip):

I’m still not sure how hard is this to implement and how long it would take as well

nagisa (Dec 13 2018 at 15:04, on Zulip):

we did release NLL into the wild, didn’t we?

nagisa (Dec 13 2018 at 15:04, on Zulip):

(with the 2018) does that mean that this is an actual stability hazard?

nikomatsakis (Dec 13 2018 at 15:04, on Zulip):

yes, it is

nikomatsakis (Dec 13 2018 at 15:04, on Zulip):

there's also not clear consensus on whether such code should be an error

nikomatsakis (Dec 13 2018 at 15:04, on Zulip):

it's sort of a complex scenario

nikomatsakis (Dec 13 2018 at 15:05, on Zulip):

but the assumption was that perhaps we want to make an error in this case to leave ourselves more room

nikomatsakis (Dec 13 2018 at 15:05, on Zulip):

NLL is currently in a kind of "transition" state, so we probably have some room to connect this to that same warning period

nikomatsakis (Dec 13 2018 at 15:05, on Zulip):

(but the justification would be that it is a soundness fix, as this would indeed cause previously compiling code to now get future-compatibility warnings)

nagisa (Dec 13 2018 at 15:06, on Zulip):

does this need any extra tags?

nagisa (Dec 13 2018 at 15:06, on Zulip):

as in it does not appear to be an actual soundness issue to me

nikomatsakis (Dec 13 2018 at 15:07, on Zulip):

heh well that depends

nikomatsakis (Dec 13 2018 at 15:07, on Zulip):

if we adopted stacked borrows, it would mean the borrow checker accepts UB

nikomatsakis (Dec 13 2018 at 15:07, on Zulip):

alternatively, we could modify the stacked borrows model to accept this

nikomatsakis (Dec 13 2018 at 15:07, on Zulip):

this is why it's debated

nikomatsakis (Dec 13 2018 at 15:07, on Zulip):

anyway I think the tags are probably sufficient

nikomatsakis (Dec 13 2018 at 15:08, on Zulip):

alternatively, we could modify the stacked borrows model to accept this

at the cost of having the definition of UB / MIR etc be correspondingly more complex

nikomatsakis (Dec 13 2018 at 15:08, on Zulip):

I guess the point is that the definition of "soundness" is not entirely fixed either

nagisa (Dec 13 2018 at 15:08, on Zulip):

okay, then it means that this issue is in fairly capable hands for now...

nikomatsakis (Dec 13 2018 at 15:08, on Zulip):

however, I think there is no soundness issue in the sense that we can create a crash today

nagisa (Dec 13 2018 at 15:09, on Zulip):

P-high: "regression: stack overflow on macosx with xcode 6.4" #55471. The short-term fix to this was to bump stack size to 32 MB, which landed on nightly via PR #56467 ; that nightly then was promoted to beta, and subsequently the bump was backported to stable via PR #56518

nagisa (Dec 13 2018 at 15:10, on Zulip):

this is related to the suggested revert of that certain PR

oli (Dec 13 2018 at 15:10, on Zulip):

We don't need to do the revert if we accept the stacker PR

nagisa (Dec 13 2018 at 15:10, on Zulip):

which Ishi Tatsuyuki agreed to (with preference that the functionality stays behind the flag)

nagisa (Dec 13 2018 at 15:10, on Zulip):

@Oli stacker is not cross-platform… yet

oli (Dec 13 2018 at 15:10, on Zulip):

we still need to do the revert because of soundness of TLS, but that's a different story

oli (Dec 13 2018 at 15:11, on Zulip):

true, but we do cover mac, win & linux

nagisa (Dec 13 2018 at 15:11, on Zulip):

I do have the portable stack manipulation crate moving along nicely (supports x86/arm/powerpc now) so there’s light at the end of the tunnel

nikomatsakis (Dec 13 2018 at 15:11, on Zulip):

I'm feeling a bit confused, which PR are we specifically proposing to revert?

nikomatsakis (Dec 13 2018 at 15:11, on Zulip):

" Bump stack size to 32MB #56467 "?

nagisa (Dec 13 2018 at 15:12, on Zulip):

https://github.com/rust-lang/rust/pull/48575

nikomatsakis (Dec 13 2018 at 15:12, on Zulip):

ah, ok, yes

nagisa (Dec 13 2018 at 15:13, on Zulip):

@Oli you said "linux" is supported, but are all architectures that linux+rust support also supported by stacker?

nagisa (Dec 13 2018 at 15:13, on Zulip):

I dont recall seeing assembly for anything other than x86 in stacker.

nikomatsakis (Dec 13 2018 at 15:13, on Zulip):

( it seemed like we did indeed have general consensus to revert that, possibly with a flag )

nagisa (Dec 13 2018 at 15:13, on Zulip):

yeah, somebody has to do the work of making the revert

oli (Dec 13 2018 at 15:13, on Zulip):

@nagisa yea it's not fully general

oli (Dec 13 2018 at 15:14, on Zulip):

Should I prioritize the revert before the stacker PR?

nagisa (Dec 13 2018 at 15:14, on Zulip):

(there is also a point release after a week, so the sooner we move here in some way, the better)

nagisa (Dec 13 2018 at 15:14, on Zulip):

stable point release, that is

nagisa (Dec 13 2018 at 15:15, on Zulip):

so depending on the complexity of the PR, we might want to backport revertion/etc to stable as well

oli (Dec 13 2018 at 15:15, on Zulip):

I can do a minimal impl (not a revert) that just adds threads by default

pnkfelix (Dec 13 2018 at 15:15, on Zulip):

/me is surprised that @pnkfelix lives just 13 minutes away from their office.

(20 minutes If they get very lucky with the metro timing)

nagisa (Dec 13 2018 at 15:15, on Zulip):

hey, welcome back! /me hands off

pnkfelix (Dec 13 2018 at 15:16, on Zulip):

@Oli I'm happy to do revert if you have too much on your plate

pnkfelix (Dec 13 2018 at 15:16, on Zulip):

(we can discuss after mtg)

nagisa (Dec 13 2018 at 15:16, on Zulip):

@pnkfelix not-covered-yet P-high are #56611 #55465 #55396

pnkfelix (Dec 13 2018 at 15:17, on Zulip):

ah right

pnkfelix (Dec 13 2018 at 15:17, on Zulip):

P-high: "1.30.0 fails to build for target powerpc-unknown-netbsd" #55465. I have posted PR #56562 to "fix" this. I put "fix" in quotes because our Cargo.lock was independently updated to include the version update anyway (but my fix should be beta-backported IMO). There was some independent discussion of (mis)use of git submodules in our source tree; there is related work to that in PR #56092

pnkfelix (Dec 13 2018 at 15:17, on Zulip):

I don't think there's really more to say there

pnkfelix (Dec 13 2018 at 15:18, on Zulip):

P-high: "rustc 1.30.0's linker flavor inference is a non-backwards compat change to -Clinker" #55396, has one work-item PR (#56349) that landed, and then the final step is in PR #56351, which has gotten approval via rfcbot so I think all that remains is to bors r+ it?

nikomatsakis (Dec 13 2018 at 15:18, on Zulip):

seems like a simple backport

nagisa (Dec 13 2018 at 15:18, on Zulip):

I just r+d the linker-flavour PR.

pnkfelix (Dec 13 2018 at 15:18, on Zulip):

@nagisa am I wrong in thinking that the doc for -C linker-flavor needs some expansion?

pnkfelix (Dec 13 2018 at 15:18, on Zulip):

And/or clarification? Maybe I'll just file some issues.

davidtwco (Dec 13 2018 at 15:19, on Zulip):

I'm happy to adjust it quickly before it lands or gets up the queue. It doesn't do a whole lot so I couldn't think of more to write.

nagisa (Dec 13 2018 at 15:19, on Zulip):

I think they are fine. When the topic is linkers, you can get as in depth as is conceivable to human

pnkfelix (Dec 13 2018 at 15:19, on Zulip):

(I assume the meaning of "linker-flavor" is to indicate what style of arguments/options should be passed to the provided -C linker)

pnkfelix (Dec 13 2018 at 15:20, on Zulip):

anyway I can again talk about this post mtg with @davidtwco

nagisa (Dec 13 2018 at 15:20, on Zulip):

(though I guess I’m not the target audience for such documentation so it is hard for me to judge its quality)

pnkfelix (Dec 13 2018 at 15:20, on Zulip):

and it certainly should not block that PR landing

davidtwco (Dec 13 2018 at 15:20, on Zulip):

(relevant topic)

pnkfelix (Dec 13 2018 at 15:21, on Zulip):

finally, the last P-high issue was one that I added after my sweep

pnkfelix (Dec 13 2018 at 15:21, on Zulip):

"Compiler panics unexpectedly in 1.31 when constructing struct with wrong syntax" #56611

pnkfelix (Dec 13 2018 at 15:21, on Zulip):

why am I not able to assign that issue to alexreg?

davidtwco (Dec 13 2018 at 15:21, on Zulip):

He needs to be in the rust-lang organization.

pnkfelix (Dec 13 2018 at 15:22, on Zulip):

hmm alexreg claimed to be in the org

davidtwco (Dec 13 2018 at 15:22, on Zulip):

Normally people are in it through some working group.

nikomatsakis (Dec 13 2018 at 15:22, on Zulip):

I can add them

nikomatsakis (Dec 13 2018 at 15:22, on Zulip):

if they're not already

davidtwco (Dec 13 2018 at 15:22, on Zulip):

Maybe he is, I just know I've seen that be the reason that you can't assign someone before.

pnkfelix (Dec 13 2018 at 15:23, on Zulip):

its weird, when I mouse over alexreg it says they are member of https://github.com/rust-lang

nikomatsakis (Dec 13 2018 at 15:23, on Zulip):

I think they were in the org

nagisa (Dec 13 2018 at 15:23, on Zulip):

assigned him for you

nagisa (Dec 13 2018 at 15:23, on Zulip):

topic over :P

pnkfelix (Dec 13 2018 at 15:23, on Zulip):

but when I click on them, I don't see rust-lang icon in the Organizations

pnkfelix (Dec 13 2018 at 15:23, on Zulip):

wait, @nagisa was able to do it but i wasn't???

nagisa (Dec 13 2018 at 15:23, on Zulip):

/me shows their github sudopowers.

pnkfelix (Dec 13 2018 at 15:23, on Zulip):

So weird!

Esteban Küber (Dec 13 2018 at 15:23, on Zulip):

Was that ticket correctly labeled as regression?

davidtwco (Dec 13 2018 at 15:23, on Zulip):

but when I click on them, I don't see rust-lang icon in the Organizations

You can choose if you show your membership to an org.

pnkfelix (Dec 13 2018 at 15:23, on Zulip):

oh. okay. hmm.

Esteban Küber (Dec 13 2018 at 15:23, on Zulip):

I mentioned my reasoning but wasn't sure if appropriate

pnkfelix (Dec 13 2018 at 15:24, on Zulip):

stable-to-stable regression seems right I think...

pnkfelix (Dec 13 2018 at 15:24, on Zulip):

(i.e. going from "reports an error" to ICE is bad)

pnkfelix (Dec 13 2018 at 15:25, on Zulip):

okay that's all the P-highs then

pnkfelix (Dec 13 2018 at 15:25, on Zulip):

beta-nominated: "Fix #56237: normalize type before deferred sizedness checking." #56282

nagisa (Dec 13 2018 at 15:26, on Zulip):

I’d like to come back to the alexreg’s fix

nagisa (Dec 13 2018 at 15:26, on Zulip):

do we want to backport it?

pnkfelix (Dec 13 2018 at 15:26, on Zulip):

I assume @nikomatsakis that you feel okay about the changes to replace higher-order lifetimes with fresh vars?

pnkfelix (Dec 13 2018 at 15:26, on Zulip):

oh sorry @nagisa

nagisa (Dec 13 2018 at 15:26, on Zulip):

the change is large, sure, but perhaps there is a small subset of their PR that is backportable?

pnkfelix (Dec 13 2018 at 15:27, on Zulip):

I'm not 100% sure that its a high priority to backport

nagisa (Dec 13 2018 at 15:27, on Zulip):

in case such PR exists (we could ask that async) would we like to do a backport?

nagisa (Dec 13 2018 at 15:28, on Zulip):

okay, that answers my question

pnkfelix (Dec 13 2018 at 15:28, on Zulip):

since the code that triggers the ICE on stable/beta would always be some sort of syntax error ... right?

pnkfelix (Dec 13 2018 at 15:28, on Zulip):

or did I misunderstand that portion of what @Esteban Küber said?

pnkfelix (Dec 13 2018 at 15:28, on Zulip):

the better thing may be to just tag it as beta-nominated

pnkfelix (Dec 13 2018 at 15:28, on Zulip):

well

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

the decision is going to be made here anyway I guess

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

I just figure its good to give absent members of T-compiler a chance to weigh in perhaps...

Esteban Küber (Dec 13 2018 at 15:29, on Zulip):

A patch that stops the ice would be small enough

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

Since it does not appear to be too important, perhaps it is not worth the bother

Esteban Küber (Dec 13 2018 at 15:29, on Zulip):

The final fix to account for the case and provide suggestions will probably take a bit longer to get right

pnkfelix (Dec 13 2018 at 15:30, on Zulip):

how about this

pnkfelix (Dec 13 2018 at 15:30, on Zulip):

we're not going to take the PR as is

pnkfelix (Dec 13 2018 at 15:30, on Zulip):

but if a small one fixes the ICE... I mean, I guess my question is whether the code that causes the ICE is likely to arise in the wild on stable/beta

pnkfelix (Dec 13 2018 at 15:31, on Zulip):

certainly the code from #56611 is an easy example of that...

nagisa (Dec 13 2018 at 15:31, on Zulip):

seems plausible to me, I definitely switch from tuple-structs to brace-structs when they get larger

pnkfelix (Dec 13 2018 at 15:31, on Zulip):

so ... maybe I can be swayed that it might be worth fixing. Except we do report a reasonable error prior to the ICE ,right?

pnkfelix (Dec 13 2018 at 15:32, on Zulip):

oh. maybe not so reasonable.

davidtwco (Dec 13 2018 at 15:32, on Zulip):

any error is more reasonable than an ICE for a beginner who stumbles upon this case getting to grips with the syntax?

mw (Dec 13 2018 at 15:32, on Zulip):

Does this only happen with Self?

Esteban Küber (Dec 13 2018 at 15:32, on Zulip):

I believe so

pnkfelix (Dec 13 2018 at 15:32, on Zulip):

I'm willing to be convinced that a backport is worth making

nagisa (Dec 13 2018 at 15:32, on Zulip):

Yes, it only happens with Self (just checked)

nagisa (Dec 13 2018 at 15:33, on Zulip):

which makes it much lower priority to backport IME

pnkfelix (Dec 13 2018 at 15:33, on Zulip):

hah

pnkfelix (Dec 13 2018 at 15:33, on Zulip):

okay then maybe lets just nip that in the bud then.

nikomatsakis (Dec 13 2018 at 15:33, on Zulip):

I'm not opposed (if PR is small), but I don't think the urgency is high. Let's just fix it.

pnkfelix (Dec 13 2018 at 15:35, on Zulip):

okay I tried to relay our general feeling as a comment on PR #55994. Feel free to correct it if you disagree. This is what I wrote:

(discussed amongst T-compiler whether it might be worthwhile to factor out a piece of this PR that just fixes the ICE from #56611 and then backporting that. The general consensus was essentially: Such a backport does not seem terribly high priority, but if someone wants to make it, the team isn't immediately opposed to such a backport.)

pnkfelix (Dec 13 2018 at 15:35, on Zulip):

okay

pnkfelix (Dec 13 2018 at 15:36, on Zulip):

I think that brings us back to: beta-nominated: "Fix #56237: normalize type before deferred sizedness checking." #56282

pnkfelix (Dec 13 2018 at 15:36, on Zulip):

I assume @nikomatsakis that you feel okay about the changes to replace higher-order lifetimes with fresh vars?

pnkfelix (Dec 13 2018 at 15:36, on Zulip):

i.e. do you feel comfortable backporting that? Or should we let it ride the trains?

nikomatsakis (Dec 13 2018 at 15:36, on Zulip):

yes. The truth is that I didn't dig that deeply into the PR (I'd like to build it locally and walk through), since I was pretty busy last week, but this approach seems "obviously more correct" than what was there before.

nikomatsakis (Dec 13 2018 at 15:37, on Zulip):

in other words, erasing late-bound regions is basically equivalent to putting in variables

nikomatsakis (Dec 13 2018 at 15:37, on Zulip):

so yeah I feel ok about a backport

pnkfelix (Dec 13 2018 at 15:37, on Zulip):

okay. I'm inclined to beta-accept then

pnkfelix (Dec 13 2018 at 15:38, on Zulip):

next ... no stable-nominations for us I think

pnkfelix (Dec 13 2018 at 15:39, on Zulip):

stable-to-beta regression: "Regression in rust-nightly 2018-12-02" #56618

pnkfelix (Dec 13 2018 at 15:39, on Zulip):

I think I missed all these in my earlier sweep

pnkfelix (Dec 13 2018 at 15:40, on Zulip):

this is tagged I-unsound. It is fixed via a patch to LLVM, whcih is in PR #56781

pnkfelix (Dec 13 2018 at 15:40, on Zulip):

so I'm going to tag as P-high

Nikita Popov (Dec 13 2018 at 15:41, on Zulip):

I've just submitted https://github.com/rust-lang/rust/pull/56781 with beta-nominated.

pnkfelix (Dec 13 2018 at 15:41, on Zulip):

The other stable-to-beta regressions have already been discussed

pnkfelix (Dec 13 2018 at 15:42, on Zulip):

Thanks @Nikita Popov

pnkfelix (Dec 13 2018 at 15:42, on Zulip):

This reminds me, @simulacrum posted a note up above, don't know if you all saw it

pnkfelix (Dec 13 2018 at 15:42, on Zulip):

@pnkfelix FYI: We're doing a point release next week; if the compiler team would like to backport anything we'll need patches by Monday (we're operating on such a quick turnaround to avoid pushing deep into the holidays).

pnkfelix (Dec 13 2018 at 15:43, on Zulip):

so that means it would be good for us to indeed decide about these back ports now so that they can happen .. hopefully tomorrow ...

nikomatsakis (Dec 13 2018 at 15:43, on Zulip):

well, specifically that was about stable backports, right?

nikomatsakis (Dec 13 2018 at 15:43, on Zulip):

vs backports to beta?

pnkfelix (Dec 13 2018 at 15:43, on Zulip):

oh good point

pnkfelix (Dec 13 2018 at 15:43, on Zulip):

we didn't have any stable-nominations

pnkfelix (Dec 13 2018 at 15:44, on Zulip):

I was going to say that this means we may want to eagerly evaluate whether to beta-accept "Update LLVM submodule" #56781

pnkfelix (Dec 13 2018 at 15:44, on Zulip):

which we may indeed want to beta-accept

pnkfelix (Dec 13 2018 at 15:45, on Zulip):

but if I understand correctly, the LLVM issues only got into beta, not stable, right?

nikomatsakis (Dec 13 2018 at 15:45, on Zulip):

based on the tags, sounds right

pnkfelix (Dec 13 2018 at 15:45, on Zulip):

cc @Nikita Popov ^

Nikita Popov (Dec 13 2018 at 15:45, on Zulip):

Yeah. Well, one of them is an old issue, but I don't think it's worth stable backporting.

Nikita Popov (Dec 13 2018 at 15:45, on Zulip):

Especially as stable still has an older LLVM baseline.

pnkfelix (Dec 13 2018 at 15:46, on Zulip):

here's my thought then on PR #56781

pnkfelix (Dec 13 2018 at 15:46, on Zulip):

lets give it a week to cook on nightly

pnkfelix (Dec 13 2018 at 15:46, on Zulip):

and then decide about whether to beta-accept next mtg?

pnkfelix (Dec 13 2018 at 15:47, on Zulip):

okay

pnkfelix (Dec 13 2018 at 15:47, on Zulip):

next up, stable-to-nightly regressions list

Esteban Küber (Dec 13 2018 at 15:47, on Zulip):

Beta accepting is needed for inclusion on next stable,right?

pnkfelix (Dec 13 2018 at 15:47, on Zulip):

@Esteban Küber it pushes it closer to landing sooner in stable, yeah?

pnkfelix (Dec 13 2018 at 15:47, on Zulip):

as in, it would be part of the next train promotion

pnkfelix (Dec 13 2018 at 15:48, on Zulip):

but that doesn't hapepn via point releases, right?

nikomatsakis (Dec 13 2018 at 15:48, on Zulip):

right, to get into a point release, it would need an add'l backport

pnkfelix (Dec 13 2018 at 15:48, on Zulip):

its still on the six week schedule, which is why I would think we can give it a week to cook

pnkfelix (Dec 13 2018 at 15:48, on Zulip):

The story would be different if we were aiming to do a backport all the way to stable.

pnkfelix (Dec 13 2018 at 15:48, on Zulip):

but it sounds like even @Nikita Popov isn't pushing for that?

Nikita Popov (Dec 13 2018 at 15:49, on Zulip):

Yeah, I don't think this is particularly urgent.

pnkfelix (Dec 13 2018 at 15:49, on Zulip):

okay

pnkfelix (Dec 13 2018 at 15:49, on Zulip):

stable-to-nightly regressions (which again I appear to have overlooked in my earlier sweep. I was clearly rushing...)

pnkfelix (Dec 13 2018 at 15:49, on Zulip):

stable-to-nightly: "nightly rustc --version hangs forever" #56736

pnkfelix (Dec 13 2018 at 15:49, on Zulip):

egad

pnkfelix (Dec 13 2018 at 15:50, on Zulip):

P-high ...

pnkfelix (Dec 13 2018 at 15:50, on Zulip):

I'm just debating about whom to assign this to

pnkfelix (Dec 13 2018 at 15:51, on Zulip):

does anyone here use this "ESET" anti-virus software for Ubuntu ?

nagisa (Dec 13 2018 at 15:51, on Zulip):

P-low IMO

nagisa (Dec 13 2018 at 15:51, on Zulip):

antivirus issues are antivirus problem

nagisa (Dec 13 2018 at 15:52, on Zulip):

/me does not see a reason to take a responsibility for people using snake oil

pnkfelix (Dec 13 2018 at 15:52, on Zulip):

oh dear

pnkfelix (Dec 13 2018 at 15:53, on Zulip):

heh

pnkfelix (Dec 13 2018 at 15:53, on Zulip):

maybe the answer is to tag it P-low and assign to @nagisa

pnkfelix (Dec 13 2018 at 15:54, on Zulip):

okay well I dont want our meeting to get hung forever on this isssue

nikomatsakis (Dec 13 2018 at 15:54, on Zulip):

this does seem like clearly an ESET bug

nikomatsakis (Dec 13 2018 at 15:54, on Zulip):

(vs rustc)

pnkfelix (Dec 13 2018 at 15:54, on Zulip):

I'm just going to move forward with no tagging

pnkfelix (Dec 13 2018 at 15:54, on Zulip):

since it may not be a T-compiler problem or a rustc problem in the first place.

nikomatsakis (Dec 13 2018 at 15:54, on Zulip):

(I'm a bit confused because I thought we weren't using jemalloc anymore? but anyway)

pnkfelix (Dec 13 2018 at 15:54, on Zulip):

I think rustc is still opting into using jemalloc ?

nagisa (Dec 13 2018 at 15:54, on Zulip):

(I'm a bit confused because I thought we weren't using jemalloc anymore? but anyway)

compiler speicifically uses jemalloc, just not the programs built by rustc by default

nikomatsakis (Dec 13 2018 at 15:55, on Zulip):

hmm

pnkfelix (Dec 13 2018 at 15:55, on Zulip):

anyway next stable-to-nightly regression: "ICE when trying to process a circular dependency between traits" #56670

pnkfelix (Dec 13 2018 at 15:55, on Zulip):

I'll mark as P-high and T-compiler

nikomatsakis (Dec 13 2018 at 15:55, on Zulip):
# Link the compiler against `jemalloc`, where on Linux and OSX it should
# override the default allocator for rustc and LLVM.
#jemalloc = false
nikomatsakis (Dec 13 2018 at 15:55, on Zulip):

(from config.toml.example)

nikomatsakis (Dec 13 2018 at 15:55, on Zulip):

anyway, not urgent

Nikita Popov (Dec 13 2018 at 15:56, on Zulip):

@nikomatsakis It's disabled for local builds by default, but CI builds use it, because performance is significantly better for rustc's use case.

pnkfelix (Dec 13 2018 at 15:56, on Zulip):

!

pnkfelix (Dec 13 2018 at 15:56, on Zulip):

that sounds like entirely the wrong work flow

pnkfelix (Dec 13 2018 at 15:56, on Zulip):

but what do i know

nikomatsakis (Dec 13 2018 at 15:57, on Zulip):

I assigned that issue to me, i'll try to investigate ( ICE when trying to process a circular dependency between traits #56670 )

pnkfelix (Dec 13 2018 at 15:57, on Zulip):

/me would prefer for the majority of the CI's to match up with the local builds...

pnkfelix (Dec 13 2018 at 15:57, on Zulip):

anyway

pnkfelix (Dec 13 2018 at 15:57, on Zulip):

no issues are tagged as waiting for our team, yay

pnkfelix (Dec 13 2018 at 15:58, on Zulip):

the only I-nominated issue is #56288 which is really nominated for T-lang

pnkfelix (Dec 13 2018 at 15:58, on Zulip):

I added some new links to our regular agenda

pnkfelix (Dec 13 2018 at 15:58, on Zulip):

for unprioritized issues

pnkfelix (Dec 13 2018 at 15:58, on Zulip):

unprioritized issues -- there are >1,000 of them

pnkfelix (Dec 13 2018 at 15:59, on Zulip):

so lets go quickly through them now :lol:

davidtwco (Dec 13 2018 at 15:59, on Zulip):

1000 issues, 1 minute remaining.

pnkfelix (Dec 13 2018 at 15:59, on Zulip):

anyway its more as a reminder for me to try to process them as part of pre-triage

pnkfelix (Dec 13 2018 at 16:00, on Zulip):

but if people have thoughts on how to better revise the links there

pnkfelix (Dec 13 2018 at 16:00, on Zulip):

to try to break them down into smaller groups where the things that are more likely to be actually P-high are more likely to be looked at first

oli (Dec 13 2018 at 16:00, on Zulip):

Maybe we can split that process among the compiler team

pnkfelix (Dec 13 2018 at 16:00, on Zulip):

well, feel free to edit the links if you have thoughts on that. :)

pnkfelix (Dec 13 2018 at 16:01, on Zulip):

there continues to be a slew of stable-to-stable regressions

pnkfelix (Dec 13 2018 at 16:01, on Zulip):

but we are out of time once again. :)

pnkfelix (Dec 13 2018 at 16:01, on Zulip):

And after I show up late, I now have to ditch immediately.

nikomatsakis (Dec 13 2018 at 16:02, on Zulip):

it may be worth considering an alternate meeting, but I'm not sure

nikomatsakis (Dec 13 2018 at 16:02, on Zulip):

thanks @pnkfelix and @nagisa =)

pnkfelix (Dec 13 2018 at 16:03, on Zulip):

an alternative meeting for processing the unpriroritized issues? maybe.

pnkfelix (Dec 13 2018 at 16:03, on Zulip):

but I would be fine with just a dedicated topic with asynchronous effort

pnkfelix (Dec 13 2018 at 16:03, on Zulip):

that might be useful then for breaking up the labor

pnkfelix (Dec 13 2018 at 16:03, on Zulip):

maybe the first step is to identify how many volunteers (N) we have to help with the processing

pnkfelix (Dec 13 2018 at 16:04, on Zulip):

and then we can just process the issues by number mod N

pnkfelix (Dec 13 2018 at 16:04, on Zulip):

but as I said, I must vamooze

Last update: Nov 16 2019 at 01:25UTC