Stream: t-compiler

Topic: weekly meeting 2018-09-13


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

FYI @T-compiler this is where this week's meeting will be held (in approximately 3 minutes)

nagisa (Sep 13 2018 at 13:58, on Zulip):

o/

nagisa (Sep 13 2018 at 13:58, on Zulip):

/me gladly closes discord tab

mw (Sep 13 2018 at 14:01, on Zulip):

o/

pnkfelix (Sep 13 2018 at 14:01, on Zulip):

okay so lets get started

pnkfelix (Sep 13 2018 at 14:01, on Zulip):

/me is dreading getting to the P-high bugs assigned to him

pnkfelix (Sep 13 2018 at 14:01, on Zulip):

P-high: http://is.gd/0ohWmp

pnkfelix (Sep 13 2018 at 14:01, on Zulip):

first up, unstable alloc crate can be used without feature gates on the 2018 edition #54006

nikomatsakis (Sep 13 2018 at 14:02, on Zulip):

oh dear

pnkfelix (Sep 13 2018 at 14:03, on Zulip):

well it may be misleading

pnkfelix (Sep 13 2018 at 14:03, on Zulip):

... or at leasts, the bug is in that we allow you to reference vec via use alloc::vec

nikomatsakis (Sep 13 2018 at 14:03, on Zulip):

yes, I read it now

nikomatsakis (Sep 13 2018 at 14:03, on Zulip):

I left a comment

nikomatsakis (Sep 13 2018 at 14:03, on Zulip):

regardless, seems like @eddyb is on it

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

So lets assign to @eddyb and move on then?

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

Next up: Macro resolution regression when a #[test] is named 'panic' #53675

nikomatsakis (Sep 13 2018 at 14:05, on Zulip):

hmm

pnkfelix (Sep 13 2018 at 14:05, on Zulip):

on the one hand, @Vadim Petrochenkov remains the right person to investigate this

nikomatsakis (Sep 13 2018 at 14:06, on Zulip):

yes, but it's unclear how much time they have etc

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

The comment about "this issue is kinda known" makes me wonder if we are supposed to revisit the priority here?

pnkfelix (Sep 13 2018 at 14:07, on Zulip):

though I don't interpret that link to that commit as implying that this is "known bug" ..

pnkfelix (Sep 13 2018 at 14:07, on Zulip):

("unknown knowns")

nikomatsakis (Sep 13 2018 at 14:08, on Zulip):

I guess I think this remains important

pnkfelix (Sep 13 2018 at 14:08, on Zulip):

Do we consider trying to open this up to other volunteers to resolve?

pnkfelix (Sep 13 2018 at 14:09, on Zulip):

well I don't want to block the meeting on it

nikomatsakis (Sep 13 2018 at 14:09, on Zulip):

maybe? it seems like somebody should try to investigate, but we ought to ping @Vadim Petrochenkov also to see if they have any further thouthts

pnkfelix (Sep 13 2018 at 14:09, on Zulip):

I'm going to assign it to myself in the hopes that I then remember to either look at it or delegate it

nikomatsakis (Sep 13 2018 at 14:10, on Zulip):

heh I was gonna say the same

nikomatsakis (Sep 13 2018 at 14:10, on Zulip):

well, we can talk later

nikomatsakis (Sep 13 2018 at 14:10, on Zulip):

I'd like to read more into resolve code

nikomatsakis (Sep 13 2018 at 14:10, on Zulip):

next issue?

pnkfelix (Sep 13 2018 at 14:10, on Zulip):

Next: Tracking issue for RFC 2093: Infer T: 'x outlives requirements on structs #44493

nikomatsakis (Sep 13 2018 at 14:10, on Zulip):

stabilization PR landed

nikomatsakis (Sep 13 2018 at 14:10, on Zulip):

I guess we can close that issue

nikomatsakis (Sep 13 2018 at 14:11, on Zulip):

or at least remove the P-high label

nikomatsakis (Sep 13 2018 at 14:11, on Zulip):

ok, closed, seems fine

Zack M. Davis (Sep 13 2018 at 14:11, on Zulip):

and the corresponding lint has just now been rebased: https://github.com/rust-lang/rust/pull/53013

pnkfelix (Sep 13 2018 at 14:11, on Zulip):

oh well

nikomatsakis (Sep 13 2018 at 14:11, on Zulip):

I was just going to mention the lint

pnkfelix (Sep 13 2018 at 14:11, on Zulip):

not all the checkboxes (in description) were checked off

pnkfelix (Sep 13 2018 at 14:12, on Zulip):

but I assume that was just an oversight?

nikomatsakis (Sep 13 2018 at 14:12, on Zulip):

I think we should just open a new issue for the lint

nikomatsakis (Sep 13 2018 at 14:12, on Zulip):

if it doesn't exist

nikomatsakis (Sep 13 2018 at 14:12, on Zulip):

(not sure though: do we have a separate issue for the 'static case?)

nikomatsakis (Sep 13 2018 at 14:12, on Zulip):

anyway i'll figure it out

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

Okay then, that's all the P-highs. (yay)

Zack M. Davis (Sep 13 2018 at 14:13, on Zulip):

lint issue is https://github.com/rust-lang/rust/issues/52042

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

next stable-to-beta regressions: https://github.com/rust-lang/rust/labels/regression-from-stable-to-beta

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

oh yes, Recent 12% perf regression in unused-warnings #52092

nikomatsakis (Sep 13 2018 at 14:14, on Zulip):

didn't we decide not to touch that?

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

I guess this is still downgraded

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

yes, but we shouldn't remove the regression label, should we?

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

Maybe in future I should skip the P-medium and P-low regressinos

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

will do

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

okay so then next up, stable-to-nightly regressions: https://github.com/rust-lang/rust/labels/regression-from-stable-to-nightly

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

first up is ICE compiling the objrs crate #54059

pnkfelix (Sep 13 2018 at 14:16, on Zulip):

I assume @Ariel Ben-Yehuda is not here

nikomatsakis (Sep 13 2018 at 14:16, on Zulip):

Maybe in future I should skip the P-medium and P-low regressinos

that's what I usually do, yeah

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

Okay I will assign this to self to followup with @Ariel Ben-Yehuda if possible, or look into if need be

nikomatsakis (Sep 13 2018 at 14:17, on Zulip):

seems like P-high

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

Oh yeah, priority

nikomatsakis (Sep 13 2018 at 14:17, on Zulip):

another possibility might be @mw

nikomatsakis (Sep 13 2018 at 14:17, on Zulip):

(to follow up)

nikomatsakis (Sep 13 2018 at 14:18, on Zulip):

not sure how much time you have, @mw ?

mw (Sep 13 2018 at 14:18, on Zulip):

yes, you can assign to me

pnkfelix (Sep 13 2018 at 14:18, on Zulip):

I'll remove the needs-bisection tag

pnkfelix (Sep 13 2018 at 14:18, on Zulip):

okay reassigning to mw

pnkfelix (Sep 13 2018 at 14:18, on Zulip):

okay next up

pnkfelix (Sep 13 2018 at 14:18, on Zulip):

we already discussed #53675

pnkfelix (Sep 13 2018 at 14:18, on Zulip):

so last is "rustc incorrectly parses attributes as macro invocations in nightly-2018-08-18 #53583"

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

@Vadim Petrochenkov says it is fixed by PR #54069

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

which is seeking T-lang review

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

okay well in any case this seems under control

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

going to mark as P-high though

nikomatsakis (Sep 13 2018 at 14:19, on Zulip):

sounds good

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

that's all the regressions

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

look at that, we're only 20minutes into the mtg, yay

pnkfelix (Sep 13 2018 at 14:21, on Zulip):

okay next up, BetaNominated: https://github.com/rust-lang/rust/issues?utf8=%E2%9C%93&q=label%3Abeta-nominated

nikomatsakis (Sep 13 2018 at 14:21, on Zulip):

so.. argh

pnkfelix (Sep 13 2018 at 14:21, on Zulip):

only one PR, and it hasn't landed in master yet

nikomatsakis (Sep 13 2018 at 14:21, on Zulip):

I did some investigating here but faile to make much progress

nagisa (Sep 13 2018 at 14:21, on Zulip):

(I want to propose thumbs-up for "yay, backport" and thumbs down for "nah, reasons")

nikomatsakis (Sep 13 2018 at 14:21, on Zulip):

the PR didn't really fix the regression in any case

pnkfelix (Sep 13 2018 at 14:22, on Zulip):

so we're talking about [WIP] Fix evaluation overflow #53687

nikomatsakis (Sep 13 2018 at 14:22, on Zulip):

(I want to propose thumbs-up for "yay, backport" and thumbs down for "nah, reasons")

(makes sense in general)

pnkfelix (Sep 13 2018 at 14:22, on Zulip):

the relevant bug is "ICE when running cargo doc on typenum at librustc/traits/structural_impls.rs:178 #52873"

nikomatsakis (Sep 13 2018 at 14:23, on Zulip):

it converts the ICE into a (probably incorrect?) error

nikomatsakis (Sep 13 2018 at 14:23, on Zulip):

but it's still kind of unclear to me what is going on

nikomatsakis (Sep 13 2018 at 14:23, on Zulip):

I should read more closely into it

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

"incorrect" as in, we should have accepted the code?

nikomatsakis (Sep 13 2018 at 14:23, on Zulip):

point is, I'm not really happy yet about landing this PR period, much less backporting it :)

nikomatsakis (Sep 13 2018 at 14:23, on Zulip):

well, we do accept the code, rustdoc just doesn't

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

or "incorrect" as in, "this is misleading as to what the error is"

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

hmm

pnkfelix (Sep 13 2018 at 14:24, on Zulip):

well maybe lets just remove the beta-nomination tag

nikomatsakis (Sep 13 2018 at 14:24, on Zulip):

but the other point is: if I am so grumpy, I need to dedicate time to fixing it

nikomatsakis (Sep 13 2018 at 14:24, on Zulip):

seems ok, but maybe also mark the issue P-high and assign to me, I think

nikomatsakis (Sep 13 2018 at 14:24, on Zulip):

I can do that

pnkfelix (Sep 13 2018 at 14:24, on Zulip):

since it sounds very unlikely that we're going to approve this PR unless it undergoes serious reviision?

pnkfelix (Sep 13 2018 at 14:26, on Zulip):

okay that's all the beta-nominations

nikomatsakis (Sep 13 2018 at 14:26, on Zulip):

ok, I also marked original issue T-compiler, P-high and assigned to me

pnkfelix (Sep 13 2018 at 14:26, on Zulip):

next, PR's waiting on our team: https://github.com/rust-lang/rust/pulls?q=is%3Aopen+is%3Apr+label%3AS-waiting-on-team+label%3AT-compiler

pnkfelix (Sep 13 2018 at 14:26, on Zulip):

ah fun this is relevant to me

nagisa (Sep 13 2018 at 14:27, on Zulip):

#54139

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

"this" being ... yes, " [WIP] migrate more of src/test/run-pass to src/test/ui/run-pass #54139 "

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

but really the actual thing to discuss might be "Implement some way to run UI tests ignoring run-pass tests #54047"

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

so let me give the background here

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

For NLL, we realized that while we were testing all sorts of diagnostics

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

we were failing to check that all the run-pass tests would continue to run-pass under NLL

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

(which was filed as "NLL: need to run the run-pass tests under compare-mode=nll #53764")

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

#53764 proposed various ways to attack

nagisa (Sep 13 2018 at 14:29, on Zulip):

what functional differences NLL can introduce into the code for run-pass to matter here?

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

it can reject your code incorrectly

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

and also it currently changes the match code gen

nikomatsakis (Sep 13 2018 at 14:29, on Zulip):

(fwiw, I also think it's good for us to check the warnings and things generated by run-pass tests)

nikomatsakis (Sep 13 2018 at 14:29, on Zulip):

(independently from NLL)

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

(which perhaps should not be made conditional on NLL)

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

((the match codegen changes))

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

Anyway

nagisa (Sep 13 2018 at 14:30, on Zulip):

okay, rejecting the code seems hardly important, but it changing the match codegen is :slight_smile:

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

I singlehandledly decided to resolve #53764 by porting run-pass into ui/

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

and @Vadim Petrochenkov pointed out that this is a UX regression for them and other developers

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

(in terms of rustc hacking UX)

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

so, I agree with that statement, and just want help deciding how to resolve

nikomatsakis (Sep 13 2018 at 14:31, on Zulip):

and @Vadim Petrochenkov pointed out that this is a UX regression for them and other developers

say a bit more?

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

So I wrote this comment https://github.com/rust-lang/rust/issues/54047#issuecomment-420562280

nikomatsakis (Sep 13 2018 at 14:31, on Zulip):

or maybe I should read

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

oh, the UX regression is jus that

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

when they do x.py test ui or whatever it is

nikomatsakis (Sep 13 2018 at 14:31, on Zulip):

it takes longer to run the tests basically?

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

they don't want to waste time on the run-pass tests

pnkfelix (Sep 13 2018 at 14:32, on Zulip):

yes I think the time regression was the main issue

nikomatsakis (Sep 13 2018 at 14:32, on Zulip):

ok

nikomatsakis (Sep 13 2018 at 14:32, on Zulip):

there are also some run-pass tests that fail on stage1 for .. some reason

nikomatsakis (Sep 13 2018 at 14:32, on Zulip):

that I don't really understand

nikomatsakis (Sep 13 2018 at 14:32, on Zulip):

and haven't investigated

pnkfelix (Sep 13 2018 at 14:32, on Zulip):

so I proposed four alternative approaches here: https://github.com/rust-lang/rust/issues/54047#issuecomment-420562280

nikomatsakis (Sep 13 2018 at 14:32, on Zulip):

but that is annoying

nagisa (Sep 13 2018 at 14:32, on Zulip):

for me half of the test suite consistently fails all the time since time immemorial

pnkfelix (Sep 13 2018 at 14:32, on Zulip):

labelled (1a), (1b), (2a), (2b)

pnkfelix (Sep 13 2018 at 14:32, on Zulip):

so maybe its best if you all go read that comment quickly

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

or maybe better still

nikomatsakis (Sep 13 2018 at 14:33, on Zulip):

hmm

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

just weigh in on the ticket, async?

nikomatsakis (Sep 13 2018 at 14:33, on Zulip):

yeah, will do, I'm torn :)

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

basically I want to know what I shoud try to do next

nagisa (Sep 13 2018 at 14:33, on Zulip):

so I just learnt to ignore those. I feel that UX is already not ideal

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

I'm willing to do the implementatino legwork

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

but I don't want to waste time implementing somethign that people will tehn reject

pnkfelix (Sep 13 2018 at 14:34, on Zulip):

I did realize that one approach we could use

pnkfelix (Sep 13 2018 at 14:34, on Zulip):

is to reinterpret the --mode argument to compiletest

nagisa (Sep 13 2018 at 14:34, on Zulip):

@pnkfelix it is not clear to me yet, why do we run run-pass tests as part of x.py test ui now?

pnkfelix (Sep 13 2018 at 14:34, on Zulip):

because they were moved into ui/run-pass/

pnkfelix (Sep 13 2018 at 14:35, on Zulip):

src/test/run-pass/ is nearly empty

pnkfelix (Sep 13 2018 at 14:35, on Zulip):

especially after PR #54139

nagisa (Sep 13 2018 at 14:35, on Zulip):

okay, so how about we make src/test/run-pass use the ui infrastructure

nagisa (Sep 13 2018 at 14:35, on Zulip):

instead

nagisa (Sep 13 2018 at 14:35, on Zulip):

then it would not break the usecase of x.py test src/test/ui

pnkfelix (Sep 13 2018 at 14:35, on Zulip):

well

pnkfelix (Sep 13 2018 at 14:35, on Zulip):

one reason is that some tests can't currently use ui/

pnkfelix (Sep 13 2018 at 14:35, on Zulip):

or at least, it wasn't reasonable for me to port some of them

pnkfelix (Sep 13 2018 at 14:35, on Zulip):

We could make a run-pass-ui

nagisa (Sep 13 2018 at 14:35, on Zulip):

mv src/test/run-pass src/test/extra-special-run-pass-tests/

nagisa (Sep 13 2018 at 14:36, on Zulip):

or yeah, the other way around

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

I don't mind moving around the remaining run-pass tests to somewhere lese

nagisa (Sep 13 2018 at 14:36, on Zulip):

would doing so resolve Petrochenkov’s concerns?

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

like run-pass-fulldeps

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

I bet it would

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

The other issue here, thought

pnkfelix (Sep 13 2018 at 14:37, on Zulip):

is that longer term, I think @nikomatsakis wants a feature-oriented layout to all of our tests

pnkfelix (Sep 13 2018 at 14:37, on Zulip):

or at least that was claimed by @Vadim Petrochenkov here: https://github.com/rust-lang/rust/issues/54047#issuecomment-420588732

nikomatsakis (Sep 13 2018 at 14:37, on Zulip):

I left a comment

nikomatsakis (Sep 13 2018 at 14:37, on Zulip):

but indeed I think that src/test/run-pass (by whatever name) is suboptimal

nikomatsakis (Sep 13 2018 at 14:38, on Zulip):

I'd prefer to have the tests grouped by what they test

nikomatsakis (Sep 13 2018 at 14:38, on Zulip):

and if you want to filter by "we expect UI output", then have a way to do that

nagisa (Sep 13 2018 at 14:38, on Zulip):

moving to that scheme would break x.py test ui anyway

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

okay well I think the approach suggested by @nagisa may be best in the short term.

nikomatsakis (Sep 13 2018 at 14:38, on Zulip):

moving to that scheme would break x.py test ui anyway

what scheme do you mean?

nagisa (Sep 13 2018 at 14:38, on Zulip):

so might as well just go with whatever is a path of least resistance now

nikomatsakis (Sep 13 2018 at 14:38, on Zulip):

I also don't want to waste a ton of time

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

it will still mean I'll have to do a little more work (to make a --compare-mode=nll analogue that applies to the new ui-type run-pass/)

nikomatsakis (Sep 13 2018 at 14:39, on Zulip):

I wouldn't object to run-pass-ui per se as a temporary thing

nagisa (Sep 13 2018 at 14:39, on Zulip):

moving to that scheme would break x.py test ui anyway

what scheme do you mean?

the per-feature scheme. You would have to specify whatever else than ui in your cmd.

nikomatsakis (Sep 13 2018 at 14:39, on Zulip):

well, it'd be subdirectories of ui?

nikomatsakis (Sep 13 2018 at 14:39, on Zulip):

I mean it would basically just be "tests"

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

we clearly need a semantic FS

nikomatsakis (Sep 13 2018 at 14:39, on Zulip):

because there is no need for anything that is not ui

nikomatsakis (Sep 13 2018 at 14:39, on Zulip):

(ok, I'm ignoring all the other random tests, like gdb or whatever)

nikomatsakis (Sep 13 2018 at 14:39, on Zulip):

maybe rename ui to rs :)

eddyb (Sep 13 2018 at 14:39, on Zulip):

those tests should be moved to ui tests

eddyb (Sep 13 2018 at 14:39, on Zulip):

and just drop the ui directory prefix

nikomatsakis (Sep 13 2018 at 14:40, on Zulip):

I think fewer "modes" of tests is good

nikomatsakis (Sep 13 2018 at 14:40, on Zulip):

in short

nagisa (Sep 13 2018 at 14:40, on Zulip):

@nikomatsakis having a hierarchcal structure is hard, because if you wanted to run nll tests specifically, you wouldn’t be able to run all nll tests from all suites (compile-fail/ui/etc).

eddyb (Sep 13 2018 at 14:40, on Zulip):

MIR tests? UI. debugger tests? UI. codegen tests? UI

nikomatsakis (Sep 13 2018 at 14:40, on Zulip):

or rather, move the modes into comments

qmx (Sep 13 2018 at 14:40, on Zulip):

this definitely was one of the pain points while being onboarded on the compiler

eddyb (Sep 13 2018 at 14:40, on Zulip):

in restrospect, "UI" is misleading. these are output tests. which all tests are, really

nikomatsakis (Sep 13 2018 at 14:40, on Zulip):

@nagisa yes, there is no perfect hierarchy, I've been thinking about // keywords in the tests as well

pnkfelix (Sep 13 2018 at 14:40, on Zulip):

so wait

pnkfelix (Sep 13 2018 at 14:40, on Zulip):

there's two things here

pnkfelix (Sep 13 2018 at 14:40, on Zulip):

there's the short term Q

eddyb (Sep 13 2018 at 14:40, on Zulip):

+1 for getting rid of the "ui" distinction and moving modes to comments

pnkfelix (Sep 13 2018 at 14:40, on Zulip):

of what should @pnkfelix do right now

pnkfelix (Sep 13 2018 at 14:40, on Zulip):

and the longer term Q

nikomatsakis (Sep 13 2018 at 14:41, on Zulip):

yep, we should separate :)

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

of how we want to structure our test suite

eddyb (Sep 13 2018 at 14:41, on Zulip):

with longer-term replacing all the modes with a more unified system

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

I want to make sure we get through other nominated stuff

nagisa (Sep 13 2018 at 14:41, on Zulip):

@pnkfelix I thought we already figured out the short term :slight_smile:

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

before we discuss longer term Q

nagisa (Sep 13 2018 at 14:41, on Zulip):

yeah lets move on then.

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

okay, good good

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

that was the one PR waiting on our team. (And I will close it.)

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

okay, I-nominated: https://github.com/rust-lang/rust/issues?utf8=%E2%9C%93&q=is%3Aopen%20label%3AT-compiler%20label%3AI-nominated%20

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

we just discussed "Implement some way to run UI tests ignoring run-pass tests #54047"

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

but we shoud put a priority on it.

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

I'll just make it P-high

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

(where the plan is for me to move ui/run-pass/ back to src/test/run-pass/, after I make the latter a ui-suite variant.)

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

any objections to P-high?

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

(strategy itself can be discussed on the ticket, asynchronously

nikomatsakis (Sep 13 2018 at 14:44, on Zulip):

@pnkfelix did we completely migrate all run-pass tests?

pnkfelix (Sep 13 2018 at 14:45, on Zulip):

No, not 100%

pnkfelix (Sep 13 2018 at 14:45, on Zulip):

but most

pnkfelix (Sep 13 2018 at 14:45, on Zulip):

or rather

pnkfelix (Sep 13 2018 at 14:45, on Zulip):

there was a PR that got us to like 99.9%

nikomatsakis (Sep 13 2018 at 14:45, on Zulip):

ok

nikomatsakis (Sep 13 2018 at 14:45, on Zulip):

I am fine with making a ui-run-pass directory or whatever

pnkfelix (Sep 13 2018 at 14:45, on Zulip):

so its an unfortunate bit of churn

nikomatsakis (Sep 13 2018 at 14:45, on Zulip):

or just run-pass, but in that case I'd prefer to kill the distinct run-pass mode

pnkfelix (Sep 13 2018 at 14:45, on Zulip):

moving back and forth and all-around

nikomatsakis (Sep 13 2018 at 14:45, on Zulip):

(I still prefer to leverage --test-args for this)

nikomatsakis (Sep 13 2018 at 14:45, on Zulip):

and it might be easier

nikomatsakis (Sep 13 2018 at 14:46, on Zulip):

anyway sorry

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

okay lets discuss outside of meeting

eddyb (Sep 13 2018 at 14:46, on Zulip):

--test-args should not be a thing IMO

nikomatsakis (Sep 13 2018 at 14:46, on Zulip):

I know we said we settled :)

eddyb (Sep 13 2018 at 14:46, on Zulip):

x.py should integrate better with compiletest

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

oh strategy can be revised, just don't want to spend time on it now

nikomatsakis (Sep 13 2018 at 14:46, on Zulip):

orthogonal!

eddyb (Sep 13 2018 at 14:46, on Zulip):

and allow passing subdirectories/exact file names

nikomatsakis (Sep 13 2018 at 14:46, on Zulip):

but I agree :)

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

next up: Regression in #[allow(deprecated)] for impl Trait return type #54045

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

egad

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

P-high?

nikomatsakis (Sep 13 2018 at 14:48, on Zulip):

I think so

pnkfelix (Sep 13 2018 at 14:48, on Zulip):

Seems P-high to me

pnkfelix (Sep 13 2018 at 14:48, on Zulip):

don't know whom to give it to

nikomatsakis (Sep 13 2018 at 14:48, on Zulip):

not clear whether @Oli and @eddyb agreed on what to do

nikomatsakis (Sep 13 2018 at 14:48, on Zulip):

(also, I think @Oli is not around just now)

eddyb (Sep 13 2018 at 14:49, on Zulip):

uhhh

pnkfelix (Sep 13 2018 at 14:49, on Zulip):

I'm going to assign it to both @Oli and @eddyb

pnkfelix (Sep 13 2018 at 14:49, on Zulip):

and let them fight

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

AFAIK @Oli is gonna be gone for a while now.

eddyb (Sep 13 2018 at 14:49, on Zulip):

I basically think the "clean" approach oli moved to causes too many problems

eddyb (Sep 13 2018 at 14:49, on Zulip):

that is, it's overall better to keep the synthetic existential type nested in the HIR

eddyb (Sep 13 2018 at 14:50, on Zulip):

(which it was before existential type had a surface syntax)

nikomatsakis (Sep 13 2018 at 14:50, on Zulip):

they should not be siblings, you mean

nikomatsakis (Sep 13 2018 at 14:50, on Zulip):

probably

eddyb (Sep 13 2018 at 14:50, on Zulip):

@Oli keeps mentioning siblings vs inside the body

nikomatsakis (Sep 13 2018 at 14:50, on Zulip):

maybe we can discuss in a separate topic, I'm curious exactly what you have in mind

eddyb (Sep 13 2018 at 14:50, on Zulip):

whereas I prefer the hir::Ty node being the parent of the existential type item

pnkfelix (Sep 13 2018 at 14:51, on Zulip):

yeah, let s move on, we're so close!

pnkfelix (Sep 13 2018 at 14:51, on Zulip):

next: ICE in libserialize with incremental build across changed rustc version #53792

pnkfelix (Sep 13 2018 at 14:51, on Zulip):

man I feel like I keep seeing issues with incremental builds

pnkfelix (Sep 13 2018 at 14:51, on Zulip):

I know I've given up using it in some contexts

nikomatsakis (Sep 13 2018 at 14:52, on Zulip):

hmm

nikomatsakis (Sep 13 2018 at 14:52, on Zulip):

I feel like we tried to make incremental robust against this case, right @mw ?

nikomatsakis (Sep 13 2018 at 14:52, on Zulip):

however

mw (Sep 13 2018 at 14:52, on Zulip):

well, incr. comp. is pervading the whole compiler, of course you will keep seeing issues related to it :)

nikomatsakis (Sep 13 2018 at 14:52, on Zulip):

we might be relying a bit on cargo for that, not sure

pnkfelix (Sep 13 2018 at 14:52, on Zulip):

anyway, seems P-high. And assign to @mw ?

nikomatsakis (Sep 13 2018 at 14:52, on Zulip):

(I use it all the time. Lovin' it.)

Zack M. Davis (Sep 13 2018 at 14:52, on Zulip):

If libserialize itself is at fault, maybe we should be moving to Serde? (given deprecation anyway: https://github.com/rust-lang-deprecated/rustc-serialize#rustc-serialize)

mw (Sep 13 2018 at 14:52, on Zulip):

i think, rustc is using the git hash and if that doesn't properly change, it fails

nagisa (Sep 13 2018 at 14:53, on Zulip):

/me goes to deal with real life, thanks all

pnkfelix (Sep 13 2018 at 14:53, on Zulip):

I just know I see various linker errors that seem to go away when I turn it off

mw (Sep 13 2018 at 14:53, on Zulip):

libserialize is not the problem

eddyb (Sep 13 2018 at 14:53, on Zulip):

serde is not doable today

eddyb (Sep 13 2018 at 14:53, on Zulip):

and won't be for another release cycle or two

mw (Sep 13 2018 at 14:53, on Zulip):

@pnkfelix did you open issues about those?

pnkfelix (Sep 13 2018 at 14:53, on Zulip):

I've opened issues

mw (Sep 13 2018 at 14:53, on Zulip):

you can assign this issue to me

pnkfelix (Sep 13 2018 at 14:54, on Zulip):

but I haven't necessarily tagged them as incremental related

nikomatsakis (Sep 13 2018 at 14:54, on Zulip):

I'm guessing @Taylor Cramer could help you reproduce if needed

mw (Sep 13 2018 at 14:54, on Zulip):

I'll take a look if the check can be made more conservative

pnkfelix (Sep 13 2018 at 14:54, on Zulip):

becasued usually by the time I've done a rebuild with incremetnal turned off

pnkfelix (Sep 13 2018 at 14:54, on Zulip):

I'm trying to get something else done

eddyb (Sep 13 2018 at 14:54, on Zulip):

(basically we can't use custom derives until https://github.com/rust-lang/rust/pull/49219 reaches beta at least, and the edition makes finishing that PR harder)

pnkfelix (Sep 13 2018 at 14:55, on Zulip):

okay well i assigned the incremental issue to @mw

pnkfelix (Sep 13 2018 at 14:55, on Zulip):

so lets move on

pnkfelix (Sep 13 2018 at 14:55, on Zulip):

next is: Mistaken "dropped here while still borrowed" with impl trait + match #52706

pnkfelix (Sep 13 2018 at 14:55, on Zulip):

I didn't get a chance to look at this yet

eddyb (Sep 13 2018 at 14:56, on Zulip):

that's just impl Trait

eddyb (Sep 13 2018 at 14:56, on Zulip):

same as the issue you closed (other than there not being a good way to say it has no destructor)

Taylor Cramer (Sep 13 2018 at 14:56, on Zulip):

@mw feel free to ping me later if you need help reproducing

pnkfelix (Sep 13 2018 at 14:56, on Zulip):

@eddyb but in this case adding that other statement fixes it?

mw (Sep 13 2018 at 14:56, on Zulip):

@Taylor Cramer thanks!

pnkfelix (Sep 13 2018 at 14:56, on Zulip):

you think that's still a destructor issue?

eddyb (Sep 13 2018 at 14:57, on Zulip):

yes, it's a scope thing

eddyb (Sep 13 2018 at 14:57, on Zulip):

any statement would work, I think?

eddyb (Sep 13 2018 at 14:57, on Zulip):

or just a semicolon, maybe

eddyb (Sep 13 2018 at 14:57, on Zulip):

yeah semicolon fixes it

nikomatsakis (Sep 13 2018 at 14:57, on Zulip):

man I hate that

pnkfelix (Sep 13 2018 at 14:57, on Zulip):

yeah okay so then I'd categorize it as one of those annoying cases where adding a semicolon does the magic

pnkfelix (Sep 13 2018 at 14:58, on Zulip):

we have an issue for that somewhere

nikomatsakis (Sep 13 2018 at 14:58, on Zulip):

we should really see if we can fix that scope issue

nikomatsakis (Sep 13 2018 at 14:58, on Zulip):

yeah, I know

eddyb (Sep 13 2018 at 14:58, on Zulip):

because the compiler thinks impl Abc + 'a can have a destructor, it's the good ol' "destructory things in block tail expression"

pnkfelix (Sep 13 2018 at 14:58, on Zulip):

e.g. "fallout from Sound Generic Drop to be addressed later #22321"

eddyb (Sep 13 2018 at 14:58, on Zulip):

yuuup

pnkfelix (Sep 13 2018 at 14:58, on Zulip):

or maybe "Temporary lifetimes sometimes yield surprising errors #46413"

nikomatsakis (Sep 13 2018 at 14:58, on Zulip):

but I agree with @eddyb's diagnosis, sounds right

pnkfelix (Sep 13 2018 at 14:58, on Zulip):

okay

pnkfelix (Sep 13 2018 at 14:58, on Zulip):

I'll write up some notes on the bug

pnkfelix (Sep 13 2018 at 14:59, on Zulip):

link it accordingly

pnkfelix (Sep 13 2018 at 14:59, on Zulip):

and then close it

pnkfelix (Sep 13 2018 at 14:59, on Zulip):

sound good to everyone?

pnkfelix (Sep 13 2018 at 14:59, on Zulip):

(alternatively I could leave it open but set to P-low)

pnkfelix (Sep 13 2018 at 14:59, on Zulip):

(but I think we've been just closing instances of things like this...)

pnkfelix (Sep 13 2018 at 15:00, on Zulip):

okay okay one last bug nominated, : "infinite recursion ICE #52701"

pnkfelix (Sep 13 2018 at 15:00, on Zulip):

we can do this! (as in, get through the whole list in a meeting...)

pnkfelix (Sep 13 2018 at 15:00, on Zulip):

so close ...

nikomatsakis (Sep 13 2018 at 15:00, on Zulip):

(but I think we've been just closing instances of things like this...)

I think we should pick a tracking issue and then close duplicates, but add links to the tracking issue

eddyb (Sep 13 2018 at 15:00, on Zulip):

I'm not seeing an ICE

nikomatsakis (Sep 13 2018 at 15:00, on Zulip):

I was just going to say the same

Wesley Wiser (Sep 13 2018 at 15:01, on Zulip):

It's hidden in the backtrace

eddyb (Sep 13 2018 at 15:01, on Zulip):

found it too

nikomatsakis (Sep 13 2018 at 15:01, on Zulip):

this version seems to ICE as well

nikomatsakis (Sep 13 2018 at 15:01, on Zulip):

you can assign to me @pnkfelix, perhaps

nikomatsakis (Sep 13 2018 at 15:02, on Zulip):

the ICE seems related to the query normalization stuff

nikomatsakis (Sep 13 2018 at 15:02, on Zulip):

maybe even related to that rustdoc failure

pnkfelix (Sep 13 2018 at 15:02, on Zulip):

okay. What priority, P-high ?

nikomatsakis (Sep 13 2018 at 15:02, on Zulip):

I guess

pnkfelix (Sep 13 2018 at 15:03, on Zulip):

okay great, that's a wrap everyone

Last update: Nov 22 2019 at 05:55UTC