Stream: t-compiler

Topic: weekly meeting 2019-11-21 #54818


pnkfelix (Nov 21 2019 at 13:07, on Zulip):

Hi @T-compiler/meeting ; the triage meeting will be starting in 1 hours 53 minutes

pnkfelix (Nov 21 2019 at 13:08, on Zulip):

I will be doing pre-triage in a parallel topic

pnkfelix (Nov 21 2019 at 13:08, on Zulip):

This week we have a WG check-in scheduled for WG-traits

pnkfelix (Nov 21 2019 at 13:09, on Zulip):

@nikomatsakis do you think you'll want to report anything? (Or identify someone else to do reporting?)

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

in any case, inspired by last week's steering meeting about WG's, I have created a dedicated zulip topic in #wg-traits for this: WG checkin notes

pnkfelix (Nov 21 2019 at 13:12, on Zulip):

(and I'll add similar note taking topics for next weeks' WG's too)

pnkfelix (Nov 21 2019 at 15:00, on Zulip):

Hi @T-compiler/meeting! Add a :wave: emoji to show you're here :)

pnkfelix (Nov 21 2019 at 15:01, on Zulip):

We'll start off with five minutes for:

pnkfelix (Nov 21 2019 at 15:01, on Zulip):

Announcements

pnkfelix (Nov 21 2019 at 15:01, on Zulip):

I'm trying something new this week: I've tried to predraft the agenda into a hackmd

pnkfelix (Nov 21 2019 at 15:02, on Zulip):

(this is inspired by @centril 's practice at the lang team mtgs, though they use paper.dropbox.com)

pnkfelix (Nov 21 2019 at 15:02, on Zulip):

my main motivation for this was to try to bias the nominations list towards people who've bribed me accordingly.

simulacrum (Nov 21 2019 at 15:02, on Zulip):

perf.rust-lang.org will be switching to a new collection server, which means that all old data will be gone from the official server (but not actually deleted) sometime in the next week

centril (Nov 21 2019 at 15:03, on Zulip):

(what's the currency for bribes? ^^ -- r+ses?)

Wesley Wiser (Nov 21 2019 at 15:03, on Zulip):

const prop on by default landed on the 19th with a nice drop on a number of perf metrics https://perf.rust-lang.org/?start=2019-11-18&end=2019-11-19&absolute=true&stat=instructions%3Au

pnkfelix (Nov 21 2019 at 15:03, on Zulip):

@simulacrum is the deletion something to celebrate or mourn ?

pnkfelix (Nov 21 2019 at 15:03, on Zulip):

(what's the currency for bribes? ^^ -- r+ses?)

work-stealing open P-high issues I think.

simulacrum (Nov 21 2019 at 15:03, on Zulip):

depends on perspective

pnkfelix (Nov 21 2019 at 15:04, on Zulip):

@simulacrum okay, I might follow up with you about this afterward

simulacrum (Nov 21 2019 at 15:04, on Zulip):

Happy to discuss further :)

pnkfelix (Nov 21 2019 at 15:04, on Zulip):

Oh, in case you all didn't see it, I wrote a blog post about minimizing rustc bugs

pnkfelix (Nov 21 2019 at 15:04, on Zulip):

its ... long. Just like the bugs you're reducing.

pnkfelix (Nov 21 2019 at 15:05, on Zulip):

(I welcome posts on blog post minimization patterns)

centril (Nov 21 2019 at 15:05, on Zulip):

/me likes shrinking easy ICEs that fit into the playground

pnkfelix (Nov 21 2019 at 15:06, on Zulip):

so lets see, lets start with beta-nominations.We have one

pnkfelix (Nov 21 2019 at 15:06, on Zulip):

“Do not ICE on trait aliases with missing obligations” #66392

pnkfelix (Nov 21 2019 at 15:07, on Zulip):

wait, heh

pnkfelix (Nov 21 2019 at 15:07, on Zulip):

we accepted this last week , and I wrote a comment saying so today

pnkfelix (Nov 21 2019 at 15:07, on Zulip):

but forgot to tag it

pnkfelix (Nov 21 2019 at 15:07, on Zulip):

so great

pnkfelix (Nov 21 2019 at 15:08, on Zulip):

we've got a bunch of stable-nominations to go through

pnkfelix (Nov 21 2019 at 15:08, on Zulip):

“Do not ICE on trait aliases with missing obligations” #66392

centril (Nov 21 2019 at 15:08, on Zulip):

feels safe

pnkfelix (Nov 21 2019 at 15:09, on Zulip):

its a little bit larger than I'd like for something that does not seem to be fixing a mission critical issue

pnkfelix (Nov 21 2019 at 15:09, on Zulip):

but if people are okay with it, i won't stop it

pnkfelix (Nov 21 2019 at 15:09, on Zulip):

@nikomatsakis do you have thoughts?

centril (Nov 21 2019 at 15:09, on Zulip):

Leave the "should we make point release?" to T-release ;)

Esteban Küber (Nov 21 2019 at 15:09, on Zulip):

Voting on my own PR feels funny

pnkfelix (Nov 21 2019 at 15:10, on Zulip):

I'm not talking about the "make point release"

centril (Nov 21 2019 at 15:10, on Zulip):

@Esteban Küber I know how you feel

nikomatsakis (Nov 21 2019 at 15:10, on Zulip):

I'm not sure

pnkfelix (Nov 21 2019 at 15:10, on Zulip):

I'm just talking about "is this worth the effort"

nikomatsakis (Nov 21 2019 at 15:10, on Zulip):

but I agree it feels safe

pnkfelix (Nov 21 2019 at 15:11, on Zulip):

okay well I guess I'll mark as stable-accepted

mw (Nov 21 2019 at 15:11, on Zulip):

this is an unstable feature?

pnkfelix (Nov 21 2019 at 15:11, on Zulip):

@mw the ICE can be triggered without the gate

pnkfelix (Nov 21 2019 at 15:12, on Zulip):

(the same conversation happened last week too. @Esteban Küber that's a vote for putting in tests explicitly)

mw (Nov 21 2019 at 15:12, on Zulip):

ok

pnkfelix (Nov 21 2019 at 15:12, on Zulip):

okay lets move along

pnkfelix (Nov 21 2019 at 15:12, on Zulip):

stable-nom "Fix ICE when trying to suggest Type<> instead of Type()" #66390

pnkfelix (Nov 21 2019 at 15:13, on Zulip):

seems less risky to me than previous one

pnkfelix (Nov 21 2019 at 15:14, on Zulip):

(which was already noted to be non-risky)

pnkfelix (Nov 21 2019 at 15:14, on Zulip):

so I'll mark stable-accepted if no one argues otherwise by end of meeting

pnkfelix (Nov 21 2019 at 15:14, on Zulip):

stable-nom “Do not ICE on recovery from unmet associated type bound obligation” #66388

pnkfelix (Nov 21 2019 at 15:15, on Zulip):

I'm staring at this diff and saying "oh god do I not know Rust"

pnkfelix (Nov 21 2019 at 15:16, on Zulip):

why did this make a difference here

nikomatsakis (Nov 21 2019 at 15:16, on Zulip):

temporary lifetime rules

nikomatsakis (Nov 21 2019 at 15:16, on Zulip):

the .borrow_mut() expires at the end of the let

centril (Nov 21 2019 at 15:16, on Zulip):

because if let desugars to match and the lifetime ends at the end of the match

pnkfelix (Nov 21 2019 at 15:16, on Zulip):

and this causes the compiler to ICE?

centril (Nov 21 2019 at 15:16, on Zulip):

whereas if has DropTemps(..) on the condition

mw (Nov 21 2019 at 15:17, on Zulip):

still weird

nikomatsakis (Nov 21 2019 at 15:17, on Zulip):

and this causes the compiler to ICE?

inside the mutate_fullfillment_errors(&mut errors); callback

nikomatsakis (Nov 21 2019 at 15:17, on Zulip):

it attempts to acquire the refcell

pnkfelix (Nov 21 2019 at 15:17, on Zulip):

aha

nikomatsakis (Nov 21 2019 at 15:17, on Zulip):

in the old code, that was a recursive ref-cell acquire

Wesley Wiser (Nov 21 2019 at 15:17, on Zulip):

The ICE Is thread 'rustc' panicked at 'already borrowed: BorrowMutError', src/libcore/result.rs:1165:5

nikomatsakis (Nov 21 2019 at 15:17, on Zulip):

subtle indeed

pnkfelix (Nov 21 2019 at 15:17, on Zulip):

classic

pnkfelix (Nov 21 2019 at 15:17, on Zulip):

okay, thanks

centril (Nov 21 2019 at 15:17, on Zulip):

still weird

Oh yeah... this is something we might want to change with an edition for if let

centril (Nov 21 2019 at 15:17, on Zulip):

very confusing indeed

nikomatsakis (Nov 21 2019 at 15:18, on Zulip):

I do think we should consider tweaking the temporary lifetime rules

pnkfelix (Nov 21 2019 at 15:18, on Zulip):

okay well similarly, lets stable-accept this unless I hear an objection by meetings end

nikomatsakis (Nov 21 2019 at 15:18, on Zulip):

but not at this moment :)

pnkfelix (Nov 21 2019 at 15:18, on Zulip):

I do think we should consider tweaking the temporary lifetime rules

sounds like a steering meeting topic

pnkfelix (Nov 21 2019 at 15:18, on Zulip):

stable-nom: “find_deprecation: deprecation attr may be ill-formed meta.” #66381

nikomatsakis (Nov 21 2019 at 15:19, on Zulip):

sounds like a steering meeting topic

maybe? not really a "compiler team" meeting thing, more of an RFC/lang team thing I think

centril (Nov 21 2019 at 15:19, on Zulip):

sounds like a steering meeting topic

(+ RFC I'd say; it's fairly subtle)

pnkfelix (Nov 21 2019 at 15:19, on Zulip):

true dat

centril (Nov 21 2019 at 15:20, on Zulip):

maybe? not really a "compiler team" meeting thing, more of an RFC/lang team thing I think

Tho T-compiler's input is appreciated

pnkfelix (Nov 21 2019 at 15:20, on Zulip):

okay, for #66381, lets stable-accept this unless I hear an objection by meetings end

nikomatsakis (Nov 21 2019 at 15:20, on Zulip):

stable-nom: “find_deprecation: deprecation attr may be ill-formed meta.” #66381

so on this one: I thnk it's safe, but I do wonder if a delay_span_bug would be appropriate

nikomatsakis (Nov 21 2019 at 15:20, on Zulip):

but since it already landed etc etc...seems fine

pnkfelix (Nov 21 2019 at 15:21, on Zulip):

is it a sign of an internal compiler problem?

centril (Nov 21 2019 at 15:21, on Zulip):

@nikomatsakis heh... https://github.com/rust-lang/rust/compare/b522dd84fcc3a80ef92a6fb254eaa0cc3237a4a6..91aadf030548214da5a8f39a1b1dbd21db125625

pnkfelix (Nov 21 2019 at 15:21, on Zulip):

I thought this nonetheless arose due to malformed input

nikomatsakis (Nov 21 2019 at 15:21, on Zulip):

it does

nikomatsakis (Nov 21 2019 at 15:22, on Zulip):

point is, it would be a bug if an error had not been reported

pnkfelix (Nov 21 2019 at 15:22, on Zulip):

okay well anyway we can move along

centril (Nov 21 2019 at 15:22, on Zulip):

@pnkfelix it shouldn't be; we do syntactic validation in ast_validation but then we don't transform

pnkfelix (Nov 21 2019 at 15:23, on Zulip):

stable-nom: “parser: don’t use unreachable!() in fn unexpected.” #66361

centril (Nov 21 2019 at 15:23, on Zulip):

( Btw, I can totally recommend https://lexi-lambda.github.io/blog/2019/11/05/parse-don-t-validate/ -- we might want to take inspiration from that in general )

pnkfelix (Nov 21 2019 at 15:24, on Zulip):

in this case ... maybe we should be more caution ...?

pnkfelix (Nov 21 2019 at 15:24, on Zulip):

or no

pnkfelix (Nov 21 2019 at 15:24, on Zulip):

the diff is strictly better, I guess, than what we had before

pnkfelix (Nov 21 2019 at 15:24, on Zulip):

I was just internally musing about the delay_span_bug note here from @Esteban Küber

pnkfelix (Nov 21 2019 at 15:24, on Zulip):

so never mind me; we can wait until we see ICE's in the wild to suggest delay_span_bug here.

pnkfelix (Nov 21 2019 at 15:25, on Zulip):

so once again, i'll mark #66361 as stable-accepted later if I don't see an objection by end of meeting

pnkfelix (Nov 21 2019 at 15:25, on Zulip):

(trying hard to be better about time todya)

pnkfelix (Nov 21 2019 at 15:26, on Zulip):

we have 11 nominated issues

pnkfelix (Nov 21 2019 at 15:26, on Zulip):

lets see how many we can get through in 20 to 25 minutes

pnkfelix (Nov 21 2019 at 15:26, on Zulip):

nominated: “sysroot spans are not printed on some targets (affected: Debian, rust-lang’s own i586; unaffected: Fedora)” #53081

pnkfelix (Nov 21 2019 at 15:27, on Zulip):

@Esteban Küber you were pushing for this, want to summarize?

eddyb (Nov 21 2019 at 15:27, on Zulip):

I've been dreading looking closer at this

Esteban Küber (Nov 21 2019 at 15:27, on Zulip):

Spans to stdlib work and do what you want, but it doesn't work in some 32bit targets

Esteban Küber (Nov 21 2019 at 15:28, on Zulip):

My understanding is due to incorrect packaging making the compiler not find std due to bad paths

pnkfelix (Nov 21 2019 at 15:28, on Zulip):

what do we want to do about this? Should we try to figure out what the source of the discrepancy is? Or should we attempt to apply some further normalization to the test output to sidestep this?

Esteban Küber (Nov 21 2019 at 15:28, on Zulip):

This makes tests that point into std fail in those targets

eddyb (Nov 21 2019 at 15:29, on Zulip):

if it's just paths being wrong that would make me very happy, I was under the assumption that this was some weird hashing bug (because we ignore existing sources at the same paths if the hashes changed)

Esteban Küber (Nov 21 2019 at 15:29, on Zulip):

Ah! If it is the hashing it would be trickier...

pnkfelix (Nov 21 2019 at 15:29, on Zulip):

I'm under impression that in some cases we just don't emit the span info at all? (Which may be an artifact of the path changing, for all i know?) Like this comment

Esteban Küber (Nov 21 2019 at 15:29, on Zulip):

An option is to explicitly ignore those platforms when testing

Esteban Küber (Nov 21 2019 at 15:30, on Zulip):

But I know petrochenkov was against that

eddyb (Nov 21 2019 at 15:30, on Zulip):

@pnkfelix the path saved in the rmeta (which may be affected by remapping? not sure how much) has to exist and the loaded contents have to match a precomputed hash, but if those conditions are met, we should always emit a snippet

pnkfelix (Nov 21 2019 at 15:30, on Zulip):

I think @Vadim Petrochenkov 's alternative suggestion there was to ignore a full target triple...

pnkfelix (Nov 21 2019 at 15:31, on Zulip):

which may or may not be what you are suggesting, @Esteban Küber ? I just know @Vadim Petrochenkov was against ignoring all of // x86

eddyb (Nov 21 2019 at 15:31, on Zulip):

there are diagnostics which check for cross-crate spans and avoid mentioning those files entirely (so they wouldn't show up in the error message at all, regardless of snippet), but the ones that don't should be consistent in their behavior

nikomatsakis (Nov 21 2019 at 15:31, on Zulip):

it seems like this problem is affecting a lot of people

pnkfelix (Nov 21 2019 at 15:32, on Zulip):

anyway I don't know if we really can resolve all these questions right now

nikomatsakis (Nov 21 2019 at 15:32, on Zulip):

I was going to say, "at the same time, I think we should split it out..."

centril (Nov 21 2019 at 15:32, on Zulip):

@eddyb seems like this subject might interact with https://github.com/rust-lang/rust/pull/66364 ?

pnkfelix (Nov 21 2019 at 15:32, on Zulip):

Can I get some a subset of the people here willing to spend some time on this??

eddyb (Nov 21 2019 at 15:32, on Zulip):

another thing I'm unclear on is whether // ignore-x86 affects x86_64 as well because that one seems dangerous to ignore

eddyb (Nov 21 2019 at 15:32, on Zulip):

(since I expect the vast majority of builds hosts to be x64)

pnkfelix (Nov 21 2019 at 15:33, on Zulip):

well here's the relevant dialogue

pnkfelix (Nov 21 2019 at 15:33, on Zulip):

I would hope // ignore-x86 does not affect x86_64, but I have not tested

eddyb (Nov 21 2019 at 15:33, on Zulip):

@centril good point, your PR would introduce situations in which the problem can occur (by not destroying the relevant information :P)

Esteban Küber (Nov 21 2019 at 15:34, on Zulip):

I can get the triple comment ignore thing going

pnkfelix (Nov 21 2019 at 15:34, on Zulip):

thank you @Esteban Küber

pnkfelix (Nov 21 2019 at 15:34, on Zulip):

next up, hopefully quicker

pnkfelix (Nov 21 2019 at 15:34, on Zulip):

“Remove pretty printing of specific nodes in AST” #66575

pnkfelix (Nov 21 2019 at 15:35, on Zulip):

this is a feature that people want to remove

pnkfelix (Nov 21 2019 at 15:35, on Zulip):

if you don't want it removed, speak up either here or on the PR

pnkfelix (Nov 21 2019 at 15:35, on Zulip):

but I don't think anyone here objects

pnkfelix (Nov 21 2019 at 15:35, on Zulip):

so lets move along

pnkfelix (Nov 21 2019 at 15:35, on Zulip):

I-nominated: “NLL Regressions in 1.40” #66517

pnkfelix (Nov 21 2019 at 15:36, on Zulip):

Here, there's work to be done

pnkfelix (Nov 21 2019 at 15:36, on Zulip):

depending on your point of view, at least

pnkfelix (Nov 21 2019 at 15:36, on Zulip):

There's one school of thought that says that we owe it to our users to try to put up PRs fixing these crates

pnkfelix (Nov 21 2019 at 15:37, on Zulip):

There's another school of thought that says that this is a largely volunteer driven project and we don't owe nobody nuthin'

centril (Nov 21 2019 at 15:37, on Zulip):

(and that these are abandonware ^^)

eddyb (Nov 21 2019 at 15:38, on Zulip):

There's another school of thought that says that this is a largely volunteer driven project and we don't owe nobody nuthin'

can't wait for Rust to become an industry standard while remaining volunteer-driven,

pnkfelix (Nov 21 2019 at 15:38, on Zulip):

what is the assertion that these are abandonware based on? lack of activity?

nikomatsakis (Nov 21 2019 at 15:38, on Zulip):

it seems like the abandonware aspect is relevant

nikomatsakis (Nov 21 2019 at 15:38, on Zulip):

presumably @lqd's comment

eddyb (Nov 21 2019 at 15:38, on Zulip):

@pnkfelix repos marked as archived, I think?

centril (Nov 21 2019 at 15:38, on Zulip):

@pnkfelix NLL regressions have had warnings that ignore --cap-lints for a long time

centril (Nov 21 2019 at 15:38, on Zulip):

if they haven't been fixed by now...

nikomatsakis (Nov 21 2019 at 15:39, on Zulip):

but I'm not clear on how many add'l projects there are beyond those

nikomatsakis (Nov 21 2019 at 15:39, on Zulip):

(I don't know that I agree that ignoring warnings indicates abandonware, but I do agree we've tried hard to let people know what's coming)

nikomatsakis (Nov 21 2019 at 15:39, on Zulip):

but archived projects seems pretty unambig :)

simulacrum (Nov 21 2019 at 15:39, on Zulip):

I feel like we should go to each crate/repository and put in the minimum effort

pnkfelix (Nov 21 2019 at 15:39, on Zulip):

Well, is capnp itself fixed?

lqd (Nov 21 2019 at 15:39, on Zulip):

yes

Pietro Albini (Nov 21 2019 at 15:39, on Zulip):

/me still thinks that a 1000 regressions crater run for a stable release is not acceptable, even when cargo update fixes most of them

lqd (Nov 21 2019 at 15:40, on Zulip):

this is a 3y old version

lqd (Nov 21 2019 at 15:40, on Zulip):

but it's still active

centril (Nov 21 2019 at 15:40, on Zulip):

(@nikomatsakis it's more like "no one has seen these warnings" rather than "actively ignoring" ^^)

eddyb (Nov 21 2019 at 15:40, on Zulip):

also this may be obvious but archived projects are frozen and I don't even know if you can open a PR on them

simulacrum (Nov 21 2019 at 15:40, on Zulip):

Archived projects I think we're fine to not try to fix

nikomatsakis (Nov 21 2019 at 15:40, on Zulip):

I feel a bit confused about the status

eddyb (Nov 21 2019 at 15:40, on Zulip):

/me should check with one of the eddyb-abandoned repos

simulacrum (Nov 21 2019 at 15:40, on Zulip):

but we should do our best to fix anything that's not archived

pnkfelix (Nov 21 2019 at 15:41, on Zulip):

actually, let me try to establish something: One issue here is what is our policy about what to do about the crates here

nikomatsakis (Nov 21 2019 at 15:41, on Zulip):

I guess the point is, there are older releases that we didn't see before

simulacrum (Nov 21 2019 at 15:41, on Zulip):

We have a list of ~80 repositories/crates that are root regressions and some of them are deprecated/archived

centril (Nov 21 2019 at 15:41, on Zulip):

But... I think that if you care deeply about filing PRs or issues then of course by all means do so...

eddyb (Nov 21 2019 at 15:41, on Zulip):

you can't even comment on issues on archived repos (as a repo owner)

simulacrum (Nov 21 2019 at 15:41, on Zulip):

My POV is similar (but less strict than @Pietro Albini's) - we should do our best to minimize stable regressions

pnkfelix (Nov 21 2019 at 15:41, on Zulip):

My other question, though, is: What remains to go through on this list? @lqd did some, but a bunch remain, right, that have not been establhsed as archived/abandonware by these metrics?

simulacrum (Nov 21 2019 at 15:41, on Zulip):

Yes

centril (Nov 21 2019 at 15:42, on Zulip):

(but as a matter of general policy I don't think filing PRs or issues should be a requirement)

simulacrum (Nov 21 2019 at 15:42, on Zulip):

(but as a matter of general policy I don't think filing PRs or issues should be a requirement)

I strongly disagree

Pietro Albini (Nov 21 2019 at 15:42, on Zulip):

how much do we care about our stability guarantee though?

simulacrum (Nov 21 2019 at 15:42, on Zulip):

If we're landing a breaking change the least that we should do is to fix things we know are broken

pnkfelix (Nov 21 2019 at 15:42, on Zulip):

(but as a matter of general policy I don't think filing PRs or issues should be a requirement)

I strongly disagree

I am not sure if the T-compiler triage meeting is the place to resolve this disagreement

Esteban Küber (Nov 21 2019 at 15:42, on Zulip):

What do we do about active crates with transitive deps on a now broken archived crate?

simulacrum (Nov 21 2019 at 15:43, on Zulip):

I understand that we won't fix everything, and I'm not trying to get that

pnkfelix (Nov 21 2019 at 15:43, on Zulip):

it seems like a more fundamental question of a core value that we need to establish consistently across the project, right?

lqd (Nov 21 2019 at 15:43, on Zulip):

@pnkfelix I'm looking at more rn (some of which are not abandoned)

simulacrum (Nov 21 2019 at 15:43, on Zulip):

So it is my understanding that our breakage policy is "don't do it" and if you have to, "go fix it"

nikomatsakis (Nov 21 2019 at 15:43, on Zulip):

I do feel like we should codify some rules here into an agreed upon procedure-- but I think it makes sense to do our best here

pnkfelix (Nov 21 2019 at 15:43, on Zulip):

What I would like to do, regarding #66517, is at least get a better grip on the to-do list

pnkfelix (Nov 21 2019 at 15:44, on Zulip):

in terms of bucketing the cases of abandonware vs non-abandonware

simulacrum (Nov 21 2019 at 15:44, on Zulip):

It sounds like @lqd is trying to put together a todo list?

pnkfelix (Nov 21 2019 at 15:44, on Zulip):

(however that criteria is determined)

simulacrum (Nov 21 2019 at 15:44, on Zulip):

I'm fine with tabling and coming back to it next week with a todo list

pnkfelix (Nov 21 2019 at 15:44, on Zulip):

@lqd do you think you are going to be able to do the whole set? I was sort of thinking/hoping we might rope other people in

simulacrum (Nov 21 2019 at 15:45, on Zulip):

to be clear it's basically clicking through and categorizing, I think it's also fine for the todo list to contain "triage this as abandonware"

pnkfelix (Nov 21 2019 at 15:45, on Zulip):

maybe lets spawn off a parallel zulip topic dedicated to the bucketing effort?

simulacrum (Nov 21 2019 at 15:45, on Zulip):

I can help out with some of them

nikomatsakis (Nov 21 2019 at 15:45, on Zulip):

should we coordinate in #t-compiler/wg-nll ?

lqd (Nov 21 2019 at 15:45, on Zulip):

unlikely that I could do it all by myself, but yeah I'm trying to thin it out

pnkfelix (Nov 21 2019 at 15:46, on Zulip):

okay @nikomatsakis made a topic here: https://rust-lang.zulipchat.com/#narrow/stream/122657-t-compiler.2Fwg-nll/topic/regressions.20in.20.2366517/near/181316916

pnkfelix (Nov 21 2019 at 15:46, on Zulip):

I-nominated: “Some features can no longer be controlled by conditional compilation” #65860

centril (Nov 21 2019 at 15:47, on Zulip):

I would prefer skipping that issue for now

centril (Nov 21 2019 at 15:47, on Zulip):

& it's not urgent

pnkfelix (Nov 21 2019 at 15:47, on Zulip):

welll, the thing I had hoped to figure out last week, was just: who and where we are going to resolve the questions here: Q1: Is it T-lang or T-compiler or both? Q2: Is asynchronous discussion sufficient (with or without an rfcbot poll), or do we need a synchronized meeting (maybe one of the Friday steering meetings)?

pnkfelix (Nov 21 2019 at 15:47, on Zulip):

that is, I don't want/need the full issue resolved. I just want to know if we're going to keep bringing it up at this meeting or not

pnkfelix (Nov 21 2019 at 15:48, on Zulip):

does anyone have any strong opinions on that matter?

centril (Nov 21 2019 at 15:48, on Zulip):

My suggestion:
- It's both teams
- I can re-nominate it at some point (to save time for now) but let's have some more async discussion

pnkfelix (Nov 21 2019 at 15:48, on Zulip):

if there are no strong opinions, I will probably let it sit as is for now

nikomatsakis (Nov 21 2019 at 15:49, on Zulip):

@centril's suggestion sounds reasonable. I myself would like to see a hackmd created that starts to summarize what's going on, which we can edit as we go to include details

centril (Nov 21 2019 at 15:49, on Zulip):

I can work on that probably, but that's going to take some time

pnkfelix (Nov 21 2019 at 15:49, on Zulip):

okay

pnkfelix (Nov 21 2019 at 15:49, on Zulip):

I'll do one more nominated issue before we move to WG-checkin

pnkfelix (Nov 21 2019 at 15:49, on Zulip):

“under latest MinGW, cannot link with C code using stdout” #47048

pnkfelix (Nov 21 2019 at 15:50, on Zulip):

(and yeah, if you're looking at the hackmd, I went out of order. sue me.)

pnkfelix (Nov 21 2019 at 15:50, on Zulip):

I just want to know what priority #47048 deserves

nikomatsakis (Nov 21 2019 at 15:50, on Zulip):

I can work on that probably, but that's going to take some time

it doesn't hve to be a work of art, bullet points and links to comments seems ok

pnkfelix (Nov 21 2019 at 15:51, on Zulip):

is this P-high? If so, is there anyone here willing to attack it?

pnkfelix (Nov 21 2019 at 15:51, on Zulip):

MinGW ain't exactly the most popular platform

pnkfelix (Nov 21 2019 at 15:52, on Zulip):

(to be honest I did not realize it was Tier1 until reading @Jonas Schievink 's comment ...)

mw (Nov 21 2019 at 15:52, on Zulip):

does the proposed target tier policy say anything about demoting a target?

nikomatsakis (Nov 21 2019 at 15:53, on Zulip):

I'm still trying to catch up

nikomatsakis (Nov 21 2019 at 15:53, on Zulip):

it seems the issue here is:

nikomatsakis (Nov 21 2019 at 15:53, on Zulip):
nikomatsakis (Nov 21 2019 at 15:53, on Zulip):
nikomatsakis (Nov 21 2019 at 15:54, on Zulip):
nikomatsakis (Nov 21 2019 at 15:54, on Zulip):
pnkfelix (Nov 21 2019 at 15:54, on Zulip):

is this in part an infrastructure issue then?

nikomatsakis (Nov 21 2019 at 15:54, on Zulip):

(For quite some time, e.g. I got most of that from @Vadim Petrochenkov's comment which dates from April)

pnkfelix (Nov 21 2019 at 15:55, on Zulip):

in terms of our own limitations about what MinGW versions we can/do test in CI ?

nikomatsakis (Nov 21 2019 at 15:55, on Zulip):

where they also mention:

It may make sense to bump the version used by Rust from 6.3.0 to something up-to-date though.

pnkfelix (Nov 21 2019 at 15:55, on Zulip):

Or, hold on, Before we talk more about this:

pnkfelix (Nov 21 2019 at 15:55, on Zulip):

@nikomatsakis did you want to say anything on behalf of WG-traits ?

centril (Nov 21 2019 at 15:55, on Zulip):

(Aside: I assigned https://github.com/rust-lang/rust/issues/66501 to @Nadrieril as well since they've been working on related things...)

nikomatsakis (Nov 21 2019 at 15:56, on Zulip):

nikomatsakis did you want to say anything on behalf of WG-traits ?

oh I didn't realize it was that time ;) I can give a few quick updates

pnkfelix (Nov 21 2019 at 15:56, on Zulip):

Lets do that then

nikomatsakis (Nov 21 2019 at 15:56, on Zulip):

that is, I missed your earlier ping :)

nikomatsakis (Nov 21 2019 at 15:56, on Zulip):

ok a few notes:

nikomatsakis (Nov 21 2019 at 15:56, on Zulip):
nikomatsakis (Nov 21 2019 at 15:57, on Zulip):
nikomatsakis (Nov 21 2019 at 15:58, on Zulip):
centril (Nov 21 2019 at 15:59, on Zulip):

is getting very, very close to landing support for trait object upcasts

@nikomatsakis speaking off... could we cherry-pick the drive-by cosmetics changes to a separate PR? feels hard to read the PR otherwise with those in

nikomatsakis (Nov 21 2019 at 15:59, on Zulip):
nikomatsakis (Nov 21 2019 at 15:59, on Zulip):

nikomatsakis speaking off... could we cherry-pick the drive-by cosmetics changes to a separate PR? feels hard to read the PR otherwise with those in

I agree they make it hard to read the PR. I've been debating what to suggest about it.

centril (Nov 21 2019 at 16:00, on Zulip):

maybe just cherry-pick yourself to a different PR or some such

nikomatsakis (Nov 21 2019 at 16:00, on Zulip):

Currently we are analzing the regressions from a crater run. The next step after that may be some sort of RFC. I'm not sure.

centril (Nov 21 2019 at 16:00, on Zulip):

(should be easy the way it is structured)

nikomatsakis (Nov 21 2019 at 16:00, on Zulip):
nikomatsakis (Nov 21 2019 at 16:01, on Zulip):

all that said, my main goal for the next few weeks I think is going to be trying to lay out the roadmap for the year ahead

centril (Nov 21 2019 at 16:01, on Zulip):

unblocking const generics in chalk?

nikomatsakis (Nov 21 2019 at 16:01, on Zulip):

in rustc, though we do have to extend chalk to support constants too

nikomatsakis (Nov 21 2019 at 16:01, on Zulip):

in some minimal way

nikomatsakis (Nov 21 2019 at 16:02, on Zulip):

ok, I gotta go

pnkfelix (Nov 21 2019 at 16:02, on Zulip):

okay well we've hit an hour (+2 minutes)

pnkfelix (Nov 21 2019 at 16:02, on Zulip):

thanks for the update on WG-traits @nikomatsakis !

pnkfelix (Nov 21 2019 at 16:02, on Zulip):

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

Last update: Dec 12 2019 at 01:30UTC