Stream: t-compiler

Topic: weekly meeting 2019-10-10 #54818


pnkfelix (Oct 10 2019 at 12:31, on Zulip):

Hi @T-compiler/meeting ; the triage meeting will be starting in 1 hours 29 minutes

pnkfelix (Oct 10 2019 at 12:32, on Zulip):

I will be doing pre-triage in a parallel topic

pnkfelix (Oct 10 2019 at 12:32, on Zulip):

This week's WG checkins are with Learning and LLVM

pnkfelix (Oct 10 2019 at 12:32, on Zulip):

I didn't even know we had a WG-learning...

pnkfelix (Oct 10 2019 at 12:33, on Zulip):

ah I see, its about providing resources for rustc developers

pnkfelix (Oct 10 2019 at 12:33, on Zulip):

@Santiago Pastorino do you think you could present an update on behalf of WG-learning?

pnkfelix (Oct 10 2019 at 12:33, on Zulip):

and @nagisa , do you think you could present an update on behalf of WG-LLVM ?

Santiago Pastorino (Oct 10 2019 at 12:34, on Zulip):

Santiago Pastorino do you think you could present an update on behalf of WG-learning?

yes

nagisa (Oct 10 2019 at 12:35, on Zulip):

I haven’t been following what was happening as part of that WG for a couple of months already, but I can try to draft a short summary, yes.

pnkfelix (Oct 10 2019 at 12:37, on Zulip):

@nagisa okay, thanks! Going forward, should we try to add an additional lead for WG-LLVM so that you don't have to worry about future checkins?

nagisa (Oct 10 2019 at 12:37, on Zulip):

I think @Nikita Popov serves that role already

nagisa (Oct 10 2019 at 12:38, on Zulip):

I’ll chat with them about adding them to the list of leads.

pnkfelix (Oct 10 2019 at 12:38, on Zulip):

great thanks!

pnkfelix (Oct 10 2019 at 14:00, on Zulip):

Hi @T-compiler/meeting

pnkfelix (Oct 10 2019 at 14:01, on Zulip):

We'll spend the first five minutes with the floor open to announcements

centril (Oct 10 2019 at 14:02, on Zulip):

I'm working on a PR to split libsyntax along data/logic lines and to make librustc not depend on some large parts (like the parser and macro stuff)

nagisa (Oct 10 2019 at 14:03, on Zulip):

So on my side, a technicality, I finally released the cross-platform version of stacker, enabling us to use it in the compiler. A couple of years more and stack overflows in the compiler will no longer be a concern.

pnkfelix (Oct 10 2019 at 14:04, on Zulip):

Oh, here's another annoucement: We have a design meeting scheduled tomorrow to discuss @Zoxc 's series of PR's for migrating rustc_interface queries.

nikomatsakis (Oct 10 2019 at 14:05, on Zulip):

I'm working on a PR to split libsyntax along data/logic lines and to make librustc not depend on some large parts (like the parser and macro stuff)

hey @centril -- very timely -- @mw and I were just chatting about trying to make more of a documented procedure around big refactorings. Part of it would be declaring an "intent to implement" and part of it would be around trying to document the design you wound up with (perhaps after doing some exploration). So thanks for announcing. :heart:

pnkfelix (Oct 10 2019 at 14:05, on Zulip):

Oh, here's another annoucement: We have a design meeting scheduled tomorrow to discuss Zoxc 's series of PR's for migrating rustc_interface queries.

@Zoxc I think you did confirm that you think you'll be awake at the time of the meeting?

nikomatsakis (Oct 10 2019 at 14:05, on Zulip):

Another thing we could talk about would be specifically the persistent dep-graph loading PR

nikomatsakis (Oct 10 2019 at 14:06, on Zulip):

i.e., if we want to narrow the topic

pnkfelix (Oct 10 2019 at 14:06, on Zulip):

which would also be in @Zoxc 's wheelhouse, but would be easier to prepare for, yes?

mw (Oct 10 2019 at 14:06, on Zulip):

Another thing we could talk about would be specifically the lazy dep-graph loading PR

I think that's orthogonal to the currently proposed topic

nikomatsakis (Oct 10 2019 at 14:06, on Zulip):

My point is mostly that Zoxc recently posted a short writeup to the PR

nikomatsakis (Oct 10 2019 at 14:06, on Zulip):

so some of the prep is done

centril (Oct 10 2019 at 14:06, on Zulip):

Yeah intent to implement sounds like a cool lightweight process

pnkfelix (Oct 10 2019 at 14:07, on Zulip):

on the subject of intent-to-implement

pnkfelix (Oct 10 2019 at 14:08, on Zulip):

I am/was planning on revising the code that handles the structural-match checking (that's the const_to_pat fun I alluded to earlier); this is related to #61188 and other bugs.

nikomatsakis (Oct 10 2019 at 14:09, on Zulip):

Probably worth pointing out my new "universes" PR, also: #65232

This is an important step towards lazy normalization, as it enables us to generate "obligations" from inside unification that don't have to be completed within the same stack frame.

pnkfelix (Oct 10 2019 at 14:10, on Zulip):

Lets call that the end of "five minutes" for announcements. (privmsg me if you want time at end for extra announcements)

nikomatsakis (Oct 10 2019 at 14:10, on Zulip):

Oh, one other thing:

In #wg-traits, we're planning on creating a weekly design meeting. Precise time not yet determined, prob Mon/Wed at 2pm. It'll be recorded. The idea is to have a space to talk over things in detail and also to help people follow along and learn more about the overall traits design.

pnkfelix (Oct 10 2019 at 14:11, on Zulip):

so with announcements "over", lets do beta-nominations

pnkfelix (Oct 10 2019 at 14:11, on Zulip):

there are four unaccepted nominations

centril (Oct 10 2019 at 14:12, on Zulip):

@nikomatsakis nice work on that lattice drawing btw, very readable / intuitive stuff

pnkfelix (Oct 10 2019 at 14:12, on Zulip):

beta-nom: "Ensure that associated async fns have unique fresh param names #65142

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

(I take it this gets the usual "its async-await so we'll prioritize backport", right?

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

looks like it

nikomatsakis (Oct 10 2019 at 14:13, on Zulip):

it's also like super trivial

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

okay beta accepted then

nagisa (Oct 10 2019 at 14:13, on Zulip):

This PR passes all of my backportability checks too

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

beta-nom: "Account for macro invocation in let mut $pat diagnostic. #65123

pnkfelix (Oct 10 2019 at 14:14, on Zulip):

/me needs to learn to script the Zulip UI

nagisa (Oct 10 2019 at 14:14, on Zulip):

Originally an ICE, small fix, tested.

pnkfelix (Oct 10 2019 at 14:15, on Zulip):

okay beta-accept I think

pnkfelix (Oct 10 2019 at 14:15, on Zulip):

beta-nom: " Fix the start/end byte positions in the compiler JSON output" #65074

pnkfelix (Oct 10 2019 at 14:16, on Zulip):

hmm

nikomatsakis (Oct 10 2019 at 14:17, on Zulip):

this is fairly involved

nikomatsakis (Oct 10 2019 at 14:17, on Zulip):

I'm not quite seeing the motivation for backport

pnkfelix (Oct 10 2019 at 14:17, on Zulip):

so is the reason to back port this ... so that rustfix acts better ?

mw (Oct 10 2019 at 14:17, on Zulip):

It also hasn't landed yet

nikomatsakis (Oct 10 2019 at 14:17, on Zulip):

I lean towards "no" myself

centril (Oct 10 2019 at 14:18, on Zulip):

given that this had to be r-ed once and it is fairly large it feels rather risky and for little gain

nikomatsakis (Oct 10 2019 at 14:18, on Zulip):

If I'm not mistaken, this is a long-standing problem, seems ok for the fix to ride the trains

pnkfelix (Oct 10 2019 at 14:18, on Zulip):

okay, lets decline to backport

pnkfelix (Oct 10 2019 at 14:18, on Zulip):

beta-nom: " extract expected return type for async fn generators" #64999

pnkfelix (Oct 10 2019 at 14:19, on Zulip):

okay so this PR injected an ICE (good thing we waited a week)

nikomatsakis (Oct 10 2019 at 14:19, on Zulip):

yep. fairly trivial one, but still.

pnkfelix (Oct 10 2019 at 14:19, on Zulip):

the fix for the ICE is in PR #65235

centril (Oct 10 2019 at 14:20, on Zulip):

imo even tho this is async-await it is sufficiently large that I would like this to ride the trains

nikomatsakis (Oct 10 2019 at 14:20, on Zulip):

I think it's fine to ride the trains

mw (Oct 10 2019 at 14:20, on Zulip):

it's a really big PR. is it that urgent?

nagisa (Oct 10 2019 at 14:20, on Zulip):

What is the impact of this PR?

nagisa (Oct 10 2019 at 14:20, on Zulip):

Looks like improvement for diagnostics

nikomatsakis (Oct 10 2019 at 14:20, on Zulip):

The impact is that we successfully compile things

centril (Oct 10 2019 at 14:20, on Zulip):

@nagisa accepting more code due to better type inference

nikomatsakis (Oct 10 2019 at 14:20, on Zulip):

It seems like "ride the trains" is the consensus

pnkfelix (Oct 10 2019 at 14:20, on Zulip):

what is workaround, if any, without it landed?

nikomatsakis (Oct 10 2019 at 14:20, on Zulip):

I'm good w/ that

nikomatsakis (Oct 10 2019 at 14:20, on Zulip):

what is workaround, if any, without it landed?

explicit coercion, or -- once it lands -- use nightly :)

centril (Oct 10 2019 at 14:21, on Zulip):
pub async fn test3() -> Box<dyn Debug> {
    let tmp : Box<dyn Debug> = Box::new("asdf");
    tmp
}
pnkfelix (Oct 10 2019 at 14:21, on Zulip):

is there a good way to document that on whatever issue is likely to come up in response to a search of our github repo?

pnkfelix (Oct 10 2019 at 14:21, on Zulip):

oh its already there

pnkfelix (Oct 10 2019 at 14:21, on Zulip):

on #60424

pnkfelix (Oct 10 2019 at 14:21, on Zulip):

okay I'm okay with that then

pnkfelix (Oct 10 2019 at 14:22, on Zulip):

thanks @centril

pnkfelix (Oct 10 2019 at 14:22, on Zulip):

okay then lets decline to backport

pnkfelix (Oct 10 2019 at 14:22, on Zulip):

that's all the beta-nominations then

nikomatsakis (Oct 10 2019 at 14:22, on Zulip):

imo even tho this is async-await it is sufficiently large that I would like this to ride the trains

(Just for the record, I'm softening on my "backport the async-await things" stance as we get further into the cycle. I suspect most people will still use nightly for their "day to day" coding experience for some time yet, to get e.g. best error messages and so forth, even if stable suffices for the CI etc.)

centril (Oct 10 2019 at 14:23, on Zulip):

seems sensible :slight_smile:

pnkfelix (Oct 10 2019 at 14:23, on Zulip):

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

nikomatsakis (Oct 10 2019 at 14:23, on Zulip):

side note, @centril or others, if you know how to extract what the error is from this azure logs I'd be appreciative :) I can't tell

pnkfelix (Oct 10 2019 at 14:23, on Zulip):

we'll skip #59064 (its an old @Zoxc one, maybe we'll talkk about it tomorrow, or maybe not)

pnkfelix (Oct 10 2019 at 14:24, on Zulip):

but PR "minimize the rust-std component" #64823 is also S-waiting-on-team

pnkfelix (Oct 10 2019 at 14:24, on Zulip):

somewhat large comment thread on the PR itself

pnkfelix (Oct 10 2019 at 14:25, on Zulip):

silly question: Is T-compiler actually the right team to decide this?

centril (Oct 10 2019 at 14:25, on Zulip):

@pnkfelix lol :D t-core seems to think so? I was thinking it should be t-release

pnkfelix (Oct 10 2019 at 14:25, on Zulip):

in any case I don't think we can really dive into this sensibly in this meeting

pnkfelix (Oct 10 2019 at 14:26, on Zulip):

but yeah, lets at least resolve that question

centril (Oct 10 2019 at 14:26, on Zulip):

( The fact that I am on t-release is a coincidence! :D )

mw (Oct 10 2019 at 14:26, on Zulip):

T-compiler does not seem the right team for this

pnkfelix (Oct 10 2019 at 14:26, on Zulip):

thumbs up if you think that this should be moved to T-release

pnkfelix (Oct 10 2019 at 14:26, on Zulip):

(that's just me adding the template as usual, I didn't vote against it)

centril (Oct 10 2019 at 14:27, on Zulip):

@nikomatsakis here you go https://dev.azure.com/rust-lang/e71b0ddf-dd27-435a-873c-e30f86eea377/_apis/build/builds/10326/logs/799

nikomatsakis (Oct 10 2019 at 14:27, on Zulip):

I think that there isn't a perfect team for this; it's sort of out of the t-compiler wheelhouse, for sure. I sometimes think we need a "product team" that encompasses cargo, rustc, dev-tools, libs, and lang, and it might be the right spot for this. But I am ok with T-release for now.

pnkfelix (Oct 10 2019 at 14:27, on Zulip):

wasnt there a tools team at some point?

nikomatsakis (Oct 10 2019 at 14:27, on Zulip):

There is a tools team, it's called dev-tools

pnkfelix (Oct 10 2019 at 14:27, on Zulip):

or is that narrower than what you descrbe? I geuss so

pnkfelix (Oct 10 2019 at 14:28, on Zulip):

okay

nikomatsakis (Oct 10 2019 at 14:28, on Zulip):

( And yes that is narrower, as it is focused on individual tools like clippy )

nikomatsakis (Oct 10 2019 at 14:28, on Zulip):

anyway I didn't mean to introduce a digression :)

pnkfelix (Oct 10 2019 at 14:28, on Zulip):

lets move it to T-release. And hey, if any of you on @T-compiler want to chime in or object, I figure you're welcome to write a comment there.

centril (Oct 10 2019 at 14:28, on Zulip):

My pitch for T-release: we make the actual releases and contain the people who know how to fix the issues here ^^

centril (Oct 10 2019 at 14:29, on Zulip):

(not myself, rather pietro, simulacrum, Josh stone, ...)

pnkfelix (Oct 10 2019 at 14:29, on Zulip):

okay. last thing before WG checkins:nominated issues

pnkfelix (Oct 10 2019 at 14:29, on Zulip):

nom: "Star Import Precedence Regression in Beta" #65090

pnkfelix (Oct 10 2019 at 14:29, on Zulip):

I nominated this because I triaged it as P-medium

pnkfelix (Oct 10 2019 at 14:29, on Zulip):

and I wanted to make sure everyone was okay with that

pnkfelix (Oct 10 2019 at 14:30, on Zulip):

#65090 is categorized as a "regression" but my understanding is that it is an old bug that is exposed due to a change in macro expansion order

pnkfelix (Oct 10 2019 at 14:31, on Zulip):

so, if you object and think this should get P-high treatment like "other regressions", feel free to comment here, or on the issue, as you like.

pnkfelix (Oct 10 2019 at 14:31, on Zulip):

last nom: "Rustdoc support API in rustc" #59790

pnkfelix (Oct 10 2019 at 14:32, on Zulip):

I nominated this last week because I wanted to understand if there was actually any actionable items here or not. cc @oli cc @eddyb ^

eddyb (Oct 10 2019 at 14:32, on Zulip):

I'm not up to date on this at all

pnkfelix (Oct 10 2019 at 14:32, on Zulip):

i'm not sure if there's anything to be up-to-date on

eddyb (Oct 10 2019 at 14:32, on Zulip):

oh it's that thing

nikomatsakis (Oct 10 2019 at 14:32, on Zulip):

Side note: I wonder if we can document the "major queries" and their role

nikomatsakis (Oct 10 2019 at 14:32, on Zulip):

in some nice way...

nikomatsakis (Oct 10 2019 at 14:33, on Zulip):

not entirely side because this seems to be a question of "which queries should rustdoc be using"

eddyb (Oct 10 2019 at 14:33, on Zulip):

idk why oli opened this issue at all. those PRs of mine have been blocked for ages on #59953 which is only now close to review/merge

mw (Oct 10 2019 at 14:33, on Zulip):

it sounds like this is actually not a problem?

pnkfelix (Oct 10 2019 at 14:33, on Zulip):

is this worth devoting a design meeting to? Seems like an overall process question to be resolved, maybe.

mw (Oct 10 2019 at 14:34, on Zulip):

is there really demand for rustdoc specific features?

nikomatsakis (Oct 10 2019 at 14:34, on Zulip):

we have at least a few

eddyb (Oct 10 2019 at 14:34, on Zulip):

(mostly because I expected #59953 to be a slam-dunk)

pnkfelix (Oct 10 2019 at 14:34, on Zulip):

idk why oli opened this issue at all. those PRs of mine have been blocked for ages on #59953 which is only now close to review/merge

My inference is that it was because @oli wants rustdoc's needs to go through an explict interface that we'd identify as for rustdoc ?

pnkfelix (Oct 10 2019 at 14:34, on Zulip):

(and I don't see how that relates to #59953 ...?)

centril (Oct 10 2019 at 14:34, on Zulip):

rustdoc_foobar seems sensible to me if it means not mkaing problems in rustc code

mw (Oct 10 2019 at 14:35, on Zulip):

I'd be fine with having queries that only rustdoc needs as long as they are marked as such and are not "hacks"

eddyb (Oct 10 2019 at 14:35, on Zulip):

@pnkfelix I think there were performance issues that #59953 would fix

pnkfelix (Oct 10 2019 at 14:35, on Zulip):

ah, @eddyb okay now I see. But I think @oli 's concern wasn't about that one speciifc bug

pnkfelix (Oct 10 2019 at 14:36, on Zulip):

it was about "how do we provide one-off hacks to rustdoc in a controlled fashion", i think?

centril (Oct 10 2019 at 14:36, on Zulip):

that was also my inference

eddyb (Oct 10 2019 at 14:36, on Zulip):

still, I disagree with @oli as I wrote in my comments on the issue

nikomatsakis (Oct 10 2019 at 14:36, on Zulip):

I do think that longer term it'd be nice if rustdoc had a more "stable" interface through a set of rustdoc_ queries -- or, ideally, some stable-ish "external" queries that are also used by other tools like clippy/IDEs -- but I guess for now it's "ok" the way it is.

pnkfelix (Oct 10 2019 at 14:36, on Zulip):

maybe we do need a design meeting here. :)

nikomatsakis (Oct 10 2019 at 14:36, on Zulip):

if we were going to do a design meeting though I think we'd be talking about that?

eddyb (Oct 10 2019 at 14:36, on Zulip):

there's no need for one-off hacks most of the time, just a tiny bit of design work

eddyb (Oct 10 2019 at 14:36, on Zulip):

rustdoc shouldn't be literally a pile of hacks, that's not a good direction to head in IMO

nikomatsakis (Oct 10 2019 at 14:37, on Zulip):

rustdoc shouldn't be literally a pile of hacks, that's not a good direction to head in IMO

I disagree :P

centril (Oct 10 2019 at 14:37, on Zulip):

@nikomatsakis it should be a pile of hacks? :D

nikomatsakis (Oct 10 2019 at 14:37, on Zulip):

Pile of hacks! Pile of hacks!

nikomatsakis (Oct 10 2019 at 14:37, on Zulip):

ok, sorry, just kidding :)

mw (Oct 10 2019 at 14:37, on Zulip):

is rustdoc rustc's bad bank?

eddyb (Oct 10 2019 at 14:37, on Zulip):

@pnkfelix anyway I don't think there's anything productive in that issue, other than we should coordinate more on rustdoc and maybe take care of it more

pnkfelix (Oct 10 2019 at 14:37, on Zulip):

well, we shouldn't sacrfice rustdoc's functionality in the name of API purity, maybe?

pnkfelix (Oct 10 2019 at 14:38, on Zulip):

(that's my way of saying "Pile of hacks!")

eddyb (Oct 10 2019 at 14:38, on Zulip):

@pnkfelix there are principled ways to do just about anything

nikomatsakis (Oct 10 2019 at 14:38, on Zulip):

I don't think "pile of hacks" is really a fair characterization :)

pnkfelix (Oct 10 2019 at 14:38, on Zulip):

classic "best enemy of good" or something

nikomatsakis (Oct 10 2019 at 14:38, on Zulip):

but regardless we should try to keep things nice

nikomatsakis (Oct 10 2019 at 14:38, on Zulip):

anyway, I'm not sure what we're talking about here :)

centril (Oct 10 2019 at 14:39, on Zulip):

Title suggestion: "How to make ad-hoc hacks less ad-hoc"

pnkfelix (Oct 10 2019 at 14:39, on Zulip):

I think its sounds like people are generally opposed to adding the convention @oli proposed

pnkfelix (Oct 10 2019 at 14:39, on Zulip):

because that would likely just encourage hacks?

pnkfelix (Oct 10 2019 at 14:39, on Zulip):

in the name of "well its labelled so its safe"

eddyb (Oct 10 2019 at 14:39, on Zulip):

we do have a couple things which are mostly (or only?) for rustdoc, and they have no prefix I don't think

pnkfelix (Oct 10 2019 at 14:39, on Zulip):

see also: how rustdoc managed to add a dependence on everybody-loops

centril (Oct 10 2019 at 14:39, on Zulip):

I don't think something that only rustdoc needs is necessarily a hack

centril (Oct 10 2019 at 14:40, on Zulip):

but it does help if it is marked clearly

centril (Oct 10 2019 at 14:40, on Zulip):

in terms of seeing the dependencies clearer

centril (Oct 10 2019 at 14:40, on Zulip):

aside: is there a stan4j rust equivalent?

eddyb (Oct 10 2019 at 14:40, on Zulip):

e.g. function argument names and pretty-printed const bodies, cross-crate

nikomatsakis (Oct 10 2019 at 14:40, on Zulip):

I think its sounds like people are generally opposed to adding the convention oli proposed

I am not :) I think the convention is fine. It's just not an excuse to land any old thing, obviously.

pnkfelix (Oct 10 2019 at 14:40, on Zulip):

okay. Lets leave #59790 open then

pnkfelix (Oct 10 2019 at 14:41, on Zulip):

it sounds like there may be action to take here

pnkfelix (Oct 10 2019 at 14:41, on Zulip):

the specific question of which action is maybe up in the air

pnkfelix (Oct 10 2019 at 14:41, on Zulip):

and I'll unnominate it, because I'm satisfied now.

pnkfelix (Oct 10 2019 at 14:42, on Zulip):

(and if anyone wants to archive their thoughts on the matter, feel free to leave a comment on that issue)

pnkfelix (Oct 10 2019 at 14:42, on Zulip):

So that's everything on the agenda I think

pnkfelix (Oct 10 2019 at 14:42, on Zulip):

well, apart from WG-checkins!

centril (Oct 10 2019 at 14:42, on Zulip):

sounds like @oli and @eddyb should battle it out? ;)

pnkfelix (Oct 10 2019 at 14:42, on Zulip):

@Santiago Pastorino you want to start with a checkin from WG-learning ?

Santiago Pastorino (Oct 10 2019 at 14:42, on Zulip):

WG-Learning update

Past

We are trying to write rustc-guide chapters from the different compiler lectures series.

We have a lot of people engaged in the group but most (maybe all?) of the team members are new to the compiler.

Some weeks ago we figured that assigning each individual volunteer to a different lecture as you can see in the project, ended in a too complex task for each one.

Present

So we decided to work collectively and we are going after the ty lecture since then. We are aiming to collectively produce a chapter out of that lecture and a lot of progress have been made and @WG-learning members seems to be happy :). This is what we have for now but the most important achievement, IMHO, is that we seemed to have figured out a way to work towards getting more content to the rustc-guide. So the future seems to be bright in this regard :).

nikomatsakis (Oct 10 2019 at 14:43, on Zulip):

oh nice! I hadn't had time to see the "WIP"

Santiago Pastorino (Oct 10 2019 at 14:44, on Zulip):

yeah, no worries when is finished or there are existential doubts we will try to reach you :)

nikomatsakis (Oct 10 2019 at 14:44, on Zulip):

how "far" are you through the lecture?

nikomatsakis (Oct 10 2019 at 14:44, on Zulip):

do you feel ready for part 2? :P

Santiago Pastorino (Oct 10 2019 at 14:45, on Zulip):

so ... I need to invest some time to finish some parts and leave more stuff to be filled by the rest of the members

Santiago Pastorino (Oct 10 2019 at 14:45, on Zulip):

we probably need to organize a bit on that regard

Santiago Pastorino (Oct 10 2019 at 14:45, on Zulip):

do you feel ready for part 2? :P

definitely, I guess you meant to dive more into Ty things?

nikomatsakis (Oct 10 2019 at 14:46, on Zulip):

yeah, I dont' remember how far that lecture got, but I'm confident there are areas we could dive into more deeply

nikomatsakis (Oct 10 2019 at 14:46, on Zulip):

anyway i'll skim chapter :)

pnkfelix (Oct 10 2019 at 14:46, on Zulip):

I should skim it too

Santiago Pastorino (Oct 10 2019 at 14:46, on Zulip):

there's more stuff to cover yeah, like TypeFoldable to mention something

nikomatsakis (Oct 10 2019 at 14:46, on Zulip):

from my quick skim over though it seems pretty great

pnkfelix (Oct 10 2019 at 14:47, on Zulip):

/me writes yet another post-it note to-do

Santiago Pastorino (Oct 10 2019 at 14:47, on Zulip):

it's not finished, but if you want to go ahead now do so, otherwise I can ping you both when I feel it's closer

pnkfelix (Oct 10 2019 at 14:47, on Zulip):

okay thanks @Santiago Pastorino !

pnkfelix (Oct 10 2019 at 14:48, on Zulip):

@nagisa do you have anything to report from WG-LLVM ?

nagisa (Oct 10 2019 at 14:48, on Zulip):

WG-LLVM

nikomatsakis (Oct 10 2019 at 14:49, on Zulip):

Whatever happened with the "insert noop to suppress optimization" PR?

nagisa (Oct 10 2019 at 14:49, on Zulip):

I haven’t been following happenings in the WG too closely lately, so I might have missed a bunch. These stood out as the most important ones.

pnkfelix (Oct 10 2019 at 14:49, on Zulip):

Whatever happened with the "insert noop to suppress optimization" PR?

is that the same as llvm.sideeffect ?

mw (Oct 10 2019 at 14:49, on Zulip):

Whatever happened with the "insert noop" PR?

I'm still waiting for that to land (behind a flag) for performance testing

nagisa (Oct 10 2019 at 14:49, on Zulip):

@nikomatsakis for the loop {} optimisation? It is still open I believe.

nikomatsakis (Oct 10 2019 at 14:49, on Zulip):

is that the same as llvm.sideeffect ?

that's the one I meant, yeah

nikomatsakis (Oct 10 2019 at 14:50, on Zulip):

Ok. Is it blocked on anything in particular?

nikomatsakis (Oct 10 2019 at 14:50, on Zulip):

I'll just go check, nm

pnkfelix (Oct 10 2019 at 14:50, on Zulip):

PR #59546: "Add llvm.sideeffect to potential infinite loops and recursions"

nagisa (Oct 10 2019 at 14:50, on Zulip):

It appears that it is not blocked at all, I should review it again.

nikomatsakis (Oct 10 2019 at 14:50, on Zulip):

Yeah, seems like it's good to go

nikomatsakis (Oct 10 2019 at 14:51, on Zulip):

Cool

mw (Oct 10 2019 at 14:51, on Zulip):

sounds like awesome things are happening in WG-LLVM, @nagisa !

matklad (Oct 10 2019 at 14:51, on Zulip):

Hey, just saw comment on https://github.com/rust-lang/rust/pull/65074#issuecomment-540621524

centril (Oct 10 2019 at 14:51, on Zulip):

Would be good to set a deadline for making the -Z flag the default

matklad (Oct 10 2019 at 14:51, on Zulip):

General consensus was that this is a long-standing problem

I am not sure this is true

pnkfelix (Oct 10 2019 at 14:51, on Zulip):

the detail about sideeffect on entry vs call is interesting

matklad (Oct 10 2019 at 14:52, on Zulip):

In particular, regression for \r\n is pretty fresh

pnkfelix (Oct 10 2019 at 14:52, on Zulip):

(this is regarding declined beta nom of PR #65074 )

pnkfelix (Oct 10 2019 at 14:52, on Zulip):

but

pnkfelix (Oct 10 2019 at 14:52, on Zulip):

before we discuss that

pnkfelix (Oct 10 2019 at 14:52, on Zulip):

@nagisa is there more from WG-LLVM ?

pnkfelix (Oct 10 2019 at 14:53, on Zulip):

no don't think so

pnkfelix (Oct 10 2019 at 14:53, on Zulip):

okay so

centril (Oct 10 2019 at 14:53, on Zulip):

Like... 6 months tops for the sideeffect thing and then I think the soundness hole needs to be closed

pnkfelix (Oct 10 2019 at 14:53, on Zulip):

regarding #65074

pnkfelix (Oct 10 2019 at 14:53, on Zulip):

@matklad what are expected benefits of backporting? Why does this need to be rushed?

matklad (Oct 10 2019 at 14:54, on Zulip):

Basically, rustfix is broken on windows at the moment

pnkfelix (Oct 10 2019 at 14:54, on Zulip):

is the main problem with use of rustfix on windows?

nikomatsakis (Oct 10 2019 at 14:54, on Zulip):

In particular, regression for \r\n is pretty fresh

Hmm.

nikomatsakis (Oct 10 2019 at 14:55, on Zulip):

Seems bad.

nikomatsakis (Oct 10 2019 at 14:55, on Zulip):

Is there a more targeted fix we could backport?

mw (Oct 10 2019 at 14:56, on Zulip):

we should at least let it be merged first and see if it works in the wild, right?

pnkfelix (Oct 10 2019 at 14:56, on Zulip):

well we've accepted some others things prior to merge.

pnkfelix (Oct 10 2019 at 14:56, on Zulip):

but usually they were trivial PRs

pnkfelix (Oct 10 2019 at 14:56, on Zulip):

(or rush situations, i.e. pending beta release)

mw (Oct 10 2019 at 14:56, on Zulip):

but there's no downside to revisiting it next week, is there?

matklad (Oct 10 2019 at 14:56, on Zulip):

We can try reveting https://github.com/rust-lang/rust/pull/62948

pnkfelix (Oct 10 2019 at 14:57, on Zulip):

We can try reveting https://github.com/rust-lang/rust/pull/62948

I was tempted to suggest this, so I'm glad you brought it up

matklad (Oct 10 2019 at 14:57, on Zulip):

we should at least let it be merged first and see if it works in the wild, right?

Right, sorry for applying the label for not yet merged commit :)

pnkfelix (Oct 10 2019 at 14:57, on Zulip):

has #62948 made it to stable?

nikomatsakis (Oct 10 2019 at 14:57, on Zulip):

We can try reveting https://github.com/rust-lang/rust/pull/62948

I'd probably prefer that (for beta)

nikomatsakis (Oct 10 2019 at 14:58, on Zulip):

(but I was assuming it hadn't reached stable)

pnkfelix (Oct 10 2019 at 14:58, on Zulip):

landed August 18th ...

nikomatsakis (Oct 10 2019 at 14:58, on Zulip):

(we need to make a tool for this)

matklad (Oct 10 2019 at 14:58, on Zulip):

I assumed it didn't, but I had to double-check

nikomatsakis (Oct 10 2019 at 14:59, on Zulip):

woah could we have a label automaticaly get added

centril (Oct 10 2019 at 14:59, on Zulip):

seems reasonable to revert on beta

nikomatsakis (Oct 10 2019 at 14:59, on Zulip):

"on-stable", "on-beta" etc...

centril (Oct 10 2019 at 14:59, on Zulip):

@nikomatsakis by what bot?

centril (Oct 10 2019 at 14:59, on Zulip):

please not homu

centril (Oct 10 2019 at 14:59, on Zulip):

homu is a pile of hacks

nikomatsakis (Oct 10 2019 at 14:59, on Zulip):

ok ok nm digression

pnkfelix (Oct 10 2019 at 14:59, on Zulip):

yep lets have rustdoc do it

nagisa (Oct 10 2019 at 14:59, on Zulip):

The same one which would nag about T-flags missing on nominated issues, probably highfive?

pnkfelix (Oct 10 2019 at 15:00, on Zulip):

anyway I too would prefer reverting #62948 on beta, if it hasn't hit stable yet

pnkfelix (Oct 10 2019 at 15:00, on Zulip):

@matklad can you look into this?

matklad (Oct 10 2019 at 15:00, on Zulip):

Yeah, I'll do!

pnkfelix (Oct 10 2019 at 15:00, on Zulip):

maybe we can just vote now

pnkfelix (Oct 10 2019 at 15:00, on Zulip):

revert PR #62948 on beta if it hasn't reached stable

pnkfelix (Oct 10 2019 at 15:01, on Zulip):

(talk about hacks, I'm having us vote before there's even a PR ...)

nikomatsakis (Oct 10 2019 at 15:01, on Zulip):

lol

pnkfelix (Oct 10 2019 at 15:01, on Zulip):

((but I assume this PR would be beta only anyway...))

centril (Oct 10 2019 at 15:01, on Zulip):

Sept. 26 Announcing Rust 1.38.0

matklad (Oct 10 2019 at 15:02, on Zulip):

Checked, it's in beta and not in stable

pnkfelix (Oct 10 2019 at 15:02, on Zulip):

okay great

centril (Oct 10 2019 at 15:02, on Zulip):

Yep that jives with the blog post date

pnkfelix (Oct 10 2019 at 15:02, on Zulip):

I'll leave a follow up note then

pnkfelix (Oct 10 2019 at 15:04, on Zulip):

okay hey that was awesome i think we actually got through everything

pnkfelix (Oct 10 2019 at 15:04, on Zulip):

thanks everybody that is how a meeting is supposed to work!

pnkfelix (Oct 10 2019 at 15:04, on Zulip):

(that is, thanks to everyone in @T-compiler/meeting !)

nikomatsakis (Oct 10 2019 at 20:20, on Zulip):

ps @Wesley Wiser -- not sure if you planned to make minutes? I was thinking of trying to write them up in blog post form and post to Inside Rust

Wesley Wiser (Oct 10 2019 at 20:27, on Zulip):

Yeah, I was going to do that this evening.

Wesley Wiser (Oct 10 2019 at 20:27, on Zulip):

Do you want them in the usual format or something different?

Wesley Wiser (Oct 10 2019 at 20:32, on Zulip):

(@nikomatsakis )

nikomatsakis (Oct 11 2019 at 13:09, on Zulip):

@Wesley Wiser actually minutes in the usual format are probably good -- I was thinking it'd be cool to have a weekly blog post from compile team with highlights from the triage meeting and design meeting (if any)

nikomatsakis (Oct 11 2019 at 13:10, on Zulip):

I made a point in this meeting of using :point_up: to designate things I thought might be good to highlight in that post this week

nikomatsakis (Oct 11 2019 at 13:10, on Zulip):

do you think you'd want to write such a blog post?

nikomatsakis (Oct 11 2019 at 13:10, on Zulip):

I don't imagine it has to be very long, more "highlights"

Wesley Wiser (Oct 11 2019 at 13:11, on Zulip):

Yeah, I'd definitely be interested!

Wesley Wiser (Oct 11 2019 at 13:11, on Zulip):

There's notes I posted last night here https://github.com/rust-lang/compiler-team/pull/197

nikomatsakis (Oct 11 2019 at 13:13, on Zulip):

merged :)

nikomatsakis (Oct 11 2019 at 13:13, on Zulip):

those are pretty close to the blog post I have in mind, indeed

Wesley Wiser (Oct 11 2019 at 13:24, on Zulip):

Ok cool! I'll go back and make sure all the :point_up: comments were included and then open a PR on blog.rlo with a slightly more cleaned up version.

Last update: Nov 20 2019 at 01:10UTC