Stream: t-compiler/meetings

Topic: [weekly meeting] 2020-06-18 #54818


Santiago Pastorino (Jun 17 2020 at 20:06, on Zulip):

Hi @T-compiler/meeting; the triage meeting will be starting in ~ 18 hours

Santiago Pastorino (Jun 17 2020 at 20:06, on Zulip):

The @WG-prioritization have done pre-triage in #t-compiler/wg-prioritization

Santiago Pastorino (Jun 17 2020 at 20:07, on Zulip):

@WG-prioritization have prepared the meeting agenda

Santiago Pastorino (Jun 17 2020 at 20:07, on Zulip):

We will have checkins from @WG-prioritization and @WG-rfc-2229

Santiago Pastorino (Jun 17 2020 at 20:08, on Zulip):

I've already filled the checkin of @WG-prioritization in the agenda

Santiago Pastorino (Jun 17 2020 at 20:08, on Zulip):

@nikomatsakis do you have something you want to share about @WG-rfc-2229?

Santiago Pastorino (Jun 18 2020 at 13:45, on Zulip):

Hi @T-compiler/meeting, triage meeting will be starting in 15 minutes

Santiago Pastorino (Jun 18 2020 at 13:45, on Zulip):

Check out the meeting agenda

pnkfelix (Jun 18 2020 at 14:01, on Zulip):

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

pnkfelix (Jun 18 2020 at 14:02, on Zulip):

we will start off with 5 minutes for ...

Announcements

nikomatsakis (Jun 18 2020 at 14:02, on Zulip):

Santiago Pastorino said:

do you have something you want to share about @WG-rfc-2229?

I'll write down some notes!

pnkfelix (Jun 18 2020 at 14:03, on Zulip):
nikomatsakis (Jun 18 2020 at 14:04, on Zulip):

So, on the topic of Illumos (compiler-team#279)

nikomatsakis (Jun 18 2020 at 14:04, on Zulip):

for RISC-V I did a full FCP

nikomatsakis (Jun 18 2020 at 14:04, on Zulip):

because it implies an ongoing commitment that is at least somewhat hard to disentangle

nikomatsakis (Jun 18 2020 at 14:04, on Zulip):

perhaps we ought to do the same for illumos, not sure

pnkfelix (Jun 18 2020 at 14:05, on Zulip):

/me just noticed for the first time how similar the initialisms FCP and MCP are

nikomatsakis (Jun 18 2020 at 14:05, on Zulip):

side note to check your boxes regarding RISC-V Tier 2 support

pnkfelix (Jun 18 2020 at 14:05, on Zulip):

This issue, of whether toolchain builds should require an FCP, is worth discussing

pnkfelix (Jun 18 2020 at 14:06, on Zulip):

perhaps until we have resolved that question, it would be better to be conservative and do an FCP on compiler-team#279 ?

nikomatsakis (Jun 18 2020 at 14:06, on Zulip):

it was discussed some in #t-compiler > approving new targets

nikomatsakis (Jun 18 2020 at 14:06, on Zulip):

and yes I decided to just do an FCP for now, I tend to think it's a reasonable thing for tier 2 and above

nikomatsakis (Jun 18 2020 at 14:06, on Zulip):

or at least for rustup distribution

nikomatsakis (Jun 18 2020 at 14:07, on Zulip):

but I could be convinced otherwise

pnkfelix (Jun 18 2020 at 14:07, on Zulip):

actually I guess the phrase "doing an FCP" is slightly ambiguous, since MCP's themselves go through an FCP after being seconded...; but you did include the adjective "full FCP" up above.

nikomatsakis (Jun 18 2020 at 14:07, on Zulip):

heh yes

nikomatsakis (Jun 18 2020 at 14:08, on Zulip):

anyway I meant "full checkoff"

nikomatsakis (Jun 18 2020 at 14:08, on Zulip):

and that is probably a better term

nikomatsakis (Jun 18 2020 at 14:08, on Zulip):

in my view the two processes are "analogous" but in one case you need one check, and in the other, you need all (minus 3 or whatever our current threshold is)

pnkfelix (Jun 18 2020 at 14:08, on Zulip):

anyway it looks like the process for such promotion, based on compiler-team#312 (the RISC-V support) is just to do @rfcbot fcp merge on the MCP itself.

pnkfelix (Jun 18 2020 at 14:09, on Zulip):

anyway we can let further discussion take place on #t-compiler > approving new targets

pnkfelix (Jun 18 2020 at 14:09, on Zulip):

other announcements

pnkfelix (Jun 18 2020 at 14:09, on Zulip):
pnkfelix (Jun 18 2020 at 14:09, on Zulip):

WG checkins

pnkfelix (Jun 18 2020 at 14:10, on Zulip):

@WG-prioritization checkin by @spastorino:

pnkfelix (Jun 18 2020 at 14:10, on Zulip):

(this shift to fully async is pretty cool)

nikomatsakis (Jun 18 2020 at 14:10, on Zulip):

that is awesome

Santiago Pastorino (Jun 18 2020 at 14:11, on Zulip):

we plan to have more triagebot commands to help us a bit more in this async process, some things happen kind of manually still but we're going there :)

pnkfelix (Jun 18 2020 at 14:11, on Zulip):

I don't know how others feel about having these various zulip threads pop up under #t-compiler/wg-prioritization , but I think its the sort of thing where if you don't care at all, then you mute the stream, and if you care and like it, then you're happy.

pnkfelix (Jun 18 2020 at 14:11, on Zulip):

So the only question is whether there's anyone who both 1. cares and 2. dislikes the current protocols (which are still under revision I think its fair to say)

nikomatsakis (Jun 18 2020 at 14:12, on Zulip):

Random side note that feels relevant-ish: Zulip has added a new feature called "Recent Topics" (press t to access it)

nikomatsakis (Jun 18 2020 at 14:12, on Zulip):

if you have a stream muted, and you participate in a particular topic,

nikomatsakis (Jun 18 2020 at 14:12, on Zulip):

this lets you keep up with it relatively easily

nikomatsakis (Jun 18 2020 at 14:13, on Zulip):

(I think it could be made more obvious, but I think it's pretty relevant to that "mute the stream" workflow)

pnkfelix (Jun 18 2020 at 14:13, on Zulip):

anyway if you're in that latter boat, do feel free to raise it with @WG-prioritization ; I think the wg is happy even with negative feedback because it shows that people care about the effort being put in.

DPC (Jun 18 2020 at 14:13, on Zulip):

from a wg-perspective, members agreed that either they want the pings (are okay with it) or are indifferent with the mentions to the wg or not in the thread.

pnkfelix (Jun 18 2020 at 14:14, on Zulip):

oh and @nikomatsakis posted a wg chekin, I'll paste it here

pnkfelix (Jun 18 2020 at 14:14, on Zulip):

@WG-rfc-2229 checkin by @nikomatsakis:

nikomatsakis (Jun 18 2020 at 14:14, on Zulip):

I want to highlight the repository bit -- it's a bit of an experiment

nikomatsakis (Jun 18 2020 at 14:14, on Zulip):

but so far it's proven fairly useful to have the tracking issue on rust-lang/rust and move the detailed refactoring work over to a separate repo

nikomatsakis (Jun 18 2020 at 14:15, on Zulip):

e.g. I have given the working group full write access, so people can edit comments and leave notes, and we have a project board, and it's also possible to "watch" the repo without being drowned out with general rust-lang/rust notifications

pnkfelix (Jun 18 2020 at 14:15, on Zulip):

tangentially related

pnkfelix (Jun 18 2020 at 14:15, on Zulip):

I did an AMA yesterday for PLDI regarding Rust

pnkfelix (Jun 18 2020 at 14:15, on Zulip):

and one of the questions at the end was "how do you keep a volunteer effort like this organized"

pnkfelix (Jun 18 2020 at 14:15, on Zulip):

my answer was essentially "we've been doing it in a manner that doesn't scale, and we're trying to fix that."

pnkfelix (Jun 18 2020 at 14:16, on Zulip):

so, yeah, :+1: to structures like what @nikomatsakis outlined above for @WG-rfc-2229

pnkfelix (Jun 18 2020 at 14:16, on Zulip):

(if you want to see the AMA, its up on youtube)

pnkfelix (Jun 18 2020 at 14:17, on Zulip):

https://youtu.be/jGgQmnPH0dQ?t=28439

pnkfelix (Jun 18 2020 at 14:18, on Zulip):

(it claims the slot is 20 minutes but I ended up talking for an hour :open_mouth: )

pnkfelix (Jun 18 2020 at 14:18, on Zulip):

okay so moving on

pnkfelix (Jun 18 2020 at 14:18, on Zulip):

Beta-nominations

T-compiler

pnkfelix (Jun 18 2020 at 14:18, on Zulip):
pnkfelix (Jun 18 2020 at 14:19, on Zulip):

I think the main reason I didn't unilaterally approve this is that I wasn't sure why we shouldn't just let it ride the trains

pnkfelix (Jun 18 2020 at 14:20, on Zulip):

reading the comment thread now, I guess it was identified as a regression.

nikomatsakis (Jun 18 2020 at 14:20, on Zulip):

oh, I see

nikomatsakis (Jun 18 2020 at 14:20, on Zulip):

but yeah if there is an active regression, it seems harmless

Santiago Pastorino (Jun 18 2020 at 14:20, on Zulip):

unsure if worth but this could even be backported to stable

Santiago Pastorino (Jun 18 2020 at 14:20, on Zulip):

it's stable to stable regression

pnkfelix (Jun 18 2020 at 14:21, on Zulip):

is it? 15 days ago the beta didn't have the problem?

pnkfelix (Jun 18 2020 at 14:21, on Zulip):

https://github.com/rust-lang/rust/issues/72933#issuecomment-637984345

Santiago Pastorino (Jun 18 2020 at 14:22, on Zulip):

I was quickly checking the issue #72933 it's tagged as stable to stable regression

pnkfelix (Jun 18 2020 at 14:22, on Zulip):

I'm assuming that this is one of those regressions where there was unintended fallout from some other library change? Or at least, AFAICT, no one pinpointed the regression to a PR?

pnkfelix (Jun 18 2020 at 14:22, on Zulip):

Santiago Pastorino said:

I was quickly checking the issue #72933 it's tagged as stable to stable regression

but ... its not? Its tagged stable-to-beta ?

pnkfelix (Jun 18 2020 at 14:23, on Zulip):

/me refreshes browser a few times

nagisa (Jun 18 2020 at 14:23, on Zulip):

I'm pretty sure the regression is due to changes to the compiler.

Santiago Pastorino (Jun 18 2020 at 14:23, on Zulip):

sorry unsure what I was looking at, nevermind

pnkfelix (Jun 18 2020 at 14:23, on Zulip):

anyway, it seems like no one objects to a backport

pnkfelix (Jun 18 2020 at 14:23, on Zulip):

so beta backport approved

nagisa (Jun 18 2020 at 14:23, on Zulip):

I guess is the pattern matching changes that implemented | patterns inside of other patterns.

pnkfelix (Jun 18 2020 at 14:24, on Zulip):
pnkfelix (Jun 18 2020 at 14:26, on Zulip):

can someone remind me what the trace! macro does? Is that provided by log, i.e. a variant of info! and warn! ?

nikomatsakis (Jun 18 2020 at 14:26, on Zulip):

even less than debug iirc

pnkfelix (Jun 18 2020 at 14:26, on Zulip):

okay

pnkfelix (Jun 18 2020 at 14:26, on Zulip):

anyway, beta approved

pnkfelix (Jun 18 2020 at 14:27, on Zulip):
Santiago Pastorino (Jun 18 2020 at 14:29, on Zulip):

I don't expect it to fail with old MinGW but who knows shrug

pnkfelix (Jun 18 2020 at 14:29, on Zulip):

how can I infer that this doesn't inject a bug?

nagisa (Jun 18 2020 at 14:29, on Zulip):

yeah I was thinking the same

nagisa (Jun 18 2020 at 14:29, on Zulip):

what about the old mingw versions

nikomatsakis (Jun 18 2020 at 14:29, on Zulip):

It built just fine on my PC with mingw-w64 6.3.0.

nikomatsakis (Jun 18 2020 at 14:29, on Zulip):

is that an older mingw version?

nikomatsakis (Jun 18 2020 at 14:29, on Zulip):

iirc we've found in the past that it is hard to support all mingw versions

pnkfelix (Jun 18 2020 at 14:30, on Zulip):

oh yeah, wasn't there a topic about that somewhere?

pnkfelix (Jun 18 2020 at 14:30, on Zulip):

in terms of what our official policy was w.r.t. which MinGW versions are supported?

pnkfelix (Jun 18 2020 at 14:30, on Zulip):

Anyway I'm personally inclined to either 1. let this wait a week (to see if anyone finds a problem on nightly with it on older MinGW), or 2. just decline to backport outright

nikomatsakis (Jun 18 2020 at 14:31, on Zulip):

yes, we discussed it, I think mati865 and Vadim Petrochenkov are two of the most knowledgeable folks in this area iirc

pnkfelix (Jun 18 2020 at 14:31, on Zulip):

i'll interpret the silence as "lets wait a week"

nikomatsakis (Jun 18 2020 at 14:31, on Zulip):

so I'm inclined to trust them

nikomatsakis (Jun 18 2020 at 14:32, on Zulip):

(but I don't have a strong opinion about backporting per se)

pnkfelix (Jun 18 2020 at 14:32, on Zulip):
pnkfelix (Jun 18 2020 at 14:34, on Zulip):

hmm why didn't i unilaterally approve this

Santiago Pastorino (Jun 18 2020 at 14:34, on Zulip):

this one is it stable to stable :P

pnkfelix (Jun 18 2020 at 14:34, on Zulip):

oh that might be why. :slight_smile:

pnkfelix (Jun 18 2020 at 14:34, on Zulip):

I did dive into the history of that file

pnkfelix (Jun 18 2020 at 14:34, on Zulip):

to try to figure out why @eddyb didn't "just" make this change when they wrote that fixme comment

pnkfelix (Jun 18 2020 at 14:35, on Zulip):

anyway it looks like everyone's cool with backporting.

eddyb (Jun 18 2020 at 14:36, on Zulip):

that doesn't look familiar, I'll check the blame

pnkfelix (Jun 18 2020 at 14:36, on Zulip):

thanks @eddyb

nikomatsakis (Jun 18 2020 at 14:36, on Zulip):

I think there may have been changes to eval in the meantime

nikomatsakis (Jun 18 2020 at 14:37, on Zulip):

but i'm not sure :)

pnkfelix (Jun 18 2020 at 14:37, on Zulip):
pnkfelix (Jun 18 2020 at 14:37, on Zulip):

(hmm, wondering again why I didn't unilaterally ... oh now I remember)

pnkfelix (Jun 18 2020 at 14:38, on Zulip):

it was my question about whether everyone understands that mir-opt-level >= 2 is known to have unsound optimizations

eddyb (Jun 18 2020 at 14:38, on Zulip):

inlining had bugs for a while that were kind of like that I guess

pnkfelix (Jun 18 2020 at 14:39, on Zulip):

(I personally think that is a mistake, in the sense that I would make a stronger opt-in for unsound stuff.)

eddyb (Jun 18 2020 at 14:39, on Zulip):

oh right that was discussed at some point

pnkfelix (Jun 18 2020 at 14:39, on Zulip):

(I understand the precedent that has been set by other compilers; I just don't think we should follow it.)

eddyb (Jun 18 2020 at 14:39, on Zulip):

have a different -Z flag for the known broken ones

pnkfelix (Jun 18 2020 at 14:39, on Zulip):

anyway we don't need to get sidetracked by that here

pnkfelix (Jun 18 2020 at 14:40, on Zulip):

eddyb said:

have a different -Z flag for the known broken ones

yeah I'd be much happier with that approach

eddyb (Jun 18 2020 at 14:40, on Zulip):

also I should've opened issues for this instead of just leaving drive-by comments on https://github.com/rust-lang/rust/pull/66282#discussion_r376730824

pnkfelix (Jun 18 2020 at 14:40, on Zulip):

or an unstable attribute to apply it more locally, even

pnkfelix (Jun 18 2020 at 14:40, on Zulip):

but anyway

eddyb (Jun 18 2020 at 14:40, on Zulip):

it's just almost impossible to leave a post-hoc review of a PR :/

pnkfelix (Jun 18 2020 at 14:40, on Zulip):

All of this is sort of moot; I'm fine with approving the backport itself

pnkfelix (Jun 18 2020 at 14:40, on Zulip):

the broader policy can be dealt with later

pnkfelix (Jun 18 2020 at 14:41, on Zulip):

beta backport approved

pnkfelix (Jun 18 2020 at 14:41, on Zulip):
pnkfelix (Jun 18 2020 at 14:42, on Zulip):

"who uses non-gnu linkers, anyway"

pnkfelix (Jun 18 2020 at 14:43, on Zulip):

beta backport approved

pnkfelix (Jun 18 2020 at 14:43, on Zulip):

(I didn't unilaterally approve this because I didn't get around to looking at it)

pnkfelix (Jun 18 2020 at 14:43, on Zulip):

libs-impl
There are no issues this time.

T-rustdoc
There are no issues this time.

pnkfelix (Jun 18 2020 at 14:44, on Zulip):

Stable-nominations

T-compiler
There are no issues this time.

libs-impl
There are no issues this time.

T-rustdoc
There are no issues this time.

pnkfelix (Jun 18 2020 at 14:44, on Zulip):

PR's S-waiting-on-team

T-compiler

pnkfelix (Jun 18 2020 at 14:44, on Zulip):

libs-impl
There are no issues this time.

pnkfelix (Jun 18 2020 at 14:44, on Zulip):

Issues of Note

Short Summary

pnkfelix (Jun 18 2020 at 14:45, on Zulip):

P-critical

T-compiler

pnkfelix (Jun 18 2020 at 14:45, on Zulip):
pnkfelix (Jun 18 2020 at 14:45, on Zulip):

(these are all well in hand)

pnkfelix (Jun 18 2020 at 14:46, on Zulip):

by the way, there was as debate on some zulip topic (@Santiago Pastorino can you find link) about whether we should demote P-critical issues to P-high once they are "on track to be resolved"

pnkfelix (Jun 18 2020 at 14:46, on Zulip):

The notable case was P-critical beta-regressions that 1. have a PR landed on nightly 2. that is approved for beta-backport (but has no backport PR yet)

pnkfelix (Jun 18 2020 at 14:47, on Zulip):

is that worth keeping at P-critical, or is it better to demote to P-high at that point

pnkfelix (Jun 18 2020 at 14:47, on Zulip):

I think it was @ecstatic-morse who said that demoting in this scenario was counter-intuitive

Santiago Pastorino (Jun 18 2020 at 14:47, on Zulip):

https://rust-lang.zulipchat.com/#narrow/stream/227806-t-compiler.2Fwg-prioritization/topic/P-critical.20issues.20for.20beta.2Fstable.20nominations

pnkfelix (Jun 18 2020 at 14:47, on Zulip):

thanks @Santiago Pastorino

pnkfelix (Jun 18 2020 at 14:48, on Zulip):

anyway I'm mostly ambivalent about whether to demote or not. The main reason to demote is to keep the set of P-critical issues small and focused, so that this synchronous meeting can likewise be short and focused

pnkfelix (Jun 18 2020 at 14:48, on Zulip):

feel free to voice your opinion on #t-compiler/wg-prioritization > P-critical issues for beta/stable nominations

Santiago Pastorino (Jun 18 2020 at 14:49, on Zulip):

I'm still with the opinion that issues mainly track what happens on master , so on master is not P-critical anymore

pnkfelix (Jun 18 2020 at 14:49, on Zulip):

last P-critical issue:

pnkfelix (Jun 18 2020 at 14:49, on Zulip):

Santiago Pastorino said:

I'm still with the opinion that issues mainly track what happens on master , so on master is not P-critical anymore

yeah we should resolve that too

Santiago Pastorino (Jun 18 2020 at 14:49, on Zulip):

Santiago Pastorino said:

I'm still with the opinion that issues mainly track what happens on master , so on master is not P-critical anymore

I've just seen earier one P-critical automatically closed because the PR had the text Closes #XYZ

pnkfelix (Jun 18 2020 at 14:49, on Zulip):

I am a little worried about closing beta-regression issues before beta itself is fixed

Santiago Pastorino (Jun 18 2020 at 14:49, on Zulip):

and it had a pending beta backport

pnkfelix (Jun 18 2020 at 14:50, on Zulip):

and while sometimes they do get closed due to the mechanism you described earlier, that doesn't mean that is what we should strive for

pnkfelix (Jun 18 2020 at 14:50, on Zulip):

anyway lets maybe let debate on that matter live on #t-compiler/wg-prioritization > P-critical issues for beta/stable nominations

pnkfelix (Jun 18 2020 at 14:51, on Zulip):

last P-critical issue:

pnkfelix (Jun 18 2020 at 14:51, on Zulip):
pnkfelix (Jun 18 2020 at 14:51, on Zulip):

oh wait I unilaterally approved beta backport here

pnkfelix (Jun 18 2020 at 14:51, on Zulip):

so again we're okay. (That's an example, I think, of that issue that could be demoted to P-high if we follow the aforementioned demotion policy)

pnkfelix (Jun 18 2020 at 14:52, on Zulip):

libs-impl
There are no issues this time.

T-rustdoc
There are no issues this time.

pnkfelix (Jun 18 2020 at 14:52, on Zulip):

Unassigned P-high regressions

Beta regressions

pnkfelix (Jun 18 2020 at 14:52, on Zulip):
pnkfelix (Jun 18 2020 at 14:53, on Zulip):

okay so this is dealt with (see above backport approval)

pnkfelix (Jun 18 2020 at 14:53, on Zulip):

Nightly regressions
There are no issues this time.

pnkfelix (Jun 18 2020 at 14:53, on Zulip):

Performance logs

Triage done by njn

Some bustage involving Cargo meant that some merges in this period didn't get measured.

Regressions:

Improvements:

pnkfelix (Jun 18 2020 at 14:54, on Zulip):

hmm. A little unfortunate about the various bits only being evaluated at a coarse grain due to bustage and/or rollups

pnkfelix (Jun 18 2020 at 14:54, on Zulip):

I don't think there's anything there we need to lose sleep over, except maybe the 3.5% regression...?

pnkfelix (Jun 18 2020 at 14:55, on Zulip):

Nominated Issues

T-compiler

pnkfelix (Jun 18 2020 at 14:55, on Zulip):

five minutes left, and four nominated issues...

nikomatsakis (Jun 18 2020 at 14:55, on Zulip):

pnkfelix said:

I don't think there's anything there we need to lose sleep over, except maybe the 3.5% regression...?

I think the logic in #71824 is gated on a feature-gate, right @ecstatic-morse ?

pnkfelix (Jun 18 2020 at 14:55, on Zulip):
Santiago Pastorino (Jun 18 2020 at 14:56, on Zulip):

pnkfelix said:

five minutes left, and four nominated issues...

don't worry that much about this, we can defer most of them for next meeting when we run out of time

pnkfelix (Jun 18 2020 at 14:56, on Zulip):

nikomatsakis said:

pnkfelix said:

I don't think there's anything there we need to lose sleep over, except maybe the 3.5% regression...?

I think that logic is gated on a feature-gate, right ecstatic-morse ?

that's assuming the regression is indeed due to #71824, which has not yet been confirmed, of course...

nikomatsakis (Jun 18 2020 at 14:56, on Zulip):

yeah I'm kind of suggesting that may be incorrect

pnkfelix (Jun 18 2020 at 14:56, on Zulip):

gotcha

pnkfelix (Jun 18 2020 at 14:57, on Zulip):

i'm going to jump to this one issue

nikomatsakis (Jun 18 2020 at 14:57, on Zulip):

Also man huge shout out to the cleanup crew

pnkfelix (Jun 18 2020 at 14:57, on Zulip):
nikomatsakis (Jun 18 2020 at 14:57, on Zulip):

stuff like this is awesome

pnkfelix (Jun 18 2020 at 14:57, on Zulip):

basically, if you object to #73265

pnkfelix (Jun 18 2020 at 14:57, on Zulip):

this is your chance to say something. :slight_smile:

mark-i-m (Jun 18 2020 at 14:58, on Zulip):

Specifically this: https://github.com/rust-lang/rust/pull/73265#issuecomment-643416610

pnkfelix (Jun 18 2020 at 14:59, on Zulip):

so you all ended up settling on std/<crate>/src/**.rs in the end

simulacrum (Jun 18 2020 at 15:00, on Zulip):

FWIW there's a process question here as well -- the MCP sort of ended/accepted with some concerns, but it's not really clear what to do about them

simulacrum (Jun 18 2020 at 15:00, on Zulip):

with FCPs, the team sort of makes a decision after FCP concludes

simulacrum (Jun 18 2020 at 15:00, on Zulip):

but with MCP it's less clear who is responsible for that

simulacrum (Jun 18 2020 at 15:00, on Zulip):

(obviously if there's no objections, there's no problem)

pnkfelix (Jun 18 2020 at 15:00, on Zulip):

maybe we should use an fcp merge checkbox here

pnkfelix (Jun 18 2020 at 15:01, on Zulip):

I know the MCP was approved, but as @simulacrum points out, there were open questiosn that are getting resolved on the PR thread itself

simulacrum (Jun 18 2020 at 15:01, on Zulip):

perhaps, yeah (see https://rust-lang.zulipchat.com/#narrow/stream/233931-t-compiler.2Fmajor-changes/topic/.60mv.20src.2Flib.7Bstd.2Ccore.2Calloc.2Ctest.2Cetc.7D.20std.2Fl.20compiler-team.23298/near/199156966 discussion)

pnkfelix (Jun 18 2020 at 15:01, on Zulip):

so that seems worth some sort of round-trip through the team

pnkfelix (Jun 18 2020 at 15:01, on Zulip):

and the only question is: How much of a roundtrip?

simulacrum (Jun 18 2020 at 15:01, on Zulip):

but I don't know if we have a explicit policy for things like this

simulacrum (Jun 18 2020 at 15:01, on Zulip):

seems good to have one though, and FCP seems reasonable

nikomatsakis (Jun 18 2020 at 15:01, on Zulip):

we don't have an explicit policy,

pnkfelix (Jun 18 2020 at 15:02, on Zulip):

Absent an explicit policy, I think we should be conservative and do an fcpbot merge here then

simulacrum (Jun 18 2020 at 15:02, on Zulip):

unresolved objections on MCP => get FCP approval before moving ahead, basically

nikomatsakis (Jun 18 2020 at 15:02, on Zulip):

I think escalating to a full check-off is reasonable

pnkfelix (Jun 18 2020 at 15:02, on Zulip):

great

nikomatsakis (Jun 18 2020 at 15:02, on Zulip):

of course I'd rather just see the proposal kind of changed to resolve the objections

simulacrum (Jun 18 2020 at 15:02, on Zulip):

(can someone kick that off on the relevant thread?)

pnkfelix (Jun 18 2020 at 15:02, on Zulip):

will do

pnkfelix (Jun 18 2020 at 15:02, on Zulip):

and with that

pnkfelix (Jun 18 2020 at 15:02, on Zulip):

meeting time is up, and we got through almost everything

pnkfelix (Jun 18 2020 at 15:03, on Zulip):

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

simulacrum (Jun 18 2020 at 15:03, on Zulip):

@nikomatsakis there's a fundamental trade-off here between the typical thing being "Cargo projects have a src/ directory in each crate" and "shortness of paths"

Vadim Petrochenkov (Jun 18 2020 at 15:03, on Zulip):

Regarding the mingw backport - https://github.com/rust-lang/rust/pull/73184#issuecomment-646078056

pnkfelix (Jun 18 2020 at 15:03, on Zulip):

and thanks to @WG-prioritization for all your work triaging and preparing the agenda

pnkfelix (Jun 18 2020 at 15:03, on Zulip):

stay safe, stay healthy :mask:

nikomatsakis (Jun 18 2020 at 15:04, on Zulip):

@simulacrum yeah I recognize this may be a case that the concerns can't be resolved

eddyb (Jun 18 2020 at 15:04, on Zulip):

@pnkfelix kinda boring but if you're still curious: https://github.com/rust-lang/rust/pull/73225#discussion_r442296635

simulacrum (Jun 18 2020 at 15:05, on Zulip):

:thumbs_up: I agree in general, just don't think it's really feasible in this case. Of course, we can resolve them by going with option B and not A, but IMO option A is better so then I'd raise an objection etc

pnkfelix (Jun 18 2020 at 15:06, on Zulip):

hey @mark-i-m i didn't check: The proposal you wrote in the comment does reflect the state of the PR, right?

simulacrum (Jun 18 2020 at 15:12, on Zulip):

I've also updated the PR description fwiw

ecstatic-morse (Jun 18 2020 at 15:31, on Zulip):

nikomatsakis said:

pnkfelix said:

I don't think there's anything there we need to lose sleep over, except maybe the 3.5% regression...?

I think the logic in #71824 is gated on a feature-gate, right ecstatic-morse ?

@nikomatsakis The regression is not due to #71824. HIR lowering and borrowck were the queries that regressed. More likely would be #72389 or #72598, which touch that code

nikomatsakis (Jun 18 2020 at 15:34, on Zulip):

I wonder if we should have some better guidelines about what to rollup=never

ecstatic-morse (Jun 18 2020 at 15:35, on Zulip):

If that's correct, how much of a perf regression should we tolerate for improved diagnostics? I don't think "no regressions" is feasible. Even adding a field to something can cause a small regression if that thing is widely copied.

nikomatsakis (Jun 18 2020 at 15:35, on Zulip):

and/or if we can handle rollups better

nikomatsakis (Jun 18 2020 at 15:35, on Zulip):

(I reviewed those but definitely perf was not on my mind as a likely consequence)

ecstatic-morse (Jun 18 2020 at 15:36, on Zulip):

It's not too hard to revert for purposes of a perf run after the fact. I don't think anyone could have reasonably expected a regression from either of those PRs

ecstatic-morse (Jun 18 2020 at 15:37, on Zulip):

(I don't know if it is either one of those for sure)

ecstatic-morse (Jun 18 2020 at 15:37, on Zulip):

It's not too hard to revert for purposes of a perf run after the fact.

Assigning that responsibility is the harder part I guess :smile:

nikomatsakis (Jun 18 2020 at 15:40, on Zulip):

True

ecstatic-morse (Jun 18 2020 at 15:40, on Zulip):

I'll do two perf runs this time. To defend the honor of #71824 XD

ecstatic-morse (Jun 18 2020 at 15:43, on Zulip):

But if we find out that one of those PRs slowed things down, do we revert in this case?

mark-i-m (Jun 18 2020 at 16:29, on Zulip):

pnkfelix said:

hey mark-i-m i didn't check: The proposal you wrote in the comment does reflect the state of the PR, right?

Sorry, I had to run... No the PR needs to be fixed/updated/squashed

mark-i-m (Jun 18 2020 at 16:40, on Zulip):

@pnkfelix there was a suggestion on the gh issue to also remove the lib prefixes (src/libcore -> std/core), which seems to have at least partial support. But it looks like a lot of boxes have already been ticked here. What's the correct thing to do here?

pnkfelix (Jun 18 2020 at 17:08, on Zulip):

well, first of all

pnkfelix (Jun 18 2020 at 17:08, on Zulip):

oh sorry

pnkfelix (Jun 18 2020 at 17:08, on Zulip):

okay so the PR doesn't reflect the proposal

pnkfelix (Jun 18 2020 at 17:08, on Zulip):

and the proposal is not finalized

pnkfelix (Jun 18 2020 at 17:08, on Zulip):

so i'll just cancel the FCP

pnkfelix (Jun 18 2020 at 17:09, on Zulip):

@mark-i-m you let me know when you have a finalized proposal ready. It might be good to update the PR to reflect that proposal, if possible, but I can understand that's a large effort that would be wasted if the proposal does not end up being approved.

pnkfelix (Jun 18 2020 at 17:09, on Zulip):

(I just expect the proposal to be approved, and am trying to streamline the approval process.)

simulacrum (Jun 18 2020 at 17:15, on Zulip):

@pnkfelix I feel like canceling fcp is going to end with us spinning a few more weeks on this needlessly, personally, but I'm okay with it if you want to

simulacrum (Jun 18 2020 at 17:15, on Zulip):

i.e. the current suggestions all seem like things that we would naturally incorporate after fcp finished

simulacrum (Jun 18 2020 at 17:16, on Zulip):

but I also see the desire to start FCP on a complete proposal

pnkfelix (Jun 18 2020 at 17:16, on Zulip):

well, its the kind of change where I'd prefer that each file (or individual directory with its immediate file children, really) gets moved at most once

pnkfelix (Jun 18 2020 at 17:16, on Zulip):

and thus I'd prefer everyone be on the same page about what's happening.

pnkfelix (Jun 18 2020 at 17:17, on Zulip):

the main drawback to delaying (even if only by a week) would be that @mark-i-m would have to maintain the branch

pnkfelix (Jun 18 2020 at 17:18, on Zulip):

so how about this 1. I cancel the FCP, saying that the proposal has written hadn't been properly finalized to the form that @mark-i-m wanted, and 2. I state, as part of the cancellation, that we will be finalizing this next thursday

pnkfelix (Jun 18 2020 at 17:19, on Zulip):

simulacrum said:

i.e. the current suggestions all seem like things that we would naturally incorporate after fcp finished

when you say "would naturally incorporate", do you mean in follow-up PR's? Or folded into this PR, after everyone had signed off on the proposed directory structure?

simulacrum (Jun 18 2020 at 17:22, on Zulip):

Folded into this PR

simulacrum (Jun 18 2020 at 17:22, on Zulip):

I think it was probably a mistake to do FCP on the PR (maybe we should have opened a new issue or so)

simulacrum (Jun 18 2020 at 17:23, on Zulip):

I think asking @mark-i-m to implement each iteration of the proposal is a bit excessive

simulacrum (Jun 18 2020 at 17:24, on Zulip):

And the proposal has not majorly changed, just people are identifying things we can cleanup "while we're at it" as they see it, I guess

simulacrum (Jun 18 2020 at 17:32, on Zulip):

I think the proposal as written by @mark-i-m in the summary comment basically matches the current end state - the main known point of contention at least is there, which is whether it's okay to add a level of nesting

pnkfelix (Jun 18 2020 at 17:33, on Zulip):

okay then. I'm willing to let the current FCP stand then

pnkfelix (Jun 18 2020 at 17:33, on Zulip):

and trust that the remaining revisions are small enough that people won't objecrt

simulacrum (Jun 18 2020 at 17:42, on Zulip):

I think we can say that next Thursday meeting we'll check (like beta backport) and if anyone feels we should re-FCP we can

mark-i-m (Jun 18 2020 at 17:45, on Zulip):

Thanks @pnkfelix and @simulacrum ! That sounds good

Last update: Nov 25 2020 at 02:45UTC