Stream: t-compiler/meetings

Topic: [weekly meeting] 2020-05-21 #54818


Santiago Pastorino (May 20 2020 at 19:00, on Zulip):

Hi @T-compiler/meeting; the triage meeting will be starting in ~ 19 hours

Santiago Pastorino (May 20 2020 at 19:01, on Zulip):

The @WG-prioritization will be doing pre-triage in #t-compiler/wg-prioritization > pre-meeting triage 2020-05-21 #54818

Santiago Pastorino (May 20 2020 at 19:02, on Zulip):

During pre-triage we will be preparing the meeting agenda

Santiago Pastorino (May 20 2020 at 19:03, on Zulip):

We will have checkins from @WG-async-foundations and @WG-diagnostics

Santiago Pastorino (May 20 2020 at 19:03, on Zulip):

@tmandry || @nikomatsakis do you have something you want to share about @WG-async-foundations ?

Santiago Pastorino (May 20 2020 at 19:04, on Zulip):

@oli || @Esteban Küber do you have something you want to share about @WG-diagnostics ?

nikomatsakis (May 20 2020 at 19:13, on Zulip):

I'll make a quick hackmd

nikomatsakis (May 20 2020 at 19:13, on Zulip):

and @tmandry and others can add things into it

Santiago Pastorino (May 20 2020 at 21:04, on Zulip):

btw, I forgot but we skipped last week so we skipped @WG-traits checkin, @nikomatsakis do you want to also do that one? :)

Santiago Pastorino (May 21 2020 at 13:32, on Zulip):

Hi @T-compiler/meeting, triage meeting will be starting in ~ 28 minutes

Santiago Pastorino (May 21 2020 at 13:32, on Zulip):

Check out the meeting agenda

nikomatsakis (May 21 2020 at 14:02, on Zulip):

Santiago Pastorino said:

btw, I forgot but we skipped last week so we skipped @WG-traits checkin, nikomatsakis do you want to also do that one? :)

https://blog.rust-lang.org/inside-rust/2020/05/18/traits-sprint-2.html =)

pnkfelix (May 21 2020 at 14:02, on Zulip):

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

pnkfelix (May 21 2020 at 14:02, on Zulip):

Lets start off with five minutes for ...

pnkfelix (May 21 2020 at 14:02, on Zulip):

Announcements

Santiago Pastorino (May 21 2020 at 14:02, on Zulip):

nikomatsakis said:

Santiago Pastorino said:

btw, I forgot but we skipped last week so we skipped @WG-traits checkin, nikomatsakis do you want to also do that one? :)

https://blog.rust-lang.org/inside-rust/2020/05/18/traits-sprint-2.html =)

that's the way checkins should really work :)

pnkfelix (May 21 2020 at 14:02, on Zulip):
pnkfelix (May 21 2020 at 14:03, on Zulip):

Santiago Pastorino said:

that's the way checkins should really work :)

(maybe, maybe not. I don't know how well links to blog posts encourage interaction during the meeting. but maybe that's fine.)

nikomatsakis (May 21 2020 at 14:03, on Zulip):
pnkfelix (May 21 2020 at 14:04, on Zulip):

(i guess a link to a blog post is just as good as a wall of text, which is what many checkins have evolved into, for good reason)

nikomatsakis (May 21 2020 at 14:04, on Zulip):

pnkfelix said:

(maybe, maybe not. I don't know how well links to blog posts encourage interaction during the meeting. but maybe that's fine.)

I think we should make blog posts, but I think we can go over the content during the meeting in more detail

nikomatsakis (May 21 2020 at 14:04, on Zulip):

or more like less detail, probably

Santiago Pastorino (May 21 2020 at 14:04, on Zulip):

blog post + wall of text here seems good :)

pnkfelix (May 21 2020 at 14:04, on Zulip):
pnkfelix (May 21 2020 at 14:05, on Zulip):

boo I keep meaning to review MCP #294 and then get distracted...

nikomatsakis (May 21 2020 at 14:05, on Zulip):

Hmm, the bot never posts a comment when FCP is complete

nikomatsakis (May 21 2020 at 14:05, on Zulip):

I just closed compiler-team#280 anyway

simulacrum (May 21 2020 at 14:06, on Zulip):

triagebot has no sense of time right now

nikomatsakis (May 21 2020 at 14:06, on Zulip):

side note, should we be adding a label to proposals that were accepted?

nikomatsakis (May 21 2020 at 14:06, on Zulip):

I wonder if we should change them to PRs

Wesley Wiser (May 21 2020 at 14:07, on Zulip):

Applying a label seems like a good idea!

pnkfelix (May 21 2020 at 14:07, on Zulip):

PR's with some sort of document that gets added to the repo, you mean @nikomatsakis ?

nikomatsakis (May 21 2020 at 14:07, on Zulip):

Yeah, instead of e.g. linking to hackmd etc

nikomatsakis (May 21 2020 at 14:07, on Zulip):

Anyway for now I am closing out the issues

pnkfelix (May 21 2020 at 14:08, on Zulip):

(sounds somewhat like some of the ideas that were being proposed for lang team)

nikomatsakis (May 21 2020 at 14:08, on Zulip):

yes..

pnkfelix (May 21 2020 at 14:08, on Zulip):

okay so lets move along the agenda

pnkfelix (May 21 2020 at 14:08, on Zulip):

Beta-nominations

No beta nominations this time for T-compiler and libs-impl.

There's one about T-rustdoc ...

pnkfelix (May 21 2020 at 14:08, on Zulip):

we are not formally in charge of T-rustdoc stuff, but I suspect we need to be

Santiago Pastorino (May 21 2020 at 14:09, on Zulip):

@pnkfelix I've added that one due to your comment about it

pnkfelix (May 21 2020 at 14:09, on Zulip):

I checked on the rustdoc chat server yesterday with @GuillaumeGomez and others; I definitely get the impression that they wont mind if we approve backports on their behalf

simulacrum (May 21 2020 at 14:09, on Zulip):

(there is a separate chat server?)

Santiago Pastorino (May 21 2020 at 14:10, on Zulip):

Discord I guess?

pnkfelix (May 21 2020 at 14:10, on Zulip):

@simulacrum I meant to write "discord server"

simulacrum (May 21 2020 at 14:10, on Zulip):

ah okay

nikomatsakis (May 21 2020 at 14:10, on Zulip):

this seems pretty harmless

nikomatsakis (May 21 2020 at 14:10, on Zulip):

in any case

pnkfelix (May 21 2020 at 14:10, on Zulip):

yeah let me do this right

pnkfelix (May 21 2020 at 14:10, on Zulip):

beta-nom: Bump pulldown-cmark #71682

Santiago Pastorino (May 21 2020 at 14:10, on Zulip):

@pnkfelix unsure if this should be discussed here but I wonder what means taking care of that? only backports? should we go over other things like nominations, p-critical, etc?

Santiago Pastorino (May 21 2020 at 14:10, on Zulip):

I can open a topic about this discussion if you want

pnkfelix (May 21 2020 at 14:11, on Zulip):

I suspect we don't have the bandwidth to deal with all of those things

pnkfelix (May 21 2020 at 14:11, on Zulip):

but yeah, lets maybe put it into a separate thread

pnkfelix (May 21 2020 at 14:13, on Zulip):

to be honest, given that this wasn't high enough priority to get backport approved by T-rustdoc itself

pnkfelix (May 21 2020 at 14:13, on Zulip):

...

pnkfelix (May 21 2020 at 14:13, on Zulip):

well it is harmless

pnkfelix (May 21 2020 at 14:13, on Zulip):

I guess I don't mind approving it

pnkfelix (May 21 2020 at 14:14, on Zulip):

given the radio silence here and on the vote tally above

pnkfelix (May 21 2020 at 14:14, on Zulip):

(apart from I assume implicit approval from @simulacrum and the ... shrug? from @nikomatsakis ?)

Wesley Wiser (May 21 2020 at 14:14, on Zulip):

I mean it seems ok to me but it would be great to see the diff of the actual code changes.

pnkfelix (May 21 2020 at 14:14, on Zulip):

anyway I'll unilaterally pack the voting booth

simulacrum (May 21 2020 at 14:14, on Zulip):

I don't vote because I'm not T-compiler?

pnkfelix (May 21 2020 at 14:15, on Zulip):

beta-accepted

simulacrum (May 21 2020 at 14:15, on Zulip):

(not sure if I should...)

Santiago Pastorino (May 21 2020 at 14:15, on Zulip):

btw, opened #t-compiler > Handling some T-rustdoc stuff

pnkfelix (May 21 2020 at 14:15, on Zulip):

ah good point @simulacrum :)

nikomatsakis (May 21 2020 at 14:15, on Zulip):

I feel like we've generally allowed votes from anyone

nikomatsakis (May 21 2020 at 14:15, on Zulip):

I view it as more of a "straw poll"

nikomatsakis (May 21 2020 at 14:15, on Zulip):

than a "vote" :)

nikomatsakis (May 21 2020 at 14:15, on Zulip):

but anyway

Santiago Pastorino (May 21 2020 at 14:15, on Zulip):

simulacrum said:

I don't vote because I'm not T-compiler?

aren't you compiler contributor?

simulacrum (May 21 2020 at 14:15, on Zulip):

Yeah, not sure. Seems fine to approve.

pnkfelix (May 21 2020 at 14:16, on Zulip):

Stable-nominations

No stable nominations this time.

nikomatsakis (May 21 2020 at 14:16, on Zulip):

In any case I didn't really read the diff but I am indeed ok with approving it :)

pnkfelix (May 21 2020 at 14:16, on Zulip):

PR's S-waiting-on-team

No PR's waiting on team this time.

pnkfelix (May 21 2020 at 14:16, on Zulip):

Issues of Note

Short Summary

There is one more P-critical issues and one less P-high issues in comparison with last week.

pnkfelix (May 21 2020 at 14:16, on Zulip):

P-critical

pnkfelix (May 21 2020 at 14:17, on Zulip):

"Unsoundness due to variance of trait objects WRT associated types" #71550

- This issue is assigned to @spastorino and has an PR open.
- We've run crater and got a bunch of results.
- Niko is investigating possibilities and started a HackMD with regressions if PR lands.

nikomatsakis (May 21 2020 at 14:18, on Zulip):

Hmm

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

to be clear: the summary lists only two packages currently

nikomatsakis (May 21 2020 at 14:18, on Zulip):

Niko "is" investigating :)

nikomatsakis (May 21 2020 at 14:18, on Zulip):

I should do more of that

Santiago Pastorino (May 21 2020 at 14:18, on Zulip):

these were more or less the notes we had for last meeting which was cancelled

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

but there are others on the report, right?

Santiago Pastorino (May 21 2020 at 14:18, on Zulip):

note that the following P-critical could be related

pnkfelix (May 21 2020 at 14:19, on Zulip):

anyway it sounds like @Santiago Pastorino and @nikomatsakis are moving forward here

pnkfelix (May 21 2020 at 14:19, on Zulip):

this is a stable-to-stable regression

pnkfelix (May 21 2020 at 14:20, on Zulip):

is the P-critical label "correct", in the sense that this is a release blocker?

nikomatsakis (May 21 2020 at 14:20, on Zulip):

it may be worth talking about it

nikomatsakis (May 21 2020 at 14:20, on Zulip):

not the label

nikomatsakis (May 21 2020 at 14:20, on Zulip):

but how to handle the issue

nikomatsakis (May 21 2020 at 14:20, on Zulip):

we can also discuss the label

pnkfelix (May 21 2020 at 14:20, on Zulip):

/me is worried about P-critical losing its meaning

nikomatsakis (May 21 2020 at 14:20, on Zulip):

I'm of the opinion that it's "critical", yes, because I think this is one of those issues that we needed to get on top of or it risks becoming much harder to fix

pnkfelix (May 21 2020 at 14:20, on Zulip):

i mean, if you want to ping this issue at every weekly meeting, we can do that

nikomatsakis (May 21 2020 at 14:20, on Zulip):

it's already almost too late

simulacrum (May 21 2020 at 14:20, on Zulip):

I don't really see it as a release blocker, it's in stable already

nikomatsakis (May 21 2020 at 14:20, on Zulip):

I kind of do, e.g. it reminded me of it right now :)

simulacrum (May 21 2020 at 14:20, on Zulip):

if it was a new regression... maybe?

pnkfelix (May 21 2020 at 14:21, on Zulip):

Okay we can leave it at P-critical then

Santiago Pastorino (May 21 2020 at 14:21, on Zulip):

the group is seeing P-critical as potentially release blockers

Santiago Pastorino (May 21 2020 at 14:21, on Zulip):

in this case it won't block it

pnkfelix (May 21 2020 at 14:21, on Zulip):

(this all goes back to whether we need a separate set of tags for "visit frequency")

nikomatsakis (May 21 2020 at 14:21, on Zulip):

I am open to changing to P-high but I do feel like it belongs to that class of unsoundness that is both (a) pretty easy to do and (b) often useful and hence likely to have people relying on it

nikomatsakis (May 21 2020 at 14:21, on Zulip):

I do agree it doesn't block the release

Santiago Pastorino (May 21 2020 at 14:22, on Zulip):

nikomatsakis said:

I am open to changing to P-high but I do feel like it belongs to that class of unsoundness that is both (a) pretty easy to do and (b) often useful and hence likely to have people relying on it

the problem is that it will get "lost" with 40 more less important issues

nikomatsakis (May 21 2020 at 14:22, on Zulip):

oh and (c) only somewhat recently introduced, not around since 1.0 or something

Santiago Pastorino (May 21 2020 at 14:22, on Zulip):

I agree with Felix that adding a Hz kind of label may be a good idea

pnkfelix (May 21 2020 at 14:22, on Zulip):

okay so given that its not a release blocker (and thus we don't need to worry so much about trying to land something in time for release on June 4th), and that we've got people working on it

pnkfelix (May 21 2020 at 14:22, on Zulip):

we can move along

pnkfelix (May 21 2020 at 14:22, on Zulip):

P-critical: "Trait object with non-static lifetime is accepted where static lifetime is expected and required" #72315

- This issue is assigned to @Matthew Jasper
- This issue may be a variation of the previous one
- Needs mcve and we need to find out if it's a dupe of the previous one.

nikomatsakis (May 21 2020 at 14:23, on Zulip):

I don't think this is related to the previous issue

nikomatsakis (May 21 2020 at 14:23, on Zulip):

when did this regress?

nikomatsakis (May 21 2020 at 14:24, on Zulip):

oh wait

nikomatsakis (May 21 2020 at 14:24, on Zulip):

I'm reading more closely

pnkfelix (May 21 2020 at 14:24, on Zulip):

I don't think we know yet? but then again, don't we now have a cargo-bisect-rustc bot, @Santiago Pastorino ?

nikomatsakis (May 21 2020 at 14:24, on Zulip):

I take it back and I believe it is a dup

pnkfelix (May 21 2020 at 14:24, on Zulip):

/me cannot remember the state of things with that

Santiago Pastorino (May 21 2020 at 14:24, on Zulip):

nikomatsakis said:

when did this regress?

I'm not sure if it's a regression

nikomatsakis (May 21 2020 at 14:24, on Zulip):

the example is very complex, the description at the top made it sound different

nikomatsakis (May 21 2020 at 14:24, on Zulip):

we should test with your PR, @Santiago Pastorino

Santiago Pastorino (May 21 2020 at 14:25, on Zulip):

nikomatsakis said:

we should test with your PR, Santiago Pastorino

yes I can do that quickly today

nikomatsakis (May 21 2020 at 14:25, on Zulip):

but it's a good example of why I think it's worth trying to fix that issue sooner rather later =)

Santiago Pastorino (May 21 2020 at 14:25, on Zulip):

pnkfelix said:

I don't think we know yet? but then again, don't we now have a cargo-bisect-rustc bot, Santiago Pastorino ?

I'm not sure if we have it ready or not yet, cc @bjorn3

Wesley Wiser (May 21 2020 at 14:25, on Zulip):

A quick check on rust.godbolt.org suggests it may have regressed in 1.35.

Santiago Pastorino (May 21 2020 at 14:26, on Zulip):

let's label regression-from-stable-to-stable then

pnkfelix (May 21 2020 at 14:26, on Zulip):

okay well the only question for me is whether this should remain assigned to @Matthew Jasper or if we should give it to someone else

bjorn3 (May 21 2020 at 14:26, on Zulip):

@Santiago Pastorino You can just leave a comment at https://github.com/bjorn3/cargo-bisect-rustc-bot/issues/1 to use it.

pnkfelix (May 21 2020 at 14:27, on Zulip):

for now we can leave it with @Matthew Jasper ; we'll be checking in on it again next week

Santiago Pastorino (May 21 2020 at 14:27, on Zulip):

I guess I can continue working on this but I probably need help to decide what to do

pnkfelix (May 21 2020 at 14:27, on Zulip):

I'm going to keep moving through agenda then

Santiago Pastorino (May 21 2020 at 14:27, on Zulip):

well I'm saying, if those are related I can take care of it but in any case I can report if my patch fixes the issue or not first

pnkfelix (May 21 2020 at 14:27, on Zulip):

Unassigned P-high regressions

pnkfelix (May 21 2020 at 14:28, on Zulip):

oh man, #70819 still isn't fixed

nikomatsakis (May 21 2020 at 14:28, on Zulip):

/me sighs

pnkfelix (May 21 2020 at 14:28, on Zulip):

let me look into that one, I think I can grab some time for it today

pnkfelix (May 21 2020 at 14:28, on Zulip):

/me self-assigns

pnkfelix (May 21 2020 at 14:28, on Zulip):

so for #70924 ....

pnkfelix (May 21 2020 at 14:30, on Zulip):

skimming the comments, I'm not 100% clear that we have a good set of reproduection steps ?

pnkfelix (May 21 2020 at 14:30, on Zulip):

oh wait, @Eric Huss may have added steps 11 days ago

Santiago Pastorino (May 21 2020 at 14:31, on Zulip):

yeah, look at the nomination in the agenda

pnkfelix (May 21 2020 at 14:31, on Zulip):

so that's something

Santiago Pastorino (May 21 2020 at 14:31, on Zulip):
pnkfelix (May 21 2020 at 14:31, on Zulip):

okay... for now I'm going to leave this unassigned

pnkfelix (May 21 2020 at 14:31, on Zulip):

(the main thing I wanted to try to address, in the part of the meeting about "unassigned issues", is to establish whether we should be working harder to assign them)

nikomatsakis (May 21 2020 at 14:31, on Zulip):

cc @eddyb

pnkfelix (May 21 2020 at 14:32, on Zulip):

though in this case, its not clear to me whether this is in fact a stable-to-beta regression?

nikomatsakis (May 21 2020 at 14:32, on Zulip):

How so?

pnkfelix (May 21 2020 at 14:32, on Zulip):

or wait

pnkfelix (May 21 2020 at 14:32, on Zulip):

well Iw as confused

pnkfelix (May 21 2020 at 14:32, on Zulip):

by the two bisections

nikomatsakis (May 21 2020 at 14:33, on Zulip):

given it was caused by " Translate the virtual /rustc/$hash prefix back to a real directory. #70642 "

pnkfelix (May 21 2020 at 14:33, on Zulip):

the claim now is that it was injected by PR #70642 ?

pnkfelix (May 21 2020 at 14:33, on Zulip):

okay

nikomatsakis (May 21 2020 at 14:33, on Zulip):

and the trigger is adding/removing the rust-src component

nikomatsakis (May 21 2020 at 14:33, on Zulip):

it seems "sorta clear" what's going wrong

nikomatsakis (May 21 2020 at 14:33, on Zulip):

i.e., the path worked and now it does't or vice versa..something something..:)

pnkfelix (May 21 2020 at 14:33, on Zulip):

/me shrugs

pnkfelix (May 21 2020 at 14:33, on Zulip):

if that's your def'n of "sorta clear"

nikomatsakis (May 21 2020 at 14:34, on Zulip):

lol it is

pnkfelix (May 21 2020 at 14:34, on Zulip):

anyway

pnkfelix (May 21 2020 at 14:34, on Zulip):

given that it is a beta regression

pnkfelix (May 21 2020 at 14:34, on Zulip):

should we be pushing harder to assign someone

pnkfelix (May 21 2020 at 14:34, on Zulip):

or ... even consider reverting PR #70642 ?

nikomatsakis (May 21 2020 at 14:35, on Zulip):

that's a good question

pnkfelix (May 21 2020 at 14:35, on Zulip):

the bugs that PR #70642 fixes do not seem as high priority as the problem that it may be injecting?

nikomatsakis (May 21 2020 at 14:35, on Zulip):

I think we should try to investigate at least a bit

pnkfelix (May 21 2020 at 14:35, on Zulip):

at the very least, reverting PR #70642 may be an easy to implement fix

nikomatsakis (May 21 2020 at 14:35, on Zulip):

but I don't know who will do it

pnkfelix (May 21 2020 at 14:35, on Zulip):

I can assign #70924 to myself and reach out to @eddyb about it

Santiago Pastorino (May 21 2020 at 14:36, on Zulip):

also separate discussion, there are a bunch of incr. comp. issues like this

nikomatsakis (May 21 2020 at 14:36, on Zulip):

(maybe we should say if somebody doesn't have a pending fix by next week or something..?)

pnkfelix (May 21 2020 at 14:36, on Zulip):

to try to evaluate whether a revert of the PR is the right thing to do for beta

pnkfelix (May 21 2020 at 14:36, on Zulip):

nikomatsakis said:

(maybe we should say if somebody doesn't have a pending fix by next week or something..?)

yeah that is also a good thing to establish as the outcome of this meeting

pnkfelix (May 21 2020 at 14:36, on Zulip):

Santiago Pastorino said:

also separate discussion, there are a bunch of incr. comp. issues like this

yeah this is an example, in my mind, of our need to revisit our strategy for testing incr comp.

Santiago Pastorino (May 21 2020 at 14:37, on Zulip):

let's form a wg :)

pnkfelix (May 21 2020 at 14:37, on Zulip):

anyway I will self-assign #70924

pnkfelix (May 21 2020 at 14:37, on Zulip):

Santiago Pastorino said:

let's form a wg :)

I keep "threatening" to do this. :)

pnkfelix (May 21 2020 at 14:38, on Zulip):

okay, so, next up

pnkfelix (May 21 2020 at 14:38, on Zulip):

Nominated Issues

T-compiler I-nominated

pnkfelix (May 21 2020 at 14:38, on Zulip):

I think @Santiago Pastorino picked out a particular order, so make sure to refer to the agenda if you want to know what's coming next

pnkfelix (May 21 2020 at 14:38, on Zulip):

the first one is what we were just discussing:

pnkfelix (May 21 2020 at 14:39, on Zulip):
pnkfelix (May 21 2020 at 14:39, on Zulip):

so we can "skip" that for this meeting; what we established above suffices

pnkfelix (May 21 2020 at 14:39, on Zulip):
pnkfelix (May 21 2020 at 14:40, on Zulip):

hmm, this is also a stable-to-beta regression

Santiago Pastorino (May 21 2020 at 14:40, on Zulip):

(just in case, I've mentioned that there's a more important issue to discuss in the next list, but I think we will have time to cover everything)

pnkfelix (May 21 2020 at 14:40, on Zulip):

but tracking down injection point is harder because it requires more human effort at each step

simulacrum (May 21 2020 at 14:41, on Zulip):

it would be interesting for me to know if this is fixed by llvm 10 which just landed

pnkfelix (May 21 2020 at 14:41, on Zulip):

@Wesley Wiser you nominated this, was that to get group discussion of the priority choice (P-medium)?

Wesley Wiser (May 21 2020 at 14:41, on Zulip):

I think there's a few questions here:

  1. How serious should we prioritize these kinds of issues which are just "I noticed generated code changed between {x} and {y}"
  2. Is this something we should be tracking as a bug or is this just expected fallout from std lib internals changing?
Wesley Wiser (May 21 2020 at 14:41, on Zulip):

(sorry, was busy typing)

pnkfelix (May 21 2020 at 14:41, on Zulip):

no, thanks, saves me the trouble of trying to guess what you wanted. :)

nikomatsakis (May 21 2020 at 14:42, on Zulip):

I feel like it'd be good to try and come up with some sort of criteria for when we care about changes to perf of generated code

simulacrum (May 21 2020 at 14:42, on Zulip):

well, we should also be clear, this is just looks bad -- I'm not sure it has much effect

pnkfelix (May 21 2020 at 14:42, on Zulip):

back in the day we did used to try to keep track of "missed optimization opportunities"

nikomatsakis (May 21 2020 at 14:42, on Zulip):

there are sometimes things we've specifically optimized

nikomatsakis (May 21 2020 at 14:42, on Zulip):

e.g., that some pattern X doesn't have panic handlers

pnkfelix (May 21 2020 at 14:42, on Zulip):

(mentally I associate those with pcwalton for some reason)

nikomatsakis (May 21 2020 at 14:43, on Zulip):

because it matters to embedded users

nikomatsakis (May 21 2020 at 14:43, on Zulip):

(or that we don't have bounds checks on iterator execution, idk)

nikomatsakis (May 21 2020 at 14:43, on Zulip):

but I don't think we have a good list or sense of what those are

Wesley Wiser (May 21 2020 at 14:43, on Zulip):

I think the embedded case is very important. For this specific issue, my inclination would just be to close because there's no evidence that this is slow or blocking an embedded use case.

Wesley Wiser (May 21 2020 at 14:44, on Zulip):

Or other use case that cares about code size.

pnkfelix (May 21 2020 at 14:44, on Zulip):

hmm

nikomatsakis (May 21 2020 at 14:44, on Zulip):

that might be a reasonable general criteria

nikomatsakis (May 21 2020 at 14:44, on Zulip):

that folks have to point to a "real world" project

pnkfelix (May 21 2020 at 14:44, on Zulip):

@Santiago Pastorino when you look at the perf stuff each week

pnkfelix (May 21 2020 at 14:44, on Zulip):

do you look at just instruction counts? Or also code size?

pnkfelix (May 21 2020 at 14:45, on Zulip):

/me thinks its just average instruction counts?

Wesley Wiser (May 21 2020 at 14:45, on Zulip):

Basically, because there's no issue reported here other than "this is not optimal", I don't want us to be on the hook for fixing every codegen change from one release to another.

pnkfelix (May 21 2020 at 14:45, on Zulip):

to play devil's advocate

pnkfelix (May 21 2020 at 14:45, on Zulip):

issues need not mean that we are "on the hook" for anything

pnkfelix (May 21 2020 at 14:46, on Zulip):

they can just be used as ways to track information

Santiago Pastorino (May 21 2020 at 14:46, on Zulip):

pnkfelix said:

/me thinks its just average instruction counts?

we use to do a quick check, mainly instruction counts but I feel we should more properly discuss what should we check :)

pnkfelix (May 21 2020 at 14:46, on Zulip):

having said that, that's the kind of attitude that led us to have 5,000+ issues

nikomatsakis (May 21 2020 at 14:46, on Zulip):

nikomatsakis said:

that folks have to point to a "real world" project

I'm curious @Wesley Wiser what you think of this standard

Wesley Wiser (May 21 2020 at 14:46, on Zulip):

That's a very good point too. But I wouldn't want to see a new issue opened anytime someone noticed this kind of thing.

nikomatsakis (May 21 2020 at 14:47, on Zulip):

it's not "enough" I think but it may be a bar for "we are going to investigate a bit more"

pnkfelix (May 21 2020 at 14:47, on Zulip):

I don't mind the idea of closing this and saying "feel free to reopen if you can show significant code size effects on some actual project."

Wesley Wiser (May 21 2020 at 14:47, on Zulip):

I'm fine with the bar being as low as "this causes an issue for me because ..."

pnkfelix (May 21 2020 at 14:47, on Zulip):

/me goes to look at the linked irlo post

pnkfelix (May 21 2020 at 14:48, on Zulip):

heh, @Josh Triplett asked them to report an issue

Wesley Wiser (May 21 2020 at 14:48, on Zulip):

If that leads to a proliferation of issues we're not going to deal with, then we can re-evaluate at that time.

pnkfelix (May 21 2020 at 14:49, on Zulip):

so what's the outcome here

pnkfelix (May 21 2020 at 14:49, on Zulip):

we close with a note like the one I wrote above?

nikomatsakis (May 21 2020 at 14:49, on Zulip):

I think pinging the author and @Josh Triplett with a request for more info seems good

pnkfelix (May 21 2020 at 14:49, on Zulip):

(and a "thanks for letting us know about the regression! We are however not going to track every code size regression?")

pnkfelix (May 21 2020 at 14:50, on Zulip):

okay so don't close yet

nikomatsakis (May 21 2020 at 14:50, on Zulip):

I can go either way on closing

Wesley Wiser (May 21 2020 at 14:50, on Zulip):

I do really like that wording :)

pnkfelix (May 21 2020 at 14:50, on Zulip):

just ping author and @Josh Triplett first

nikomatsakis (May 21 2020 at 14:50, on Zulip):

but I'd probably start with a ping and see if they respond

pnkfelix (May 21 2020 at 14:50, on Zulip):

okay

pnkfelix (May 21 2020 at 14:50, on Zulip):

lets move along

pnkfelix (May 21 2020 at 14:50, on Zulip):

libs-impl I-nominated

pnkfelix (May 21 2020 at 14:50, on Zulip):
nikomatsakis (May 21 2020 at 14:51, on Zulip):

OK so

nikomatsakis (May 21 2020 at 14:51, on Zulip):

I suspect the reason we don't see a panic is because

nikomatsakis (May 21 2020 at 14:51, on Zulip):

the stdlib is built optimized and dynamically linked

nikomatsakis (May 21 2020 at 14:52, on Zulip):

so it never gets overflow tests :(

pnkfelix (May 21 2020 at 14:52, on Zulip):

okay, but what way is best to address this?

pnkfelix (May 21 2020 at 14:52, on Zulip):

use checked_add where necessary in stdlib?

Esteban Küber (May 21 2020 at 14:52, on Zulip):

It shouldn't even panic in the first place

nikomatsakis (May 21 2020 at 14:52, on Zulip):

@Hanna Kruppe's suggestion seems right

pnkfelix (May 21 2020 at 14:53, on Zulip):

Esteban Küber said:

It shouldn't even panic in the first place

wait that does not match @Hanna Kruppe 's claim?

nikomatsakis (May 21 2020 at 14:53, on Zulip):

Esteban Küber said:

It shouldn't even panic in the first place

that's kind of T-libs call I guess

nikomatsakis (May 21 2020 at 14:53, on Zulip):

we've traditionally been strict about out-of-bounds ranges though, haven't we?

pnkfelix (May 21 2020 at 14:53, on Zulip):

/me did not realize that @Esteban Küber 's nomination was in part questioning whether this should panic

pnkfelix (May 21 2020 at 14:54, on Zulip):

so, is the first question to throw this over to T-libs about whether this should panic?

nikomatsakis (May 21 2020 at 14:54, on Zulip):

I think so

pnkfelix (May 21 2020 at 14:54, on Zulip):

and if they say it should, then we attack question of how to implement that properly?

pnkfelix (May 21 2020 at 14:54, on Zulip):

okay

nikomatsakis (May 21 2020 at 14:54, on Zulip):

Basically "tell us what behavior you expect" :)

Esteban Küber (May 21 2020 at 14:55, on Zulip):

If we change the impls of drain to accept this off by one case we can make it work as the original code intended, otherwise itd be much nicer for it to be a compile time error instead of a panic. If we want it to be a panic then the patch is minimal

pnkfelix (May 21 2020 at 14:55, on Zulip):

of course, now there's the question of how to adjust the labels to make this nomination for T-libs clear

pnkfelix (May 21 2020 at 14:56, on Zulip):

I guess I could remove T-compiler and libs-impl, and add T-libs

simulacrum (May 21 2020 at 14:56, on Zulip):

we could also just ping libs, I don't think libs has meetings so nominations is a bit useless I imagine?

simulacrum (May 21 2020 at 14:56, on Zulip):

not sure

pnkfelix (May 21 2020 at 14:56, on Zulip):

true

pnkfelix (May 21 2020 at 14:56, on Zulip):

okay so we can just ignore the labelling then

simulacrum (May 21 2020 at 14:56, on Zulip):

I imagine opening a thread in #t-libs is maybe more effective too

Santiago Pastorino (May 21 2020 at 14:56, on Zulip):

pnkfelix said:

I guess I could remove T-compiler and libs-impl, and add T-libs

actually this should have been just libs-impl, not T-compiler I guess

nikomatsakis (May 21 2020 at 14:56, on Zulip):

heh we definitely need a "protocol" for summoning libs

Esteban Küber (May 21 2020 at 14:56, on Zulip):

We _can_ make a quick patch to panic in the meantime

pnkfelix (May 21 2020 at 14:57, on Zulip):

I'll write a comment that links to the discussion here

pnkfelix (May 21 2020 at 14:57, on Zulip):

and try to reach out to them

pnkfelix (May 21 2020 at 14:57, on Zulip):

simulacrum said:

I imagine opening a thread in #t-libs is maybe more effective too

probably this way, or including this way

pnkfelix (May 21 2020 at 14:57, on Zulip):
nikomatsakis (May 21 2020 at 14:57, on Zulip):

(side note, I wonder if it'd be nice to have a way to say @rustbot zulip or something, have it kick off a thread in the suitable stream and then leave a comment in the GH issue with a link to it...)

pnkfelix (May 21 2020 at 14:59, on Zulip):

are we going to try to fix #71359 ?

nikomatsakis (May 21 2020 at 14:59, on Zulip):

I didn't think we were :)

pnkfelix (May 21 2020 at 14:59, on Zulip):

there are statements that it was injected by PR #70896

pnkfelix (May 21 2020 at 15:00, on Zulip):

I'm not clear on whether the benefit from PR #70896 justifies the regression

nikomatsakis (May 21 2020 at 15:01, on Zulip):

cc @cuviper

nikomatsakis (May 21 2020 at 15:01, on Zulip):

and perhaps @scottmcm, who reviewed the PR

pnkfelix (May 21 2020 at 15:01, on Zulip):

I'll write a note on PR #70896 for now

nikomatsakis (May 21 2020 at 15:01, on Zulip):

My impression is that the regression is fairly niche but that may be incorrect.

nikomatsakis (May 21 2020 at 15:02, on Zulip):

I can certainly imagine people creating huge chains

pnkfelix (May 21 2020 at 15:02, on Zulip):

okay left note

pnkfelix (May 21 2020 at 15:02, on Zulip):
pnkfelix (May 21 2020 at 15:03, on Zulip):

oh no we're over time

pnkfelix (May 21 2020 at 15:03, on Zulip):

lets leave #67599 for next week

pnkfelix (May 21 2020 at 15:03, on Zulip):

we have three checkin's scheduled, presumably due to our meeting last week being cancelled at last minute

pnkfelix (May 21 2020 at 15:04, on Zulip):

we already saw what WG-traits posted, thanks to @nikomatsakis

pnkfelix (May 21 2020 at 15:04, on Zulip):

is there anything from WG-async-foundations, @nikomatsakis ?

nikomatsakis (May 21 2020 at 15:04, on Zulip):

Yes, from this hackmd:

pnkfelix (May 21 2020 at 15:06, on Zulip):

aside: wasn't the name "Stream" associated with "streaming iterators" which are blocked on GATs?

pnkfelix (May 21 2020 at 15:07, on Zulip):

obviously I'm out of the loop regarding async stuff, so maybe that's taken over the communal terminology

pnkfelix (May 21 2020 at 15:07, on Zulip):

anyway

pnkfelix (May 21 2020 at 15:07, on Zulip):

the last scheduled check in is from WG-diagnostics

pnkfelix (May 21 2020 at 15:08, on Zulip):

@Esteban Küber if you have time, feel free to post something here

nikomatsakis (May 21 2020 at 15:08, on Zulip):

pnkfelix said:

aside: wasn't the name "Stream" associated with "streaming iterators" which are blocked on GATs?

I'm trying to change that terminology

pnkfelix (May 21 2020 at 15:08, on Zulip):

but I feel pretty bad about letting this run so far past the hour

nikomatsakis (May 21 2020 at 15:08, on Zulip):

to "attached" or someting else

nikomatsakis (May 21 2020 at 15:08, on Zulip):

(that is, "Streaming iterators" etc)

nikomatsakis (May 21 2020 at 15:08, on Zulip):

but yes, the term has been overloaded

nikomatsakis (May 21 2020 at 15:08, on Zulip):

I"m also kind of afk now sorry :)

Esteban Küber (May 21 2020 at 15:08, on Zulip):

It's hard for me because my updates are "we landed some improvements or tweaks" :)

nikomatsakis (May 21 2020 at 15:09, on Zulip):

Side note

pnkfelix (May 21 2020 at 15:09, on Zulip):

Cooperators

Santiago Pastorino (May 21 2020 at 15:09, on Zulip):

pnkfelix said:

but I feel pretty bad about letting this run so far past the hour

the agenda was quite long due to the last meeting cancellation (next time I think I can add a mark that goes like, if we have time cover these next things)

nikomatsakis (May 21 2020 at 15:09, on Zulip):

We talked about having a kdin of running log, where we'd have a topic in the appropriate Zulip stream that we harvest for updates

nikomatsakis (May 21 2020 at 15:09, on Zulip):

instead of pinging people

nikomatsakis (May 21 2020 at 15:09, on Zulip):

but I think we kind of never got that off the ground

pnkfelix (May 21 2020 at 15:09, on Zulip):

yes that was an idea that was put forth

Esteban Küber (May 21 2020 at 15:09, on Zulip):

But yeah, in the last 12 weeks a lot of improvements on assoc types and lifetime errors that should make life easier for newcomers

nikomatsakis (May 21 2020 at 15:10, on Zulip):

might be kind of fun to have a log of diagnostic improvements and PRs

nikomatsakis (May 21 2020 at 15:10, on Zulip):

though I am curious @Esteban Küber whether there is still discussion of extracting a library for diagnostic formatting or any of that sort of thing (seems fine if not...)

Esteban Küber (May 21 2020 at 15:10, on Zulip):

annotate-snippets is in the repo now

pnkfelix (May 21 2020 at 15:11, on Zulip):

okay. lets call this meeting done

nikomatsakis (May 21 2020 at 15:11, on Zulip):

and somewhat separately I've been thinking that we sohuld be paying attention to how diagnostics appera in vscode---

Esteban Küber (May 21 2020 at 15:11, on Zulip):

But haven't put the time into switching to it by default yet

nikomatsakis (May 21 2020 at 15:11, on Zulip):

I was hearing some reports, which I have to verify, that some of the work we did on async-await errors

pnkfelix (May 21 2020 at 15:11, on Zulip):

thanks to everyone in @T-compiler/meeting for attending! Stay safe, stay healthy

nikomatsakis (May 21 2020 at 15:11, on Zulip):

doesn't show up in vscode

nikomatsakis (May 21 2020 at 15:11, on Zulip):

anyway I have to run too thanks @pnkfelix and @WG-prioritization !

Esteban Küber (May 21 2020 at 15:12, on Zulip):

Ive been of the thought that we might customize the vsc plugin to display a webview with the ASCII art. I find it much more readable than the truncated tooltip labels

Esteban Küber (May 21 2020 at 15:12, on Zulip):

And you can't really customize tooltips beyond basic md

pnkfelix (May 21 2020 at 15:14, on Zulip):

i wonder if we can get a mermaid rendering plugin into vs code

pnkfelix (May 21 2020 at 15:14, on Zulip):

and then start emitting ```mermaid blocks

pnkfelix (May 21 2020 at 15:15, on Zulip):

(I guess at that point one might as well start emitting SVG directly ...)

Josh Triplett (May 21 2020 at 17:26, on Zulip):

@nikomatsakis @pnkfelix Pong. I would suggest that whether we continue to track a regression depends on how severe we consider it. "We generate one extra instruction" isn't necessarily a bug worth tracking. "We fail to vectorize something we vectorized before", on the other hand, is severe enough to be worth tracking.

Josh Triplett (May 21 2020 at 17:27, on Zulip):

In this case, I don't think the code size issue alone is critical, but generating panic handlers when we didn't before seems problematic.

Josh Triplett (May 21 2020 at 17:28, on Zulip):

And I suggested reporting it because I wondered if fixing it might be as simple as an #[inline] hint in the right place (and then appropriate testing to make sure that doesn't cause unexpected problems / perf issues).

Notification Bot (May 21 2020 at 18:07, on Zulip):

This topic was moved by simulacrum to #t-compiler > tracking performance regressions

Last update: Nov 25 2020 at 02:15UTC