Stream: t-compiler

Topic: weekly meeting 2019-04-25 #54818


nagisa (Apr 25 2019 at 12:10, on Zulip):

@pnkfelix is having some time off so this week’s meeting will be headed by yours truly.

nagisa (Apr 25 2019 at 12:12, on Zulip):

As per this, wg-rfc-2229 and wg-self-profile are expected to check-in this week.

mw (Apr 25 2019 at 12:19, on Zulip):

@Wesley Wiser do you want to give an update of self-profiling?

nikomatsakis (Apr 25 2019 at 12:25, on Zulip):

I can update rfc-2229

Wesley Wiser (Apr 25 2019 at 13:08, on Zulip):

@mw Sure, I can give the update.

Wesley Wiser (Apr 25 2019 at 13:08, on Zulip):

I'll also write up notes for this week's meeting

mw (Apr 25 2019 at 13:08, on Zulip):

@Wesley Wiser Thanks!

nagisa (Apr 25 2019 at 13:52, on Zulip):

@T-compiler/meeting meeting in ~7 minutes.

nagisa (Apr 25 2019 at 14:00, on Zulip):

Okay, lets begin. First, announcements.

nagisa (Apr 25 2019 at 14:00, on Zulip):

There are a number of ongoing FCPs (http://rfcbot.rs) which many haven’t looked at yet. Please read through them or raise concerns when you have the time.

nagisa (Apr 25 2019 at 14:00, on Zulip):

The trains roll over in about a month on May 23rd.

nagisa (Apr 25 2019 at 14:00, on Zulip):

any other announcements?

nikomatsakis (Apr 25 2019 at 14:01, on Zulip):

Contributor RFC has entered FCP.

nagisa (Apr 25 2019 at 14:02, on Zulip):

Okay. Great!

nikomatsakis (Apr 25 2019 at 14:02, on Zulip):

Also, @Santiago Pastorino and I are trying to lead up a new working group, #t-compiler/wg-learning

nikomatsakis (Apr 25 2019 at 14:02, on Zulip):

The focus will be trying to fill out the rustc-guide and API docs

nikomatsakis (Apr 25 2019 at 14:03, on Zulip):

This problem means we'll be asking folks from @T-compiler to record video calls about components they know well; and then others will try to turn that into text.

nikomatsakis (Apr 25 2019 at 14:03, on Zulip):

At least, that's the idea =)

nikomatsakis (Apr 25 2019 at 14:03, on Zulip):

We'll see how well it works!

nagisa (Apr 25 2019 at 14:04, on Zulip):

Perfect. Is there anything else?

nagisa (Apr 25 2019 at 14:04, on Zulip):

/me waits some 15 more seconds

nagisa (Apr 25 2019 at 14:05, on Zulip):

So I have looked through most of the nominated issues during triage and was left with two groups of the issues: those that are P-high and need an assignee to investigate it further and those that were nominated for actual discussion here.

nagisa (Apr 25 2019 at 14:05, on Zulip):

ICE while bootstrapping (./x.py build) #2 #60228 is an ICE in metadata decoding

nagisa (Apr 25 2019 at 14:06, on Zulip):

this is an issue that feels familiar, especially when in combination with incremental, but it also is something that I remember being difficult to reproduce

nagisa (Apr 25 2019 at 14:06, on Zulip):

Obviously a P-high, since it is an ICE, but we need to assign somebody to look and give a shot at reproduction.

nagisa (Apr 25 2019 at 14:07, on Zulip):

Any takers?

nikomatsakis (Apr 25 2019 at 14:07, on Zulip):

I'm a bit confused by this one

nikomatsakis (Apr 25 2019 at 14:07, on Zulip):

Like, do we have any idea what the "trigger" is?

nikomatsakis (Apr 25 2019 at 14:08, on Zulip):

(Is it something in their configuration?)

nikomatsakis (Apr 25 2019 at 14:08, on Zulip):

I guess figuring that out is the work to be done

nagisa (Apr 25 2019 at 14:08, on Zulip):

Most likely no idea. In my memory these kinds of bugs occur over a number of incremental builds, possibly with compiler upgrades in between.

nikomatsakis (Apr 25 2019 at 14:08, on Zulip):

@nagisa if the main work to be done at this point is trying to reproduce, I thnk we should note that -- I was planning on trying to make a "call for participation" thread on internals that we can populate every week, and I think identifying bugs that need this kind of "external investiation" would be particularly useful

nagisa (Apr 25 2019 at 14:08, on Zulip):

at least that was historically the case for these kinds of issues.

nikomatsakis (Apr 25 2019 at 14:09, on Zulip):

Most likely no idea. In my memory these kinds of bugs occur over a number of incremental builds, possibly with compiler upgrades in between.

ok, that's...going to be hard to reproduce

centril (Apr 25 2019 at 14:09, on Zulip):

I did hit an ICE that was similar an hour or so ago and it was resolved with clean

mw (Apr 25 2019 at 14:10, on Zulip):

over a number of incremental builds, possibly with compiler upgrades in between

That's basically guaranteed to fail. ./x.py should try to detect this case and do a clean

nagisa (Apr 25 2019 at 14:10, on Zulip):

I see @nikomatsakis left a comment here.

varkor (Apr 25 2019 at 14:11, on Zulip):

getting x.py to handle this automatically would be very useful

nikomatsakis (Apr 25 2019 at 14:11, on Zulip):

Yeah, I was just asking them to give a bit more details about how reliably they are able to reproduce the failure.

nagisa (Apr 25 2019 at 14:11, on Zulip):

I think that is good enough for what knowledge we have available to us currently.

nagisa (Apr 25 2019 at 14:12, on Zulip):

I’m inclined to leave this I-nominated, but would also be fine with just marking it P-high.

nikomatsakis (Apr 25 2019 at 14:12, on Zulip):

Agreed.

nagisa (Apr 25 2019 at 14:12, on Zulip):

any preferences in either direction?

nikomatsakis (Apr 25 2019 at 14:12, on Zulip):

Slight preference to P-high

nikomatsakis (Apr 25 2019 at 14:12, on Zulip):

though I feel like maybe we want some more labels :) but leave it be for now

centril (Apr 25 2019 at 14:12, on Zulip):

(Here's the backtrace: https://gist.github.com/Centril/be64d49ccc758cab4c11d921bc9eb1f6)

nikomatsakis (Apr 25 2019 at 14:13, on Zulip):

Slight preference to P-high

I would however want to add a comment like "Marking P-high so can keep track of this, but we are still trying to ascertain just what's going on."

nagisa (Apr 25 2019 at 14:13, on Zulip):

Slight preference to P-high

I would however want to add a comment like "Marking P-high so can keep track of this, but we are still trying to ascertain just what's going on."

Done.

nagisa (Apr 25 2019 at 14:13, on Zulip):

The next bug I had in my list is ICE on higher-trait bounds #60218 which @nikomatsakis already self-assigned for investigation.

nagisa (Apr 25 2019 at 14:14, on Zulip):

Thus lets not dwell on it for too long and move on.

nagisa (Apr 25 2019 at 14:14, on Zulip):

cannot borrow as mutable because it is also borrowed as immutable (likely regression) #60136 is an interesting regression that we want to fix. Obviously P-high again, but unclear whom to assign.

nagisa (Apr 25 2019 at 14:16, on Zulip):

We have ~month to fix this before it hits stable.

nikomatsakis (Apr 25 2019 at 14:16, on Zulip):

It seemed like @Matthew Jasper had a theory as to what was going on

centril (Apr 25 2019 at 14:16, on Zulip):

(aside: that diagnostic is really hard to read... everything is smushed together)

nagisa (Apr 25 2019 at 14:17, on Zulip):

Probably the next good step here is to bisect. Although @Matthew Jasper linked a PR, it is not obvious whether it is really the cause.

nikomatsakis (Apr 25 2019 at 14:17, on Zulip):

Bisection or -- even better -- a small test case

nagisa (Apr 25 2019 at 14:18, on Zulip):

Reduction would be great as well, Given what web-dom is, it should be pretty easy to do.

nagisa (Apr 25 2019 at 14:19, on Zulip):

I guess I’ll just turn to RNG and pick one of the people who are not currently present

nagisa (Apr 25 2019 at 14:19, on Zulip):

/me jests.

centril (Apr 25 2019 at 14:19, on Zulip):

Good incentive model to show up :P

nagisa (Apr 25 2019 at 14:21, on Zulip):

Okay, I’ll just mark P-high for now without assignee

Esteban Küber (Apr 25 2019 at 14:21, on Zulip):

Let's remember to check up on it next week...

nagisa (Apr 25 2019 at 14:21, on Zulip):

error: internal compiler error: unexpected region in query response: ReScope(Destruction(219)) #60127 I have absolutely no idea what is going on here, but it is a proper ICE, and therefore P-high.

nikomatsakis (Apr 25 2019 at 14:22, on Zulip):

another one where it would be great to investigate / try to get a more narrow reproduction

nikomatsakis (Apr 25 2019 at 14:22, on Zulip):

probably that falls at least partly to me

centril (Apr 25 2019 at 14:22, on Zulip):

@varkor r? https://github.com/rust-lang/rust/issues/60263

nikomatsakis (Apr 25 2019 at 14:22, on Zulip):

there is at least some kind of zip file that reproduces it

nagisa (Apr 25 2019 at 14:23, on Zulip):

Indeed.

nikomatsakis (Apr 25 2019 at 14:23, on Zulip):

@nagisa sigh, assign me, i'll try to look a bit into that too, or maybe to find someone to pawn it off onto :P

nikomatsakis (Apr 25 2019 at 14:23, on Zulip):

I suspect it's related to the universes PR or some such thing

varkor (Apr 25 2019 at 14:24, on Zulip):

@centril: could you assign me (I'm on mobile and not logged into GH right now)?

centril (Apr 25 2019 at 14:24, on Zulip):

done

nagisa (Apr 25 2019 at 14:24, on Zulip):

rustc incorrectly warn #[no_mangle] is not used when doing incremental build #60050 felt P-medium to me. Although it is a… regression… in behaviour, it technically does not break any code that does not use #[deny(warnings)] ;)

nagisa (Apr 25 2019 at 14:25, on Zulip):

We do have a minimal test case this time around

nikomatsakis (Apr 25 2019 at 14:25, on Zulip):

I can...sort of imagine how this comes about

mw (Apr 25 2019 at 14:25, on Zulip):

a bisection would be great here

nagisa (Apr 25 2019 at 14:26, on Zulip):

I can run rustc-bisect script for this, but will not be able to investigate any more than that

nikomatsakis (Apr 25 2019 at 14:26, on Zulip):

sounds like it would be helpful

nagisa (Apr 25 2019 at 14:26, on Zulip):

Does anybody object to me making this P-medium?

oli (Apr 25 2019 at 14:26, on Zulip):

@nagisa assign to me with the bisection result

nagisa (Apr 25 2019 at 14:28, on Zulip):

That’s all for the issues which needed assignees

nagisa (Apr 25 2019 at 14:28, on Zulip):

We also have 3 issues which were nominated for discussion

nagisa (Apr 25 2019 at 14:28, on Zulip):

-Clinker-plugin-lto doesn't work without extra manual work #60059

nagisa (Apr 25 2019 at 14:28, on Zulip):

Nominating for (hopefully brief) discussion in the @rust-lang/compiler meeting -- this is holding up Firefox's efforts to use ThinLTO to eliminate glue code. It would be good to get it fixed or, at minimum, to specify what the correct behavior is and figure out who to ping about it (I'm not that familiar with these options).

nikomatsakis (Apr 25 2019 at 14:29, on Zulip):

So this was brought to my attention by the FF team

nikomatsakis (Apr 25 2019 at 14:29, on Zulip):

I don't really understand what this issue is all about

mw (Apr 25 2019 at 14:29, on Zulip):

yeah, this is mostly a usability problem, I'd say

nagisa (Apr 25 2019 at 14:29, on Zulip):

it seems to me that -Clinker-plugin-lto flag has all the means necessary to operate in a way that does not cause these issues

mw (Apr 25 2019 at 14:29, on Zulip):

rustc doesn't always know which linker it uses

mw (Apr 25 2019 at 14:30, on Zulip):

and so it relies on the user (who should know) to pass some additional flags

nagisa (Apr 25 2019 at 14:30, on Zulip):

Generally when it uses gcc or clang as a linker, those handle the nitty gritty themselves.

mw (Apr 25 2019 at 14:30, on Zulip):

but all that can be rather complicated and isn't well documented

nagisa (Apr 25 2019 at 14:30, on Zulip):

in all other cases we do have -Clinker-flavour.

nikomatsakis (Apr 25 2019 at 14:30, on Zulip):

but all that can be rather complicated and isn't well documented

is this mostly a matter of documenting ?

nikomatsakis (Apr 25 2019 at 14:31, on Zulip):

Or is it that we could do things in a simpler way for people too?

mw (Apr 25 2019 at 14:31, on Zulip):

well, I think we should try to enumerate the possible cases and let rustc detect them, if that's feasible

nagisa (Apr 25 2019 at 14:31, on Zulip):

Given https://github.com/rust-lang/rust/issues/60059#issuecomment-484717436 I feel like we could just not allow -Clinker-plugin-lto without an argument, I’m not sure what the point of the flag is without the argument.

mw (Apr 25 2019 at 14:31, on Zulip):

I'm not sure that's always possible

nikomatsakis (Apr 25 2019 at 14:32, on Zulip):

See also this comment by glandium:

FWIW, we have a workaround for Firefox, which is that rust is not involved in linking at all. This is more about the general case than Firefox itself.

which suggests that this is not, for FF at least, a blocker.

nikomatsakis (Apr 25 2019 at 14:32, on Zulip):

I'm not sure how to track this sort of thing though

mw (Apr 25 2019 at 14:33, on Zulip):

it's a bit complicated. if LLD is used then we don't need the plugin. but only clang can use LLD, and so on...

nikomatsakis (Apr 25 2019 at 14:33, on Zulip):

but it sounds like a starting point would be documentation

mw (Apr 25 2019 at 14:33, on Zulip):

yes

mw (Apr 25 2019 at 14:33, on Zulip):

and finding out if rustc knows which linker it is using

mw (Apr 25 2019 at 14:34, on Zulip):

then we could try to make rustc smart

nikomatsakis (Apr 25 2019 at 14:34, on Zulip):

so, my sense is, this is medium priority -- but maybe a good place to try to get some contributors interested in trying to document the complexity here

nagisa (Apr 25 2019 at 14:35, on Zulip):

I have written a short summary and removed the I-nominated.

nikomatsakis (Apr 25 2019 at 14:35, on Zulip):

thanks @nagisa

nagisa (Apr 25 2019 at 14:35, on Zulip):

We still have 2 topics for discussion and 2 backports to get to in 10 minutes

nagisa (Apr 25 2019 at 14:35, on Zulip):

Lets get backports over with first, then continue discussing things for the rest of the short time that is not dedicated to check-ins.

nikomatsakis (Apr 25 2019 at 14:35, on Zulip):

(JFYI, there's not a lot to say for rfc-2229 wg)

nagisa (Apr 25 2019 at 14:36, on Zulip):

Temporarily accept [i|u][32|size] suffixes on a tuple index and warn #60186

/poll Backport Temporarily accept [i|u][32|size] suffixes on a tuple index and warn #60186
yay
nay

nikomatsakis (Apr 25 2019 at 14:36, on Zulip):

(well, actually, I do have some questions where I wouldn't mind getting feedback from team)

nagisa (Apr 25 2019 at 14:36, on Zulip):

/poll Backport Revert "compile crates under test w/ -Zemit-stack-sizes" #59911
yay
nay

centril (Apr 25 2019 at 14:36, on Zulip):

(can we use one poll format :joy: )

nikomatsakis (Apr 25 2019 at 14:36, on Zulip):

What is the protocol here :)

oli (Apr 25 2019 at 14:37, on Zulip):

emojis due to mobile apps not showing polls

nikomatsakis (Apr 25 2019 at 14:37, on Zulip):

I guess "vote in poll" or (if on mobile) use emojis

nikomatsakis (Apr 25 2019 at 14:37, on Zulip):

ok

nagisa (Apr 25 2019 at 14:37, on Zulip):

Oh, I didn’t know polls do not show up on mobile

centril (Apr 25 2019 at 14:38, on Zulip):

@Esteban Küber are you saying yes and no simultaneously?

nagisa (Apr 25 2019 at 14:38, on Zulip):

Let me redo this

Esteban Küber (Apr 25 2019 at 14:38, on Zulip):

Was putting the options to make it easier for others

oli (Apr 25 2019 at 14:38, on Zulip):

both look positive to me @nagisa, seems ok to backport both

nikomatsakis (Apr 25 2019 at 14:38, on Zulip):

emojis due to mobile apps not showing polls

do we know if there is an issue filedabout this with Zulip?

UPDATE: Yes, there is

nikomatsakis (Apr 25 2019 at 14:38, on Zulip):

/me goes to check

nikomatsakis (Apr 25 2019 at 14:39, on Zulip):

polls are super handy, after all :)

nagisa (Apr 25 2019 at 14:39, on Zulip):

The polls seem positive to me as well, but I’m willing to hear if there are any opponents

nagisa (Apr 25 2019 at 14:39, on Zulip):

I’m willing to redo the votes if necessary

centril (Apr 25 2019 at 14:39, on Zulip):

Seems clear in favor of backport I think

centril (Apr 25 2019 at 14:39, on Zulip):

for both

Esteban Küber (Apr 25 2019 at 14:40, on Zulip):

I can start using the laptop going forward if I'm the only one on mobile

nagisa (Apr 25 2019 at 14:40, on Zulip):

Okay, I marked both as accepted.

nagisa (Apr 25 2019 at 14:41, on Zulip):

Lets get to the check-ins and return to discussions later if we have time left.

nagisa (Apr 25 2019 at 14:41, on Zulip):

@nikomatsakis mic to you

nikomatsakis (Apr 25 2019 at 14:41, on Zulip):

I can start using the laptop going forward if I'm the only one on mobile

fwiw, they've marked the relevant bug as high priority

nikomatsakis (Apr 25 2019 at 14:41, on Zulip):

OK, so, the rfc-2229 WG...

nikomatsakis (Apr 25 2019 at 14:42, on Zulip):

(cc @WG-rfc-2229)

nikomatsakis (Apr 25 2019 at 14:42, on Zulip):

I forget if people know what RFC 2229 was -- we are working on implementing the RFC that proposed modifying closures to not capture entire variables, but specific sub-paths

nikomatsakis (Apr 25 2019 at 14:43, on Zulip):

this seems to require a fair amount of refactoring

nikomatsakis (Apr 25 2019 at 14:43, on Zulip):

some of which @blitzerr already did, some of which is still in progress

centril (Apr 25 2019 at 14:44, on Zulip):

@nagisa aside: seems there's another PR/commit in need of backport: https://github.com/rust-lang/rust/pull/59974#issuecomment-486701034

nikomatsakis (Apr 25 2019 at 14:44, on Zulip):

right now, I would say the current status is that @blitzerr has a PR that is "almost done" (the UI output on some tests regresses, and we've been stuck trying to fix that for a while, since they've not had time to investigate). That PR adjusts how the "generics" for closures work to give us more flexibility (right now it's quite hard-coded to capturing individual upvars)

nikomatsakis (Apr 25 2019 at 14:45, on Zulip):

Meanwhile, there are still some big steps left. @ange and I recently had a long discussion for example about how to refactor the ExprUseVisitor, probably making a version tailored to the needs of upvar capture (since that code will go away once the AST borrowck goes away)

nikomatsakis (Apr 25 2019 at 14:45, on Zulip):

There are also some pending refactoring bugs opened, with light mentoring instructions

nikomatsakis (Apr 25 2019 at 14:45, on Zulip):

That's roughly the current status I guess; but one area where I wouldn't mind some feedback is

nikomatsakis (Apr 25 2019 at 14:45, on Zulip):

that I've been wondering how to prioritize this work.

nikomatsakis (Apr 25 2019 at 14:46, on Zulip):

I've not been putting in a lot of time but it seems clear that to really get this done it's going to require some dedicated attention to get the designs right

nikomatsakis (Apr 25 2019 at 14:46, on Zulip):

I'm contemplating trying to "finish up" some of the refactorings that @blitzerr did and maybe try to pause things, so we can better focus on compilation time and other priorities.

nikomatsakis (Apr 25 2019 at 14:47, on Zulip):

(But documenting the designs and thoughts for later.)

nikomatsakis (Apr 25 2019 at 14:47, on Zulip):

Anyway, that's it.

nikomatsakis (Apr 25 2019 at 14:47, on Zulip):

(I haven't really talked to the rest of @WG-rfc-2229 about those doubts yet, so sorry if they come as a surprise=)

oli (Apr 25 2019 at 14:47, on Zulip):

Just a comprehension question: is this being implemented for ast and mir borrowck?

nikomatsakis (Apr 25 2019 at 14:48, on Zulip):

I hadn't given a lot of thought to AST borrowck

nikomatsakis (Apr 25 2019 at 14:48, on Zulip):

Most of the work has been dedicated to changing the MIR desugaring

oli (Apr 25 2019 at 14:48, on Zulip):

ah great!

nikomatsakis (Apr 25 2019 at 14:48, on Zulip):

Which would "just work" for MIR borrowck -- modulo errors, which is a big caveat

nikomatsakis (Apr 25 2019 at 14:48, on Zulip):

Just a comprehension question: is this being implemented for ast and mir borrowck?

but this does get me thinking that waiting for AST borrowck to die is not the worst idea

centril (Apr 25 2019 at 14:49, on Zulip):

Depends on how much the wait is ;) (do we have any ideas about that...?)

nikomatsakis (Apr 25 2019 at 14:49, on Zulip):

I think the idea of refactoring ExprUseVisitor etc is pretty appealing, I'd love to get rid of that code (+ MemCategorization) and replace it with something new and shiny

nikomatsakis (Apr 25 2019 at 14:49, on Zulip):

So a middle ground would be to try and slowly make progress on that

nikomatsakis (Apr 25 2019 at 14:49, on Zulip):

Might be a good "internship" project :)

nikomatsakis (Apr 25 2019 at 14:50, on Zulip):

anyway, looking at time, maybe we can move on, but if people have thoughts to share async, we can open another topic, I'd love to hear them

oli (Apr 25 2019 at 14:50, on Zulip):

note that clippy uses ExprUseVisitor a lot, so some coordination there would be great

ange (Apr 25 2019 at 14:50, on Zulip):

Might be a good "internship" project :)

still need someone to mentor it though

nagisa (Apr 25 2019 at 14:50, on Zulip):

I see. Thanks for the update @nikomatsakis! @Wesley Wiser mic’s in your hands now.

nikomatsakis (Apr 25 2019 at 14:50, on Zulip):

note that clippy uses ExprUseVisitor a lot, so some coordination there would be great

ah, interesting, I did not know that

Wesley Wiser (Apr 25 2019 at 14:50, on Zulip):

Sure!

nikomatsakis (Apr 25 2019 at 14:51, on Zulip):

@oli I'm going to open a topic in #t-compiler/wg-rfc-2229 and ask you some questions :)

Wesley Wiser (Apr 25 2019 at 14:51, on Zulip):

wg-self-profile has been making progress on our "minimum viable product" tracking issue (#58967):

Wesley Wiser (Apr 25 2019 at 14:51, on Zulip):

That's everything I'm aware of unless @mw has anything to add

mw (Apr 25 2019 at 14:52, on Zulip):

sgtm

centril (Apr 25 2019 at 14:52, on Zulip):

re. self-profile, @oli pointed me towards https://github.com/rust-lang/measureme/tree/master/summarize for a PR of mine...

I found that the docs are really lacking and could use some love for folks like me who are total newbs wrt. profiling

mw (Apr 25 2019 at 14:52, on Zulip):

@Wesley Wiser do you still have the flame graphs svg around?

Wesley Wiser (Apr 25 2019 at 14:52, on Zulip):

https://rust-lang.zulipchat.com/user_uploads/4715/Y66yrgBYGpAw4mk0xXR5-LIH/rustc.svg

centril (Apr 25 2019 at 14:53, on Zulip):

So if you could produce a more thorough guide that'd be nice

mw (Apr 25 2019 at 14:53, on Zulip):

@centril yes, docs are part of our MVP milestone

Wesley Wiser (Apr 25 2019 at 14:53, on Zulip):

@centril Yeah, that's something I hope to get to soon.

nikomatsakis (Apr 25 2019 at 14:53, on Zulip):

I found that the docs are really lacking and could use some love for folks like me who are total newbs wrt. profiling

this seems like a great opportunity to reach out to some of those newer folks who've been showing up -- for example, @Alice Ryhl or @Amanjeev Sethi

nagisa (Apr 25 2019 at 14:54, on Zulip):

Nice svg, although the caveat is probably are that labeling (function/samples) does not match the actual things being counted

nagisa (Apr 25 2019 at 14:54, on Zulip):

great, thanks!

nagisa (Apr 25 2019 at 14:54, on Zulip):

I encourage to fill an issue re documentation against measureme repository

mw (Apr 25 2019 at 14:54, on Zulip):

Nice svg, although the caveat is probably are that labeling (function/samples) does not match the actual things being counted

what do you mean exactly?

nagisa (Apr 25 2019 at 14:55, on Zulip):

Nice svg, although the caveat is probably are that labeling (function/samples) does not match the actual things being counted

what do you mean exactly?

I would guess that in this instance the things being counted are queries rather than functions and their invocation counts rather than external profile counters.

nagisa (Apr 25 2019 at 14:55, on Zulip):

Okay, we have 5 minutes

nagisa (Apr 25 2019 at 14:55, on Zulip):

we can fit in 1 more discussion into this meeting

Esteban Küber (Apr 25 2019 at 14:56, on Zulip):

I have two relatively quick things

nagisa (Apr 25 2019 at 14:56, on Zulip):

Go ahead.

Esteban Küber (Apr 25 2019 at 14:56, on Zulip):

https://github.com/rust-lang/rust/pull/60118

mw (Apr 25 2019 at 14:56, on Zulip):

I would guess that in this instance the things being counted are queries rather than functions and their invocation counts rather than external profile counters.

The numbers here should be actual execution time, not invocation counts. So that should be pretty useful data.

Esteban Küber (Apr 25 2019 at 14:57, on Zulip):

Policy around complicating rustc for delayed improved diagnostics, @Vadim Petrochenkov wanted this discussed

centril (Apr 25 2019 at 14:57, on Zulip):

https://github.com/rust-lang/measureme/issues/30

Esteban Küber (Apr 25 2019 at 14:57, on Zulip):

What should an acceptable threshold be so I don't constantly push against it :smile:

nikomatsakis (Apr 25 2019 at 14:57, on Zulip):

lol

nagisa (Apr 25 2019 at 14:58, on Zulip):

Yeah, this is a larger policy question and is not quick at all. There will obviously as many opinions as there are contributors.

Esteban Küber (Apr 25 2019 at 14:58, on Zulip):

And https://github.com/rust-lang/rust/issues/45600 after a change I made we can start relying on Mir borrow checker for match guard checking

nagisa (Apr 25 2019 at 14:58, on Zulip):

Ideally we would not be pushing against anything when working diagnostics at all.

nikomatsakis (Apr 25 2019 at 14:58, on Zulip):

This does sound complex. I wonder if it would be well-suited to "design meeting" (or maybe steering meeting) discussion.

nikomatsakis (Apr 25 2019 at 14:58, on Zulip):

I'm trying to decide what an answer looks like

mw (Apr 25 2019 at 14:59, on Zulip):

sounds like a steering meeting topic

centril (Apr 25 2019 at 14:59, on Zulip):

ISTM that in places like typeck, we are rendering diagnostics too early with incomplete information

Esteban Küber (Apr 25 2019 at 14:59, on Zulip):

I know the conversation is not quick, but rather wanted to put it on the table for next meeting

Esteban Küber (Apr 25 2019 at 14:59, on Zulip):

Ok

nikomatsakis (Apr 25 2019 at 14:59, on Zulip):

I guess I can imagine heuristics of how "local" the considerations are

nikomatsakis (Apr 25 2019 at 14:59, on Zulip):

I was hoping today to describe the design meeting procedure4

nikomatsakis (Apr 25 2019 at 14:59, on Zulip):

and create a template for proposals

nikomatsakis (Apr 25 2019 at 14:59, on Zulip):

and start to reach out to people

nikomatsakis (Apr 25 2019 at 14:59, on Zulip):

let me add this @Esteban Küber to my list of design meeting ideas

centril (Apr 25 2019 at 14:59, on Zulip):

@nagisa hmm, there are still 3 i-nominated issues, https://github.com/rust-lang/rust/pulls?q=is%3Aopen+is%3Apr+label%3AI-nominated+label%3AT-compiler

Esteban Küber (Apr 25 2019 at 14:59, on Zulip):

I've recently been adding maps and sets with extra info for later stages of the compiler

Esteban Küber (Apr 25 2019 at 15:00, on Zulip):

To add more detail to errors it seems to be ok

nagisa (Apr 25 2019 at 15:00, on Zulip):

@centril yes, 2 of them are discussions we did not get to

Esteban Küber (Apr 25 2019 at 15:00, on Zulip):

It gets murkier when roofing errors

Esteban Küber (Apr 25 2019 at 15:00, on Zulip):

To add more detail to errors it seems to be ok

Esteban Küber (Apr 25 2019 at 15:00, on Zulip):

To add more detail to errors it seems to be ok

centril (Apr 25 2019 at 15:00, on Zulip):

Bummer; was hoping you'd get through https://github.com/rust-lang/rust/pull/60167 :slight_smile:

Esteban Küber (Apr 25 2019 at 15:00, on Zulip):

*eliding

Esteban Küber (Apr 25 2019 at 15:00, on Zulip):

It gets murkier when eliding errors

varkor (Apr 25 2019 at 15:00, on Zulip):

https://github.com/rust-lang/rust/pull/60167 doesn't necessarily need to be discussed: I just wanted everyone to be aware of it (introducing a 3k line limit in the tidy check)

nikomatsakis (Apr 25 2019 at 15:01, on Zulip):

I would be ok with that

varkor (Apr 25 2019 at 15:01, on Zulip):

so if there are any objections, make them soonish (maybe by the end of the week?) :)

nikomatsakis (Apr 25 2019 at 15:01, on Zulip):

/poll Should we warn on files with >3K lines?

varkor (Apr 25 2019 at 15:02, on Zulip):

or we could just do it now :grinning_face_with_smiling_eyes:

nikomatsakis (Apr 25 2019 at 15:02, on Zulip):

I just figured we'd take a quick temperature :)

Alice Ryhl (Apr 25 2019 at 15:02, on Zulip):

Oh hey I was mentioned

nikomatsakis (Apr 25 2019 at 15:02, on Zulip):

Oh hey I was mentioned

Yes :) I was saying that the #t-compiler/wg-self-profile group could maybe use some help improving the documentation for their feature, and I remembered seeing you join the group

nagisa (Apr 25 2019 at 15:03, on Zulip):

Either way, I gotta run to walk my dog. Thanks everyone for participating and putting up with my inexperienced chair-occupying. See you next time @pnkfelix is on vacation! :slight_smile:

Alice Ryhl (Apr 25 2019 at 15:03, on Zulip):

Well I am currently on the way to the monthly Copenhagen rust group meeting, so that sounds like convenient timing

nikomatsakis (Apr 25 2019 at 15:04, on Zulip):

Probably best would be to talk to @Wesley Wiser (and maybe @mw) over in #t-compiler/wg-self-profile

centril (Apr 25 2019 at 15:04, on Zulip):

@varkor Seems the conclusion is to go ahead and warn then

Amanjeev Sethi (Apr 25 2019 at 15:05, on Zulip):

Well I am currently on the way to the monthly Copenhagen rust group meeting, so that sounds like convenient timing

please let me know if I can help with something (but I am total newbie)

centril (Apr 25 2019 at 15:06, on Zulip):

(I took the liberty of leaving a comment to that effect on the PR)

Wesley Wiser (Apr 25 2019 at 15:06, on Zulip):

We should probably take this to #t-compiler/wg-self-profile at this point, but I'd personally love some help from beginners to make sure our docs are good

centril (Apr 25 2019 at 15:09, on Zulip):

@Wesley Wiser things that would be nice to mention:

1. Start with git clone for the measure me repo itself and progress to cargo install;
2. Tell the user how to setup a rustc toolchain and run summarize for that.
3. And also how to deal with diffs and such

Last update: Nov 16 2019 at 02:10UTC