Hi @WG-traits =) Sync meeting in a few minutes
Was there an hour shift somewhere in the world?
Usually it’s 8pm my time I think
DST began yesterday
So we lost an hour
This will last for a few weeks
I'm trying to remember
Did we have a paper or anything like that where we are kind of tracking things?
at one time we did
seems like we don't have the condensed version I am thinking of
yeah, maybe we can adapt that paper doc
I'd basically like to be evolving so that we track our 'active projects' and current goals within that context
well anyway, plans for the coming week were:
In terms of the first point: I did yet more investigation. I feel like I could walk through the current state of normalization but haven't had time to schedule an actual call.
I am however feeling kind of curious by now if lazy norm is truly a blocker for const generics as has been claimed
One thing that is a problem is that, the way the code presently works, there are some steps we are not taking precisely to avoid cycles
so it makes hard to observe the cycle
since it does't occur :)
so I was thinking about trying to port back some commits from a PR by @eddyb so I could observe the problem first hand
I am interested in potentially collaborating with someone on this instead
Maybe that would be someone here at this meeting, or maybe worth putting out a bit of a call -- basically looking for someone to work with me on this investiation, doing some coding, syncing up semi-regularly
(Specifically what would be involved at this point is modifying the
predicates_of implementations for anonymous constants; it will require a bit of experimentation)
Don't everybody volunteer at once :stuck_out_tongue_wink:
Anyway it might just be you + me @tmandry :)
OK, well, let's move on. I'll come back to that point.
Explore the use cases for GATs (aturon, centril)
lol, I was thinking I might have bandwidth for it but not sure when yet
I believe this happened, right @centril / @Aaron Turon? Over in this topic, and I think there is some form of document that I've failed to read
Here is a link to the document, which is a dropbox paper
I'm not sure the next steps here
I think it's probably to do some kind of review of these use cases, I thikn @scalexm did a bit of that already which is great
I guess I'll make a "work item" for myself at least on that, to do the review
Discuss RLS 2.0 type checker and contemplate how to integrate chalk
into it (nikomatsakis, scalexm)
So we had this meeting, and there was a plan to do another talk this week to cover some of how the Lark type checker worked
it occurs to me that I never scheduled this call :)
There is video available of the previous one
So I guess scheduling another chat is good here, though I think that beyond Lark we should also start talking about (and probably doing some hacking!) around the chalk integration question.
@scalexm how much time / interest do you have in hacking here
I'm trying to think what would be a good first PR
maybe trying to replicate something like what the rustc integration currently does?
i.e., to start by not sharing definition of types etc
I definitely have a lot of interest in this, don’t know yet about time for hacking
or would it be better to start by trying to port to some common foundation
I'm guessing it would be better to start integrating chalk-engine
but with liberal use of
panic! where needed
I’m not sure about trying to replicate the what rustc does
It seems like a lot of complex code duplicated
well we know we want to share the lowering code right
maybe the thing to do is to schedule some time to try and sketch out what that means
i.e., I would be game to try and put one or two hours into reviewing what it might look like
and formulate a kind of roadmap / plan
any better ideas? :)
why don't we do this -- I will do an open call and spend some time on it. Anyone who wants can join, if just to "chat along"
That sounds good
/me wants to do some coding dang it
If we settle on something regarding sharing code, I’d be game for trying to hack a bit on that part
ok, sounds good
@Florian Diebold may also be interested
maybe we'll do that in place of the Lark review, although I still want to do that. I wouldn't mind pushing that till next week since I didn't schedule anything yet and I sort of want some time to prep
The next two items:
- Maybe discuss other chalk improvements? (e.g., benchmarking harness) (nikomatsakis)
- Maybe discuss the "lifetime capture" problem in impl Trait? (nikomatsakis, alexreg)
Did not happen.
However, @Alexander Regueiro and I have been talking about the "associated type bounds", so maybe we should add that as a work item
@Alexander Regueiro do you happen to be around?
@nikomatsakis I'm going to work on
let_chains a bit this week; If I have time over I'll work incrementally towards generic closures and
(as RFCs, several... like 4 of them maybe)
@centril maybe say a bit more about what you have planned when you say "work incrementally towards" -- an RFC?
@centril I have a question. If I were to create a repository for the traits WG, which I've been meaning to do forever
how would you feel about posting some of those draft RFCs to the repo?
I'd enjoy being able to skim over them at my liesure
and I'd like to have a place to consolidate some of this planning etc
@nikomatsakis sure, why not... I elaborated in https://rust-lang.zulipchat.com/#narrow/stream/144729-t-compiler.2Fwg-traits/topic/Generic.20bounds.20.26.20closures
I'd like to move generally towards a model of rfcs as "collaborative docs" that a working group si working on
ok, cool, i'll take a look
ok, so that's another work item to list
As far as "associated type bounds", I'll just put down "further progress" for now and mention @Alexander Regueiro =)
(I think I mentioned this before, i've been using :point_up: to help myself find key comments when making minutes later =)
I make no promises of actually having a ready RFC for
for<'a: 'b> this week tho =P
let_chains may or may not take a bunch of time
ok, sound good, anybody got anything else to bring up?
I make no promises of actually having a ready RFC for
for<'a: 'b>this week tho =P
let_chainsmay or may not take a bunch of time
no hurry, I think the actual work item woudl be to upload drafts to a working group repo :)
and for me to create said repo, perhaps
minor point about creating a traits repo: I can watch the repo without getting notifications for everything in t-compiler repo
sounds like a "point in favor"?
@nikomatsakis I meant the draft -- I make no promises wrt. that ^,-
I meant the draft -- I make no promises wrt. that ^,-
I know, my point was, no worries :) nobody is asking for a promise :)
you do a ton :)
OK, cool, meeting adjourned then. :)
that was quick :rabbit:
/me shoots for 30 minute meetings
Just wrote up a quick write-up from the meeting. These are my collected "next steps" from our conversation:
generics_ofquery etc for constants so we can
@nikomatsakis so does the
generics_of thing take the place of the call you had mentioned about normalization today + lazy norm?
probably, yes. I could definitely stilld o that
but I still dont' quite have all the "dots together"
i.e., I can talk about how system works, but I can't yet explain what the problem is that @eddyb was encountering and if I see an easy fix
maybe I can get @eddyb to listen to the talk and they may have some insights :)
@nikomatsakis sorry, bad time for me again this Monday. But yes, ATB well along the way.
Have we decided yet if lazy normalisation is definitely a blocker for fixing type alias bounds btw?
Also, what’s the progress on Chalk integration and lifetimes?