Stream: t-compiler

Topic: steering meeting 2019-01-17


nikomatsakis (Jan 17 2019 at 16:16, on Zulip):

Reminder: Steering meeting is scheduled for tomorrow.

I put a post on internals to discuss possible agenda items.

nikomatsakis (Jan 17 2019 at 16:17, on Zulip):

( cc @T-compiler/meeting )

pnkfelix (Jan 18 2019 at 12:49, on Zulip):

(and IIUC the meeting will take place about 3 hours and 10 minutes from now)

nikomatsakis (Jan 18 2019 at 14:38, on Zulip):

I posted some further thoughts on the thread.

nikomatsakis (Jan 18 2019 at 14:40, on Zulip):

In particular, this post makes an effort to boil down all hands planning into something semi-manageable

nikomatsakis (Jan 18 2019 at 14:44, on Zulip):

@T-compiler/meeting -- steering meeting in ~15 minutes

nikomatsakis (Jan 18 2019 at 15:01, on Zulip):

Hello @T-compiler/meeting, steering meeting now. :wave:

nikomatsakis (Jan 18 2019 at 15:02, on Zulip):

First thing to figure out: our agenda. =)

I propose that we talk about the upcoming Rust All Hands. What I would specifically like to get out of the meeting is a sense for "which specific topics we think are most important to discuss at the all hands". I can take this it to the planning sessions, where we will kind of balance between the various teams and figure out how many meetings we can accommodate given the rooms available etc. I've posted a comment on internals with some initial thoughts.

I'm open to other things we might want to talk about, however, if people have suggestions. And/or, if my list above is perfect, we might have time for more things. =)

nagisa (Jan 18 2019 at 15:03, on Zulip):

isn’t there a document with ideas for topics somewhere?

pnkfelix (Jan 18 2019 at 15:03, on Zulip):

do you mean the hackmd.io page?

nagisa (Jan 18 2019 at 15:04, on Zulip):

yeah

nikomatsakis (Jan 18 2019 at 15:04, on Zulip):

Yes, this was an attempt to distill that into something smaller

nikomatsakis (Jan 18 2019 at 15:04, on Zulip):

but please do review the full list

nikomatsakis (Jan 18 2019 at 15:05, on Zulip):

Also worth maybe just checking in on: I wrote this in my post:

... Relate this design to the key problems we are trying to tackle ...

but it might be worth stating explicitly what we see as the P1 problems we want to tackle.

nikomatsakis (Jan 18 2019 at 15:06, on Zulip):

I don't hear a lot of alternative proposals for agenda items, so maybe we talk about All Hands planning?

mw (Jan 18 2019 at 15:06, on Zulip):

:+1:

pnkfelix (Jan 18 2019 at 15:06, on Zulip):

/me had sort of figured that anything that was big enough to discuss in this meeting was big enough to be delayed to discussion at All Hands meeting

nikomatsakis (Jan 18 2019 at 15:07, on Zulip):

OK, so, to start out, let me toss out a list of "Priority 1" problems that I see. I'm curious what other think.

pnkfelix (Jan 18 2019 at 15:07, on Zulip):

(e.g. "Are there any fires we need to put out?" isn't really right for this meeting)

mw (Jan 18 2019 at 15:08, on Zulip):

I think that all things on @nikomatsakis list are P1

mw (Jan 18 2019 at 15:08, on Zulip):

I don't know if it is exhaustive :)

nikomatsakis (Jan 18 2019 at 15:08, on Zulip):

( Meta-note: one of the things that I think we ought to discuss is what to do with this meeting in the long term. I've sort of imagined it as evolving into a periodic check-in on "how are we doing with our plan? does plan need adjusting?" but I guess we'll see )

pnkfelix (Jan 18 2019 at 15:08, on Zulip):

if we had to prioritize within that list, or throw things out ...

nikomatsakis (Jan 18 2019 at 15:08, on Zulip):

It's interesting to compare that list to some of the things I anticipate us doing next year

pnkfelix (Jan 18 2019 at 15:08, on Zulip):

I guess "too hard to do anything" would be the first thing I'd downgrade

pnkfelix (Jan 18 2019 at 15:08, on Zulip):

if I had to

pnkfelix (Jan 18 2019 at 15:08, on Zulip):

even though I'm the first to agree that its too hard to do anything

nikomatsakis (Jan 18 2019 at 15:09, on Zulip):

e.g., I think chalk is important, but it doesn't obviously tackle one of those problems, so that raises the question of why I think so. It's related to "too hard to do things", I guess -- basically, there are language features we need to lay the foundation to support

nikomatsakis (Jan 18 2019 at 15:09, on Zulip):

I guess "too hard to do anything" would be the first thing I'd downgrade

interesting, I don't feel that way. I feel like the technical debt is approaching unmanageable levels

nikomatsakis (Jan 18 2019 at 15:10, on Zulip):

e.g., the situation around type and region inference is really baroque right now, with a lot of systems layered on one another

pnkfelix (Jan 18 2019 at 15:10, on Zulip):

well, which one would you throw out if you had to throw out one?

nikomatsakis (Jan 18 2019 at 15:10, on Zulip):

heh, hard question, I don't know

pnkfelix (Jan 18 2019 at 15:10, on Zulip):

(my second choice would be "compilation time" -- i.e., continue kicking that can down the road)

nikomatsakis (Jan 18 2019 at 15:10, on Zulip):

I guess i'm not sure how important it is to throw any out :)

pnkfelix (Jan 18 2019 at 15:10, on Zulip):

I'm just worried about having too broad a set of topics

nikomatsakis (Jan 18 2019 at 15:11, on Zulip):

well, I see these as 'things we aim to improve as part of the "2021 goals"'

pnkfelix (Jan 18 2019 at 15:11, on Zulip):

and/or lacking a good way to acquire laser focus to the most important things

nikomatsakis (Jan 18 2019 at 15:11, on Zulip):

but ok, it is a good question to have:

nikomatsakis (Jan 18 2019 at 15:11, on Zulip):

how much do we focus on e.g. improving RLS if it comes at cost of technical debt etc?

nikomatsakis (Jan 18 2019 at 15:11, on Zulip):

I remember @mw saying, for example, they would like to write docs, but it always feels like lower priority

mw (Jan 18 2019 at 15:12, on Zulip):

it does, and that's not good

mw (Jan 18 2019 at 15:12, on Zulip):

but some of that is just bad habits

nikomatsakis (Jan 18 2019 at 15:12, on Zulip):

I'm not sure if there's a great way to change it, but setting ourselves some explicit goals for documentation might be a start

nikomatsakis (Jan 18 2019 at 15:13, on Zulip):

(something to ponder: I wonder if we should try to make specific goals of that kind, that we can check-in on? e.g., "enable #[deny(missing_docs)] for rustc by summer" or something)

Wesley Wiser (Jan 18 2019 at 15:13, on Zulip):

Is there a concept of "doc coverage" like test coverage but for docs?

nikomatsakis (Jan 18 2019 at 15:13, on Zulip):

there's certainly a concept of it =)

Wesley Wiser (Jan 18 2019 at 15:14, on Zulip):

We could perhaps aim to hit x% of rustc covered and fail PRs that push that below the threshold or something

mw (Jan 18 2019 at 15:14, on Zulip):

can we come up with some kind of concrete incentive for writing docs?

nikomatsakis (Jan 18 2019 at 15:14, on Zulip):

I think if I were going to measure I would say two things:

nikomatsakis (Jan 18 2019 at 15:14, on Zulip):

can we come up with some kind of concrete incentive for writing docs?

I will mail candy to every rustc-guide contributor

mw (Jan 18 2019 at 15:14, on Zulip):

:)

nikomatsakis (Jan 18 2019 at 15:15, on Zulip):

I'm actually semi-serious :)

mw (Jan 18 2019 at 15:15, on Zulip):

but really: when I implement some optimization (or review one of @Zoxc's PRs) I see a lot of green on perf.rlo.

nikomatsakis (Jan 18 2019 at 15:15, on Zulip):

yeah

nikomatsakis (Jan 18 2019 at 15:15, on Zulip):

well, this relates to @Wesley Wiser's question, I guess

mw (Jan 18 2019 at 15:15, on Zulip):

some way to see some progress when writing docs would maybe help

mw (Jan 18 2019 at 15:15, on Zulip):

yes, it does

nikomatsakis (Jan 18 2019 at 15:15, on Zulip):

i.e., if we could have some page charting the documentation "metrics"?

nikomatsakis (Jan 18 2019 at 15:16, on Zulip):

good question to think about

nikomatsakis (Jan 18 2019 at 15:16, on Zulip):

Well, so, to bring this back to all hands:

nikomatsakis (Jan 18 2019 at 15:16, on Zulip):

is this the sort of thing we should be talking about?

davidtwco (Jan 18 2019 at 15:16, on Zulip):

Would blocking PRs that regress % of compiler documented by some threshold be worthwhile or too much pain?

nikomatsakis (Jan 18 2019 at 15:17, on Zulip):

is this the sort of thing we should be talking about?

it would probably fall under the topic I listed as "Compiler team governance and apprenticeship"

pnkfelix (Jan 18 2019 at 15:17, on Zulip):

IMO, We should either figure out the list of P1 things before the All-Hands, or make finalizing such a list an early part of the All-Hands

nikomatsakis (Jan 18 2019 at 15:17, on Zulip):

Would blocking PRs that regress % of compiler documented by some threshold be worthwhile or too much pain?

I personally think it'd be fine, but we probably have to lay some grounwork first. I am serious about deny(missing_docs)

nikomatsakis (Jan 18 2019 at 15:17, on Zulip):

IMO, We should either figure out the list of P1 things before the All-Hands, or make finalizing such a list an early part of the All-Hands

I was just going to circle back to that

QuietMisdreavus (Jan 18 2019 at 15:17, on Zulip):

/me begins pondering a new doc-coverage pass in rustdoc :thinking:

nikomatsakis (Jan 18 2019 at 15:17, on Zulip):

like, do we agree on the list? what does it mean to "finalize" it?

mw (Jan 18 2019 at 15:17, on Zulip):

I think we should try to figure out the list of P1 issues now

nikomatsakis (Jan 18 2019 at 15:18, on Zulip):

I agree, it's something that'd be great to have as "groundwork" going in

mw (Jan 18 2019 at 15:18, on Zulip):

I think the current list will keep us busy enough

mw (Jan 18 2019 at 15:18, on Zulip):

so if it's not missing anything critical, I think we can keep it

nikomatsakis (Jan 18 2019 at 15:18, on Zulip):

I'm going to paste my list into that hackmd document as a starting point; the only thing I think is semi-missing is maybe some language features, but maybe that's too specific

pnkfelix (Jan 18 2019 at 15:19, on Zulip):

I'm still a little worried that the current list is too broad, but I don't mind starting big and figuring out how to refine it later if need be.

nagisa (Jan 18 2019 at 15:19, on Zulip):

One thing that I would like to result out of All Hands is a slew of technical design RFCs. To give an example, either list has a MIR 2.0, but it is not at all clear what it ought to be at all. Sitting down on a round table, not necessarily with all of T-compiler present, and nailing down what would help to achieve the MIR 2.0 goals (goals should be figured out before the all hands) would be nice.

nikomatsakis (Jan 18 2019 at 15:20, on Zulip):

So this is an interesting question

centril (Jan 18 2019 at 15:20, on Zulip):

fwiw I really love the idea of focusing on docs and @davidtwco's suggestion; I also think that big/important language features mostly fall out of foundational work like chalk, etc.

nikomatsakis (Jan 18 2019 at 15:20, on Zulip):

Do we expect to leave the All Hands with technical design documents

nikomatsakis (Jan 18 2019 at 15:20, on Zulip):

or with .. what exactly

nikomatsakis (Jan 18 2019 at 15:20, on Zulip):

I feel like walking out with an outline is a good goal

nikomatsakis (Jan 18 2019 at 15:21, on Zulip):

at least for some topics

nikomatsakis (Jan 18 2019 at 15:21, on Zulip):

we clearly won't write the full text :)

mw (Jan 18 2019 at 15:21, on Zulip):

an outline and a plan who writes the full text ideally :)

nikomatsakis (Jan 18 2019 at 15:21, on Zulip):

yes, this last part feels key to me: the plan. Specifically, by plan, I mean kind of "a set of people"

nikomatsakis (Jan 18 2019 at 15:21, on Zulip):

and/or a structure to support them

pnkfelix (Jan 18 2019 at 15:22, on Zulip):

some stakeholders for some topics won't be at the All-Hands, right?

pnkfelix (Jan 18 2019 at 15:22, on Zulip):

(narrow topics, now, not the broad ones in our P1 list)

pnkfelix (Jan 18 2019 at 15:22, on Zulip):

I don't know if that limits us

pnkfelix (Jan 18 2019 at 15:22, on Zulip):

its probably fine

nagisa (Jan 18 2019 at 15:22, on Zulip):

Yeah, I don’t think I’ve seen e.g. @eddyb on a participant list.

nikomatsakis (Jan 18 2019 at 15:23, on Zulip):

Quoting @nagisa:

Sitting down on a round table, not necessarily with all of T-compiler present, and nailing down what would help

If one of the limitations is the amount of "full team meetings" we can do, maybe it's a good idea to identify "breakout groups" to sketch out a preliminary design, and we can use some of the full team meetings to dig with a wider setting?

pnkfelix (Jan 18 2019 at 15:23, on Zulip):

oh dear, I was thinking of @Vadim Petrochenkov 's absence

nikomatsakis (Jan 18 2019 at 15:23, on Zulip):

Yeah, I don’t think I’ve seen e.g. @eddyb on a participant list.

they will be there

nikomatsakis (Jan 18 2019 at 15:23, on Zulip):

but @Vadim Petrochenkov and @Zoxc (to name two) will not

pnkfelix (Jan 18 2019 at 15:23, on Zulip):

Yeah, I don’t think I’ve seen e.g. @eddyb on a participant list.

they will be there

phew

mw (Jan 18 2019 at 15:24, on Zulip):

we can try to find proxies for the people who won't be there

nikomatsakis (Jan 18 2019 at 15:25, on Zulip):

maybe it's a good idea to identify "breakout groups" to sketch out a preliminary design, and we can use some of the full team meetings to dig with a wider setting?

(to elaborate: maybe monday afternoon (say), is "breakout time", where we sketch out a set of design points (e.g., MIR 2.0, etc) and people go off to cafes or something in smaller groups. Then Tuesdsay morning we discuss those things? I feel like there are discussions where I would like to review and understand what's being said, but I don't have to be part of the initial feedback.)

mw (Jan 18 2019 at 15:25, on Zulip):

e.g. I can try and sync up with Zoxc, so I know their point of view

nikomatsakis (Jan 18 2019 at 15:26, on Zulip):

actually, one could imagine even that we spend some time coming up with initial proposals, and then sort of present them with minimal questions, and then have more of the "follow-up feedback" take place during lunch or something, and then decide (on the fly?) which things to dig into in more detail later on.

nikomatsakis (Jan 18 2019 at 15:27, on Zulip):

I still feel though that the question of RLS etc is really a central one

nikomatsakis (Jan 18 2019 at 15:27, on Zulip):

that involves a lot of key stakeholders

pnkfelix (Jan 18 2019 at 15:27, on Zulip):

so you want to do the proposals before we have the presentations of "here how this works" (for RLS + rust-analyzer) ?

nikomatsakis (Jan 18 2019 at 15:27, on Zulip):

it's sort of "what does the backbone of the compiler look like"

nikomatsakis (Jan 18 2019 at 15:27, on Zulip):

basically I'm wondering if the RLS + rustc question is something in its own category :)

nikomatsakis (Jan 18 2019 at 15:27, on Zulip):

or if it's "just another technical topic to reach consensus on"

mw (Jan 18 2019 at 15:28, on Zulip):

I think that all of the P1 problems are closely related to the compiler's architecture going forward

matklad (Jan 18 2019 at 15:29, on Zulip):

@nikomatsakis you've talked about review of RLS and rust-analyzer. Does it mean "offline review (docs/screencast/presentation) before" All Hands?

nikomatsakis (Jan 18 2019 at 15:29, on Zulip):

I meant "online"

nikomatsakis (Jan 18 2019 at 15:29, on Zulip):

but that's a good question

nikomatsakis (Jan 18 2019 at 15:29, on Zulip):

offline might be better

nikomatsakis (Jan 18 2019 at 15:29, on Zulip):

if people actually watch :)

pnkfelix (Jan 18 2019 at 15:30, on Zulip):

if it can be created in time. (And yes, if people actually watch)

nikomatsakis (Jan 18 2019 at 15:30, on Zulip):

I was imagining for example that Monday afternoon we could have a 2hr slot where we review RLS / rust-analyzer

nikomatsakis (Jan 18 2019 at 15:30, on Zulip):

(and maybe a bit more?)

nikomatsakis (Jan 18 2019 at 15:30, on Zulip):

discuss over dinner etc, and tuesday morning (say) try to further the conversation to some form of consensus

matklad (Jan 18 2019 at 15:31, on Zulip):

I'd be ok with either on-line or off-line review. I personally have time to prep offline thing before all-hands. Should I do it?

Offline has a benefit: everyone can watch it

davidtwco (Jan 18 2019 at 15:31, on Zulip):

I'd definitely appreciate if there were some reading resources or things to watch before the All Hands that would provide some context for conversations that will be happening.

nikomatsakis (Jan 18 2019 at 15:31, on Zulip):

I say do it

davidtwco (Jan 18 2019 at 15:32, on Zulip):

(as I don't want to end up asking too many questions and distract from whatever topic is being discussed)

nikomatsakis (Jan 18 2019 at 15:32, on Zulip):

Maybe we can drill a bit into the RLS questions for a second

nikomatsakis (Jan 18 2019 at 15:33, on Zulip):

in particular, i'm trying to wrap my head around "what are the questions to answer exactly"

matklad (Jan 18 2019 at 15:33, on Zulip):

"how we make completion work" is a good high-level one

nikomatsakis (Jan 18 2019 at 15:33, on Zulip):

like, rust-analyzer is basically proposing a second compiler focused on the IDE, that hopefully shares code and reaches convergence -- so I guess one question is about whether to adopt a strategy like that (is that what we really need to do), or whether it's better for us to try and work forward from what we have incrementally

nikomatsakis (Jan 18 2019 at 15:34, on Zulip):

"how we make completion work" is a good high-level one

right, and this is another angle: sort of starting from a technical review of what we want

nikomatsakis (Jan 18 2019 at 15:34, on Zulip):

and then once we have a vague design, trying to decide how to get there most efficiently?

pnkfelix (Jan 18 2019 at 15:34, on Zulip):

"that hopefully shares code"

matklad (Jan 18 2019 at 15:34, on Zulip):

Yep: "how to make completions" is basically "what is the end design". How do we get there is a separate topic.

nikomatsakis (Jan 18 2019 at 15:35, on Zulip):

so do we think it's better to start by spooling out the "end design" first?

pnkfelix (Jan 18 2019 at 15:35, on Zulip):

Surely the only way we would manage to share code between rust-analyzer and rustc would be to actively work towards that goal from both sides, no?

nikomatsakis (Jan 18 2019 at 15:35, on Zulip):

I can imagine that being more fruitful

nikomatsakis (Jan 18 2019 at 15:35, on Zulip):

Surely the only way we would manage to share code between rust-analyzer and rustc would be to actively work towards that goal from both sides, no?

yes, I think so

matklad (Jan 18 2019 at 15:35, on Zulip):

btw, some code sharing already happens: we use ena for unification.

nikomatsakis (Jan 18 2019 at 15:35, on Zulip):

(I mean I can see a world where we just focus our resources on pushing RLS forward -- maybe rust-analyzer does some work to enable code sharing, but it's not a major priority for rust team per se)

nikomatsakis (Jan 18 2019 at 15:36, on Zulip):

I personally like the idea of having shared libraries, though.

mw (Jan 18 2019 at 15:36, on Zulip):

me too

nikomatsakis (Jan 18 2019 at 15:36, on Zulip):

so do we think it's better to start by spooling out the "end design" first?

but I do want to hear what people think about this. is this the right place to start?

mw (Jan 18 2019 at 15:36, on Zulip):

I think it is

nikomatsakis (Jan 18 2019 at 15:37, on Zulip):

maybe then we can get some specific questions, like:

matklad (Jan 18 2019 at 15:37, on Zulip):

I think so. The goal of rust-analyzer currently is exactly to probe the "end design"

nikomatsakis (Jan 18 2019 at 15:37, on Zulip):

maybe we can make a little meeting, @Aleksey Kladov, to elaborate those? perhaps @Igor Matuszewski can join, and whomever else wants

nikomatsakis (Jan 18 2019 at 15:37, on Zulip):

(after this meeting)

matklad (Jan 18 2019 at 15:37, on Zulip):

sure!

nikomatsakis (Jan 18 2019 at 15:37, on Zulip):

ok, I feel pretty good about this

nikomatsakis (Jan 18 2019 at 15:38, on Zulip):

the more I think about it, though, the more I feel like maybe the IDE questions are "just another technical topic" in some sense

nikomatsakis (Jan 18 2019 at 15:38, on Zulip):

by this I mean:

nikomatsakis (Jan 18 2019 at 15:38, on Zulip):

it makes sense to start with some amount of "heads down" to sketch out, and then come together to review?

nikomatsakis (Jan 18 2019 at 15:39, on Zulip):

however, I would for sure want e.g. @eddyb in that conversation, so it kind of overlaps with some of the other things we might design (e.g., MIR 2.0)

nikomatsakis (Jan 18 2019 at 15:39, on Zulip):

which merely suggests to me that we might have 2 rounds of "breakout + review" or something

nikomatsakis (Jan 18 2019 at 15:40, on Zulip):

basically I'm wondering if there are "breakout topics" where the key person would not want to be part of the IDE design discussion =)

mw (Jan 18 2019 at 15:40, on Zulip):

I think we should have the meta goal to decouple compiler internals as much as possible from each other

mw (Jan 18 2019 at 15:40, on Zulip):

because that scales better with increased team size

nikomatsakis (Jan 18 2019 at 15:42, on Zulip):

seems true

nikomatsakis (Jan 18 2019 at 15:43, on Zulip):

well, ok, so we have some handle of how to approach RLS. I don't yet know if this breakout idea is good, but let's table it. After the RLS technical question, what do we think is "next most useful thing to tear into", maybe?

mw (Jan 18 2019 at 15:44, on Zulip):

I'd pick organizational questions

nikomatsakis (Jan 18 2019 at 15:44, on Zulip):

My thoughts are probably either:

Igor Matuszewski (Jan 18 2019 at 15:44, on Zulip):

@nikomatsakis having an IDE review meeting would be great, yes

nikomatsakis (Jan 18 2019 at 15:44, on Zulip):

I'd pick organizational questions

can we drill a bit into this maybe?

Igor Matuszewski (Jan 18 2019 at 15:45, on Zulip):

Should I make an architecture overview for how RLS is currently set up?

nikomatsakis (Jan 18 2019 at 15:45, on Zulip):

I feel like I now have a fairly concrete idea of what the "agenda" for an RLS-like meeting might be, and also a semi-plan to make it more specific, but maybe we can get a concrete idea for the "agenda" of an organizational meeting

nikomatsakis (Jan 18 2019 at 15:45, on Zulip):

Should I make an architecture overview for how RLS is currently set up?

that would be great

Igor Matuszewski (Jan 18 2019 at 15:45, on Zulip):

(although have in mind that ideally we'd like to move off the current architecture into on-demand one)

nikomatsakis (Jan 18 2019 at 15:46, on Zulip):

yep, but I think it's very useful to have an overview of what happens now

pnkfelix (Jan 18 2019 at 15:46, on Zulip):

@nikomatsakis : are you asking for Agenda items for an organizational meeting ?

nikomatsakis (Jan 18 2019 at 15:46, on Zulip):

there may be problems that were encountered that are easily overlooked, etc

nikomatsakis (Jan 18 2019 at 15:46, on Zulip):

@nikomatsakis : are you asking for Agenda items for an organizational meeting ?

sort of, yes.

pnkfelix (Jan 18 2019 at 15:46, on Zulip):

e.g. How do envision an ideal team(s) structure when we get to 2021 ?

nikomatsakis (Jan 18 2019 at 15:46, on Zulip):

I guess I jsut want to discuss what kinds of questions we are trying to answer and/or problems we are trying to solve

nikomatsakis (Jan 18 2019 at 15:46, on Zulip):

right

nikomatsakis (Jan 18 2019 at 15:47, on Zulip):

so, for example, I've been trying to talk up this idea of working groups that are operating semi-independently but checking in during this steering meeting

mw (Jan 18 2019 at 15:47, on Zulip):

it would again be good to have action items at the end

pnkfelix (Jan 18 2019 at 15:47, on Zulip):

(and how tightly do we couple the team structure to the factoring of the compiler into separate libraries)

nikomatsakis (Jan 18 2019 at 15:47, on Zulip):

yeah, so, it seems like... depending on what we are trying to achieve, the agenda shifts a bit :)

pnkfelix (Jan 18 2019 at 15:48, on Zulip):

That last one presupposes the goal

nikomatsakis (Jan 18 2019 at 15:48, on Zulip):

i.e., if we know that we want to break compiler into repositories, then there might be questions about how to organize it etc. If we want to go with mono-repo, perhaps questions are different.

pnkfelix (Jan 18 2019 at 15:49, on Zulip):

okay. But team structuring seems to be a topic we'd like to tackle

nikomatsakis (Jan 18 2019 at 15:49, on Zulip):

it is actually kind of hard, I've started/stopped three messages already :)

nikomatsakis (Jan 18 2019 at 15:49, on Zulip):

trying to decide what are the "root goals" etc

mw (Jan 18 2019 at 15:50, on Zulip):

I think the compiler is already too big for any one person to have a complete grasp on. honoring this fact by adapting the architecture would also allow more focused teams.

nikomatsakis (Jan 18 2019 at 15:50, on Zulip):

I guess my root goal is to

nagisa (Jan 18 2019 at 15:50, on Zulip):

well, raising efficiency is an obvious one.

nikomatsakis (Jan 18 2019 at 15:50, on Zulip):

a lot of the things I want to do follow from "grow the team" though :)

nikomatsakis (Jan 18 2019 at 15:51, on Zulip):

I should probably add

too =)

nikomatsakis (Jan 18 2019 at 15:51, on Zulip):

ok, so, we are running short on time

centril (Jan 18 2019 at 15:51, on Zulip):

maintain => increase ;)

pnkfelix (Jan 18 2019 at 15:52, on Zulip):

Now now @centril , increasing quality is surely a P2 goal. :wink:

nikomatsakis (Jan 18 2019 at 15:53, on Zulip):

I feel like we made some good progress on the concept of how to approach IDE discusson.:

it seems like we consider organizational matters the next most important thing to discuss as a group. I feel a bit less clear on what the "agenda" will be, but maybe we'll feel our way, or else maybe we should have some more pre-discussion about it. I would at least be up for trying to prepare a kind of straw-person proposal.

nikomatsakis (Jan 18 2019 at 15:53, on Zulip):

I wonder -- maybe I can make a little "poll" for other technical matters and people can vote? I'm still curious about whether some kind of "breakout/review" structure could be a good way to e.g. have some details about MIR 2.0 emerging and so on.

nikomatsakis (Jan 18 2019 at 15:54, on Zulip):

in other words, I still hope we can leave the all hands with a pretty good idea of what 2019 holds in store for the compiler team

pnkfelix (Jan 18 2019 at 15:55, on Zulip):

maybe we need to actually make that a separate goal list

pnkfelix (Jan 18 2019 at 15:55, on Zulip):

2021 goals vs 2019 goals

pnkfelix (Jan 18 2019 at 15:55, on Zulip):

or something

nikomatsakis (Jan 18 2019 at 15:55, on Zulip):

yes, they are distinct

nikomatsakis (Jan 18 2019 at 15:56, on Zulip):

ideally one is sort of a subset of the other :)

pnkfelix (Jan 18 2019 at 15:56, on Zulip):

(or path towards)

nikomatsakis (Jan 18 2019 at 15:56, on Zulip):

right

nikomatsakis (Jan 18 2019 at 15:56, on Zulip):

maybe it's more realistic to emerge with a plan for how to formulate 2019 goals

nikomatsakis (Jan 18 2019 at 15:56, on Zulip):

ah well, we'll see

mw (Jan 18 2019 at 15:57, on Zulip):

or we go in the opposite direction in 2019 and let the yoyo effect catapult us towards perfection

nikomatsakis (Jan 18 2019 at 15:57, on Zulip):

(that reminds me, @mw, one topic I do think would be great is sync'ing up with @Josh Stone about the rayon-rustc fork etc)

QuietMisdreavus (Jan 18 2019 at 15:58, on Zulip):

or we go in the opposite direction in 2019 and let the yoyo effect catapult us towards perfection

a gravity assist by orbiting around our current position, i like it :thinking:

mw (Jan 18 2019 at 15:58, on Zulip):

yes, @QuietMisdreavus gets it

nikomatsakis (Jan 18 2019 at 16:02, on Zulip):

...and that's a wrap, I think! Thanks all! :heart:

blitzerr (Jan 21 2019 at 19:12, on Zulip):

@nikomatsakis I was just reading through this async, and I was thinking we should also focus on some sort of tracing library like dtrace. Debug-ability with minimum effort will lead to wide adoption of the language for large scale systems. I am not suggesting this be in rustc. It might be part of the toolchain rust ships with and compiler provides functionality if the tool needs it. Definitely not P1. Just a thought, I don't know what people think ?

Last update: Nov 16 2019 at 01:35UTC