Stream: t-compiler/wg-meta

Topic: meeting 2020.04.02


nikomatsakis (Apr 02 2020 at 18:04, on Zulip):

Hey @T-compiler/WG-meta -- meeting time =)

Santiago Pastorino (Apr 02 2020 at 18:04, on Zulip):

@T-compiler/WG-meta :wave:

Santiago Pastorino (Apr 02 2020 at 18:04, on Zulip):

ohhh :)

DPC (Apr 02 2020 at 18:04, on Zulip):

:wave:

nikomatsakis (Apr 02 2020 at 18:05, on Zulip):

I'm not sure what folks have on their minds, I'd like to maybe use this time to move that Major Change Proposal RFC to "completion" status

nikomatsakis (Apr 02 2020 at 18:06, on Zulip):

Did we ever post a status update in the compiler team meeting this morning? I've been "on the go" since it ended and haven't had a chance to check :)

Santiago Pastorino (Apr 02 2020 at 18:06, on Zulip):

I have another thing also, but maybe moving that forward is the best thing to do

Santiago Pastorino (Apr 02 2020 at 18:06, on Zulip):

nikomatsakis said:

Did we ever post a status update in the compiler team meeting this morning? I've been "on the go" since it ended and haven't had a chance to check :)

we did

nikomatsakis (Apr 02 2020 at 18:06, on Zulip):

Santiago Pastorino said:

I have another thing also, but maybe moving that forward is the best thing to do

what's the other thing?

nikomatsakis (Apr 02 2020 at 18:06, on Zulip):

I'm not sure if there are really any open questions left for the MCP RFC

Santiago Pastorino (Apr 02 2020 at 18:07, on Zulip):

I've mentioned it last meeting

Santiago Pastorino (Apr 02 2020 at 18:07, on Zulip):

the ideas stated in #t-compiler > active working groups

pnkfelix (Apr 02 2020 at 18:07, on Zulip):

@oli mentioned today that they don't know what "next steps" are for a MCP

Santiago Pastorino (Apr 02 2020 at 18:07, on Zulip):

in particular, I've summarized stuff here compiler-team#264

pnkfelix (Apr 02 2020 at 18:08, on Zulip):

but that could be because the RFC itself hasn't been merged

nikomatsakis (Apr 02 2020 at 18:09, on Zulip):

I'm writing some quick motivation on the RFC as we talk

nikomatsakis (Apr 02 2020 at 18:09, on Zulip):

I'm skimming over compiler-team#264 --

nikomatsakis (Apr 02 2020 at 18:09, on Zulip):

I basically agree with those points :)

nikomatsakis (Apr 02 2020 at 18:10, on Zulip):

maybe a good thing for today wouldl be to look at the list of working groups and kind of "finalize" the category we think they'd be in?

nikomatsakis (Apr 02 2020 at 18:10, on Zulip):

and/or have a few sentences of what those categories mean?

Santiago Pastorino (Apr 02 2020 at 18:10, on Zulip):

yes

Santiago Pastorino (Apr 02 2020 at 18:11, on Zulip):

maybe we can split working groups and all the different areas of the compiler that we want to track?

Santiago Pastorino (Apr 02 2020 at 18:11, on Zulip):

nikomatsakis said:

and/or have a few sentences of what those categories mean?

hmm what do you mean exactly by categories?

Santiago Pastorino (Apr 02 2020 at 18:12, on Zulip):

I think from the checklist I want:

nikomatsakis (Apr 02 2020 at 18:12, on Zulip):

maybe start with the categories? I see there as being three useful categories:

Santiago Pastorino (Apr 02 2020 at 18:12, on Zulip):

Go deeper and breakdown every compiler area and have dedicated chats for each one

Santiago Pastorino (Apr 02 2020 at 18:12, on Zulip):

and

Santiago Pastorino (Apr 02 2020 at 18:12, on Zulip):

Split out some areas of the compiler from working groups like Diagnostics, LLVM and Traits.

Santiago Pastorino (Apr 02 2020 at 18:12, on Zulip):

and a possible next step is to fix a bit better the experts map but that will take more time

nikomatsakis (Apr 02 2020 at 18:12, on Zulip):

I think we use working groups for all 3 of those right now

nikomatsakis (Apr 02 2020 at 18:13, on Zulip):

Santiago Pastorino said:

and a possible next step is to fix a bit better the experts map but that will take more time

yeah the expert map feels related to areas

Santiago Pastorino (Apr 02 2020 at 18:13, on Zulip):

exactly

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

did you make a list of working groups somewhere, @Santiago Pastorino ?

Santiago Pastorino (Apr 02 2020 at 18:14, on Zulip):

from those 3 things I think projects is fine for WG

Santiago Pastorino (Apr 02 2020 at 18:14, on Zulip):

areas shouldn't be included

Santiago Pastorino (Apr 02 2020 at 18:14, on Zulip):

and I have doubts about project groups

Santiago Pastorino (Apr 02 2020 at 18:15, on Zulip):

I meant, project groups aren't just areas?

Santiago Pastorino (Apr 02 2020 at 18:15, on Zulip):

the list is just the table here https://rust-lang.github.io/compiler-team/

nikomatsakis (Apr 02 2020 at 18:16, on Zulip):

I see "project groups" as being a subset of "projects", in some sense, and totally distinct from areas

nikomatsakis (Apr 02 2020 at 18:16, on Zulip):

project groups have weekly meetings, roadmaps, and plans

Santiago Pastorino (Apr 02 2020 at 18:16, on Zulip):
        "Async/Await",
        "Diagnostics",
        "Learning",
        "LLVM",
        "Meta",
        "MIR Optimizations",
        "Parallel Rustc",
        "Polonius",
        "Polymorphization",
        "RLS 2.0",
        "Self-Profile",
        "Traits",
Santiago Pastorino (Apr 02 2020 at 18:16, on Zulip):

WGs

nikomatsakis (Apr 02 2020 at 18:17, on Zulip):

areas don't necessarily have those things -- they may have vague plans, but they're not as organized

nikomatsakis (Apr 02 2020 at 18:17, on Zulip):

they might have "owners" but not "leads"

Santiago Pastorino (Apr 02 2020 at 18:17, on Zulip):

yeah

Santiago Pastorino (Apr 02 2020 at 18:17, on Zulip):

well you could have areas that are not organized and areas that are :)

nikomatsakis (Apr 02 2020 at 18:17, on Zulip):

I think that #t-compiler/wg-learning is probably a project group, though it might eventually become an area

Santiago Pastorino (Apr 02 2020 at 18:18, on Zulip):

but unsure if it's important what I'm saying :)

nikomatsakis (Apr 02 2020 at 18:18, on Zulip):

yeah idk sometimes the line is blurry ;)

Santiago Pastorino (Apr 02 2020 at 18:18, on Zulip):

nikomatsakis said:

I think that #t-compiler/wg-learning is probably a project group, though it might eventually become an area

yes

nikomatsakis (Apr 02 2020 at 18:18, on Zulip):

but I think a project group should be able to have an "end state" that it is shooting for

Santiago Pastorino (Apr 02 2020 at 18:18, on Zulip):

ideally all the areas should work as currently all project groups are working

Santiago Pastorino (Apr 02 2020 at 18:18, on Zulip):

with meetings, etc, etc

nikomatsakis (Apr 02 2020 at 18:18, on Zulip):

I think for dev-guide, there is some notion of the dev-guide being "complete", i.e., it has an overview, and covers the major area of the compiler, and that's the 'project' part of it

nikomatsakis (Apr 02 2020 at 18:19, on Zulip):

Santiago Pastorino said:

with meetings, etc, etc

I'm not sure this is true

nikomatsakis (Apr 02 2020 at 18:19, on Zulip):

I think e.g, the "MIR borrowck area" doesn't need regular meetings

nikomatsakis (Apr 02 2020 at 18:19, on Zulip):

the code is basically "done", it just needs some maintaince and maybe one-off meetings to resolve issues

Santiago Pastorino (Apr 02 2020 at 18:20, on Zulip):

the thing is ... should MIR borrowck be the area or should just be MIR?

Santiago Pastorino (Apr 02 2020 at 18:20, on Zulip):

anyway, I guess this is not important :)

nikomatsakis (Apr 02 2020 at 18:20, on Zulip):

I would say MIR borrowck, MIR is too broad

Santiago Pastorino (Apr 02 2020 at 18:20, on Zulip):

the 3 levels you've defined is fine

Santiago Pastorino (Apr 02 2020 at 18:20, on Zulip):

so we need to split areas out of WGs, right?

Santiago Pastorino (Apr 02 2020 at 18:21, on Zulip):

what we currently have is

Santiago Pastorino (Apr 02 2020 at 18:21, on Zulip):
        "Async/Await",
        "Learning",
        "Meta",
        "MIR Optimizations",
        "Parallel Rustc",
        "Polonius",
        "Polymorphization",
        "RLS 2.0",
        "Self-Profile",
        "Traits",
Santiago Pastorino (Apr 02 2020 at 18:21, on Zulip):

WGs

Santiago Pastorino (Apr 02 2020 at 18:22, on Zulip):

and LLVM, Diagnostics as areas

Santiago Pastorino (Apr 02 2020 at 18:22, on Zulip):

I meant, I think we should remove LLVM and Diagnostics from WGs

Santiago Pastorino (Apr 02 2020 at 18:22, on Zulip):

and define all the areas of the compiler

nikomatsakis (Apr 02 2020 at 18:22, on Zulip):

project groups

areas

it's complicated

nikomatsakis (Apr 02 2020 at 18:23, on Zulip):

this is my take on the above btw :)

Santiago Pastorino (Apr 02 2020 at 18:23, on Zulip):

yes

Santiago Pastorino (Apr 02 2020 at 18:23, on Zulip):

agreed

nikomatsakis (Apr 02 2020 at 18:24, on Zulip):

I think it'd be useful to have #t-compiler/traits and #t-compiler/traits/project-chalk, something like that

nikomatsakis (Apr 02 2020 at 18:24, on Zulip):

it might actually help because right now #wg-traits is (imo) a bit "broad"

nikomatsakis (Apr 02 2020 at 18:25, on Zulip):

or -- more to the point -- it's de facto more of a "chalk discussion" area, but the name suggests it'd be more

Santiago Pastorino (Apr 02 2020 at 18:25, on Zulip):

yeah that sounds good

nikomatsakis (Apr 02 2020 at 18:25, on Zulip):

I feel like that means other non-chalk-related discussions kind of get "lesser status" than perhaps they ought to

Santiago Pastorino (Apr 02 2020 at 18:25, on Zulip):

what should we do first?

nikomatsakis (Apr 02 2020 at 18:25, on Zulip):

curious what you think, @pnkfelix, does the above kind of fit your expectations?

Santiago Pastorino (Apr 02 2020 at 18:25, on Zulip):

I think maybe the first step is to fix the working group situation

nikomatsakis (Apr 02 2020 at 18:26, on Zulip):

Santiago Pastorino said:

what should we do first?

I would say -- let's try to write-up a kind of proposal and spell out the implications -- doesn't have to be super long, probably start from my notes above ?

nikomatsakis (Apr 02 2020 at 18:26, on Zulip):

and then sketch out "here is what the zulip channels would look like"

Santiago Pastorino (Apr 02 2020 at 18:26, on Zulip):

I meant, I'd just remove from the list what should not be included as a WG

nikomatsakis (Apr 02 2020 at 18:26, on Zulip):

you mean from the list on the compiler-team page?

nikomatsakis (Apr 02 2020 at 18:26, on Zulip):

I think that's fine, we can do that

pnkfelix (Apr 02 2020 at 18:26, on Zulip):

I'm not sure what my expectations are

Santiago Pastorino (Apr 02 2020 at 18:27, on Zulip):

nikomatsakis said:

you mean from the list on the compiler-team page?

yes

pnkfelix (Apr 02 2020 at 18:27, on Zulip):

though it seems like, in general, Project Groups should be striving to become Areas. :)

nikomatsakis (Apr 02 2020 at 18:27, on Zulip):

yes, that's probably true

nikomatsakis (Apr 02 2020 at 18:27, on Zulip):

(or replace an area)

pnkfelix (Apr 02 2020 at 18:27, on Zulip):

right

nikomatsakis (Apr 02 2020 at 18:27, on Zulip):

(or get integrated into an area)

pnkfelix (Apr 02 2020 at 18:27, on Zulip):

e.g. AST-borrowck was an area. NLL was a project. Now MIR-borrowck is an area.

nikomatsakis (Apr 02 2020 at 18:27, on Zulip):

I think e.g. you could have a project-inlining or something in the "MIR Optimizations" area

nikomatsakis (Apr 02 2020 at 18:28, on Zulip):

basically I feel like projects can go "inside" areas (but maybe not, if they don't neatly fit...)

nikomatsakis (Apr 02 2020 at 18:28, on Zulip):

perhaps, if they don't neatly fit, then this suggests they will become a "fresh area"

pnkfelix (Apr 02 2020 at 18:28, on Zulip):

but the big idea I was trying to put forward, which I'm not sure if anyones articulated,

nikomatsakis (Apr 02 2020 at 18:28, on Zulip):

e.g., polonius

pnkfelix (Apr 02 2020 at 18:28, on Zulip):

is that project groups should be striving to make themselves unnecessary

nikomatsakis (Apr 02 2020 at 18:29, on Zulip):

that plus, I think: have a defined scope at any given time, though I think it's ok for the scope to "extend itself" over time

nikomatsakis (Apr 02 2020 at 18:29, on Zulip):

this is where I was vacillating for "wg-meta"

Santiago Pastorino (Apr 02 2020 at 18:29, on Zulip):

pnkfelix said:

is that project groups should be striving to make themselves unnecessary

yeah it is in that sense that I was thinking that both things are the same or at least very similar

nikomatsakis (Apr 02 2020 at 18:29, on Zulip):

I think the line gets blurriest for those things that are trying to do work that might become a "new area" in the end

nikomatsakis (Apr 02 2020 at 18:29, on Zulip):

polonius, rls 2.0, learning/rustc-dev-guide

nikomatsakis (Apr 02 2020 at 18:30, on Zulip):

(if the area was "borrow check" then I guess polonius is clearly a kind of "project within that", with the goal of replacing MIR borrowck)

Santiago Pastorino (Apr 02 2020 at 18:30, on Zulip):

maybe the question is what we consider important to have and why?

Santiago Pastorino (Apr 02 2020 at 18:30, on Zulip):

I meant, maybe it's not really important if wg-meta is a wg or an area

nikomatsakis (Apr 02 2020 at 18:30, on Zulip):

I think we should make a simple rule:

nikomatsakis (Apr 02 2020 at 18:30, on Zulip):

if you have regular meetings, you are a project :)

Santiago Pastorino (Apr 02 2020 at 18:31, on Zulip):

what I feel the lack of is mainly areas and things around github to contemplate those

pnkfelix (Apr 02 2020 at 18:31, on Zulip):

or a team

nikomatsakis (Apr 02 2020 at 18:31, on Zulip):

or some "structure" beyond "jump in to the channel and say hi"

Santiago Pastorino (Apr 02 2020 at 18:31, on Zulip):

like being able to ping a team

nikomatsakis (Apr 02 2020 at 18:31, on Zulip):

Santiago Pastorino said:

what I feel the lack of is mainly areas and things around github to contemplate those

yes

nikomatsakis (Apr 02 2020 at 18:31, on Zulip):

so maybe establishing the areas is most urgent, that makes sense

Santiago Pastorino (Apr 02 2020 at 18:31, on Zulip):

yes

Santiago Pastorino (Apr 02 2020 at 18:31, on Zulip):

areas and people that belong to those

nikomatsakis (Apr 02 2020 at 18:32, on Zulip):

speaking of the expert map,

Santiago Pastorino (Apr 02 2020 at 18:32, on Zulip):

so from issues it's easy to cc @t-compiler/llvm or @t-compiler/diagnostics and things like that

nikomatsakis (Apr 02 2020 at 18:32, on Zulip):

I think we should reflect the areas in the rust-lang/team repo in some way

nikomatsakis (Apr 02 2020 at 18:32, on Zulip):

in particular it's the one that manages github teams

Santiago Pastorino (Apr 02 2020 at 18:32, on Zulip):

yes

nikomatsakis (Apr 02 2020 at 18:32, on Zulip):

we might do it like "ICEbreaker" groups, where the main thing is to be able to "ping" people from that area

Santiago Pastorino (Apr 02 2020 at 18:32, on Zulip):

that's what the issue I've opened says

nikomatsakis (Apr 02 2020 at 18:32, on Zulip):

and also it's easy for people to add themselves

Santiago Pastorino (Apr 02 2020 at 18:32, on Zulip):

Define a github group so we can ping the maintainers of each area, this could be integrated in team repo and maybe the whole experts map can be integrated as maintainer attribute into team repo

nikomatsakis (Apr 02 2020 at 18:33, on Zulip):

(because there aren't real implications for access rights, you don't become an org member, etc, so the bar is low)

nikomatsakis (Apr 02 2020 at 18:33, on Zulip):

or they can be "real teams", where the bar is mildly higher

nikomatsakis (Apr 02 2020 at 18:33, on Zulip):

I'd be fine either way really

Santiago Pastorino (Apr 02 2020 at 18:33, on Zulip):

I'd kill experts map and add areas and maintainers

Santiago Pastorino (Apr 02 2020 at 18:34, on Zulip):

I can talk to @Pietro Albini to see how this can be implemented and if he feels that's a good idea

nikomatsakis (Apr 02 2020 at 18:34, on Zulip):

nikomatsakis said:

or they can be "real teams", where the bar is mildly higher

by real teams, I basically mean the equivalent of "compiler contributors" -- i.e., you get nominated, and when you're added, you become added to github teams and other things

Santiago Pastorino (Apr 02 2020 at 18:34, on Zulip):

@nikomatsakis @pnkfelix do we want to make this step then? "kill experts and add areas/maintainers to team"

nikomatsakis (Apr 02 2020 at 18:34, on Zulip):

it seems like we mostly have the tools we need, we just have to decide what it means to be added to an area

Pietro Albini (Apr 02 2020 at 18:34, on Zulip):

cc @simulacrum

nikomatsakis (Apr 02 2020 at 18:35, on Zulip):

Santiago Pastorino said:

nikomatsakis pnkfelix do we want to make this step then? "kill experts and add areas/maintainers to team"

I think that's good, but I'm not sure just how we want to model yet

Pietro Albini (Apr 02 2020 at 18:35, on Zulip):

I'd like to just use github's CODEOWNERS

nikomatsakis (Apr 02 2020 at 18:35, on Zulip):

what's that

nikomatsakis (Apr 02 2020 at 18:35, on Zulip):

but that sounds plausible :)

Pietro Albini (Apr 02 2020 at 18:35, on Zulip):

https://help.github.com/en/github/creating-cloning-and-archiving-repositories/about-code-owners

nikomatsakis (Apr 02 2020 at 18:36, on Zulip):

btw, I think that -- if library-ification ever develops as I hope -- then these "areas" would map to real libraries eventually

Santiago Pastorino (Apr 02 2020 at 18:36, on Zulip):

it would be even nice if we have a set of directories that belong to an area

Pietro Albini (Apr 02 2020 at 18:36, on Zulip):

yep that's what codeowners does

Santiago Pastorino (Apr 02 2020 at 18:36, on Zulip):

so when a PR comes we may do some heuristic and cc the appropriate team

Santiago Pastorino (Apr 02 2020 at 18:36, on Zulip):

ahh cool :)

nikomatsakis (Apr 02 2020 at 18:36, on Zulip):

yeah, that sounds good

nikomatsakis (Apr 02 2020 at 18:36, on Zulip):

I do agree this should map to the files in the code

Pietro Albini (Apr 02 2020 at 18:36, on Zulip):

and we could integrate it in the triagebot auto-assignment when it will be implemented

Santiago Pastorino (Apr 02 2020 at 18:37, on Zulip):

yep

nikomatsakis (Apr 02 2020 at 18:37, on Zulip):

very interesting

nikomatsakis (Apr 02 2020 at 18:37, on Zulip):

ok, so, let's make "figure out areas and how they sdhould work" the immediate goal, and the question of how to "recategorize" non-working groups is sort of secondary follow-up?

Santiago Pastorino (Apr 02 2020 at 18:38, on Zulip):

:+1:

Santiago Pastorino (Apr 02 2020 at 18:39, on Zulip):

I guess I could start a little document with areas, directories and possible maintainers

Santiago Pastorino (Apr 02 2020 at 18:39, on Zulip):

but I remember we had some difficulties filling up experts map

Santiago Pastorino (Apr 02 2020 at 18:39, on Zulip):

anyway, will start with something and share

DPC (Apr 02 2020 at 19:12, on Zulip):

is the meeting still on? :P

Santiago Pastorino (Apr 02 2020 at 19:21, on Zulip):

no :)

centril (Apr 02 2020 at 22:36, on Zulip):

I feel like const-eval is also in "it's complicated" territory

centril (Apr 02 2020 at 22:37, on Zulip):

while it is an area, it's also like async/await, with specific goals that need to be achieved

centril (Apr 02 2020 at 22:37, on Zulip):

like control flow and whatnot

Last update: Jan 22 2021 at 13:00UTC