Stream: t-compiler/meetings

Topic: [steering meeting] roadmap 2020-05-29 compiler-team#287


pnkfelix (May 29 2020 at 13:43, on Zulip):

Hi @T-compiler/meeting ; we'll be having a steering meeting to discuss the 2020-2021 roadmap in about 17 minutes.

nikomatsakis (May 29 2020 at 14:00, on Zulip):

Hello @T-compiler/meeting -- roadmap discussion starting imminently -- base hackmd document

But let's kick it off with a few minutes for

Announcements

pnkfelix (May 29 2020 at 14:00, on Zulip):
nikomatsakis (May 29 2020 at 14:00, on Zulip):

* Also I believe that before we truly begin there was an item on the agenda (#72545)

pnkfelix (May 29 2020 at 14:00, on Zulip):

(I don't know if that's an annoucment .. jinx!)

pnkfelix (May 29 2020 at 14:01, on Zulip):

also I probably should have discussed this other bug yesterday: 'rustc' panicked at 'failed to lookup SourceFile in new context' #70924

pnkfelix (May 29 2020 at 14:01, on Zulip):

which is a stable-to-beta regression that ... I think I now have a fix for ... but I don't know whether I should be rushing for a backport, or just let it go into stable for now

simulacrum (May 29 2020 at 14:02, on Zulip):

it's a pretty bad incremental bug

simulacrum (May 29 2020 at 14:03, on Zulip):

I'd backport probably if we have good confidence in the fix not breaking more (even if we're unsure it fixes it)

pnkfelix (May 29 2020 at 14:03, on Zulip):

@eddyb 's theory is that it only is exposed when you add/remove the rust-src component

pnkfelix (May 29 2020 at 14:03, on Zulip):

which I would have thought happens very rarely

pnkfelix (May 29 2020 at 14:03, on Zulip):

(updates to that component may or may not also affect it, to be clear. Specifically updates that add/remove file paths)

pnkfelix (May 29 2020 at 14:03, on Zulip):

but it may not be as rare as you might think, because RLS and rust-analyzer install that compoment

eddyb (May 29 2020 at 14:04, on Zulip):

the problem isn't the files inside, it's the presence of the component as a whole

pnkfelix (May 29 2020 at 14:04, on Zulip):

anyway I don't want to railroad the meeting

pnkfelix (May 29 2020 at 14:05, on Zulip):

if @simulacrum thinks its worth prioritizing for the beta, then I can coordinate with @simulacrum and hopefully @eddyb on getting my fix in

nikomatsakis (May 29 2020 at 14:05, on Zulip):

sounds good

nikomatsakis (May 29 2020 at 14:05, on Zulip):

I guess let's discuss #72545

simulacrum (May 29 2020 at 14:07, on Zulip):

what is there to discuss? a revert seems like a clear choice?

simulacrum (May 29 2020 at 14:07, on Zulip):

(why would we not?)

nikomatsakis (May 29 2020 at 14:07, on Zulip):

I'm trying to refresh my memory of what this even is

pnkfelix (May 29 2020 at 14:07, on Zulip):

actually I cannot remember why I said we needed to discuss this today instead of next week

nikomatsakis (May 29 2020 at 14:07, on Zulip):

ok, right, so we changed some details of procedural macro , which broke some crates, the consensus seems to be that

nikomatsakis (May 29 2020 at 14:08, on Zulip):

I also think that a revert plus a more "peaceful" landing makes sense

nikomatsakis (May 29 2020 at 14:09, on Zulip):

I'm not sure @pnkfelix if there are timing reasons we ought to be discussing this now, I guess reverting gets increasingly difficut

simulacrum (May 29 2020 at 14:09, on Zulip):

it seems like there's not anything urgent here? maybe we should table it till thursday or so?

simulacrum (May 29 2020 at 14:09, on Zulip):

(I can r+ a revert myself I guess?)

pnkfelix (May 29 2020 at 14:09, on Zulip):

Is anyone objecting to a revert (and more peaceful landing)?

nikomatsakis (May 29 2020 at 14:09, on Zulip):

Well I don't know if @Aaron Hill or @Vadim Petrochenkov are here =)

nikomatsakis (May 29 2020 at 14:09, on Zulip):

Anyway, ok, let's at least move it to a parallel topic I guess and we can discuss a bit async

nikomatsakis (May 29 2020 at 14:10, on Zulip):

and/or next week

pnkfelix (May 29 2020 at 14:10, on Zulip):

okay. Lets maybe try to move forward with a revert and loop them in on the appropriate PR and/or issue.

nikomatsakis (May 29 2020 at 14:10, on Zulip):

(as usual, I'll repeat what I often say at times like this, which is that I think a "no judgement revert" is something we should probably be doing more often)

nikomatsakis (May 29 2020 at 14:10, on Zulip):

(that said, it does seem like the ecosystem is largely adapting)

nikomatsakis (May 29 2020 at 14:11, on Zulip):

Right, so, for the main meeting...

nikomatsakis (May 29 2020 at 14:12, on Zulip):

So we have this hackmd that tries to sketch out a roadmap of sorts. I think the most useful structure for this meeting would be to look at the "top goals"

nikomatsakis (May 29 2020 at 14:12, on Zulip):

and to use them to judge a bit what "major projects" we think we should be focusing on

nikomatsakis (May 29 2020 at 14:13, on Zulip):

it's a bit tricky to manage this, since the nature of open source seems to lend itself to a lot of ongoing things, but I think there is a general sense (it was clear in the survey, for one thing) that we are spreading ourselves too thin

pnkfelix (May 29 2020 at 14:13, on Zulip):

there's nothing in the top goals about incremental. That doesn't surprise me, since I didn't write any down

nikomatsakis (May 29 2020 at 14:13, on Zulip):

and advancing too many things at once

nikomatsakis (May 29 2020 at 14:13, on Zulip):

I certainly feel that a bit, I'd like to pick some things, get them done, then do the next things...

pnkfelix (May 29 2020 at 14:13, on Zulip):

at least, not in the "2021 (Edition)" part

nikomatsakis (May 29 2020 at 14:13, on Zulip):

yep, that is one of things I want to talk about!

nikomatsakis (May 29 2020 at 14:13, on Zulip):

there are at least two projects I think worth discussing

pnkfelix (May 29 2020 at 14:14, on Zulip):

also, @Nicholas Nethercote had mentioned that debugging is a pain... I think they said debugging rustc itself, right?

pnkfelix (May 29 2020 at 14:14, on Zulip):

(or were they talking about debugging rust code embedded in Firefox?)

nikomatsakis (May 29 2020 at 14:14, on Zulip):

put another way, I think it's clear we need to put some energy behind ways to improve compilation time, but that is a multi-headed beast and I think it's good to talk about which thing is worth focusing on first

nikomatsakis (May 29 2020 at 14:15, on Zulip):

(one thing I've also learned is that it's not always best to try and write down what "should happen" and sometimes better to start from "what do people want to pursue", and try to arrange those things into order (and maybe cull some of them))

nikomatsakis (May 29 2020 at 14:15, on Zulip):

anyway, should we talk about the goals and incremental?

pnkfelix (May 29 2020 at 14:15, on Zulip):

you probably won't be surprised to hear me say this, (again), but given that triage work and wg-prioritzation is on here

pnkfelix (May 29 2020 at 14:16, on Zulip):

we could just add a "spin up wg-incr-comp"; which you know well is something I'm planning to spend time on in very near future.

pnkfelix (May 29 2020 at 14:16, on Zulip):

so maybe that's a simple way to say "that's what I want to pursue"

nikomatsakis (May 29 2020 at 14:17, on Zulip):

OK so

nikomatsakis (May 29 2020 at 14:17, on Zulip):

I'm not sure how much we can/should try to narrow down goals -- probably not in this meeting --

nikomatsakis (May 29 2020 at 14:18, on Zulip):

I mean within that wg --

nikomatsakis (May 29 2020 at 14:18, on Zulip):

but I think there is some tension between pushing on incremental and (e.g.) pushing on parallel compilation, and I think we shouldn't do both at once

nikomatsakis (May 29 2020 at 14:19, on Zulip):

but rather try to see if we can't build momentum, ideally by having multiple leads to a WG, behind one effort

nikomatsakis (May 29 2020 at 14:19, on Zulip):

does that make sense?

nikomatsakis (May 29 2020 at 14:19, on Zulip):

I also think it woudl be good to talk a bit about what we mean when we say "wg-incr-comp", I guess

simulacrum (May 29 2020 at 14:19, on Zulip):

IMO incremental is at this point more likely to see wins and should be prioritized over parallel. not doing work is always going to be better than (maybe) doing it in parallel

nikomatsakis (May 29 2020 at 14:20, on Zulip):

I tend to agree with that, but I'm also thinking about it from a "scenarios" POV

pnkfelix (May 29 2020 at 14:20, on Zulip):

yes. Long term parallel will need to be addressed. But short term incremental is going to provide more wins for more people in more contexts.

nikomatsakis (May 29 2020 at 14:20, on Zulip):

I think there are kind of 3 scenarios

nikomatsakis (May 29 2020 at 14:21, on Zulip):

anecdotally, I've been hearing more and more about CI builds, but then I'm not sure that parallel helps with that effort anyway, since most CI systems are not so beefy

nikomatsakis (May 29 2020 at 14:21, on Zulip):

incremental seems to clearly help with the other two

Esteban K├╝ber (May 29 2020 at 14:21, on Zulip):

Stateful CI can potentially benefit from incremental builds

nikomatsakis (May 29 2020 at 14:21, on Zulip):

(I think it's important that it dovetails with the work around rust-analyzer and so forth, but that's probably for later)

nikomatsakis (May 29 2020 at 14:22, on Zulip):

yeah, that makes sense, though I don't know of a lot of people doing that (I guess maybe things like sccache)

pnkfelix (May 29 2020 at 14:22, on Zulip):

yes, if incremental compiles were reliable enough that stateful CI could actually trust its results

pnkfelix (May 29 2020 at 14:22, on Zulip):

then presumably you'd get faster feedback about erroneous states

nikomatsakis (May 29 2020 at 14:23, on Zulip):

ok so it seems like the idea of trying to create a working group focused on incremental is "popular", so to speak

nikomatsakis (May 29 2020 at 14:23, on Zulip):

I'll add it to the list, though I think it'd be productive to think about the "Goals for 2021" (what do we expect to accomplish)

Wesley Wiser (May 29 2020 at 14:23, on Zulip):

I think it's interesting that we have cargo build and cargo build --release but no cargo build --ci. I don't think we even have official documentation about what settings you might want to tweak in a CI environment.

Wesley Wiser (May 29 2020 at 14:24, on Zulip):

(Not trying to jump to solutions but I think there's a gap in terms of our CI support)

pnkfelix (May 29 2020 at 14:25, on Zulip):

to be fair, my own philosophy is that every deviation of a CI environment from that of a "typical dev env" represents a ... well, maybe not flaw, but certainly a stumbling block

Wesley Wiser (May 29 2020 at 14:25, on Zulip):

Stuff works so it's not like anything is currently broken but learning what to tweak to minimize CI time for your project is kind of an arcane art.

matklad (May 29 2020 at 14:26, on Zulip):

(Not trying to jump to solutions but I think there's a gap in terms of our CI support)

The biggest one is that caching /target is painful. Ie, I belive the biggest impact for improving CI times would be via a certain Cargo feature (changing layout of the /target dir to separate deps from your project)

Wesley Wiser (May 29 2020 at 14:26, on Zulip):

I think there's certainly merit to that idea. I think there's some simple things people could probably do to get better CI performance though. If you don't do stateful CI, turning off incremental compilation for cargo build can save some time and disk space.

pnkfelix (May 29 2020 at 14:27, on Zulip):

matklad said:

(Not trying to jump to solutions but I think there's a gap in terms of our CI support)

The biggest one is that caching /target is painful. Ie, I belive the biggest impact for improving CI times would be via a certain Cargo feature (changing layout of the /target dir to separate deps from your project)

do you have a link for this?

nikomatsakis (May 29 2020 at 14:27, on Zulip):

This is an interesting topic, and probably worth digging into, though one intersting thoght about is that it's kind of "cross-cutting" beyond rustc

pnkfelix (May 29 2020 at 14:27, on Zulip):

the caching /target probably relates to sccache

nikomatsakis (May 29 2020 at 14:27, on Zulip):

I do agree that making small tweaks to improve and simplify rust's integration into CI could be very impactful on people

nikomatsakis (May 29 2020 at 14:27, on Zulip):

but probably the details are worth moving to a separate topic?

nikomatsakis (May 29 2020 at 14:27, on Zulip):

I'm curious to discuss a few other sort of "roadmap-y" items

pnkfelix (May 29 2020 at 14:27, on Zulip):

another potential roadmap item, maybe

pnkfelix (May 29 2020 at 14:28, on Zulip):

or, wait

pnkfelix (May 29 2020 at 14:28, on Zulip):

/me needs to go review something, never mind

nikomatsakis (May 29 2020 at 14:28, on Zulip):

Maybe

matklad (May 29 2020 at 14:28, on Zulip):

@pnkfelix https://github.com/rust-lang/cargo/issues/5885

nikomatsakis (May 29 2020 at 14:28, on Zulip):

One thing I wanted to bring up then is

parallel compilation

nikomatsakis (May 29 2020 at 14:28, on Zulip):

maybe we could/should have a separate discussion, but my sense is that this is not on our upcoming roadmap

nikomatsakis (May 29 2020 at 14:29, on Zulip):

this I guess comes back to the question of how much we want to push on many things at once :)

pnkfelix (May 29 2020 at 14:29, on Zulip):

so if we don't put parallel comp on the roadmap

pnkfelix (May 29 2020 at 14:30, on Zulip):

then the biggest potential problem I could see is that we make many steps backwards

pnkfelix (May 29 2020 at 14:30, on Zulip):

i.e. I don't think its an issue if we wait until 2021 or 2023 to attack parallel comp in earnest

nikomatsakis (May 29 2020 at 14:31, on Zulip):

right, so previously we had this plan to kind of "turn it on", monitor for regressions, etc

pnkfelix (May 29 2020 at 14:31, on Zulip):

but if we make our starting situation significantly worse in the meantime, that would be unfortunate

nikomatsakis (May 29 2020 at 14:31, on Zulip):

in some sense we were "quite close"

nikomatsakis (May 29 2020 at 14:31, on Zulip):

but in another sense I feel like we never got momentum somehow

nikomatsakis (May 29 2020 at 14:31, on Zulip):

I guess I'm putting it out there that I want to pull back

nikomatsakis (May 29 2020 at 14:31, on Zulip):

like, personally

nikomatsakis (May 29 2020 at 14:31, on Zulip):

I wasn't all that involved in parallelization anyway

simulacrum (May 29 2020 at 14:31, on Zulip):

IMO parallel doesn't need to be on the roadmap to potentially make progress

nikomatsakis (May 29 2020 at 14:32, on Zulip):

I don't know if I agree with taht

nikomatsakis (May 29 2020 at 14:32, on Zulip):

I think that's generally true

nikomatsakis (May 29 2020 at 14:32, on Zulip):

the roadmap doesn't need to be exhaustive

nikomatsakis (May 29 2020 at 14:32, on Zulip):

but parallelization feels to me like a fairly large effort

simulacrum (May 29 2020 at 14:32, on Zulip):

if we actually want wins, sure

simulacrum (May 29 2020 at 14:32, on Zulip):

but if it's just about making sure it doesn't regress too much, eh, maybe not so much

simulacrum (May 29 2020 at 14:32, on Zulip):

like I don't think we lose anything leaving the support we have today in

nikomatsakis (May 29 2020 at 14:33, on Zulip):

well I guess there are 3 questions

nikomatsakis (May 29 2020 at 14:33, on Zulip):

yes ok so leaving the current support "in stasis" seems viable

nikomatsakis (May 29 2020 at 14:33, on Zulip):

I don't know that parallelization really interferes that much at this point

nikomatsakis (May 29 2020 at 14:33, on Zulip):

so I guess I do think it's plausible it could move forward, but there's definitely some opportunity cost

Santiago Pastorino (May 29 2020 at 14:34, on Zulip):

nikomatsakis said:

well I guess there are 3 questions

not driving it but I'd be willing to be engaged with it

nikomatsakis (May 29 2020 at 14:34, on Zulip):

(i.e. say that @simulacrum and @Santiago Pastorino were to focus on it (since you two were heavily involved), that's time you're not spending on something else, and maybe it'd be better to put that energy into extending and improving incremental)

simulacrum (May 29 2020 at 14:34, on Zulip):

sure, yes.

Santiago Pastorino (May 29 2020 at 14:34, on Zulip):

yes!

simulacrum (May 29 2020 at 14:34, on Zulip):

I personally want to spend time getting our profiling into shape

nikomatsakis (May 29 2020 at 14:35, on Zulip):

what is profiling?

simulacrum (May 29 2020 at 14:35, on Zulip):

perf.rlo

nikomatsakis (May 29 2020 at 14:35, on Zulip):

interesting, ok

simulacrum (May 29 2020 at 14:35, on Zulip):

like, we have a good start, but I think to support incremental we can do more

simulacrum (May 29 2020 at 14:35, on Zulip):

(Along with other projects etc)

nikomatsakis (May 29 2020 at 14:35, on Zulip):

one thing I did want to also ask was about self-profiling

nikomatsakis (May 29 2020 at 14:35, on Zulip):

which seems related

simulacrum (May 29 2020 at 14:36, on Zulip):

yep, I see that as fitting into my desire to work with perf.rlo, to be clear

pnkfelix (May 29 2020 at 14:36, on Zulip):

@simulacrum do you think there are broad shortcomings in perf.rlo besides "support incremental" (not 100% I know what that means, beyond increasing our range of incremental benchmarks maybe)

nikomatsakis (May 29 2020 at 14:36, on Zulip):

obviously there is more things that could be done there, but I have the sense it reached a good point, and I wasn't sure how much momentum there was to push to another milestone (I'm especially thinking about improving the ability to guide people to "here are things to change to improve compilation time")

simulacrum (May 29 2020 at 14:36, on Zulip):

e.g. disk space is entirely missing

simulacrum (May 29 2020 at 14:37, on Zulip):

mir-opt's efforts on codegen units don't get surfaced well

simulacrum (May 29 2020 at 14:37, on Zulip):

I don't think there's that much but I think it is a good amount of work

nikomatsakis (May 29 2020 at 14:37, on Zulip):

OK, so, to back up, it seems like there is some "rough consensus" around the idea of parallel compilation being in "stasis" -- and in particular that if we were going to drive it, I think it would require a real assessment of the plan and to have enough committed people

nikomatsakis (May 29 2020 at 14:38, on Zulip):

we can now turn to profiling, that's an interesting addition to the roadmap list that I didn't have :)

nikomatsakis (May 29 2020 at 14:38, on Zulip):

also, one more bit of context I didn't give, I'm sort of hoping that this meeting (and some follow-up) will provide enough fodder to be able to revise our list of "working groups" and create more focused "project groups" listing

nikomatsakis (May 29 2020 at 14:39, on Zulip):

it feels like we've gotten closer but not yet "arrived" at the point where we can actually point people to the list of "ongoing major projects" that we are working on and who's working on them and if they would like help...

simulacrum (May 29 2020 at 14:39, on Zulip):

I think profiling work is more about helping other teams -- e.g., maybe we should be showing why we failed to incr-comp more than just "well we did"

simulacrum (May 29 2020 at 14:40, on Zulip):

(s/teams/project groups/)

nikomatsakis (May 29 2020 at 14:42, on Zulip):

I guess this is a combination of gathering more data

nikomatsakis (May 29 2020 at 14:42, on Zulip):

and exposing it

nikomatsakis (May 29 2020 at 14:42, on Zulip):

do you also see "end users" as "in scope"?

nikomatsakis (May 29 2020 at 14:42, on Zulip):

sounds like that would be more of a secondary focus

simulacrum (May 29 2020 at 14:42, on Zulip):

secondary focus

nikomatsakis (May 29 2020 at 14:43, on Zulip):

ok

nikomatsakis (May 29 2020 at 14:44, on Zulip):

Haev you talked much to @eddyb :)

nikomatsakis (May 29 2020 at 14:44, on Zulip):

I know they've been doing some investigations here..

simulacrum (May 29 2020 at 14:44, on Zulip):

yeah, we've been talking a bit

simulacrum (May 29 2020 at 14:44, on Zulip):

I think profiling is well in hand to some extent

nikomatsakis (May 29 2020 at 14:44, on Zulip):

OK, well, I'm not sure what more there is to say, but I've added it to the list, maybe you want to add some notion of Goal(s) there?

simulacrum (May 29 2020 at 14:44, on Zulip):

yep

nikomatsakis (May 29 2020 at 14:45, on Zulip):

so we're 45 minutes in

nikomatsakis (May 29 2020 at 14:45, on Zulip):

one thing I realize that's not on this list

nikomatsakis (May 29 2020 at 14:45, on Zulip):

I've been looking into trying to make a plan around "finish up impl Trait "

nikomatsakis (May 29 2020 at 14:45, on Zulip):

whatever that means ;)

nikomatsakis (May 29 2020 at 14:46, on Zulip):

it seems pretty clearly though to be on a list of "almost done and really useful" features

nikomatsakis (May 29 2020 at 14:46, on Zulip):

it's going to require some effort though, and I don't know how realistic it is to balance with other efforts

oli (May 29 2020 at 14:46, on Zulip):

I still have a bunch of bugs there assigned to me that are definitive blockers

nikomatsakis (May 29 2020 at 14:46, on Zulip):

Yeah, I've been looking over the list and doing some experimentation

pnkfelix (May 29 2020 at 14:46, on Zulip):

and you don't have much free time in short term, right @oli ?

oli (May 29 2020 at 14:47, on Zulip):

yea, but I've also been avoiding them forever

nikomatsakis (May 29 2020 at 14:47, on Zulip):

@Matthew Jasper and I are supposed to find a time to discuss as well, maybe @oli we'll cc you in :)

nikomatsakis (May 29 2020 at 14:47, on Zulip):

I didn't really plan to pin this on you though

nikomatsakis (May 29 2020 at 14:47, on Zulip):

I figured you'd done your shift

oli (May 29 2020 at 14:47, on Zulip):

if we get some structure there and a plan, I want to be in

nikomatsakis (May 29 2020 at 14:47, on Zulip):

ok so --

nikomatsakis (May 29 2020 at 14:47, on Zulip):

I think my ideal role would be to try and form structure

nikomatsakis (May 29 2020 at 14:47, on Zulip):

and step back from the actual coding :)

nikomatsakis (May 29 2020 at 14:47, on Zulip):

so maybe that works well

nikomatsakis (May 29 2020 at 14:48, on Zulip):

I was going to bring up MIR optimizations

nikomatsakis (May 29 2020 at 14:48, on Zulip):

(but I guess I'm curious if anybody else attending has some idea of a project they'd like to talk a bit about)

Wesley Wiser (May 29 2020 at 14:49, on Zulip):

I'd like to hear your/the team's thoughts about MIR optimization but I'd just like to throw out that I'd like to see more public planning about how to align what is being worked on with the overall Rust yearly roadmap.

nikomatsakis (May 29 2020 at 14:50, on Zulip):

So, the overall roadmap for this year was basically "pay down debt"

Wesley Wiser (May 29 2020 at 14:50, on Zulip):

I think one of the things I really liked about the 2018 impl period was that it created a really smooth ramp to get people involved with making the project goals happen.

nikomatsakis (May 29 2020 at 14:50, on Zulip):

But I am trying to think about how to "knit together" some of the plans from various groups

nikomatsakis (May 29 2020 at 14:50, on Zulip):

and I think it's itneresting that even in this discussion we hit on things that overlap with cargo

nikomatsakis (May 29 2020 at 14:50, on Zulip):

e.g., how to improve the CI integration experience

Wesley Wiser (May 29 2020 at 14:51, on Zulip):

Even if there were not so great parts about it like the timebox for implementation. If we could find a way to capture the spirit of that without the negative effects, I think we could get a lot more community involvement like we did then.

Wesley Wiser (May 29 2020 at 14:51, on Zulip):

Anyway, that's my digression :slight_smile:

nikomatsakis (May 29 2020 at 14:52, on Zulip):

It's not really a digression, it seems useful, and I'm happy to talk a bit more about it. It's kind of been a goal for me, though, but I figured a good place to start would be getting a better handle on each team's perspective.

nikomatsakis (May 29 2020 at 14:52, on Zulip):

One thing worth noting is that a lot of the goals on the compiler team list (GATs, impl Trait, etc) are also goals for lang team and require active design work, but not all,

nikomatsakis (May 29 2020 at 14:53, on Zulip):

the other thing I realize that's not on this list at all is some stuff like

nikomatsakis (May 29 2020 at 14:53, on Zulip):

diagnostics

nikomatsakis (May 29 2020 at 14:53, on Zulip):

async-await polish and impl. improvements

nikomatsakis (May 29 2020 at 14:54, on Zulip):

@Wesley Wiser do you have more thoughts on what such a project-wide structure might look like?

nikomatsakis (May 29 2020 at 14:55, on Zulip):

during the impl period in particular, we had a big push around breaking down into a list of groups, and each group having leadership and goals

nikomatsakis (May 29 2020 at 14:55, on Zulip):

that was kind of "cross-cutting"

Wesley Wiser (May 29 2020 at 14:55, on Zulip):

I think maybe one thing that comes to mind for me was that there was a fairly large "marketing" push that advertised a bunch of different projects you could get involved with.

nikomatsakis (May 29 2020 at 14:55, on Zulip):

yeah, it seems like you're talking as much about the impl period as about "Rust 2018" itself

nikomatsakis (May 29 2020 at 14:56, on Zulip):

and this definitely fits with what I was saying before about "do less but with more followthrough"

Wesley Wiser (May 29 2020 at 14:56, on Zulip):

I think it created an impression, for me, that there was certainly somewhere you could help out

nikomatsakis (May 29 2020 at 14:56, on Zulip):

although during impl period we may have tried "do more with more followthrough" :)

Wesley Wiser (May 29 2020 at 14:57, on Zulip):

nikomatsakis said:

yeah, it seems like you're talking as much about the impl period as about "Rust 2018" itself

I think this is definitely true

nikomatsakis (May 29 2020 at 14:57, on Zulip):

so, to step back, if we were to collect this list of projects, maybe group some things to gether to some extent, and combine it with some projects from a few other groups,

nikomatsakis (May 29 2020 at 14:57, on Zulip):

and then do some blog posts announcing an effort to focus on these things,

nikomatsakis (May 29 2020 at 14:57, on Zulip):

and of course actually have the "org backbone" to drive it

davidtwco (May 29 2020 at 14:57, on Zulip):

Wesley Wiser said:

I think it created an impression, for me, that there was certainly somewhere you could help out

It feels like more people from the impl period (myself included) have stuck around than those who've shown up during other periods.

nikomatsakis (May 29 2020 at 14:57, on Zulip):

that's kind of what you're talking about it seems

nikomatsakis (May 29 2020 at 14:58, on Zulip):

interesting

nikomatsakis (May 29 2020 at 14:58, on Zulip):

I definitely have the impression that involvement in successful working groups has been a good way to get people to "stick" (though not the only way by any means)

Wesley Wiser (May 29 2020 at 14:58, on Zulip):

I think that would help a lot! One other thing that stuck out to me was that by naming "leaders" or whatever, I knew that they had signed up to be pestered with questions and that it was ok to message them.

nikomatsakis (May 29 2020 at 14:58, on Zulip):

a sense of comraderye (a word I apparently cannot spell)

nikomatsakis (May 29 2020 at 14:59, on Zulip):

So I'm not sure @Wesley Wiser if you read the area proposal

Wesley Wiser (May 29 2020 at 14:59, on Zulip):

I think a lot of people are hesitant to just start messaging people unless they know that they're open to being asked questions

nikomatsakis (May 29 2020 at 14:59, on Zulip):

but I also wonder how much that fits into this

nikomatsakis (May 29 2020 at 14:59, on Zulip):

i.e., I think that in some cases

nikomatsakis (May 29 2020 at 14:59, on Zulip):

like mir optimizations perhaps :)

nikomatsakis (May 29 2020 at 14:59, on Zulip):

it's not so much about concrete goals as having areas

pnkfelix (May 29 2020 at 15:00, on Zulip):

both matter I suspect

nikomatsakis (May 29 2020 at 15:00, on Zulip):

and it'd be helpful to have a bit more "structure" in some of those (but I am wary because organizational work is often the hardest to truly sustain)

pnkfelix (May 29 2020 at 15:00, on Zulip):

since @Wesley Wiser brought up leaders

pnkfelix (May 29 2020 at 15:00, on Zulip):

I just wanted to mention

nikomatsakis (May 29 2020 at 15:00, on Zulip):

time check: 1 hour :)

pnkfelix (May 29 2020 at 15:00, on Zulip):

if you're interested in being a co-lead of wg-incr-comp

pnkfelix (May 29 2020 at 15:00, on Zulip):

feel free to speak up or PM me

pnkfelix (May 29 2020 at 15:01, on Zulip):

(which I know may sounds funny given that in another topic I mentioned how I sometimes get annoyed by PM's... its complicated...)

Santiago Pastorino (May 29 2020 at 15:01, on Zulip):

I think I'd like to join wg-incr-comp :)

pnkfelix (May 29 2020 at 15:02, on Zulip):

worst case scenario we'll force someone to be a leader by picking their userrname out of the commit log.

pnkfelix (May 29 2020 at 15:02, on Zulip):

@nikomatsakis do you think the document we have here could be the basis for a roadmap post?

Wesley Wiser (May 29 2020 at 15:03, on Zulip):

Anyway, I guess to sum up my feelings:

  1. I think a blog post followed up with tweets/reddit posts/hackernews/etc that lists the projects we want help with would be great and
  2. Identifying the people in each group that are happy to help new people and answer questions makes that a lot more inviting.
nikomatsakis (May 29 2020 at 15:03, on Zulip):

To answer your question @pnkfelix, definitely, though I'm looking over the list and my first reaction is that it's "too damn long"

nikomatsakis (May 29 2020 at 15:04, on Zulip):

like, it's a lot of things being pushed in parallel..

nikomatsakis (May 29 2020 at 15:04, on Zulip):

tl;dr yes I think it's the basis but I think we should think a bit more (not too much more) about if we can cull the list down or struture it more to pull out the things we want to highlight (maybe the other projects continue on some pace)

pnkfelix (May 29 2020 at 15:04, on Zulip):

nikomatsakis said:

To answer your question pnkfelix, definitely, though I'm looking over the list and my first reaction is that it's "too damn long"

cue Jimmy McMillan gif

nikomatsakis (May 29 2020 at 15:05, on Zulip):

I feel like we ought to be saying "no" more often, and I feel like I'm sort of responsible for some of that. I mean I guess if I look at the projects I'd like to push along

it's kind of a lot =)

nikomatsakis (May 29 2020 at 15:06, on Zulip):

otoh there are is overlap between them...

nikomatsakis (May 29 2020 at 15:06, on Zulip):

I have to think on it a bit :)

pnkfelix (May 29 2020 at 15:06, on Zulip):

is RFC 2229 actually overlapping?

pnkfelix (May 29 2020 at 15:06, on Zulip):

and the main reason to push for that is due to the semantics change in drop order, right?

pnkfelix (May 29 2020 at 15:07, on Zulip):

(and therefore we may want/need to tie it to an edition boundary. Or maybe I am misremembering our lang team conversations on that)

nikomatsakis (May 29 2020 at 15:07, on Zulip):

pnkfelix said:

and the main reason to push for that is due to the semantics change in drop order, right?

two reasons

nikomatsakis (May 29 2020 at 15:07, on Zulip):

semantic changes, but more importantly a dedicated and promising group of folks showed up who want to work on it :)

pnkfelix (May 29 2020 at 15:07, on Zulip):

oh

pnkfelix (May 29 2020 at 15:07, on Zulip):

yeah that's a good reason to push. :)

nikomatsakis (May 29 2020 at 15:08, on Zulip):

the overlap is primarily gats + chalk

nikomatsakis (May 29 2020 at 15:08, on Zulip):

impl Trait maybe a bit but I think it's fairly independent

nikomatsakis (May 29 2020 at 15:08, on Zulip):

I think the answer is that it's worth investing time in making a plan around impl Trait but then we have to take a look at how much that plan can move forward with me only checking in periodically

nikomatsakis (May 29 2020 at 15:08, on Zulip):

(which seems quite probable)

nikomatsakis (May 29 2020 at 15:10, on Zulip):

anyway I think meeting is over sorry folks we're kind of rambling :)

Wesley Wiser (May 29 2020 at 15:11, on Zulip):

Is it worth trying to prioritize those related projects and just focus on getting one done? We've mentioned before that having one person lead a bunch of different projects isn't great so maybe we should make the goal to get one project from each leader "done" this year?

Wesley Wiser (May 29 2020 at 15:11, on Zulip):

We can still support people working on the other projects of course, but the priority would be given to which ever project is part of the roadmap.

nikomatsakis (May 29 2020 at 15:11, on Zulip):

So, for "my" set above, that is exactly what I'm thinking about =)

Last update: Jul 03 2020 at 17:20UTC