Stream: t-compiler

Topic: design meeting 2019-10-04


nikomatsakis (Oct 04 2019 at 13:27, on Zulip):

Hi @T-compiler/meeting! In about 30 minutes, we'll have our design meeting for this week. The theme is

Roadmap 2020

In some pre-discussion, @pnkfelix and I decided that the best place to start would be to focus on enumerating the biggest challenges that we see for rustc and the compiler team at this moment. These challenges can be technical but also organizational.

We started a hackmd document that we will use to collect notes. We've seeded the document with our initial thoughts on the biggest challenges, but we'd like feedback from y'all.

So, please:

You can do this now, but we'll also spend a bit of time in the beginning of the meeting.

pnkfelix (Oct 04 2019 at 13:32, on Zulip):

should we provide guidance as to how many votes people should use? I guess if someone really cares about sometnign they could make dummy accounts to vote more than once on an item

eddyb (Oct 04 2019 at 13:32, on Zulip):

/me frantically searches for their Tor scripts

nikomatsakis (Oct 04 2019 at 13:59, on Zulip):

I had intended for it to be "approval voting" -- vote for as many as you like, once per item

nikomatsakis (Oct 04 2019 at 13:59, on Zulip):

but...

pnkfelix (Oct 04 2019 at 14:01, on Zulip):

approval voting is probably fine

pnkfelix (Oct 04 2019 at 14:02, on Zulip):

I was vaguely wondering if we should encourage to vote for e.g. half of the topics

pnkfelix (Oct 04 2019 at 14:02, on Zulip):

(or "at most half")

nikomatsakis (Oct 04 2019 at 14:03, on Zulip):

hi @T-compiler/meeting! It's 10:03am (at least here in Boston...) and that means it's time for....

Roadmap 2020 design meeting

nikomatsakis (Oct 04 2019 at 14:04, on Zulip):

To start, let's have some

announcements

nikomatsakis (Oct 04 2019 at 14:04, on Zulip):

the first "announcement" is about this meeting -- the plan for today (if you missed it) is to discuss challenges, and I have some notes above

nikomatsakis (Oct 04 2019 at 14:05, on Zulip):

in particular, people should be reading the list of challenges and voting -- feel free to add your own ideas!

pnkfelix (Oct 04 2019 at 14:05, on Zulip):

(if you add your own ideas, make sure to put it on the voting page!)

nikomatsakis (Oct 04 2019 at 14:06, on Zulip):

yep

nikomatsakis (Oct 04 2019 at 14:06, on Zulip):

the plan is to kind of use the votes to guide more detailed discussion, I thnk

nikomatsakis (Oct 04 2019 at 14:09, on Zulip):

ok, how are folks doing?

pnkfelix (Oct 04 2019 at 14:10, on Zulip):

I may be the only one that added an item

nikomatsakis (Oct 04 2019 at 14:10, on Zulip):

I see @pnkfelix added "paying for labor" to the list (fyi)

nikomatsakis (Oct 04 2019 at 14:10, on Zulip):

/me almost spelled that "labour" -- what has become of me.

pnkfelix (Oct 04 2019 at 14:10, on Zulip):

(and its arguably an item that belongs somewhere other than this meeting. i.e. this is not an issue for "just" 2020/2021)

nikomatsakis (Oct 04 2019 at 14:10, on Zulip):

too much time spent reading about brexit or something

nikomatsakis (Oct 04 2019 at 14:11, on Zulip):

yeah, my sense is that it's not really something that belongs on the compiler team roadmap per se, but I'm always interested to hear about it, and I think it could inform other roadmaps

pnkfelix (Oct 04 2019 at 14:11, on Zulip):

I think its a topic that we sometimes shy away from discussing

pnkfelix (Oct 04 2019 at 14:11, on Zulip):

and yet it seems vitally important ...

nikomatsakis (Oct 04 2019 at 14:11, on Zulip):

(also, I want to repeat that people who are interested in contracting or in issuing contracts should contact me privately; I'm trying to get a better sense of what people are looking for and how we can try to connect people who want to issue contracts)

Esteban Küber (Oct 04 2019 at 14:12, on Zulip):

It would be nice to have more people working full time on the compiler

nikomatsakis (Oct 04 2019 at 14:12, on Zulip):

yeah

nikomatsakis (Oct 04 2019 at 14:12, on Zulip):

yeah, my sense is that it's not really something that belongs on the compiler team roadmap per se, but I'm always interested to hear about it, and I think it could inform other roadmaps

to be clear this doesn't mean don't vote for it :)

nikomatsakis (Oct 04 2019 at 14:12, on Zulip):

ok let's get started

pnkfelix (Oct 04 2019 at 14:13, on Zulip):

heh I like these votes

nikomatsakis (Oct 04 2019 at 14:13, on Zulip):

the number one item in terms of problems by far is

compilation times remain stubbornly high

pnkfelix (Oct 04 2019 at 14:13, on Zulip):

565843

nikomatsakis (Oct 04 2019 at 14:13, on Zulip):

in the hackmd doc, I wrote the following

nikomatsakis (Oct 04 2019 at 14:13, on Zulip):

(well, I wrote some of that)

Santiago Pastorino (Oct 04 2019 at 14:14, on Zulip):

I wanted to add something around team growth, maybe it's a bit late already ...

nikomatsakis (Oct 04 2019 at 14:14, on Zulip):

ps I'm going to try and keep the hackmd up to date as we go

nikomatsakis (Oct 04 2019 at 14:14, on Zulip):

I wanted to add something around team growth, maybe it's a bit late already ...

go ahead and finish

nikomatsakis (Oct 04 2019 at 14:14, on Zulip):

the main question is -- I guess two-fold --

nikomatsakis (Oct 04 2019 at 14:14, on Zulip):

is there more to say about compilation time as a problem and do we want to kind of discuss some of what might be done to try and address it?

eddyb (Oct 04 2019 at 14:15, on Zulip):

small note: incremental isn't all the way back to parsing, that could have a nice effect in theory, especially with pipelining and everything else
but this is a significant thing to tackle and will likely not happen without coordination

nikomatsakis (Oct 04 2019 at 14:15, on Zulip):

One thing that I think we never really successfully did

pnkfelix (Oct 04 2019 at 14:15, on Zulip):

is there more to say about compilation time as a problem and do we want to kind of discuss some of what might be done to try and address it?

is it worth differentiating bootstrap time from overall compile times ?

pnkfelix (Oct 04 2019 at 14:16, on Zulip):

obviously fixing the latter has a very broad effect

nikomatsakis (Oct 04 2019 at 14:16, on Zulip):

is to establish clear benchmark(s) that we think is meaningful that we can measure and optimize for

Santiago Pastorino (Oct 04 2019 at 14:16, on Zulip):

I wanted to add something around team growth, maybe it's a bit late already ...

added it and added to the voting board

nikomatsakis (Oct 04 2019 at 14:16, on Zulip):

is it worth differentiating bootstrap time from overall compile times ?

yeah, let's talk about that a bit actually, I think that's a good question

pnkfelix (Oct 04 2019 at 14:16, on Zulip):

but focusing on the former (bootstrap times) may yield better return in short- medium- term

eddyb (Oct 04 2019 at 14:16, on Zulip):

is to establish clear benchmark(s) that we think is meaningful that we can measure and optimize for

getting a deja-vu, as if this has been said every year for the past 3-4 years

nikomatsakis (Oct 04 2019 at 14:17, on Zulip):

exactly :P

pnkfelix (Oct 04 2019 at 14:17, on Zulip):

we definitely have a set of benchmarks. dunno if its a good set

pnkfelix (Oct 04 2019 at 14:17, on Zulip):

but people use perf.rlo

simulacrum (Oct 04 2019 at 14:17, on Zulip):

I do think that setting a strict goal of "make the compiler itself bootstrap faster" may be reasonable. It's likely to help other big projects out there, and it'll help us go faster too.

nikomatsakis (Oct 04 2019 at 14:18, on Zulip):

yeah, I think we've made great progress in the sense that perf.rlo has become much better

nikomatsakis (Oct 04 2019 at 14:18, on Zulip):

I thnk what i'm advocating for is picking specific crates/scenarios (maybe from that list!) and saying "let's do that" -- but indeed maybe bootstrapping is the right one to start with

nikomatsakis (Oct 04 2019 at 14:18, on Zulip):

I know that there's been a lot of attention paid to CI specifically and where time is spent

nikomatsakis (Oct 04 2019 at 14:19, on Zulip):

I feel like building rustc incrementally still feels prety slow a ot of the time, even if you use --keep-stage

nikomatsakis (Oct 04 2019 at 14:19, on Zulip):

and of course that is tricky and confusing

nikomatsakis (Oct 04 2019 at 14:19, on Zulip):

so my first question would be something like "incremental bootstrap most of all?" (distinct from CI) and "maybe find a way to make a (95%?) reliable, automatic keep-stage?"

simulacrum (Oct 04 2019 at 14:20, on Zulip):

I personally think that my priorities would be:

  1. non incremental bootstrap
  2. incremental bootstrap
  3. keep stage improvements
nikomatsakis (Oct 04 2019 at 14:21, on Zulip):

interesting

simulacrum (Oct 04 2019 at 14:21, on Zulip):

in that order primarily because we're already really pretty fast at a check build on the compiler. codegen/LLVM time improvements are where we need to make the most improvement, largely, and that is shared between non-incr/incr

nikomatsakis (Oct 04 2019 at 14:21, on Zulip):

well, that depends

nikomatsakis (Oct 04 2019 at 14:22, on Zulip):

but that's interesting

nikomatsakis (Oct 04 2019 at 14:22, on Zulip):

(i.e., it's plausible that better codegen-unit distribution could help)

nikomatsakis (Oct 04 2019 at 14:22, on Zulip):

I know that e.g. @Eric Huss has done some cool work on visualizing bootstrap/CI times, are those easily accessible?

simulacrum (Oct 04 2019 at 14:22, on Zulip):

https://mark.rousskov.org/parallel-compiler-data/master/rustc-compiler-opt-timing.html

nikomatsakis (Oct 04 2019 at 14:22, on Zulip):

anyway, we don't want to go in too deep technically probably

nikomatsakis (Oct 04 2019 at 14:23, on Zulip):

Maybe a good idea would be: come up with some concrete-ish goal (sounds like we have a candidate -- improve LLVM time around rustc bootstrap) and identify if there are any measurements needed to validate or guide that.

varkor (Oct 04 2019 at 14:23, on Zulip):

that editing a comment has the same effect as modifying a function wrt forcing recompilation is regularly a frustration — incremental still isn't fine-grained enough

eddyb (Oct 04 2019 at 14:24, on Zulip):

that's where moving incremental all the way back to parsing would solve, I think

nikomatsakis (Oct 04 2019 at 14:25, on Zulip):

do you think?

varkor (Oct 04 2019 at 14:25, on Zulip):

(tracked by #47389)

simulacrum (Oct 04 2019 at 14:25, on Zulip):

That's true -- in some sense, if we were better at incremental, then we could avoid codegen time. Right now even a "comment edit" means we re-codegen things I believe

nikomatsakis (Oct 04 2019 at 14:25, on Zulip):

it seems like x.py build -i still takes more time x.py check

nikomatsakis (Oct 04 2019 at 14:25, on Zulip):

when all you've done is modify a comment

nikomatsakis (Oct 04 2019 at 14:25, on Zulip):

and I feel like that shouldn't be

nikomatsakis (Oct 04 2019 at 14:25, on Zulip):

but maybe that's not true

nikomatsakis (Oct 04 2019 at 14:25, on Zulip):

That's true -- in some sense, if we were better at incremental, then we could avoid codegen time. Right now even a "comment edit" means we re-codegen things I believe

it shouldn't be true -- well, modulo spans, that may be part of it

simulacrum (Oct 04 2019 at 14:26, on Zulip):

Yeah this is "buggy" for the most part, not correct -- but I believe it is the state of the world today :)

nikomatsakis (Oct 04 2019 at 14:26, on Zulip):

still, it seems like what I can extract from this is maybe a few things

nikomatsakis (Oct 04 2019 at 14:27, on Zulip):

should we turn to the next topic?

eddyb (Oct 04 2019 at 14:27, on Zulip):

I guess it's two-part:

  1. don't do any work past the HIR when only a comment changed (targeted Span improvements may help)
  2. move incremental back so you don't need to spend entire seconds getting to HIR
eddyb (Oct 04 2019 at 14:27, on Zulip):

(Zulip really hates me right now, the lag is unbearable)

nikomatsakis (Oct 04 2019 at 14:28, on Zulip):

next most voted was

tech debt around the trait system

and the text from the hackmd is

nikomatsakis (Oct 04 2019 at 14:29, on Zulip):

(side note: does anyone know if there is a bug filed with zulip about their handling of bullet lists?! it seems to have regressed significantly...)

nikomatsakis (Oct 04 2019 at 14:29, on Zulip):

I think what might be helpful here would be to discuss this in more specific terms -- have I called out the right real goals, for example?

nikomatsakis (Oct 04 2019 at 14:30, on Zulip):

(It'd probably be good to prioritize somewhat within those, as well)

eddyb (Oct 04 2019 at 14:31, on Zulip):

lazy norm also blocks any interesting usage of type/const parameters in const expressions inside types (e.g. [u8; size_of::<T>()] or Foo<{N + 1}>)
(note that this is not limited to const-generics, it includes things that work on stable when you have a concrete type instead of a type parameter!)

nikomatsakis (Oct 04 2019 at 14:32, on Zulip):

So we could talk a bit about what's the problem there. We've made .. relatively little progress on this this year, though we have made some. I think a lot of that is because my time has been limited. This remains true, though in the last few weeks I've managed to block out bigger chunks of time.

nikomatsakis (Oct 04 2019 at 14:33, on Zulip):

It's also because there's just a lot to organize and it's hard to get traction

nikomatsakis (Oct 04 2019 at 14:33, on Zulip):

There's no want of folks interested

pnkfelix (Oct 04 2019 at 14:33, on Zulip):

lazy norm, If i understand correctly, would also address the problem that right now we have various random iCE's that arise because we're missing normalize() calls at "just the right places"

nikomatsakis (Oct 04 2019 at 14:33, on Zulip):

I think a lot of the problems from rustc recur in a microcosm

nikomatsakis (Oct 04 2019 at 14:33, on Zulip):

lazy norm, If i understand correctly, would also address the problem that right now we have various random iCE's that arise because we're missing normalize() calls at "just the right places"

yes, that's a goal

nikomatsakis (Oct 04 2019 at 14:33, on Zulip):

I think a lot of the problems from rustc recur in a microcosm

specifically, laying out designs and getting feedback

nikomatsakis (Oct 04 2019 at 14:34, on Zulip):

I had blocked out time today to try and setup (e.g.) a recurring "design meeting" where we focus on some topic each week and dig in deep

nikomatsakis (Oct 04 2019 at 14:34, on Zulip):

anyway, maybe going a bit deep for roadmap discussion :)

nikomatsakis (Oct 04 2019 at 14:34, on Zulip):

I'm just braindumping

nikomatsakis (Oct 04 2019 at 14:35, on Zulip):

one thing I will say is that iterating with rust-analyzer has been good and exposed various weaknesses

nikomatsakis (Oct 04 2019 at 14:35, on Zulip):

but one of the hard things to balance is "incremental progress on existing system" and "chalk"

pnkfelix (Oct 04 2019 at 14:35, on Zulip):

should we plan to discuss, in some future meeting, whether we want to focus on swapping in a new trait subsystem (chalk) or try to ...

pnkfelix (Oct 04 2019 at 14:35, on Zulip):

okay yeah you just summarized it

nikomatsakis (Oct 04 2019 at 14:36, on Zulip):

the last time the wg-traits went deep on that topic, we concluded that it would be hard to upgade existing system to realy solve e.g. gats

nikomatsakis (Oct 04 2019 at 14:36, on Zulip):

however, I think that for lazy norm it is plausible (and I was planning to put time into exploring that today)

pnkfelix (Oct 04 2019 at 14:36, on Zulip):

thinking back on NLL, we got a lot of leverage from moving to MIR-borrowck. But even there, we did not go whole hog on trying to swap in polonius

pnkfelix (Oct 04 2019 at 14:36, on Zulip):

So I don't actually know what that analogy tells us

pnkfelix (Oct 04 2019 at 14:38, on Zulip):

How long, or how much man power, would you say have invested in trying to upgrade existing system?

nikomatsakis (Oct 04 2019 at 14:38, on Zulip):

ok I don't feel like "new information" is coming to light here, next topic? :)

nikomatsakis (Oct 04 2019 at 14:38, on Zulip):

(let me know if I'm rail-roading...)

pnkfelix (Oct 04 2019 at 14:38, on Zulip):

Well wait, I don't know what we've concluded from this. Is it just that we all agree that this is a challenge?

nikomatsakis (Oct 04 2019 at 14:38, on Zulip):

heh indeed I don't think we concluded much

nikomatsakis (Oct 04 2019 at 14:39, on Zulip):

well ok so

nikomatsakis (Oct 04 2019 at 14:39, on Zulip):

I think a good goal for the year for traits would be

nikomatsakis (Oct 04 2019 at 14:39, on Zulip):

sorry, let me rephrase that

pnkfelix (Oct 04 2019 at 14:39, on Zulip):

if the answer is just: "we should have a follow-up meeting to discuss traits challenges", I'm okay with that conclusion

nikomatsakis (Oct 04 2019 at 14:40, on Zulip):

I think a good immediate goal would be to kind of decide the extent to which we will go "all in" on chalk next year -- and what we will try to solve on existing system first

nikomatsakis (Oct 04 2019 at 14:40, on Zulip):

and I think we should make that exploration our main immediate goal, and try to propose a more concrete plan in (say) a month or so, which might indeed make it onto the roadmap

nikomatsakis (Oct 04 2019 at 14:40, on Zulip):

or just be useful :)

nikomatsakis (Oct 04 2019 at 14:41, on Zulip):

we have gained some new data in this regard (thanks e.g. to r-a exploration but also a few other things)

eddyb (Oct 04 2019 at 14:42, on Zulip):

I'm not sure even sure lazy normalization for type-level constant expressions is that tied to traits, tbh

nevermind the problem is lazy normalization for everything is needed to break cycles that happen to affect constant expressions in particular, my memory is just bad

eddyb (Oct 04 2019 at 14:42, on Zulip):

(deleted)

nikomatsakis (Oct 04 2019 at 14:43, on Zulip):

let's move to another thing, but I think I agree

nikomatsakis (Oct 04 2019 at 14:43, on Zulip):

or at least I think that we can make progress around there :)

nikomatsakis (Oct 04 2019 at 14:44, on Zulip):

next most voted:

nikomatsakis (Oct 04 2019 at 14:44, on Zulip):
nikomatsakis (Oct 04 2019 at 14:45, on Zulip):

I'd be curious to hear what "upvoters" were expressing, because I could see many things to talk about here

nikomatsakis (Oct 04 2019 at 14:46, on Zulip):

if the answer is just: "we should have a follow-up meeting to discuss traits challenges", I'm okay with that conclusion

I wrote this btw as the 'traits conclusion'

pnkfelix (Oct 04 2019 at 14:46, on Zulip):

i posed the "pay for labor" topic because it strikes me as a potential existential threat. nrc quipped years ago that (paraphrasing) we'll be fine without more compiler developers, but without users Rust is dead in the water.

pnkfelix (Oct 04 2019 at 14:47, on Zulip):

I believed that at the time. But I think now we have more users.

nikomatsakis (Oct 04 2019 at 14:47, on Zulip):

I guess I can give my current perspective

Vadim Petrochenkov (Oct 04 2019 at 14:47, on Zulip):

So much paradoxes here.
- Full-time employed people often work for sprint goals and reports rather than quality.
- Some people that are supposed to be full-time employed appear and do something like once a month and busy with something unobservable the rest of the time.

nikomatsakis (Oct 04 2019 at 14:48, on Zulip):

I think we may well make progress on some sort of foundation, but I don't think it will hire a lot of compiler devs. I think the situation we really want is to have more companies doing funding in more diverse ways.

nikomatsakis (Oct 04 2019 at 14:48, on Zulip):

One thing I've been thinking a lot about

nikomatsakis (Oct 04 2019 at 14:48, on Zulip):

is how we as a team can help to connect people with grants/contracting and make it easy for people to get grants or contracts

nikomatsakis (Oct 04 2019 at 14:48, on Zulip):

so as an example, I've wondered if

nikomatsakis (Oct 04 2019 at 14:48, on Zulip):

when we come up with projects

eddyb (Oct 04 2019 at 14:49, on Zulip):

- Some people that are supposed to be full-time employed appear and do something like once a month and busy with something unobservable the rest of the time.

sounds like something that keeps happening to me and I did want to bring up a related thing

nikomatsakis (Oct 04 2019 at 14:49, on Zulip):

if we can describe the steps needed, that could help people develop a grant application to work on it, etc. But that's also kind of a lot of work (but also very similar to the work of mentoring, so...)

nikomatsakis (Oct 04 2019 at 14:49, on Zulip):

/me stops

pnkfelix (Oct 04 2019 at 14:51, on Zulip):

Full-time employed people often work for sprint goals and reports rather than quality.

@Vadim Petrochenkov are you saying here that it would be a bad idea to try to figure out compensation systems, because it would open up people trying to game the system to get the money?

eddyb (Oct 04 2019 at 14:51, on Zulip):

due to the haphazard work distribution, I can end up trying to do up to a dozen things at a time and humans aren't known for their excellent multitasking skills

eddyb (Oct 04 2019 at 14:52, on Zulip):

I've noticed I do much better on single shorter-term focused projects, but even then there's some amount of getting dragged into a different part of rustc instead of focusing on the contract

pnkfelix (Oct 04 2019 at 14:52, on Zulip):

@Vadim Petrochenkov or was that more a general statement about the kinds of work you see (either here or across FOSS) when comparing paid- vs volunteer effort ?

Vadim Petrochenkov (Oct 04 2019 at 14:53, on Zulip):

@pnkfelix
Nah, it's more that people are overloaded and often have to do what they have to rather than want to.

Vadim Petrochenkov (Oct 04 2019 at 14:53, on Zulip):

Taking shortcuts and decreasing the quality.

eddyb (Oct 04 2019 at 14:53, on Zulip):

I have no idea what the solution is but I feel like the resources available would be more effectively used if we planned for it better. I guess I'm just rambling here

Santiago Pastorino (Oct 04 2019 at 14:53, on Zulip):

doesn't that also depends on the goals you may have in a contract?

pnkfelix (Oct 04 2019 at 14:53, on Zulip):

i see, the "have to" there is as in, what's good for the employee rather than what's good for the project itself.

nikomatsakis (Oct 04 2019 at 14:54, on Zulip):

due to the haphazard work distribution, I can end up trying to do up to a dozen things at a time and humans aren't known for their excellent multitasking skills

yes this

Vadim Petrochenkov (Oct 04 2019 at 14:54, on Zulip):

(These are the reasons why I'm personally afraid that if I turn the hobby into a job and start contracting I immediately lose all the motivation and productivity and leave the project soon.)

nikomatsakis (Oct 04 2019 at 14:55, on Zulip):

OK, well, we're coming in at 55 minutes

Santiago Pastorino (Oct 04 2019 at 14:56, on Zulip):

(These are the reasons why I'm personally afraid that if I turn the hobby into a job and start contracting I immediately lose all the motivation and productivity and leave the project soon.)

I feel completely the opposite tbh, if I can turn a hobby into my day to day job I'd feel realization

davidtwco (Oct 04 2019 at 14:56, on Zulip):

I voted for it because I would jump at the opportunity to work on rustc full time (though, if such a position existed, there are far more qualified people that should get it first), but nothing like that exists. Contracting doesn't appeal as much to me as full-time employment.

nikomatsakis (Oct 04 2019 at 14:57, on Zulip):

Yeah, that makes sense.

eddyb (Oct 04 2019 at 14:57, on Zulip):

tbqh, Rust hasn't gotten much specific from me being paid, with some small exceptions (like the Rust 2018 crunch) - to the extent that it would be more honest to have the contract say "do things around rustc for one more year"

eddyb (Oct 04 2019 at 14:58, on Zulip):

so we should maybe try to figure out, about goals (wrt contracting):

nikomatsakis (Oct 04 2019 at 14:58, on Zulip):

are there some conclusions we can draw from this? :)

pnkfelix (Oct 04 2019 at 14:58, on Zulip):

well: Is paying for labor a challenge we need to worry about in short term (as in, next year, or even next two years)

pnkfelix (Oct 04 2019 at 14:59, on Zulip):

I'm tempted to say, based on gut feeling, that the status quo will suffice for now

pnkfelix (Oct 04 2019 at 14:59, on Zulip):

but I think we should nonetheless discuss it on occasion going forward

nikomatsakis (Oct 04 2019 at 14:59, on Zulip):

Yeah. I also do think this will be a theme that the project as a whole should be working on, and it's helpful to keep talking about it.

nikomatsakis (Oct 04 2019 at 15:00, on Zulip):

(I approve of bringing it out into the open, regardless)

pnkfelix (Oct 04 2019 at 15:00, on Zulip):

i would be interested to know

pnkfelix (Oct 04 2019 at 15:00, on Zulip):

about people who did were volunteers and then had to stop due to their jobs

pnkfelix (Oct 04 2019 at 15:01, on Zulip):

Such as students who were able to participate but then lacked time/will after getting gainful employment.

pnkfelix (Oct 04 2019 at 15:01, on Zulip):

Basically I'm curious how many of those we've had, and whether we could have done something to keep them engaged in project.

pnkfelix (Oct 04 2019 at 15:01, on Zulip):

but its not a data-set I'm prepared to acquire in -1 minutes remaining in meeting

nikomatsakis (Oct 04 2019 at 15:03, on Zulip):

OK, well, I guess that's a wrap

nikomatsakis (Oct 04 2019 at 15:03, on Zulip):

TBH my main take-away from this meeting is that the problems we identified are the main ones :)

nikomatsakis (Oct 04 2019 at 15:04, on Zulip):

I guess i'll highlight this

nikomatsakis (Oct 04 2019 at 15:04, on Zulip):
nikomatsakis (Oct 04 2019 at 15:04, on Zulip):

looking back over the docs, @simulacrum's comment is interesting

nikomatsakis (Oct 04 2019 at 15:04, on Zulip):

anyway, I gotta run, thanks <3 for participating

Last update: Nov 16 2019 at 01:30UTC