Stream: t-compiler/meetings

Topic: planning meeting 2020.09.04


mark-i-m (Sep 03 2020 at 22:15, on Zulip):

So for tomorrow, it would be helpful if interested parties could read/skim through this before the meeting: https://hackmd.io/b-FKKh9DSnah-lj1NVKaAw

pnkfelix (Sep 04 2020 at 13:56, on Zulip):

this is the first year we've done a contributor survey, right? I ask because my first instinct reading over the above link was to ask myself "hmm how did this compare with last year?"

nikomatsakis (Sep 04 2020 at 13:58, on Zulip):

I believe it's the first year we've tried it

mark-i-m (Sep 04 2020 at 14:01, on Zulip):

I'm here but my internet is acting up... just a sec

mark-i-m (Sep 04 2020 at 14:03, on Zulip):

@pnkfelix yes, to my knowledge this is the first year

nikomatsakis (Sep 04 2020 at 14:05, on Zulip):

Hello @T-compiler/meeting -- we're going to be discussing the rust contributor survey today. We're running a bit slow but let's give a few minutes for ...

Announcements

nikomatsakis (Sep 04 2020 at 14:05, on Zulip):
mark-i-m (Sep 04 2020 at 14:07, on Zulip):
DPC (Sep 04 2020 at 14:07, on Zulip):

thanks mark :smile:

nikomatsakis (Sep 04 2020 at 14:09, on Zulip):

OK, shall we get started?

mark-i-m (Sep 04 2020 at 14:09, on Zulip):

What would be the best way to proceed?

pnkfelix (Sep 04 2020 at 14:09, on Zulip):

what's best way to attack this; perhaps review the high level results?

nikomatsakis (Sep 04 2020 at 14:09, on Zulip):

(actually I am going to relocate to another room)

nikomatsakis (Sep 04 2020 at 14:10, on Zulip):

but that sounds like a good way to start

pnkfelix (Sep 04 2020 at 14:10, on Zulip):

(which is what the document does in the first place)

pnkfelix (Sep 04 2020 at 14:10, on Zulip):

The summary here is excellent

mark-i-m (Sep 04 2020 at 14:11, on Zulip):

Yeah, I was going to point to that

pnkfelix (Sep 04 2020 at 14:11, on Zulip):

We have 409 responses, broken down into 164 people who never had a PR merged, 169 with <1 PR/month, and 76 >= 1 PR/month

pnkfelix (Sep 04 2020 at 14:11, on Zulip):

Its really good that we get so much response from the "never had a PR merged group"

pnkfelix (Sep 04 2020 at 14:11, on Zulip):

well, actually, there's an interesting question there

pnkfelix (Sep 04 2020 at 14:12, on Zulip):

in that, what is more important to focus on:

pnkfelix (Sep 04 2020 at 14:12, on Zulip):
  1. Getting people from the "never had a PR merged" into one of the otehr groups, or ...
pnkfelix (Sep 04 2020 at 14:12, on Zulip):
  1. Getting the people in "< 1 PR/month" into the ">= 1 PR/month"
mark-i-m (Sep 04 2020 at 14:12, on Zulip):

I would say both

pnkfelix (Sep 04 2020 at 14:12, on Zulip):

I can see arguments on both sides of that

mark-i-m (Sep 04 2020 at 14:12, on Zulip):

It's a train

nikomatsakis (Sep 04 2020 at 14:12, on Zulip):

My take is: while I care about both, I think we have more trouble with the "latter stages"

nikomatsakis (Sep 04 2020 at 14:13, on Zulip):

i.e., the hardest thing is to retain folks after the first PR or two

nikomatsakis (Sep 04 2020 at 14:13, on Zulip):

(though I am really interested in making that first PR much easier)

mark-i-m (Sep 04 2020 at 14:13, on Zulip):

I view it slightly differently: there is only a small population that will ever become long-term contributors to begin with

pnkfelix (Sep 04 2020 at 14:13, on Zulip):

and if the experience was difficult to get that one PR landed, that can be a big reason why

mark-i-m (Sep 04 2020 at 14:13, on Zulip):

and the challenge is to get as many of them

DPC (Sep 04 2020 at 14:13, on Zulip):

i think if the first PR is seen as "less workload" people will be more willing to contribute again

nikomatsakis (Sep 04 2020 at 14:14, on Zulip):

yeah, I was pondering that too :)

pnkfelix (Sep 04 2020 at 14:14, on Zulip):

anyway this distinction of which audience we wish to focus on

nikomatsakis (Sep 04 2020 at 14:14, on Zulip):

well, I guess I agree with both in any case ;)

pnkfelix (Sep 04 2020 at 14:14, on Zulip):

is just something we should keep in the back of our minds

pnkfelix (Sep 04 2020 at 14:14, on Zulip):

I do agree that we need to consider both groups

pnkfelix (Sep 04 2020 at 14:14, on Zulip):

but there are definitely some decisions that will focus more on one group or another

DPC (Sep 04 2020 at 14:15, on Zulip):

also fixing some of the issues for group 1, might have a side-effect of fixing the 2nd group partially

Wesley Wiser (Sep 04 2020 at 14:15, on Zulip):

My impression is that a lot of the work we desperately need to do is just completely infeasible for people to do one-off so while attracting new contributors is good, we need to make the ramp from one-time contribution to regular contributor smoother.

(At least, on the compiler side of things)

pnkfelix (Sep 04 2020 at 14:15, on Zulip):

(e.g. the initial landing page stuff is something that probably benefits the newcomers most)

nikomatsakis (Sep 04 2020 at 14:15, on Zulip):

agreed, let me take one more stab at saying what I was going to say: I do feel that historically we've focused more on the front of the funnel, I guess, than on the middle.

nikomatsakis (Sep 04 2020 at 14:16, on Zulip):

(anyway I think we should just dive in I guess)

nikomatsakis (Sep 04 2020 at 14:16, on Zulip):

that said I do feel that stuff like 'one clera source for information' will benefit newcomers and experienced folks alike

pnkfelix (Sep 04 2020 at 14:16, on Zulip):

so the summary spelled out the high level feedback from specific groups

matklad (Sep 04 2020 at 14:16, on Zulip):

also fixing some of the issues for group 1, might have a side-effect of fixing the 2nd group partially

There's also effect int he opposite direction: recurring contributors are mentors for the first-time contributors.

pnkfelix (Sep 04 2020 at 14:17, on Zulip):

by the way, depending on which view you use in https://hackmd.io/b-FKKh9DSnah-lj1NVKaAw?view#Summary , the comments on the right-hand side may or may not be visible.

pnkfelix (Sep 04 2020 at 14:17, on Zulip):

specifically, if you use the hybrid two column view, then you need to turn comments on explicitly

pnkfelix (Sep 04 2020 at 14:18, on Zulip):

(there's a :speech_bubble: at the top you click to do it)

mark-i-m (Sep 04 2020 at 14:18, on Zulip):

So I guess one thing that would benefit the first two groups a lot is streamline the process a bit:

mark-i-m (Sep 04 2020 at 14:18, on Zulip):
Oliver (Sep 04 2020 at 14:19, on Zulip):

Yeah my TL;DR notes are:

mark-i-m (Sep 04 2020 at 14:19, on Zulip):
pnkfelix (Sep 04 2020 at 14:19, on Zulip):

mark-i-m said:

does turning on incremental mode in config.toml help enough here? And if so, should we be advertising it more prominently?

Santiago Pastorino (Sep 04 2020 at 14:19, on Zulip):

I think what we really need is some kind of structure here, like, experts that mentor regular contributors to become experts, regular contributors that mentor people that landed one PR to become regular contributors, and first time PR people that mentor people that never landed a PR so they can do that. I kind of feel like an structure like that is what's better to try to build.

pnkfelix (Sep 04 2020 at 14:20, on Zulip):

We cannot turn incremental on by default I suspect (unfortunately)

nikomatsakis (Sep 04 2020 at 14:20, on Zulip):

The structure that @Santiago Pastorino mentioned is what we often aim to achieve with working groups.

mark-i-m (Sep 04 2020 at 14:20, on Zulip):

pnkfelix said:

mark-i-m said:

  • Making the compile/debug/edit cycle faster

does turning on incremental mode in config.toml help enough here? And if so, should we be advertising it more prominently?

Compared to most other projects, no. I think most people are expecting something like "compiles in 5 minutes on a normal machine"

simulacrum (Sep 04 2020 at 14:20, on Zulip):

I think we could perhaps, I'd be interested in hearing why you think we can't, but I've not historically seen incremental as a big win

DPC (Sep 04 2020 at 14:21, on Zulip):

currently for e-mentor issues, the person contributing has to contact the assigned/reviewer to get mentoring instructions. if we could have mentoring notes if possible on these issues, it would be easier for people to contribute

mark-i-m (Sep 04 2020 at 14:21, on Zulip):

Yep that's another thing people said

nikomatsakis (Sep 04 2020 at 14:21, on Zulip):

So to give a bit of structure, I propose we go forward by those three groups, devoting 10 minutes to each

mark-i-m (Sep 04 2020 at 14:21, on Zulip):

That even short notes on the issues by experts were helpful

nikomatsakis (Sep 04 2020 at 14:21, on Zulip):

and then 10 minutes for final conclusions?

nikomatsakis (Sep 04 2020 at 14:22, on Zulip):

I will keep a timer :)

nikomatsakis (Sep 04 2020 at 14:22, on Zulip):

Never had a PR merged

nikomatsakis (Sep 04 2020 at 14:22, on Zulip):

DPC said:

currently for e-mentor issues, the person contributing has to contact the assigned/reviewer to get mentoring instructions. if we could have mentoring notes if possible on these issues, it would be easier for people to contribute

nikomatsakis (Sep 04 2020 at 14:22, on Zulip):

mark-i-m said:

That even short notes on the issues by experts were helpful

nikomatsakis (Sep 04 2020 at 14:22, on Zulip):

I'm just highlighting those two comments because they were quite on point

nikomatsakis (Sep 04 2020 at 14:22, on Zulip):

something I've learned also is

nikomatsakis (Sep 04 2020 at 14:22, on Zulip):
nikomatsakis (Sep 04 2020 at 14:22, on Zulip):
nikomatsakis (Sep 04 2020 at 14:22, on Zulip):

both because you never write it

nikomatsakis (Sep 04 2020 at 14:23, on Zulip):

and because I think it's helpful for people to poke about

nikomatsakis (Sep 04 2020 at 14:23, on Zulip):

the closer you get to sending a diff, the less they will have to experiment and learn

nikomatsakis (Sep 04 2020 at 14:23, on Zulip):

one other thing I've noticed is that this is a great way to find places where our docs are lacking,

mark-i-m (Sep 04 2020 at 14:23, on Zulip):

Interesting -- that's a good insight

pnkfelix (Sep 04 2020 at 14:23, on Zulip):

nikomatsakis said:

the closer you get to sending a diff, the less they will have to experiment and learn

that's interesting

nikomatsakis (Sep 04 2020 at 14:23, on Zulip):

since I try to always link to rustc-dev-guide and to e.g. our API docs

pnkfelix (Sep 04 2020 at 14:24, on Zulip):

sometimes a diff is really important, but probably only for people who are already experts

Santiago Pastorino (Sep 04 2020 at 14:24, on Zulip):

and also an important thing is to try to teach people how to experiment

Santiago Pastorino (Sep 04 2020 at 14:24, on Zulip):

there are some things in the guide but unsure if what we have is enough

nikomatsakis (Sep 04 2020 at 14:24, on Zulip):

(one thing I will say is that more and more often there IS an appropriate page in the rustc-dev-guide to link to, which is awesome and credit goes to @WG-rustc-dev-guide for that)

nikomatsakis (Sep 04 2020 at 14:24, on Zulip):

yeah I mean it always needs more work ...

nikomatsakis (Sep 04 2020 at 14:25, on Zulip):

the other thing I am curious about --

nikomatsakis (Sep 04 2020 at 14:25, on Zulip):

a common theme is "PR reviews"

nikomatsakis (Sep 04 2020 at 14:25, on Zulip):

i.e., how fast are they

nikomatsakis (Sep 04 2020 at 14:25, on Zulip):

did that come up in the survey? I can't remember

mark-i-m (Sep 04 2020 at 14:25, on Zulip):

I think less so for new contributors (can't remember off the top of my head)

pnkfelix (Sep 04 2020 at 14:25, on Zulip):

(I assume we didn't ask how many of our contributors are using gdb (or lldb) as part of their workflow with rustc?)

DPC (Sep 04 2020 at 14:25, on Zulip):

that's important point because often people have a weekend to spare and are looking for something to work on that fits within that

mark-i-m (Sep 04 2020 at 14:25, on Zulip):

I think it matters more to frequent contributors

mark-i-m (Sep 04 2020 at 14:26, on Zulip):

Except in one case:

Joshua Nelson (Sep 04 2020 at 14:26, on Zulip):

pnkfelix said:

(I assume we didn't ask how many of our contributors are using gdb (or lldb) as part of their workflow with rustc?)

gdb is extremely painful to use when running rustc on the standard library/rustc itself

mark-i-m (Sep 04 2020 at 14:26, on Zulip):

when contributing to no-rustc/std, e.g. cargo, often PRs sit there untouched forever

Joshua Nelson (Sep 04 2020 at 14:26, on Zulip):

because it goes through bootstrap

mark-i-m (Sep 04 2020 at 14:26, on Zulip):

which is discouraging

pnkfelix (Sep 04 2020 at 14:26, on Zulip):

nikomatsakis said:

did that come up in the survey? I can't remember

I saw it mentioned here: https://hackmd.io/b-FKKh9DSnah-lj1NVKaAw?view#Takeaway-4

pnkfelix (Sep 04 2020 at 14:27, on Zulip):

@Joshua Nelson you mean because people don't know to use --verbose to see the actual command line to invoke first?

pnkfelix (Sep 04 2020 at 14:27, on Zulip):

(Which wouldn't surprise me; I just want to make sure I understand)

Joshua Nelson (Sep 04 2020 at 14:27, on Zulip):

pnkfelix said:

Joshua Nelson you mean because people don't know to use --verbose to see the actual command line to invoke first?

no, because --verbose is not sufficient to replicate what bootstrap is doing

Joshua Nelson (Sep 04 2020 at 14:27, on Zulip):

you need to also have various environment variables set

pnkfelix (Sep 04 2020 at 14:27, on Zulip):

oh right, those darn env variables

nikomatsakis (Sep 04 2020 at 14:27, on Zulip):

yeah that drives me crazy

pnkfelix (Sep 04 2020 at 14:27, on Zulip):

(Argh!)

Joshua Nelson (Sep 04 2020 at 14:27, on Zulip):

I spent like a week on it before giving up

nikomatsakis (Sep 04 2020 at 14:27, on Zulip):

we used to have a mode for that, @nagisa added something nifty..

pnkfelix (Sep 04 2020 at 14:27, on Zulip):

I've had bugs open forever to address that

Joshua Nelson (Sep 04 2020 at 14:27, on Zulip):

and I contribute ~fairly regularly

nikomatsakis (Sep 04 2020 at 14:27, on Zulip):

...but it stopped working at some point

nikomatsakis (Sep 04 2020 at 14:28, on Zulip):

one other observation I thought was interesting (though not unexpected) was the lack of E-mentor issues

nikomatsakis (Sep 04 2020 at 14:28, on Zulip):

it occurs to me that this is something we could definitely try to organize around, at least in theory

Joshua Nelson (Sep 04 2020 at 14:28, on Zulip):

nikomatsakis said:

one other observation I thought was interesting (though not unexpected) was the lack of E-mentor issues

I've been trying to help with that :)

Joshua Nelson (Sep 04 2020 at 14:28, on Zulip):

https://github.com/rust-lang/rust/issues?q=is%3Aopen+is%3Aissue+label%3AE-mentor

mark-i-m (Sep 04 2020 at 14:28, on Zulip):

Yep, that was definitely a big theme: unable to find something to work on

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

See also https://hackmd.io/b-FKKh9DSnah-lj1NVKaAw?view#Takeaway-7

mark-i-m (Sep 04 2020 at 14:29, on Zulip):

also among the second group, but we'll get there...

nikomatsakis (Sep 04 2020 at 14:29, on Zulip):

I feel like the ideal would be to have a mentoring group that (a) writes up some tips for good mentoring instructions and (b) tries to identify and add mentoring instructions to issues, but I worry that this might fail because it turns out that it requires kind of specialized knoweldge to do (b)

bjorn3 (Sep 04 2020 at 14:29, on Zulip):

Joshua Nelson said:

you need to also have various environment variables set

Doesn't the rustc wrapper output the whole invocation of the real rustc when using -vv?

nikomatsakis (Sep 04 2020 at 14:29, on Zulip):

my hope was that project groups could do it within their project, and sometimes that works, but it's hard to get to that state, you have to bootstrap the group enough to have enough bandwidth

mark-i-m (Sep 04 2020 at 14:29, on Zulip):

nikomatsakis said:

I feel like the ideal would be to have a mentoring group that (a) writes up some tips for good mentoring instructions and (b) tries to identify and add mentoring instructions to issues, but I worry that this might fail because it turns out that it requires kind of specialized knoweldge to do (b)

yes, I think this is one place where t-compiler needs to commit to doing it directly

nikomatsakis (Sep 04 2020 at 14:30, on Zulip):

yeah although mostly I meant

DPC (Sep 04 2020 at 14:30, on Zulip):

i've been working on the 2nd part a bit, sending issues to various newcomers to work on

nikomatsakis (Sep 04 2020 at 14:30, on Zulip):

I might be interested in getting a group of experienced folks together to spend an hour a week doing it or something

nikomatsakis (Sep 04 2020 at 14:30, on Zulip):

might be hard to sustain ...

mark-i-m (Sep 04 2020 at 14:30, on Zulip):

DPC said:

i've been working on the 2nd part a bit, sending issues to various newcomers to work on

How well has that worked?

mark-i-m (Sep 04 2020 at 14:30, on Zulip):

(genuinely curious)

nikomatsakis (Sep 04 2020 at 14:31, on Zulip):

1 more minute till we shift groups :)

Joshua Nelson (Sep 04 2020 at 14:31, on Zulip):

bjorn3 said:

Joshua Nelson said:

you need to also have various environment variables set

Doesn't the rustc wrapper output the whole invocation of the real rustc when using -vv?

not with env variables: https://gist.github.com/jyn514/0e4930b457532ec87a5e7c2cf80f491b

mark-i-m (Sep 04 2020 at 14:31, on Zulip):

I think some sort of regular way of tracking/creating e-mentor issues would be helpful

Joshua Nelson (Sep 04 2020 at 14:31, on Zulip):

maybe this is specific to rustdoc :shrug:

DPC (Sep 04 2020 at 14:31, on Zulip):

so far i can say around 60% of them have managed to contribute successfully and around 20% have managed to make more than 1 contribution as per my guesswork

mark-i-m (Sep 04 2020 at 14:31, on Zulip):

even if it was just to mention the number of e-mentor issues in the Thursday meeting and assign someone to create 5 more

mark-i-m (Sep 04 2020 at 14:32, on Zulip):

that way the load is a bit more diffuse

nikomatsakis (Sep 04 2020 at 14:32, on Zulip):

Contribute < 1 PR / month

nikomatsakis (Sep 04 2020 at 14:32, on Zulip):

mark-i-m said:

even if it was just to mention the number of e-mentor issues in the Thursday meeting and assign someone to create 5 more

that's an interesting idea

nikomatsakis (Sep 04 2020 at 14:32, on Zulip):

let's start with like 2

nikomatsakis (Sep 04 2020 at 14:32, on Zulip):

:)

nikomatsakis (Sep 04 2020 at 14:32, on Zulip):

(I should have said 1...)

DPC (Sep 04 2020 at 14:32, on Zulip):

i think it would be better to first look at existing e-mentor issues before adding more to the list

pnkfelix (Sep 04 2020 at 14:32, on Zulip):

mark-i-m said:

I think some sort of regular way of tracking/creating e-mentor issues would be helpful

and maybe measuring periodically how much progress we're making with: 1. identifying/tagging E-mentor issues, 2. writing up mentorship instructions, and 3. matching it with volunteer to assist with it.

nikomatsakis (Sep 04 2020 at 14:32, on Zulip):

So actually

nikomatsakis (Sep 04 2020 at 14:32, on Zulip):

the meaning of E-mentor in my opinion

nikomatsakis (Sep 04 2020 at 14:32, on Zulip):

should not be "someone is willing to mentor"

nikomatsakis (Sep 04 2020 at 14:33, on Zulip):

it should be "there are mentoring instructions"

nikomatsakis (Sep 04 2020 at 14:33, on Zulip):

since I think the former is kind of not that important ..

nikomatsakis (Sep 04 2020 at 14:33, on Zulip):

so I guess looking at E-mentor that are not yet having instructions makes sense

mark-i-m (Sep 04 2020 at 14:33, on Zulip):

DPC said:

i think it would be better to first look at existing e-mentor issues before adding more to the list

So actually, I think this is a good place to start with group 2

nikomatsakis (Sep 04 2020 at 14:33, on Zulip):

anyway, are there any things that are specific to 1 PR/month?

Joshua Nelson (Sep 04 2020 at 14:33, on Zulip):

nikomatsakis said:

anyway, are there any things that are specific to 1 PR/month?

churn in the contributing process maybe

pnkfelix (Sep 04 2020 at 14:33, on Zulip):

(well knowing who the expert is to contact on something might be semi-helpful. but written instructions are definitely better.)

Joshua Nelson (Sep 04 2020 at 14:33, on Zulip):

like documentation that moves

Joshua Nelson (Sep 04 2020 at 14:34, on Zulip):

and changes in bootstrap (though hopefully not too many more of those)

nikomatsakis (Sep 04 2020 at 14:34, on Zulip):

there are 7 takeaways :)

mark-i-m (Sep 04 2020 at 14:34, on Zulip):

I think mostly this group is feeling the growth pains

mark-i-m (Sep 04 2020 at 14:34, on Zulip):
mark-i-m (Sep 04 2020 at 14:34, on Zulip):
nikomatsakis (Sep 04 2020 at 14:35, on Zulip):
mark-i-m (Sep 04 2020 at 14:35, on Zulip):
Joshua Nelson (Sep 04 2020 at 14:35, on Zulip):

nikomatsakis said:

that was @pickfire I think

nikomatsakis (Sep 04 2020 at 14:35, on Zulip):

in general I think that our team/WG organization needs work :)

mark-i-m (Sep 04 2020 at 14:35, on Zulip):

It was anonymous, so I don't know

pnkfelix (Sep 04 2020 at 14:35, on Zulip):

nikomatsakis said:

every bullet like this seems like it corresponds to something we need to add to CONTRIBUTING.md or elsewhere

pnkfelix (Sep 04 2020 at 14:36, on Zulip):

(not necessarily CONTRIBUTING.md; I noted on the hackmd that for some reason that was cited as the most used document, but I don't understand how it could have been, given how little it contains.)

Joshua Nelson (Sep 04 2020 at 14:36, on Zulip):

mark-i-m said:

It was anonymous, so I don't know

oh I was just referring to https://rust-lang.zulipchat.com/#narrow/stream/122651-general/topic/Rust.20teams.20for.20public

mark-i-m (Sep 04 2020 at 14:36, on Zulip):

There was also a proposal at some point to make a contributors landing page... contribute.rlo or something

DPC (Sep 04 2020 at 14:36, on Zulip):

what if instead of people finding work, we flip it, we have a place where people can ask for something they can work on, and we distribute some of the work?

Wesley Wiser (Sep 04 2020 at 14:36, on Zulip):

Some of our working group memberships are "opt-in" and some are "by-invitation". Which I think definitely adds to the confusion.

mark-i-m (Sep 04 2020 at 14:36, on Zulip):

that could help to document everything and funnel people to the right places

pnkfelix (Sep 04 2020 at 14:37, on Zulip):

DPC said:

what if instead of people finding work, we flip it, we have a place where people can ask for something they can work on, and we distribute some of the work?

I don't quite understand the distinction you are drawing here

nikomatsakis (Sep 04 2020 at 14:37, on Zulip):

I've always imagined that we would have a clear page that's like "here are the groups", but it's harder to create than it seems, and hard to keep up to date

pnkfelix (Sep 04 2020 at 14:37, on Zulip):

Wesley Wiser said:

Some of our working group memberships are "opt-in" and some are "by-invitation". Which I think definitely adds to the confusion.

I didn't know some were by-invitation!

mark-i-m (Sep 04 2020 at 14:37, on Zulip):

I didn't either until I got the survey feedback

nikomatsakis (Sep 04 2020 at 14:37, on Zulip):

well, I think what @Wesley Wiser meant is kind of "some are not looking to mentor folks"

pnkfelix (Sep 04 2020 at 14:38, on Zulip):

I had thought they were all a "show up and see what's happening" flavor

nikomatsakis (Sep 04 2020 at 14:38, on Zulip):

or don't have bandwidth to add more people

pnkfelix (Sep 04 2020 at 14:38, on Zulip):

hmm

nikomatsakis (Sep 04 2020 at 14:38, on Zulip):

in the past I think you and I had described a "project" vs "project group" distinction, @pnkfelix

Wesley Wiser (Sep 04 2020 at 14:38, on Zulip):

Not trying to point fingers or anything but, for example when we started wg-prioritization it was "if you're interested, add yourself to the team repo"

nikomatsakis (Sep 04 2020 at 14:38, on Zulip):

i.e., it's worth tracking projects, but sometimes it just takes 1 or 2 people to pursue and that's the right number

pnkfelix (Sep 04 2020 at 14:39, on Zulip):

Wesley Wiser said:

Not trying to point fingers or anything but, for example when we started wg-prioritization it was "if you're interested, add yourself to the team repo"

is that somehow invite-only?

mark-i-m (Sep 04 2020 at 14:39, on Zulip):

oh, I took it as an example of the opposite

pnkfelix (Sep 04 2020 at 14:39, on Zulip):

It sounds to me to be "anyone can join, they just need to jump over this hurdle"

DPC (Sep 04 2020 at 14:39, on Zulip):

pnkfelix said:

DPC said:

what if instead of people finding work, we flip it, we have a place where people can ask for something they can work on, and we distribute some of the work?

I don't quite understand the distinction you are drawing here

so currently if people want to contribute, they often have to search the issue tracker for something to work on, which can be daunting because a) you need to find an issue that needs work b) they need to know the current status of the issue, which isn't easy in most cases (especially for older issues)

nikomatsakis (Sep 04 2020 at 14:40, on Zulip):

I guess that Takeaway 10 concerns PR reviews, but it seems like it's quite intermingled with other parts of the project that are kind of "out of scope"

Wesley Wiser (Sep 04 2020 at 14:40, on Zulip):

It sounds to me to be "anyone can join, they just need to jump over this hurdle"

What I meant was, in comparison to other WGs where you do some work and then the team leads offer you membership.

nikomatsakis (Sep 04 2020 at 14:40, on Zulip):

so actually maybe we should spend a bit of time on this point -- membership

nikomatsakis (Sep 04 2020 at 14:41, on Zulip):

although we're out of time for this section

DPC (Sep 04 2020 at 14:41, on Zulip):

what if instead if we could promote a place where people can say "pass me any issue to work on" and people who are familiar with the rust teams/issues can send them an issue to work on

nikomatsakis (Sep 04 2020 at 14:41, on Zulip):

but I am interested to hear what folks think about it

Wesley Wiser (Sep 04 2020 at 14:41, on Zulip):

(Again, not trying to point fingers or say one or the other is wrong but the policy is different for some wgs than others)

nikomatsakis (Sep 04 2020 at 14:41, on Zulip):

I've gone back and forth on how to track membership within WGs, or even whether to do so

mark-i-m (Sep 04 2020 at 14:41, on Zulip):

I kind of like @DPC's idea

pnkfelix (Sep 04 2020 at 14:41, on Zulip):

I'm wondering what people think of the T-compiler vs T-compiler/meeting structure

matklad (Sep 04 2020 at 14:41, on Zulip):

it should be "there are mentoring instructions"

Late to the party but we did rename E-menotor to E-has-instructions because of this in rust-analyzer

nikomatsakis (Sep 04 2020 at 14:41, on Zulip):

for some groups (e.g., wg-traits), membrship can be tied to repo rights

pnkfelix (Sep 04 2020 at 14:41, on Zulip):

where we have the group of "leaders" or "decision makers" in T-compiler

mark-i-m (Sep 04 2020 at 14:41, on Zulip):

It seems like organizing the available work is also a problem

nikomatsakis (Sep 04 2020 at 14:41, on Zulip):

but I would like some way for folks to "sign up" that is always opt-in, maybe along with higher levels that correspond to "Committer rights"

pnkfelix (Sep 04 2020 at 14:41, on Zulip):

but anyone can join T-compiler/meeting to just stay abreast of what's going on

nikomatsakis (Sep 04 2020 at 14:42, on Zulip):

yeah, so this T-compiler vs T-compiler/meeting distinction is kind of what I'm getting at

Joshua Nelson (Sep 04 2020 at 14:42, on Zulip):

Wesley Wiser said:

It sounds to me to be "anyone can join, they just need to jump over this hurdle"

What I meant was, in comparison to other WGs where you do some work and then the team leads offer you membership.

this sounds odd to me because I've only been on teams that are invite-only

nikomatsakis (Sep 04 2020 at 14:42, on Zulip):

@Santiago Pastorino and I had talked at some point about having the option in the team repo so that there is a 'notify' group for every team

Joshua Nelson (Sep 04 2020 at 14:43, on Zulip):

like, what is the distinction of being on a team if you jsut added your name to the document?

nikomatsakis (Sep 04 2020 at 14:43, on Zulip):

and then there are "members" that are invite only

Joshua Nelson (Sep 04 2020 at 14:43, on Zulip):

you can still make contributions if you're not an official member

nikomatsakis (Sep 04 2020 at 14:43, on Zulip):

@Joshua Nelson I think the answer is that you would get pings and things

mark-i-m (Sep 04 2020 at 14:43, on Zulip):

pnkfelix said:

but anyone can join T-compiler/meeting to just stay abreast of what's going on

That takes a lot of time and energy... I imagine only frequent contributors do that

nikomatsakis (Sep 04 2020 at 14:43, on Zulip):

but also it's a kind of 'commitment'

nikomatsakis (Sep 04 2020 at 14:43, on Zulip):

anyway, @DPC floated this idea that we haven't discussed much,

Wesley Wiser (Sep 04 2020 at 14:43, on Zulip):

I know a concern that has come up before is with the rust-lang.org/governance page and the Rust org logo. WG members get listed on that page and get the organization invite so essentially you can add yourself to those things without doing anything.

nikomatsakis (Sep 04 2020 at 14:44, on Zulip):

maybe we should turn to that and then move to the last set of contributors?

nikomatsakis (Sep 04 2020 at 14:44, on Zulip):

/me (eyes the clock)

Wesley Wiser (Sep 04 2020 at 14:44, on Zulip):

But also, that (the wg-prioritization thing) happened before we introduced the ping notification groups concept

DPC (Sep 04 2020 at 14:44, on Zulip):

( i feel membership deserves a separate discussion )

nikomatsakis (Sep 04 2020 at 14:44, on Zulip):

I do too

nikomatsakis (Sep 04 2020 at 14:44, on Zulip):

let's cut it off for now because I think the role of this meeting is more to survey (no pun intended) :)

pnkfelix (Sep 04 2020 at 14:45, on Zulip):

mark-i-m said:

pnkfelix said:

but anyone can join T-compiler/meeting to just stay abreast of what's going on

That takes a lot of time and energy... I imagine only frequent contributors do that

This is probably true; the @_T-compiler/meeting group is probably not as big as I had thought when I wrote that.

nikomatsakis (Sep 04 2020 at 14:45, on Zulip):

anyway I guess I'll say that @DPC I like the idea of having a place where people can request work, I think in general it fits with having a more organized set of folks devoted to 'newcomer experience'

nikomatsakis (Sep 04 2020 at 14:46, on Zulip):

which is an idea I'd like to dig into more, but prob in a separate meeting :)

mark-i-m (Sep 04 2020 at 14:46, on Zulip):

I have seen people do that a bit on zulip too, but yeah, moving on

Joshua Nelson (Sep 04 2020 at 14:46, on Zulip):

pnkfelix said:

mark-i-m said:

pnkfelix said:

but anyone can join T-compiler/meeting to just stay abreast of what's going on

That takes a lot of time and energy... I imagine only frequent contributors do that

This is probably true; the @_T-compiler/meeting group is probably not as big as I had thought when I wrote that.

as a point of interest, I wouldn't have looked at this channel if I hadn't gotten pinged

pnkfelix (Sep 04 2020 at 14:46, on Zulip):

A place where people can request work sounds good

mark-i-m (Sep 04 2020 at 14:46, on Zulip):

So the last group is frequent contributors

nikomatsakis (Sep 04 2020 at 14:46, on Zulip):

Contribute >= 1/month

pnkfelix (Sep 04 2020 at 14:46, on Zulip):

I don't know if zulip is the best place to host such a "work request channel"

nikomatsakis (Sep 04 2020 at 14:47, on Zulip):

The compiler is by far the most frequently contributed-to codebase, 7/10 reporting that they contributed to the compiler

mark-i-m (Sep 04 2020 at 14:47, on Zulip):

The primary pain point here seems to be "ergonomics"

mark-i-m (Sep 04 2020 at 14:47, on Zulip):

of various things

nikomatsakis (Sep 04 2020 at 14:47, on Zulip):

Improving the rustc-dev-guide helps frequent contributors the most.

mark-i-m (Sep 04 2020 at 14:47, on Zulip):

Work done to help new contributors will also help frequent contributors

This was surprising to me

nikomatsakis (Sep 04 2020 at 14:47, on Zulip):

that is interesting and not whatI expected

mark-i-m (Sep 04 2020 at 14:47, on Zulip):

you beat me to it

nikomatsakis (Sep 04 2020 at 14:47, on Zulip):

but it makes sense

nikomatsakis (Sep 04 2020 at 14:47, on Zulip):

(oh, sorry)

nikomatsakis (Sep 04 2020 at 14:47, on Zulip):

I'll let you keep going :)

nikomatsakis (Sep 04 2020 at 14:48, on Zulip):

Work done to help new contributors will also help frequent contributors

cough core team rustconf keynote cough :)

Joshua Nelson (Sep 04 2020 at 14:48, on Zulip):

+1 to better compile times, this is my #1 complaint and the #1 complaint of several other people I've worked with

mark-i-m (Sep 04 2020 at 14:48, on Zulip):

So yeah, I guess getting more content into the guide would be a good thing to push -- and that seems to already be a trend

mark-i-m (Sep 04 2020 at 14:49, on Zulip):

Yes, I'll add that it prevents some people from getting into the freq contrib group

nikomatsakis (Sep 04 2020 at 14:49, on Zulip):

PRs often lack context, why created?

mark-i-m (Sep 04 2020 at 14:49, on Zulip):

e.g. I don't have a machine powerful enough to build the compiler in < 2 hours

Joshua Nelson (Sep 04 2020 at 14:49, on Zulip):

https://github.com/rust-lang/rust/issues/65031 seems like a good starting point for that

nikomatsakis (Sep 04 2020 at 14:49, on Zulip):

that one is interesting because it comes from the "other direction"

DPC (Sep 04 2020 at 14:50, on Zulip):

not sure if it helps, but @Joshua Nelson in the past suggested a s-waiting-on-help label where in we know that an author needs help from the reviewer/associated team

Joshua Nelson (Sep 04 2020 at 14:51, on Zulip):

the meeting I discussed it at decided that the use cases were covered by waiting-on-review

Joshua Nelson (Sep 04 2020 at 14:51, on Zulip):

which, fair, but better labelling always helps

Joshua Nelson (Sep 04 2020 at 14:51, on Zulip):

I mostly opened it in response to triage being over-eager to switch to waiting-on-author though

Joshua Nelson (Sep 04 2020 at 14:51, on Zulip):

which might be a separate issue

nikomatsakis (Sep 04 2020 at 14:51, on Zulip):

compilation time seems like both the biggest but also the hardest problem

pnkfelix (Sep 04 2020 at 14:51, on Zulip):

so are there things we can do specifically for building the compiler that would improve the edit-compile-debug cycle

Wesley Wiser (Sep 04 2020 at 14:51, on Zulip):

mark-i-m said:

e.g. I don't have a machine powerful enough to build the compiler in < 2 hours

Does that have to be true today? Assuming you have a decent machine (maybe i5 equivalent with 8gb of RAM) I feel like you should be able to build a stage 1 rustc in less than two hours. Just by having the appropriate settings in config.toml.

nikomatsakis (Sep 04 2020 at 14:51, on Zulip):

we've taken some steps to address it lately thought

mark-i-m (Sep 04 2020 at 14:52, on Zulip):

Yes, I could also imagine that being a thing that's tracked/triaged somewhere -- # of open -need-help PRs

nikomatsakis (Sep 04 2020 at 14:52, on Zulip):

such as changing x.py to stage1

pnkfelix (Sep 04 2020 at 14:52, on Zulip):

there used to be standard tricks involving keep-stage

pnkfelix (Sep 04 2020 at 14:52, on Zulip):

but maybe that's not necessary with the recent changes to x.py that niko just mentioned?

nikomatsakis (Sep 04 2020 at 14:52, on Zulip):

it is also true that the more we are able to create standalone libraries, the less of a problem that is

DPC (Sep 04 2020 at 14:52, on Zulip):

@Wesley Wiser are people aware of the appropriate settings?

mark-i-m (Sep 04 2020 at 14:52, on Zulip):

Wesley Wiser said:

mark-i-m said:

e.g. I don't have a machine powerful enough to build the compiler in < 2 hours

Does that have to be true today? Assuming you have a decent machine (maybe i5 equivalent with 8gb of RAM) I feel like you should be able to build a stage 1 rustc in less than two hours. Just by having the appropriate settings in config.toml.

I tend to run ./x.py check and then use CI

Joshua Nelson (Sep 04 2020 at 14:52, on Zulip):

pnkfelix said:

there used to be standard tricks involving keep-stage

it would be good not to need tricks

Wesley Wiser (Sep 04 2020 at 14:52, on Zulip):

I don't know. Am I totally off base there?

Joshua Nelson (Sep 04 2020 at 14:52, on Zulip):

and that doesn't help with a stage 1 build

Joshua Nelson (Sep 04 2020 at 14:53, on Zulip):

which is normally the blocker

pnkfelix (Sep 04 2020 at 14:53, on Zulip):

Joshua Nelson said:

pnkfelix said:

there used to be standard tricks involving keep-stage

it would be good not to need tricks

yes I agree with this. I just want to understand what the people are doing

mark-i-m (Sep 04 2020 at 14:53, on Zulip):

And there are some PRs, especially large refactors that require running everything

nikomatsakis (Sep 04 2020 at 14:53, on Zulip):

I find I no longer use the various tricks, but I also have a fairly powerful desktop that can build rustc much faster. I don't even attempt to build it on my laptop, I admit.

Joshua Nelson (Sep 04 2020 at 14:53, on Zulip):

mark-i-m said:

Wesley Wiser said:

mark-i-m said:

e.g. I don't have a machine powerful enough to build the compiler in < 2 hours

Does that have to be true today? Assuming you have a decent machine (maybe i5 equivalent with 8gb of RAM) I feel like you should be able to build a stage 1 rustc in less than two hours. Just by having the appropriate settings in config.toml.

I tend to run ./x.py check and then use CI

this only works for refactoring things :/

pnkfelix (Sep 04 2020 at 14:53, on Zulip):

specifically, there were periods of time where "experts" were regularly using keep-stage to get quick feedback cycles, and I was unaware of it. and then I would start using it but it would periodically break

Joshua Nelson (Sep 04 2020 at 14:53, on Zulip):

if you're adding new functionality check is not enough, you need to make sure it works

pnkfelix (Sep 04 2020 at 14:54, on Zulip):

(because you couldn't rely on it to work properly in response to all changes )

nikomatsakis (Sep 04 2020 at 14:54, on Zulip):

5 minutes left

DPC (Sep 04 2020 at 14:54, on Zulip):

i think we could slightly push the idea that "CI is your friend" and it is fine to rely on the CI at times

nikomatsakis (Sep 04 2020 at 14:54, on Zulip):

yeah, that seems like a useful workflow, in lieu of having some way to give people acess to distributed build clusters or something :)

pnkfelix (Sep 04 2020 at 14:54, on Zulip):

(also, this comes back to my question about whether people can/should/do use incremental=true in their config.toml )

Joshua Nelson (Sep 04 2020 at 14:55, on Zulip):

DPC said:

i think we could slightly push the idea that "CI is your friend" and it is fine to rely on the CI at times

again, this only works for refactoring

pnkfelix (Sep 04 2020 at 14:55, on Zulip):

Joshua Nelson said:

DPC said:

i think we could slightly push the idea that "CI is your friend" and it is fine to rely on the CI at times

again, this only works for refactoring

? why do you say that?

Joshua Nelson (Sep 04 2020 at 14:55, on Zulip):

not opposed to it, but it doesn't fix the issue

nikomatsakis (Sep 04 2020 at 14:55, on Zulip):

so, I guess in the last 5 minutes, we can discuss any final conclusions? I'd like to maybe recap the most interesting or promising points that were raised

mark-i-m (Sep 04 2020 at 14:55, on Zulip):

Joshua Nelson said:

if you're adding new functionality check is not enough, you need to make sure it works

If you have good tests, then the CI does make sure it "works"

Joshua Nelson (Sep 04 2020 at 14:55, on Zulip):

mark-i-m said:

Joshua Nelson said:

if you're adding new functionality check is not enough, you need to make sure it works

If you have good tests, then the CI does make sure it "works"

15 minute turnaround times are too slow

pnkfelix (Sep 04 2020 at 14:55, on Zulip):

relying the CI will e.g. run the test suite, right?

DPC (Sep 04 2020 at 14:55, on Zulip):

Joshua Nelson said:

DPC said:

i think we could slightly push the idea that "CI is your friend" and it is fine to rely on the CI at times

again, this only works for refactoring

yeh won't work for all cases, but at least for some of them

mark-i-m (Sep 04 2020 at 14:55, on Zulip):

(but also, yeah, I do a lot of refactoring PRs :stuck_out_tongue:)

Joshua Nelson (Sep 04 2020 at 14:55, on Zulip):

there's no way you can get a feature done in less than a week with that

Joshua Nelson (Sep 04 2020 at 14:55, on Zulip):

and it's also really frustrating

pnkfelix (Sep 04 2020 at 14:55, on Zulip):

oh okay you're just saying that the CI turnaround time is too long

DPC (Sep 04 2020 at 14:55, on Zulip):

and then conflicts :stuck_out_tongue:

pnkfelix (Sep 04 2020 at 14:55, on Zulip):

that I agree with

nikomatsakis (Sep 04 2020 at 14:56, on Zulip):

nikomatsakis said:

so, I guess in the last 5 minutes, we can discuss any final conclusions? I'd like to maybe recap the most interesting or promising points that were raised

Joshua Nelson (Sep 04 2020 at 14:56, on Zulip):

no I'm saying that pushing to CI is too slow

Joshua Nelson (Sep 04 2020 at 14:56, on Zulip):

like, I wouldn't contribute so frequently if i couldn't build rustc locally

Joshua Nelson (Sep 04 2020 at 14:56, on Zulip):

regardless of how fast CI was

DPC (Sep 04 2020 at 14:56, on Zulip):

( i don't know i generally push at night before sleeping, and wake up to CI errors, so i don't care how long it takes :stuck_out_tongue: )

mark-i-m (Sep 04 2020 at 14:56, on Zulip):

Joshua Nelson said:

mark-i-m said:

Joshua Nelson said:

if you're adding new functionality check is not enough, you need to make sure it works

If you have good tests, then the CI does make sure it "works"

15 minute turnaround times are too slow

It's better than 30 minutes or more ... :confused:

Wesley Wiser (Sep 04 2020 at 14:56, on Zulip):

If you're working on a feature, the pain of the initial stage 1 build with incremental seems worth the relatively quick turn around times there after.

Joshua Nelson (Sep 04 2020 at 14:56, on Zulip):

DPC said:

( i don't know i generally push at night before sleeping, and wake up to CI errors, so i don't care how long it takes :P )

this is what I mean by adding features taking a week

nikomatsakis (Sep 04 2020 at 14:57, on Zulip):

please take compilation time discussion to a separate topic I think

mark-i-m (Sep 04 2020 at 14:57, on Zulip):

pnkfelix said:

oh okay you're just saying that the CI turnaround time is too long

It's not really a CI problem... this is an abuse of CI

mark-i-m (Sep 04 2020 at 14:58, on Zulip):

as in this is a workaround for another problem

Wesley Wiser (Sep 04 2020 at 14:58, on Zulip):

Take away: The rustc dev guide is immensely helpful to a wide range of contributors and we should continue pushing for people to contribute to it. Especially when they make changes to the compiler.

pnkfelix (Sep 04 2020 at 14:58, on Zulip):

(the CI discussion is part of the compilation time one, lets move it elsewhere)

mark-i-m (Sep 04 2020 at 14:58, on Zulip):

oh, sorry

nikomatsakis (Sep 04 2020 at 14:58, on Zulip):

I think one takeaway for me was the general idea of trying to map out the contribution pathways at various levels in more detail -- I like the idea of actively monitoring stats like E-mentor bugs and having effot put to increasing the number -- but I also wonder if we hav the bandwidth for it

DPC (Sep 04 2020 at 14:58, on Zulip):

the triage-wg can help in monitoring e-mentor issues if needed

nikomatsakis (Sep 04 2020 at 14:58, on Zulip):

but I think this covers both like actively increasing E-mentor, @DPC's suggestions of ways to request work, etc.

pnkfelix (Sep 04 2020 at 14:58, on Zulip):

if we could programmatically extract the stats

pnkfelix (Sep 04 2020 at 14:58, on Zulip):

then monitoring E-mentor wolud be reasonable.

nikomatsakis (Sep 04 2020 at 14:59, on Zulip):

This is not a take-away, but one other thing: I've notcied that #new members is a pretty great and active channel, and people often get useful advice there.

pnkfelix (Sep 04 2020 at 14:59, on Zulip):

but that would require some way to tell if mentoring instructions are actually added...

nikomatsakis (Sep 04 2020 at 14:59, on Zulip):

(I think #new members shows an appetite for a place for people to show up, introduce themselves, and get some help in getting oriented)

mark-i-m (Sep 04 2020 at 15:00, on Zulip):

I would propose one other thing: the compile time problems seem to point at librarification

Wesley Wiser (Sep 04 2020 at 15:00, on Zulip):

Do you mean for librarification or against?

mark-i-m (Sep 04 2020 at 15:00, on Zulip):

that is, people who contributed to things like chalk often mentioned that it was a lot easier

Wesley Wiser (Sep 04 2020 at 15:00, on Zulip):

Got it

mark-i-m (Sep 04 2020 at 15:00, on Zulip):

for librarification

nikomatsakis (Sep 04 2020 at 15:00, on Zulip):

I think it also helps with other things, like large intimidating codebase etc

nikomatsakis (Sep 04 2020 at 15:00, on Zulip):

or at least it can

Santiago Pastorino (Sep 04 2020 at 15:01, on Zulip):

nikomatsakis said:

This is not a take-away, but one other thing: I've notcied that #new members is a pretty great and active channel, and people often get useful advice there.

one first good step we're making is that people often say hi there and they always get very nice and encouraging welcoming messages, which is at least encouraging to make new people feel good, next step is well ... what we are talking about :), turn them into contributors :)

matklad (Sep 04 2020 at 15:01, on Zulip):

I would propose one other thing: the compile time problems seem to point at librarification

Im gonna write a whole Rust in 2021 blog post about this....

nikomatsakis (Sep 04 2020 at 15:01, on Zulip):

All right, we are at time

DPC (Sep 04 2020 at 15:01, on Zulip):

it also gives an implicit license for people to experiment more freely

nikomatsakis (Sep 04 2020 at 15:01, on Zulip):

I want to give a huge thank you to @mark-i-m for both making this survey happen and doing such a thorough job of analyzing the results

mark-i-m (Sep 04 2020 at 15:01, on Zulip):

Also a bit of a stretch: but I can also imagine it being an argument for being able to build the compiler, std, etc separately. e.g. moving std/core/alloc/etc to another repo and making them buildable with beta (which I had discussed with @simulacrum at some point)

pnkfelix (Sep 04 2020 at 15:01, on Zulip):

do we have any action items now? Probably nothing explicit

nikomatsakis (Sep 04 2020 at 15:01, on Zulip):

I suspect we're going to be looking at this data for some time yet to test hypotheses

pnkfelix (Sep 04 2020 at 15:01, on Zulip):

Oh yes, thank you so much @mark-i-m

nikomatsakis (Sep 04 2020 at 15:02, on Zulip):

I don't think we have concrete action items, but I think that's ok, I kind of expected mostly to brainstorm in this meeting

Wesley Wiser (Sep 04 2020 at 15:02, on Zulip):

I think there's also some problems with it (librification). If you have good hardware, it's great. If you're memory constrained, having all those link jobs running at the time hits swap and your compile times go sky high.

nikomatsakis (Sep 04 2020 at 15:02, on Zulip):

I would like someone to write minutes for this meeting

Joshua Nelson (Sep 04 2020 at 15:02, on Zulip):

Wesley Wiser said:

I think there's also some problems with it (librification). If you have good hardware, it's great. If you're memory constrained, having all those link jobs running at the time hits swap and your compile times go sky high.

you can turn that down with -j2 though

nikomatsakis (Sep 04 2020 at 15:02, on Zulip):

I think it's more important than most

nikomatsakis (Sep 04 2020 at 15:02, on Zulip):

(I can do that if nobody has time though)

nikomatsakis (Sep 04 2020 at 15:03, on Zulip):

but I think we should assign someone...

Wesley Wiser (Sep 04 2020 at 15:03, on Zulip):

I would volunteer but I owe one or two meeting notes already and I don't think I have the time to get to this soon.

nikomatsakis (Sep 04 2020 at 15:03, on Zulip):

heh I owe a bunch too :P

nikomatsakis (Sep 04 2020 at 15:04, on Zulip):

maybe I should not have volunteered :P

nikomatsakis (Sep 04 2020 at 15:04, on Zulip):

but if nobody else signs up I'll just do it right now, as I said I think this one is more important

pnkfelix (Sep 04 2020 at 15:04, on Zulip):

by the way, here is a topic I just created to discuss compile times: #t-compiler > improving bootstrap times for contributor fun and profit

Santiago Pastorino (Sep 04 2020 at 15:07, on Zulip):

nikomatsakis said:

but if nobody else signs up I'll just do it right now, as I said I think this one is more important

I'd volunteer too but don't have time neither :(

Santiago Pastorino (Sep 04 2020 at 15:07, on Zulip):

I was wondering actually, wouldn't it make sense for following meetings to define someone that takes note beforehand?

Santiago Pastorino (Sep 04 2020 at 15:07, on Zulip):

or even have that written with the proposal

Santiago Pastorino (Sep 04 2020 at 15:07, on Zulip):

it may be easier to take notes as we go

Santiago Pastorino (Sep 04 2020 at 15:07, on Zulip):

at least for me

nikomatsakis (Sep 04 2020 at 15:07, on Zulip):

I do think it makes sense, but somehow it rarely happens :)

Santiago Pastorino (Sep 04 2020 at 15:07, on Zulip):

right :)

DPC (Sep 04 2020 at 15:07, on Zulip):

t-compiler/wg-notetakers :grinning:

nikomatsakis (Sep 04 2020 at 15:07, on Zulip):

Santiago Pastorino said:

it may be easier to take notes as we go

I used to try that

nikomatsakis (Sep 04 2020 at 15:07, on Zulip):

and sometimes I do

nikomatsakis (Sep 04 2020 at 15:08, on Zulip):

but it's also kind of nice to read back over the conversation

nikomatsakis (Sep 04 2020 at 15:08, on Zulip):

anyway, we should break this into a separate topic probably ;)

Santiago Pastorino (Sep 04 2020 at 15:08, on Zulip):

yep :+1:

nikomatsakis (Sep 04 2020 at 15:08, on Zulip):

it does really help when people add :point_up: and so forth

nikomatsakis (Sep 04 2020 at 15:08, on Zulip):

or any emoji really

DPC (Sep 04 2020 at 15:08, on Zulip):

maybe we could discuss this in wg-meta

Joshua Nelson (Sep 04 2020 at 16:45, on Zulip):

oh related to the survey: someone on discord said they specifically avoided PR-ing rust-lang/rust if they could avoid it, let me go ask them why

Joshua Nelson (Sep 04 2020 at 16:45, on Zulip):

(maybe @Lokathor ?)

Joshua Nelson (Sep 04 2020 at 17:04, on Zulip):

yay they said they'll write a blog post about it sometime soon :heart: will link it when they do

Joshua Nelson (Sep 06 2020 at 19:48, on Zulip):

Joshua Nelson said:

yay they said they'll write a blog post about it sometime soon :heart: will link it when they do

ok, I spent a lot of today on this.

The tl;dr is that we do a lot of things wrong, and there are a lot of reasons they avoid PRs: https://hackmd.io/RHyLd-27QXOR5AHa-oa6Fg?view. This is not super favorable to the rust project, so I'm hitting the highlights here in a less frustrated tone. Apologies for the wall of text.

Some other pieces of knowledge:

Lokathor (Sep 06 2020 at 19:53, on Zulip):

I am available for any further follow-up questions people have about what I wrote. Each major section of the document (background, survey, meeting) was written on a separate day, so a few points might be less organized than the ideal.

Jubilee (Sep 06 2020 at 22:12, on Zulip):

Even people who think they are available for mentorship should basically assume that a new person is _terrified to ask_ (or too stubborn, or w/e... the same motivational energy gets felt in different ways) and is going to be lucky to be able to summon the mental energy to overcome their conditioning to ask once.

Joshua Nelson (Sep 06 2020 at 22:32, on Zulip):

S-request-for-help would also make it easier to do drive-by reviews I think

Joshua Nelson (Sep 06 2020 at 22:32, on Zulip):

I can tell the difference between 'needs knowledge of the compiler' and 'needs help getting started'

Jubilee (Sep 06 2020 at 22:52, on Zulip):

which git GUI do you use, Lokathor? Like, "I dunno, it came with (XYZ)" is fine, I just need that much.

Lokathor (Sep 06 2020 at 22:54, on Zulip):

I use both the official Github for Windows Desktop, and also the VS Code git plugin (on the macbook).

Joshua Nelson (Sep 06 2020 at 23:02, on Zulip):

Lokathor said:

I use both the official Github for Windows Desktop, and also the VS Code git plugin (on the macbook).

FYI this might be helpful https://docs.github.com/en/desktop/contributing-and-collaborating-using-github-desktop/syncing-your-branch#rebasing-your-project-branch-onto-another-branch

Jubilee (Sep 06 2020 at 23:08, on Zulip):

@Lokathor I know you're on Windows mostly but do you have the Git Bash for Windows environment on that and can run a shell script? Someone mentioned removing x.py from the equation entirely in favor of a shell script and as rustbuild has grown stronger (harder, better, faster), that has become increasingly plausible.

Not much of a win if it just splits into a shell script and a Powershell script tho'.

Lokathor (Sep 06 2020 at 23:09, on Zulip):

Might I suggest cargo-make or cargo-xtask or similar "teach rust to do this thing" paths forward.

Lokathor (Sep 06 2020 at 23:10, on Zulip):

though I do have some sort of msys shell thing I think, not working in cmd.exe or powershell would probably be a bad move overall.

Joshua Nelson (Sep 06 2020 at 23:10, on Zulip):

I mentioned this in DMS too - I don't really see getting rid of x.py being feasible

Joshua Nelson (Sep 06 2020 at 23:10, on Zulip):

adding a cargo xbuild alias seems easy enough

Joshua Nelson (Sep 06 2020 at 23:11, on Zulip):

but it's never going to replace src/bootstrap/bootstrap.py

Jubilee (Sep 06 2020 at 23:11, on Zulip):

fwiw I agree, I just wanted to know because uh... removing Python, if it would be a strict improvement, would be a good idea _even for Unix users_. But if it's not, it's not.

Jubilee (Sep 06 2020 at 23:13, on Zulip):

( I have had to fix people's Python packaging hells enough times I view it as suspicious as a keystone dependency. :P ...but I mean, at least it's not Ruby Bundler Hell. )

simulacrum (Sep 06 2020 at 23:22, on Zulip):

Removing x.py seems entirely feasible -- removing bootstrap does not. But making the entry point be something like cargo run -- check as alternative to ./x.py check is definitely feasible.

Jubilee (Sep 06 2020 at 23:24, on Zulip):

"bootstrap" or bootstrap.py tho'

simulacrum (Sep 06 2020 at 23:24, on Zulip):

Python.

simulacrum (Sep 06 2020 at 23:25, on Zulip):

(At least as a required part of Rust. I'm not sure what's needed to build LLVM).

Jubilee (Sep 06 2020 at 23:26, on Zulip):

Rust's Python runs at least without any package deps, right? Hm.

simulacrum (Sep 06 2020 at 23:27, on Zulip):

Anyway, let's move this to separate thread(s) if you want to discuss further, I don't want to pollute the meeting thread

Joshua Nelson (Sep 09 2020 at 01:13, on Zulip):

opened https://github.com/rust-lang/rust/issues/76503 in response to Lokathor's post and some other ideas I've had bouncing around

Joshua Nelson (Sep 30 2020 at 17:51, on Zulip):

I took another look at the post now that some changes are landed. The things left are:

Needs implementation

Needs design

Needs decision

Not feasible

Joshua Nelson (Sep 30 2020 at 17:53, on Zulip):

IMO the biggest roadblocks are review latency and compile times, which are unfortunately also the hardest to fix

Joshua Nelson (Sep 30 2020 at 17:55, on Zulip):

I had a discussion with wg-triage about pinging reviewers more often a while ago, but it never went anywhere: https://rust-lang.zulipchat.com/#narrow/stream/242269-t-release.2Ftriage/topic/Criteria.20for.20triage.20to.20look.20at.20a.20PR

Joshua Nelson (Oct 27 2020 at 18:35, on Zulip):

I wrote up meeting notes: https://hackmd.io/daPDR2Q7Ryey0Q2gqgSM2w

Last update: Nov 25 2020 at 01:30UTC