Stream: t-compiler

Topic: steering meeting 2019.03.22 #58850


nikomatsakis (Mar 22 2019 at 13:56, on Zulip):

Dear @T-compiler/meeting, steering meeting in five minutes.

nikomatsakis (Mar 22 2019 at 13:57, on Zulip):

We've been discussing the agenda in another topic, and here is the plan.

nikomatsakis (Mar 22 2019 at 13:58, on Zulip):

In terms of the major topic, I'll be creating a poll to select it (see below). While we're voting on the poll, we'll begin with 5-15min of announcements. The idea is that people can add something that they think is cool, and maybe we'll discuss it a bit.

nikomatsakis (Mar 22 2019 at 13:58, on Zulip):

The two proposed topics for today are:

We've got a lot of good projects planned, but I'm concerned that we need to also be sure we are understand the problems people face in everyday compiling. Among other things, I suspect there is still low-hanging fruit to be picked, but also I'm not convinced we have a good picture of what consumes the time when compiling (and it might not even be in the compiler). I'd like us to be able to say "not only are we working on RLS 2.0, but we've made serde 5% faster" (or whatever). Discuss.

Can we try to enumerate what are some of things the compiler team needs to be tracking? What communication is (or should) be happening? I'd like us to be able to come up with a good list of all the things we think should be happening, so that we can compare against our existing processes, and try to ensure they are covered (or adjust the processes as needed).

nikomatsakis (Mar 22 2019 at 13:58, on Zulip):

/poll Topic of the day

nikomatsakis (Mar 22 2019 at 13:59, on Zulip):

(Or add your own, if you have a good idea)

nikomatsakis (Mar 22 2019 at 14:02, on Zulip):

# Announcements

nikomatsakis (Mar 22 2019 at 14:03, on Zulip):

Let's get started @T-compiler/meeting -- anybody have any announcements they'd like to share?

nikomatsakis (Mar 22 2019 at 14:03, on Zulip):

I feel like I had some but now I have to think what they were :)

oli (Mar 22 2019 at 14:04, on Zulip):

@WG-diagnostics is planning to use the annotate-snippets crate instead of our internal diagnostics API and get rid of our in-tree diagnostics wild growth

nikomatsakis (Mar 22 2019 at 14:05, on Zulip):

Ah, I guess I'll share one. In the @WG-meta meeting yesterday, we were discussing the plan to open an RFC concerning adding a new "role" for the compiler team. The current plan is to call this "regular contributors" (as in, contributing on a regular basis). Though part of my hope is that we'll bikeshed the name a bit. this is part of an overall effort to identify various levels of contribution -- anyway, if you're interested in that topic, please come and discuss in #t-compiler/wg-meta. @davidtwco were planning on working on the RFC together.

nikomatsakis (Mar 22 2019 at 14:05, on Zulip):

is planning to use the annotate-snippets crate instead of our internal diagnostics API and get rid of our in-tree diagnostics wild growth

ooh, nice! That crate is not part of the rust-lang org, though, right?

Zoxc (Mar 22 2019 at 14:06, on Zulip):

I almost have a series of PRs pushing queries back to parsing ready. @mw will have some reviewing work when he gets back =P

nikomatsakis (Mar 22 2019 at 14:06, on Zulip):

(I'm all for sharing, but would like to ensure we also have the ability to push bug fixes etc for a crucial bit of our infrastructure)

oli (Mar 22 2019 at 14:06, on Zulip):

nope, it's a complete rewrite by @zbraniecki , we might have to ask them to move it into rust-lang

nikomatsakis (Mar 22 2019 at 14:07, on Zulip):

I almost have a series of PRs pushing queries back to parsing ready.

Wow, very cool. Do you have any kind of sketch of the overall query setup? (e.g., how fine-grained they are and so forth)

Zoxc (Mar 22 2019 at 14:08, on Zulip):

I'm starting out with very course grained queries (one of parsing, one for macro expansion + name resolution), which are non-incremental.

nikomatsakis (Mar 22 2019 at 14:08, on Zulip):

Makes sense.

nikomatsakis (Mar 22 2019 at 14:09, on Zulip):

Reminder: there is a a poll for selecting the major topic here (in case you came late and skipped to the bottom)

Zoxc (Mar 22 2019 at 14:09, on Zulip):

They're basically the same queries as rustc_interface had.

nikomatsakis (Mar 22 2019 at 14:11, on Zulip):

Question: Are any of you going to Rust LATAM next week? (I know @Santiago Pastorino will be there =)

nikomatsakis (Mar 22 2019 at 14:11, on Zulip):

If so, come up and say hi...

Zoxc (Mar 22 2019 at 14:12, on Zulip):

I think replacing rustc_metadata with storing query results would be a good way to gain compilation speed, but it might not be too short-term, and probably want @mw around to chat about that. Not sure we'd have any people to work on it either right now

nikomatsakis (Mar 22 2019 at 14:12, on Zulip):

Yeah. Sounds good though!

nikomatsakis (Mar 22 2019 at 14:13, on Zulip):

OK, let's turn to the main topic. Based on the poll, let's discuss

Can we try to enumerate what are some of things the compiler team needs to be tracking? What communication is (or should) be happening? I'd like us to be able to come up with a good list of all the things we think should be happening, so that we can compare against our existing processes, and try to ensure they are covered (or adjust the processes as needed).

pnkfelix (Mar 22 2019 at 14:13, on Zulip):

niko you had already prepared a pair of lists

pnkfelix (Mar 22 2019 at 14:13, on Zulip):

are they in a shared doc somewhere, or shall I make one?

davidtwco (Mar 22 2019 at 14:13, on Zulip):

One small announcement from me, I have a working branch locally that does the variant/ctor id separation for enum variants that unblocks RFC 2008. Still need to tidy it up a little, but I'll have a PR up soon.

nikomatsakis (Mar 22 2019 at 14:13, on Zulip):

So, I've created a paper doc here with some notes

nikomatsakis (Mar 22 2019 at 14:13, on Zulip):

One small announcement from me, I have a working branch locally that does the variant/ctor id separation for enum variants that unblocks RFC 2008. Still need to tidy it up a little, but I'll have a PR up soon.

wow, nice!!

nikomatsakis (Mar 22 2019 at 14:14, on Zulip):

You've been working on that for a while =)

nikomatsakis (Mar 22 2019 at 14:14, on Zulip):

I had kind of given up hope a mere mortal could do it.

nikomatsakis (Mar 22 2019 at 14:14, on Zulip):

So, I've created a paper doc here with some notes

So yeah @pnkfelix this was my doc, but I didn't quite finish organizing it

nikomatsakis (Mar 22 2019 at 14:14, on Zulip):

I'll transcribe some of the stuff here

davidtwco (Mar 22 2019 at 14:14, on Zulip):

You've been working on that for a while =)

RFC 2008 impl was my first contribution, I've attempted to do this refactoring once every two months since then.

nikomatsakis (Mar 22 2019 at 14:14, on Zulip):

I think the easiest way to do this is probably to let people chat

nikomatsakis (Mar 22 2019 at 14:15, on Zulip):

And I'll come back and try to transcribe later (or maybe others can help out)

nikomatsakis (Mar 22 2019 at 14:15, on Zulip):

One thing I recently noticed is that Zulip has the ability to "star
" a comment

nikomatsakis (Mar 22 2019 at 14:15, on Zulip):

So, instead of using emoji, I say "star" comments that seem like they are key pointed for a summary

nikomatsakis (Mar 22 2019 at 14:15, on Zulip):

(It's very helpful when skimming later)

Aaron Turon (Mar 22 2019 at 14:15, on Zulip):

are the stars public? they might just be private favorites

davidtwco (Mar 22 2019 at 14:16, on Zulip):

are the stars public? they might just be private favorites

This is how that works.

nikomatsakis (Mar 22 2019 at 14:16, on Zulip):

Oh, can you all not see them?

nikomatsakis (Mar 22 2019 at 14:16, on Zulip):

never mind then :)

nikomatsakis (Mar 22 2019 at 14:16, on Zulip):

let's use the :point_up: emoji (point-up)

nikomatsakis (Mar 22 2019 at 14:16, on Zulip):

it's just annoying to put an emoji on your own comment :)

nikomatsakis (Mar 22 2019 at 14:16, on Zulip):

/me often thinks that what he says is key ;)

nikomatsakis (Mar 22 2019 at 14:17, on Zulip):

Anyway, so, there's actually a good list of things in the dropbox paper already

nikomatsakis (Mar 22 2019 at 14:17, on Zulip):

Let me just say a bit more about what I was hoping to get out of this

nikomatsakis (Mar 22 2019 at 14:17, on Zulip):

Basically, I was thinking that -- both as part of the meta working group and the larger governance WG effort -- there is a need to have some idea what the processes should be doing

nikomatsakis (Mar 22 2019 at 14:18, on Zulip):

And often I think we bundle up a lot of things together with a vague term like "the WG lead should keep the WG running"

nikomatsakis (Mar 22 2019 at 14:18, on Zulip):

But what does that mean

nikomatsakis (Mar 22 2019 at 14:18, on Zulip):

I also suspect some of those roles (notably "lead") can be broken apart into finer-grained roles

nikomatsakis (Mar 22 2019 at 14:18, on Zulip):

Plus of course I'm concerned that there are things we kind of aren't doing that maybe we should be (e.g., looking over older bugs and things)

nikomatsakis (Mar 22 2019 at 14:19, on Zulip):

To toss out one of those things from the doc (I see people are editing in there, too, which is good):

where do we go with design discussions and tricky questions? (e.g., settling the question of 'gcx etc)

pnkfelix (Mar 22 2019 at 14:19, on Zulip):

so some of these questions

pnkfelix (Mar 22 2019 at 14:19, on Zulip):

seem like things that ideally one would want all T-compiler members present

pnkfelix (Mar 22 2019 at 14:20, on Zulip):

or at least the most ... prolific? productive? members

nikomatsakis (Mar 22 2019 at 14:20, on Zulip):

yeah I'm feeling a bit torn between whether we should try to solve or just enumerate :)

pnkfelix (Mar 22 2019 at 14:20, on Zulip):

which in turn requires either an asynchronous discussion and decision process

pnkfelix (Mar 22 2019 at 14:20, on Zulip):

or finding a time for a synchronous discussion+decision

nikomatsakis (Mar 22 2019 at 14:21, on Zulip):

I've become somewhat skeptical of async discussions of this kind

nikomatsakis (Mar 22 2019 at 14:21, on Zulip):

I just think we don't tend to get a ton of engagement

pnkfelix (Mar 22 2019 at 14:21, on Zulip):

oh yes, sorry, I wasn't sure what you were hoping for when you brought up 'gcx as an example

nikomatsakis (Mar 22 2019 at 14:21, on Zulip):

Well, actually, I think it's not a bad idea to spend some time talking about how we can solve it

pnkfelix (Mar 22 2019 at 14:21, on Zulip):

i.e. did "where do we go mean" : which meeting should it be attached to, if any?

nikomatsakis (Mar 22 2019 at 14:21, on Zulip):

maybe we can pick a few "key topics" as we go and spend.. 10-15 minutes on each one

nikomatsakis (Mar 22 2019 at 14:22, on Zulip):

yeah so

nikomatsakis (Mar 22 2019 at 14:22, on Zulip):

I guess I just mean "how should we decide what to do"

nikomatsakis (Mar 22 2019 at 14:22, on Zulip):

I think we prob need some sort of mix of sync + async

nikomatsakis (Mar 22 2019 at 14:23, on Zulip):

I feel like it would be useful to have some kind of dedicated time where we can discuss technical design questions

nikomatsakis (Mar 22 2019 at 14:23, on Zulip):

I could imagine e.g. picking a topic ("should we go with 'tcx") and having some meeting where we dive into it, coming up with specific questions to answer -- then async do investigations -- and then in next meeting we review results. The idea being to keep driving forward to a conclusion.

nikomatsakis (Mar 22 2019 at 14:23, on Zulip):

Maybe we can't discuss another topic will we decide :P

nikomatsakis (Mar 22 2019 at 14:24, on Zulip):

Right now I guess I feel like we open pRs but not a lot happens?

nikomatsakis (Mar 22 2019 at 14:26, on Zulip):

Another thought (a meta thought): perhaps we need to focus on things at a time. i.e., in the paper doc, we have a big list of stuff, but maybe it's good to organize it into clearer categories.

nikomatsakis (Mar 22 2019 at 14:27, on Zulip):

I'm thinking of:

maybe?

nikomatsakis (Mar 22 2019 at 14:27, on Zulip):

let me take a stab :)

nikomatsakis (Mar 22 2019 at 14:30, on Zulip):

does that seem helpful? (see here)

nikomatsakis (Mar 22 2019 at 14:31, on Zulip):

(I was thinking if we start to get some categories, then we can (here, in Zulip maybe) focus on a category at a time)

pnkfelix (Mar 22 2019 at 14:32, on Zulip):

/me is watching the text flow

nikomatsakis (Mar 22 2019 at 14:33, on Zulip):

just hacking on paper is itself kind of useful :)

nikomatsakis (Mar 22 2019 at 14:34, on Zulip):

I'm not sure where this fits in exactly

nikomatsakis (Mar 22 2019 at 14:34, on Zulip):

it's kind of a triage issue

nikomatsakis (Mar 22 2019 at 14:34, on Zulip):

but I wonder if performance deserves its own section somehow

pnkfelix (Mar 22 2019 at 14:34, on Zulip):

ideally compiler performance was/is has its own dedicated strike team

pnkfelix (Mar 22 2019 at 14:34, on Zulip):

and they should, ideally, be experts on the benchmark suite for the compiler

pnkfelix (Mar 22 2019 at 14:35, on Zulip):

a reason I wrote the line item about measuring performance and curating benchmarks

davidtwco (Mar 22 2019 at 14:35, on Zulip):

Perhaps too meta, but one idea might be: how do we best have discussions aimed at solving the types of problems that we're trying to solve right now? i.e. how could we do things like this meeting better or more effectively? It was mentioned before trying to get engagement and have compiler team folk around.

pnkfelix (Mar 22 2019 at 14:35, on Zulip):

was that I spent time recently on #11084

nikomatsakis (Mar 22 2019 at 14:35, on Zulip):

Yeah, that feels like it is sort of circling around the same question I raised earlier, @davidtwco

nikomatsakis (Mar 22 2019 at 14:35, on Zulip):

but on a different sort of "problem"

pnkfelix (Mar 22 2019 at 14:36, on Zulip):

all I wanted to point out is that sometimes someone reports a benchmark

pnkfelix (Mar 22 2019 at 14:36, on Zulip):

and then we identify "ah the perfomance issue went away"

pnkfelix (Mar 22 2019 at 14:36, on Zulip):

but the benchmark doesn't necessarily get curated. and then a subsequent regression has to be tracked down, if one wants to understand what happened.

nikomatsakis (Mar 22 2019 at 14:37, on Zulip):

yeah, sometimes we add "regression tests" to our perf suite, but not always

pnkfelix (Mar 22 2019 at 14:38, on Zulip):

and to be fair (and as you stated), if you add every little benchmark to a suite, it can overstate the importance of that benchmark.

pnkfelix (Mar 22 2019 at 14:38, on Zulip):

especially if one adds the 100 variations of the benchmark that were originally developed for investigatory purporses

pnkfelix (Mar 22 2019 at 14:38, on Zulip):

and then that gets misinterpreted later as "look at all the benchmark PR X regressed"

Zoxc (Mar 22 2019 at 14:38, on Zulip):

I'm not sure when you are referring to compile time or performance of emitted code here. Compiler performance is a bit ambiguous

nikomatsakis (Mar 22 2019 at 14:38, on Zulip):

Perhaps too meta, but one idea might be: how do we best have discussions aimed at solving the types of problems that we're trying to solve right now?

I've been circling around this in many contexts and I'm not sure what the answer is yet. For example, I feel like in the traits working group, we have a need to have regular "design discussions", both to spread knowledge and to figure out tricky stuff. At the compiler team level, I feel the same way. I definitely feel that trying to have these discussions be synchronous is a big win -- video may be helpful too, though I'm not convinced. Zulip often seems to scale great.

pnkfelix (Mar 22 2019 at 14:39, on Zulip):

I'm not sure when you are referring to compile time or performance of emitted code here. Compiler performance is a bit ambiguous

ah that is fair. I was speaking of performance of emitted code.

pnkfelix (Mar 22 2019 at 14:39, on Zulip):

but we should (and do) track performance of the compiler itself as well.

nikomatsakis (Mar 22 2019 at 14:39, on Zulip):

One thing I've thought about is having a "standing meeting time" (kind of like this, but maybe more frequent) and a shared agenda, and letting people add things to it when they are having questions. BAsically a time that the "busy experts" commit to being around to try and work through something.

nikomatsakis (Mar 22 2019 at 14:40, on Zulip):

ah that is fair. I was speaking of performance of emitted code.

argh, I forgot that indeed we don't even track this right now.

nikomatsakis (Mar 22 2019 at 14:40, on Zulip):

@Adam Perry was working on lolbench

nikomatsakis (Mar 22 2019 at 14:40, on Zulip):

but it never got "official"

nikomatsakis (Mar 22 2019 at 14:40, on Zulip):

not really sure of its status even?

pnkfelix (Mar 22 2019 at 14:41, on Zulip):

the page is still getting updated

pnkfelix (Mar 22 2019 at 14:42, on Zulip):

though I do not know if the updates reflect useful info, or are just a robot grinding with no real info being added.

nikomatsakis (Mar 22 2019 at 14:43, on Zulip):

the page is still getting updated

which page are you referring to?

qmx (Mar 22 2019 at 14:43, on Zulip):

@nikomatsakis you used to do office hours, did that worked well?

pnkfelix (Mar 22 2019 at 14:43, on Zulip):

https://lolbench.rs/

pnkfelix (Mar 22 2019 at 14:43, on Zulip):

it has entries through 2019-03-21. But they have been empty for a long time

nikomatsakis (Mar 22 2019 at 14:44, on Zulip):

you used to do office hours, did they work well?

an interesting question. I've been feeling very sad that I gave those up. I keep meaning to start again. But I was thinking of trying a different format. Instead of short 30 min sessions with lots of folks, I was going to try something more committed -- where I would pick one person per month with a hard problem. And we would spend an hour each week attacking it.

nikomatsakis (Mar 22 2019 at 14:44, on Zulip):

I was also going to make these open meetings.

nikomatsakis (Mar 22 2019 at 14:44, on Zulip):

And considering proposing to the lang team that we rebrand this as a "lang team effort", maybe rotating duty

nikomatsakis (Mar 22 2019 at 14:44, on Zulip):

I've considered similar things around compiler contributions

pnkfelix (Mar 22 2019 at 14:44, on Zulip):

I was trying to switch the NLL WG meetings to take an office hour format

pnkfelix (Mar 22 2019 at 14:45, on Zulip):

but the last two meetings were like ghost towns, so maybe that isn't working or isn't worthwhile

nikomatsakis (Mar 22 2019 at 14:45, on Zulip):

My reasoning was that -- well, sometimes the problems were easy, and that's fine, but other times, I culdn't find a satisfactory answer. And I'd like to dig more into those problems, they are probably areas where Rust could really use improvement.

pnkfelix (Mar 22 2019 at 14:45, on Zulip):

i certainly like the idea of WG-directed office hour meetings

pnkfelix (Mar 22 2019 at 14:46, on Zulip):

where the main participants agree to synchronize on a shared time to be present and be actively reading the ongoing chat session

Wesley Wiser (Mar 22 2019 at 14:47, on Zulip):

but the last two meetings were like ghost towns, so maybe that isn't working or isn't worthwhile

Not trying to ask a loaded question, but is there a lot of NLL work left to do? Perhaps as NLL has "shipped" in Rust 2018, there's less interest from contributors to show up to the meeting?

pnkfelix (Mar 22 2019 at 14:47, on Zulip):

yes that is certainly a reasonable conclusion

pnkfelix (Mar 22 2019 at 14:48, on Zulip):

(there is work left to do, especially with diagnostics, but its possible that those issues can be taken care of without a dedicated working group)

nikomatsakis (Mar 22 2019 at 14:48, on Zulip):

i certainly like the idea of WG-directed office hour meetings

I think one problem is

davidtwco (Mar 22 2019 at 14:48, on Zulip):

That's mostly why I've attended less actively, less ongoing work and there are other working groups where I'll be able to learn more (having participated in NLL for a while and learning a bunch there already).

nikomatsakis (Mar 22 2019 at 14:48, on Zulip):

there has to be enough people answering

nikomatsakis (Mar 22 2019 at 14:49, on Zulip):

i.e., if the WG is very one-sided, with a lot of folks asking and one or two people answering, it's hard to make an "office hour" format work, unless I suppose everybody is learning from each answer

nikomatsakis (Mar 22 2019 at 14:50, on Zulip):

So... the paper doc is starting to look pretty good to me

pnkfelix (Mar 22 2019 at 14:50, on Zulip):

Yes, now that I think @nikomatsakis has largely sorted the paper doc

pnkfelix (Mar 22 2019 at 14:50, on Zulip):

the next Q is: Which of these things are we doing, which are we failing to do ...

pnkfelix (Mar 22 2019 at 14:51, on Zulip):

@nikomatsakis : "Closing random ICEs ..." means fixing, right?

nikomatsakis (Mar 22 2019 at 14:51, on Zulip):

lol, correct

nikomatsakis (Mar 22 2019 at 14:53, on Zulip):

There is definitely some overlap in some of the categories

nikomatsakis (Mar 22 2019 at 14:53, on Zulip):

which... is probably ok

nikomatsakis (Mar 22 2019 at 14:53, on Zulip):

e.g., I think that under the "running a WG" label, I put "updating the team as a whole"

nikomatsakis (Mar 22 2019 at 14:53, on Zulip):

but I also put that under "intercommunication"

nikomatsakis (Mar 22 2019 at 14:53, on Zulip):

in part because I think there are two parts to it ...

nikomatsakis (Mar 22 2019 at 14:54, on Zulip):

e.g., I think there is work that @pnkfelix is presently doing around the triage meeting, of figuring out which teams to solicit updates from

nikomatsakis (Mar 22 2019 at 14:54, on Zulip):

and of course the work of the team to listen and read

nikomatsakis (Mar 22 2019 at 14:54, on Zulip):

I'll confess that when updates are overly async I tend to ignore them :P

nikomatsakis (Mar 22 2019 at 14:54, on Zulip):

/me feels bad about it, at least

pnkfelix (Mar 22 2019 at 14:54, on Zulip):

I'll confess that when updates are overly async I tend to ignore them :P

what does this mean?

pnkfelix (Mar 22 2019 at 14:55, on Zulip):

i.e. you see the ones at the triage meeting

pnkfelix (Mar 22 2019 at 14:55, on Zulip):

but if a team just sends out a message at a random time, you don't notice?

pnkfelix (Mar 22 2019 at 14:55, on Zulip):

or if its just that some PR lands and that's all the notice you got?

nikomatsakis (Mar 22 2019 at 14:55, on Zulip):

I don't know

nikomatsakis (Mar 22 2019 at 14:55, on Zulip):

I mean that e.g. I don't read "this week in rust"

pnkfelix (Mar 22 2019 at 14:55, on Zulip):

ah

nikomatsakis (Mar 22 2019 at 14:55, on Zulip):

if I just get some e-mails, I probably would ignore them

nikomatsakis (Mar 22 2019 at 14:55, on Zulip):

I mean I'll read them

nikomatsakis (Mar 22 2019 at 14:55, on Zulip):

but not as closely as at a meeting

nikomatsakis (Mar 22 2019 at 14:56, on Zulip):

and maybe late:)

nikomatsakis (Mar 22 2019 at 14:56, on Zulip):

maybe the answer is "niko needs to adult better"

pnkfelix (Mar 22 2019 at 14:56, on Zulip):

I mean that e.g. I don't read "this week in rust"

I don't either, usually, and I feel sort of bad about it too

nikomatsakis (Mar 22 2019 at 14:56, on Zulip):

but I sort of suspect that it's very common :)

nikomatsakis (Mar 22 2019 at 14:56, on Zulip):

or maybe it's not, maybe it's more because some of us are multitasking a lot of things

pnkfelix (Mar 22 2019 at 14:56, on Zulip):

TWIR is a resource that some people definitely care about

pnkfelix (Mar 22 2019 at 14:56, on Zulip):

so I do wonder if we should be leveraging it more in some way

pnkfelix (Mar 22 2019 at 14:56, on Zulip):

not just in terms of reading it

nikomatsakis (Mar 22 2019 at 14:56, on Zulip):

let me be more specific

pnkfelix (Mar 22 2019 at 14:56, on Zulip):

but also feeding info into it

nikomatsakis (Mar 22 2019 at 14:56, on Zulip):

I think that if people were e.g. posting regular updates to internals

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

I would be unlikely to see them

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

e-mail is perhaps better

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

but only marginally so

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

at the same time, I recognize that making these things purely sync is ungreat

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

that said, I could certainly imagine something like this

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

at the start of this meeting

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

we had the "announcements" period

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

I could imagine a shared "reading and ask questions" period

nikomatsakis (Mar 22 2019 at 14:57, on Zulip):

where people come with (prepared) notes and updates

nikomatsakis (Mar 22 2019 at 14:58, on Zulip):

and maybe we track if we aren't getting updates from a particular team in a long time

nikomatsakis (Mar 22 2019 at 14:58, on Zulip):

(and these updates become something we can post)

nikomatsakis (Mar 22 2019 at 14:58, on Zulip):

where people come with (prepared) notes and updates

but we don't type them into zulip, I mean

nikomatsakis (Mar 22 2019 at 14:58, on Zulip):

rather we post a link to them, and we all agree to either read them before, or read them live, and ask questions :)

nikomatsakis (Mar 22 2019 at 14:59, on Zulip):

(I feel like weekly is a burden, but that might be a good fit for steering meetings)

nikomatsakis (Mar 22 2019 at 15:00, on Zulip):

anyway, our time here is almost up

nikomatsakis (Mar 22 2019 at 15:00, on Zulip):

but one nice property from this meeting is that the summary is already written :)

pnkfelix (Mar 22 2019 at 15:00, on Zulip):

I like the starting announcement period, I think that worked well today

nikomatsakis (Mar 22 2019 at 15:00, on Zulip):

I also liked it

pnkfelix (Mar 22 2019 at 15:00, on Zulip):

I will now subvert the proces

pnkfelix (Mar 22 2019 at 15:00, on Zulip):

by saying

nikomatsakis (Mar 22 2019 at 15:00, on Zulip):

I was happy with the "meta structure" of this meeting, in terms of using a poll + general announcements

pnkfelix (Mar 22 2019 at 15:00, on Zulip):

I wish I had remembered to mention https://oli-obk.github.io/fixmeh/

pnkfelix (Mar 22 2019 at 15:00, on Zulip):

which is a cool little hack

oli (Mar 22 2019 at 15:02, on Zulip):

about that. Would making a graph about the total age of all FIXMEs send the wrong message or be otherwise undesirable?

pnkfelix (Mar 22 2019 at 15:03, on Zulip):

how would you compute a FIXME's age? Based on the last time its source line was touched according to git blame ?

oli (Mar 22 2019 at 15:03, on Zulip):

I was actually thinking about starting tracking at a certain date and just diff the list that I extracted

pnkfelix (Mar 22 2019 at 15:03, on Zulip):

oh I see

oli (Mar 22 2019 at 15:03, on Zulip):

the other thing is too fickle

pnkfelix (Mar 22 2019 at 15:04, on Zulip):

I don't have any objection to such a graph.

Vadim Petrochenkov (Mar 22 2019 at 15:04, on Zulip):

I like the starting announcement period

I'm a bit late, but here's an announcement.
I have a few issues with ICEs/regressions/questions assigned + pending reviews including the updated eager expansion RFC.
Once I'm done with that I'll be implementing name resolution for attributes in non-macro positions (a remaining piece of macro modularization).

nikomatsakis (Mar 22 2019 at 15:06, on Zulip):

@Vadim Petrochenkov cool!

Adam Perry (Mar 22 2019 at 16:13, on Zulip):

@pnkfelix the anomalous results sections are collapsed by default but results are still collected

@nikomatsakis lolbench continues to run, i have a sponsor for the hardware but haven’t migrated to the new boxes yet. i’d be happy to do a quick project status rundown if the compiler team is interested

Adam Perry (Mar 22 2019 at 16:15, on Zulip):

tldr is that i believe it’s still producing useful data, a try mode is the next big thing missing for productivity, and i’d be very happy to try to help make it officialer

Adam Perry (Mar 22 2019 at 16:20, on Zulip):

if compiler folks think that contributors would appear i would also be happy to work on mentoring instructions for a variety of tasks and list lolbench in twir

pnkfelix (Mar 22 2019 at 17:31, on Zulip):

@Adam Perry I explored the ui a bit and could not figure out how to uncollapse the results, FYI

Adam Perry (Mar 22 2019 at 17:38, on Zulip):

oh no

Adam Perry (Mar 22 2019 at 17:38, on Zulip):

i’ll take a look

nikomatsakis (Mar 22 2019 at 21:04, on Zulip):

if compiler folks think that contributors would appear i would also be happy to work on mentoring instructions for a variety of tasks and list lolbench in twir

I think we could make contributors appear

nikomatsakis (Mar 22 2019 at 21:16, on Zulip):

Minutes: https://github.com/rust-lang/compiler-team/pull/47

pnkfelix (Mar 25 2019 at 15:01, on Zulip):

Adam Perry I explored the ui a bit and could not figure out how to uncollapse the results, FYI

hmm, I now do see the [+] on the left hand side to uncollapse the results

pnkfelix (Mar 25 2019 at 15:01, on Zulip):

I'm pretty sure I couldn't see that before. I don't think my window was too thin, but I'm not sure.

pnkfelix (Mar 25 2019 at 15:02, on Zulip):

oh actually, it probably was too thin. I didn't realize how easily it is to make a window that is so thin that you only see a very slight sliver of black from the right most ] in [+]

Last update: Nov 16 2019 at 02:05UTC