Stream: t-compiler

Topic: planning meeting 2019.06.07


nikomatsakis (Jun 07 2019 at 14:08, on Zulip):

Hello @T-compiler/meeting -- apologies for running a few minutes late! We have a scheduled planning meeting today, with the goal of deciding what meetings to have over the next 3 weeks.

I'm still walking to my desk -- so please ,make any :horn: announcements :horn: now =)

nikomatsakis (Jun 07 2019 at 14:10, on Zulip):

I don't think we have a lot of proposals

nikomatsakis (Jun 07 2019 at 14:11, on Zulip):

But I think we could entertain some "last minute" proposals -- e.g., the 'gcx thing -- and schedule them too

pnkfelix (Jun 07 2019 at 14:11, on Zulip):

non-tech proposal: Our Bug Assignment process. Potential change: start "automatically" assigning bugs after they've been unassigned for N days (rather than waiting for volunteers to take them)

nikomatsakis (Jun 07 2019 at 14:11, on Zulip):

(Maybe a follow-up to parallelization?)

nikomatsakis (Jun 07 2019 at 14:12, on Zulip):

OK well let's start by collecting proposals

nikomatsakis (Jun 07 2019 at 14:12, on Zulip):

Here is the link to meeting proposals

pnkfelix (Jun 07 2019 at 14:13, on Zulip):

okay so I should create an issue there, right?

nikomatsakis (Jun 07 2019 at 14:13, on Zulip):

yeah

pnkfelix (Jun 07 2019 at 14:13, on Zulip):

I guess this might fall into the maintenance and triage one

nikomatsakis (Jun 07 2019 at 14:13, on Zulip):

seems like it

nikomatsakis (Jun 07 2019 at 14:13, on Zulip):

but maybe not

pnkfelix (Jun 07 2019 at 14:14, on Zulip):

okay. Maybe we can plan to reschedule "Maintenance and triage" #90 for end of June?

pnkfelix (Jun 07 2019 at 14:14, on Zulip):

/me goes to look if he would end up unexpectedly skipping the meeting again if we did that

nikomatsakis (Jun 07 2019 at 14:15, on Zulip):

we could schedule anytime in next 3 weeks I thik. One consideration is that the mozilla all hands is coming up

nikomatsakis (Jun 07 2019 at 14:15, on Zulip):

There is

pnkfelix (Jun 07 2019 at 14:16, on Zulip):

right, but June 21st overlaps the All-Hands, right?

nikomatsakis (Jun 07 2019 at 14:16, on Zulip):

Right

nikomatsakis (Jun 07 2019 at 14:16, on Zulip):

So maybe we x that one out

nikomatsakis (Jun 07 2019 at 14:16, on Zulip):

Since you and I at least are not available or it will be difficult

pnkfelix (Jun 07 2019 at 14:16, on Zulip):

okay, I'm proposing we reschedule "Maintenance and triage" for June 28th.

nikomatsakis (Jun 07 2019 at 14:17, on Zulip):

that works

pnkfelix (Jun 07 2019 at 14:17, on Zulip):

and I would presumably actually write a document in time for that.

nikomatsakis (Jun 07 2019 at 14:17, on Zulip):

we can discuss and make plans during all hands :)

pnkfelix (Jun 07 2019 at 14:17, on Zulip):

I won't be at all-hands

nikomatsakis (Jun 07 2019 at 14:17, on Zulip):

June 14 discuss 'gcx? maybe an update on parallelization?

pnkfelix (Jun 07 2019 at 14:17, on Zulip):

babies

nikomatsakis (Jun 07 2019 at 14:17, on Zulip):

I won't be at all-hands

:'(

nikomatsakis (Jun 07 2019 at 14:17, on Zulip):

forgot about that

pnkfelix (Jun 07 2019 at 14:17, on Zulip):

are @eddyb or @Zoxc around right now?

nikomatsakis (Jun 07 2019 at 14:18, on Zulip):

Don't see them in the :wave: list :)

pnkfelix (Jun 07 2019 at 14:19, on Zulip):

we could nontheless tentatively schedule discussion of 'gcx

nikomatsakis (Jun 07 2019 at 14:19, on Zulip):

agreed

nikomatsakis (Jun 07 2019 at 14:20, on Zulip):

I'm trying to think if there's anything else that would be good to discuss

pnkfelix (Jun 07 2019 at 14:21, on Zulip):

well

pnkfelix (Jun 07 2019 at 14:21, on Zulip):

I do have one thing

nikomatsakis (Jun 07 2019 at 14:21, on Zulip):

ok, so proposed schedule for next 3 weeks:

pnkfelix (Jun 07 2019 at 14:21, on Zulip):

Do we need an WG-incr-comp?

nikomatsakis (Jun 07 2019 at 14:21, on Zulip):

oh, you had to post that while I was typing :P

pnkfelix (Jun 07 2019 at 14:21, on Zulip):

and/or do we need more testing, like fuzzing, of incr-comp

nikomatsakis (Jun 07 2019 at 14:21, on Zulip):

ah, because of the bugs rising up there?

pnkfelix (Jun 07 2019 at 14:22, on Zulip):

yes, we have a lot of bugs and I think we may need a strike force to either fix the bugs or build up our test suite

eddyb (Jun 07 2019 at 14:22, on Zulip):

I am on my phone, on which Zulip is not great

nikomatsakis (Jun 07 2019 at 14:23, on Zulip):

We're discussing @eddyb what to discuss next week. Would you be available then (at this time)?

eddyb (Jun 07 2019 at 14:23, on Zulip):

but my plan, in short, is to introduce a type alias to allow gradually moving over

pnkfelix (Jun 07 2019 at 14:23, on Zulip):

maybe my A-incr-comp thoughts are too ill-formed to discuss this month

pnkfelix (Jun 07 2019 at 14:23, on Zulip):

and besides, @mw is on leave

pnkfelix (Jun 07 2019 at 14:24, on Zulip):

so maybe I'll just spend some time thinking about it

eddyb (Jun 07 2019 at 14:24, on Zulip):

@nikomatsakis I technically can, just need a reminder on thursday or so, I think?

eddyb (Jun 07 2019 at 14:25, on Zulip):

so my plan was to transition TyCtxt<'_, 'tcx, 'tcx>-using code first, to TyCx<'tcx> (name doesn't matter as long as it's unique for easy replacement later)

nikomatsakis (Jun 07 2019 at 14:25, on Zulip):

OK -- my sense @pnkfelix is that the incremental discussion is perhaps higher value, it's not so much technical I imagine as a problem of finding the right person to push and organize the effort?

pnkfelix (Jun 07 2019 at 14:25, on Zulip):

i suppose

pnkfelix (Jun 07 2019 at 14:25, on Zulip):

the technical matter

nikomatsakis (Jun 07 2019 at 14:25, on Zulip):

(which doesn't mean we shouldn't discuss it)

pnkfelix (Jun 07 2019 at 14:26, on Zulip):

might be about how to go about testing

nikomatsakis (Jun 07 2019 at 14:26, on Zulip):

so my plan was to transition TyCtxt<'_, 'tcx, 'tcx>-using code first, to TyCx<'tcx> (name doesn't matter as long as it's unique for easy replacement later)

this makes sense

eddyb (Jun 07 2019 at 14:26, on Zulip):

land that then start moving 'gcx,'tcx code one InferCtxt user at a time

pnkfelix (Jun 07 2019 at 14:26, on Zulip):

i.e. how we can maximize bang-for-buck with testing effort w.r.t. incr-comp

eddyb (Jun 07 2019 at 14:26, on Zulip):

by making InferCtxt creation allow both 'tcx,'tcx and 'gcx,'tcx

nikomatsakis (Jun 07 2019 at 14:26, on Zulip):

I'm wondering if it would be a good time to discuss other naming convention questions -- e.g., going from 'tcx to something else, or the "query-context"

nikomatsakis (Jun 07 2019 at 14:27, on Zulip):

(if we did schedule a "transition meeting", I mean)

eddyb (Jun 07 2019 at 14:27, on Zulip):

@nikomatsakis I want to make it shaped like that except without the name changes

pnkfelix (Jun 07 2019 at 14:27, on Zulip):

clearly we need to keep the character count the same

pnkfelix (Jun 07 2019 at 14:27, on Zulip):

so that rustfmt will make the same choices about line breaks

pnkfelix (Jun 07 2019 at 14:27, on Zulip):

so I vote for 'tcx_tcx_tcx

nikomatsakis (Jun 07 2019 at 14:27, on Zulip):

lol

nikomatsakis (Jun 07 2019 at 14:27, on Zulip):

ok, so proposed schedule for next 3 weeks:

eddyb (Jun 07 2019 at 14:27, on Zulip):

and using unique names to facilitate easy renaming later

nikomatsakis (Jun 07 2019 at 14:28, on Zulip):

seems like we agreed to that?

pnkfelix (Jun 07 2019 at 14:28, on Zulip):

sounds good to me

eddyb (Jun 07 2019 at 14:28, on Zulip):

anyway basically code used with both 'tcx,'tcx and 'gcx,'tcx should be changed last

eddyb (Jun 07 2019 at 14:28, on Zulip):

that's the trick to doing it not all in one PR

eddyb (Jun 07 2019 at 14:29, on Zulip):

IMO

nikomatsakis (Jun 07 2019 at 14:29, on Zulip):

ok ok. I was just debating if it would make sense to talk next week about the current situation (i.e., regressions, p-high bugs) as separate from the general system

nikomatsakis (Jun 07 2019 at 14:29, on Zulip):

anyway basically code used with both 'tcx,'tcx and 'gcx,'tcx should be changed last

I feel like the aim of a meeting ought to be to help design the end-point also, though

eddyb (Jun 07 2019 at 14:29, on Zulip):

ttyl, gotta kill Zulip before it drains my battery (feel free to ping me on Discord, that is not a powerhog)

nikomatsakis (Jun 07 2019 at 14:29, on Zulip):

so that we can start documenting it etc

nikomatsakis (Jun 07 2019 at 14:29, on Zulip):

ok, ttyl

eddyb (Jun 07 2019 at 14:30, on Zulip):

sure, discussing qx: QueryCx<'q> or other such ideas would be better than just the lifetimes

nikomatsakis (Jun 07 2019 at 14:30, on Zulip):

sigh :) I like how it's only the 2nd planning meeting and we already failed my goal to have proposals before the meeting

nikomatsakis (Jun 07 2019 at 14:30, on Zulip):

but I'll try to draw up something

pnkfelix (Jun 07 2019 at 14:30, on Zulip):

ok ok. I was just debating if it would make sense to talk next week about the current situation (i.e., regressions, p-high bugs) as separate from the general system

what was this in response to? The Q of discussing A-incr-comp ?

eddyb (Jun 07 2019 at 14:30, on Zulip):

one of my more recent crazytown "ideas" is mx: MirCx<'m>

nikomatsakis (Jun 07 2019 at 14:30, on Zulip):

not directly in response to anything but yeah just a thought

eddyb (Jun 07 2019 at 14:30, on Zulip):

but less viable if the entire compilation is inside it

pnkfelix (Jun 07 2019 at 14:32, on Zulip):

(in theory the proposal for next week's meeting is supposed to be available now, right?)

nikomatsakis (Jun 07 2019 at 14:38, on Zulip):

(in theory the proposal for next week's meeting is supposed to be available now, right?)

yes, I'm writing it now :)

nikomatsakis (Jun 07 2019 at 14:38, on Zulip):

oops I realize we never confirmed3 that "um we're done here"

nikomatsakis (Jun 07 2019 at 14:38, on Zulip):

but I'm assuming that's true

nikomatsakis (Jun 07 2019 at 14:38, on Zulip):

Agreed? :)

nikomatsakis (Jun 07 2019 at 14:41, on Zulip):

:thank_you: to everyone :)

nikomatsakis (Jun 07 2019 at 14:46, on Zulip):

I just made a meeting proposal re: 'gcx, tcx -- hackmd link here. I restricted editing because "world editable" has been biting us lately, but I'll invite anyone who asks. I already invited @eddyb to add details =)

nikomatsakis (Jun 07 2019 at 14:51, on Zulip):

created calendar events, updated schedule, etc

Zoxc (Jun 07 2019 at 16:24, on Zulip):

https://github.com/rust-lang/rust/pull/57173 will likely conflict heavily with a transition await from 'gcx

mark-i-m (Jun 07 2019 at 17:03, on Zulip):

Just a humble observation, but it seems like having 25% of meetings be planning meetings is too much overhead. Would it not be more efficient for someone to asynchronously ping the relevant people every 2 weeks asking what they want to discuss (if anything) and how important it is (i.e. what it is blocking) and any relevant background info (or perhaps even have a google form that feeds a spreadsheet of possible discussion topics)? Then at the time of the meeting, just pick the top thing on the priority queue and discuss it? It just seems like a lot of these process things have fallen to pnkfelix and niko recently, and they already seem oversubscribed (though perhaps that's just my perception).

Last update: Nov 22 2019 at 04:30UTC