Stream: t-compiler/wg-meta

Topic: focused and efficient triage compiler-team#247


nikomatsakis (Feb 20 2020 at 14:51, on Zulip):

So @pnkfelix, @Santiago Pastorino and I were thinking it'd be good to discuss a bit in the runup to the meeting for compiler-team#247 -- scheduled for Feb 28. I'm mainly creating the Zulip topic at this second.

nikomatsakis (Feb 20 2020 at 14:52, on Zulip):

My expectation here is that the goal should be to move some of what we now call the pre-triage process into a kind of working group -- that is, the work of prioritizing, tracking, and so forth. The output of this group would be a "triage report" that guides the triage meeting, indicating backports that have to be reviewed, critical priority issues that seems to be stalled, and other things that merit focused eyeballs of the entire team.

nikomatsakis (Feb 20 2020 at 14:52, on Zulip):

I think some interesting questions are how best to engage with the wg-triage effort -- is this the same thing? is it distinct? overlapping?

nikomatsakis (Feb 20 2020 at 14:52, on Zulip):

and how best to create guidelines etc that allow us to pull in contributors and distribute the work

nikomatsakis (Feb 20 2020 at 14:53, on Zulip):

I know @DPC has indicated that e.g. folks in #wg-triage don't necessarily have the context to make some of these calls, which makes sense.

DPC (Feb 20 2020 at 16:11, on Zulip):

I like the idea of a seperate working group. It could be a mix of some of other compiler folk and some people who have some knowledge of the compiler/team structure

nikomatsakis (Feb 27 2020 at 13:49, on Zulip):

So we haven't really had a lot of discussion here in run-up to meeting tomorrow

nikomatsakis (Feb 27 2020 at 13:49, on Zulip):

I'm looking back over the hackmd that I prepared

nikomatsakis (Feb 27 2020 at 13:49, on Zulip):
nikomatsakis (Feb 27 2020 at 13:53, on Zulip):

Maybe we can start to answer some of these questions? I haven't seen anyone really arguing with the basic premise

nikomatsakis (Feb 27 2020 at 13:53, on Zulip):

so maybe we can sharpen the proposal a bit

nikomatsakis (Feb 27 2020 at 14:00, on Zulip):

how does this sound?

nikomatsakis (Feb 27 2020 at 14:00, on Zulip):
Santiago Pastorino (Feb 27 2020 at 14:10, on Zulip):

sounds good to me

Santiago Pastorino (Feb 27 2020 at 14:10, on Zulip):

one thing that I wasn't sure and kind of confuses me now meanwhile I read

Santiago Pastorino (Feb 27 2020 at 14:10, on Zulip):

Inputs: the rust-lang repository =)

Santiago Pastorino (Feb 27 2020 at 14:11, on Zulip):

what does that mean?

Santiago Pastorino (Feb 27 2020 at 14:11, on Zulip):

github.com/rust-lang or github.com/rust-lang/rust ?

Santiago Pastorino (Feb 27 2020 at 14:12, on Zulip):

given it says repository I'm inclined to rust-lang/rust but given it says rust-lang it made me think about the whole organization :)

Santiago Pastorino (Feb 27 2020 at 14:12, on Zulip):

I think we are talking about rust-lang/rust but aren't there more things to check?

Santiago Pastorino (Feb 27 2020 at 14:12, on Zulip):

in any case I'd clarify a bit

nikomatsakis (Feb 27 2020 at 14:12, on Zulip):

yes so

nikomatsakis (Feb 27 2020 at 14:12, on Zulip):

what I meant is

nikomatsakis (Feb 27 2020 at 14:12, on Zulip):

the role of this group is to "monitor what is happening" -- but I think there's overlap here with the release team triage

nikomatsakis (Feb 27 2020 at 14:13, on Zulip):

it's probably a good idea to try and sharpen the inputs a bit

nikomatsakis (Feb 27 2020 at 14:13, on Zulip):

for example, right now, I think that @pnkfelix starts by surveying bugs tagged as I-nominated

nikomatsakis (Feb 27 2020 at 14:13, on Zulip):

he doesn't look at each new issue as it is filed, that work is done more by release team triage folks afaik

nikomatsakis (Feb 27 2020 at 14:14, on Zulip):

so perhaps the inputs to this "compiler team triage" are more narrow -- we are looking specifically at bugs with certain labels

simulacrum (Feb 27 2020 at 14:14, on Zulip):

I can confirm the "new bugs triaged by release team"

pnkfelix (Feb 27 2020 at 14:14, on Zulip):

Right. It’s the links on the agenda issue for the triage meeting

pnkfelix (Feb 27 2020 at 14:15, on Zulip):

Focusing first on unprioritized issues tagged as nominated or as regressions

nikomatsakis (Feb 27 2020 at 14:16, on Zulip):

PS I'm editing the hackmd

nikomatsakis (Feb 27 2020 at 14:16, on Zulip):

trying to make concrete some intuitions

nikomatsakis (Feb 27 2020 at 14:16, on Zulip):

let me update that part though

nikomatsakis (Feb 27 2020 at 14:16, on Zulip):

one thing I was looking at now is the "priority labels" and what they mean

nikomatsakis (Feb 27 2020 at 14:18, on Zulip):

Severity of bugs section

nikomatsakis (Feb 27 2020 at 14:19, on Zulip):

I also edited the "inputs" etc to

But more narrowly, we propose

* Create a "working group" for compiler team triage
    * Inputs:
        * bugs nominated by the release team triage effort
    * Output:
        * pings to ICE-breaker groups where appropriate to improve issues
        * blog posts that call for help on sticky issues where appropriate
            * critical issues that are not "on track" to being fixed before they hit stable
                * this could be because nobody is looking at them
                * or because progress is too slow
            * issues that need discussion by team to resolve disagreements
nikomatsakis (Feb 27 2020 at 14:21, on Zulip):

@simulacrum are we using GH projects, milestones, etc these days?

nikomatsakis (Feb 27 2020 at 14:22, on Zulip):

at some point we were using milestones I think to track 'things that should be fixed by release X'

simulacrum (Feb 27 2020 at 14:24, on Zulip):

no, not really

simulacrum (Feb 27 2020 at 14:24, on Zulip):

milestones are sort of used by Centril I think

simulacrum (Feb 27 2020 at 14:24, on Zulip):

but they're pretty annoying to maintain

simulacrum (Feb 27 2020 at 14:25, on Zulip):

(particularly with the short release cycles we have that aren't tied to issues being fixed)

nikomatsakis (Feb 27 2020 at 14:27, on Zulip):

@simulacrum ok

nikomatsakis (Feb 27 2020 at 14:27, on Zulip):

it's not clear what they offer over dedicated labels, I suppose

nikomatsakis (Feb 27 2020 at 14:27, on Zulip):

I'm trying to write down criteria for "Critical" bugs (what we today call P-high, and maybe should just keep doing so)

nikomatsakis (Feb 27 2020 at 14:28, on Zulip):

Critical bugs, at the broadest level, are those that will actively affect substantial portions of the ecosystem, harm production users, or existing users.

nikomatsakis (Feb 27 2020 at 14:30, on Zulip):

(but see end of doc for more notes)

nikomatsakis (Feb 27 2020 at 14:33, on Zulip):

I don't remember, @pnkfelix, you had a rather exhaustive write-up before, right? did you try to write out conditions that make a bug "critical" etc?

pnkfelix (Feb 27 2020 at 14:35, on Zulip):

I don’t know where that write up would be, if it exists...

pnkfelix (Feb 27 2020 at 14:35, on Zulip):

(Memory is shot)

pnkfelix (Feb 27 2020 at 14:36, on Zulip):

My intuition had been: is this a bug that a user gets no useful immediate feedback as to its existence and/or how to work around it

pnkfelix (Feb 27 2020 at 14:36, on Zulip):

So eg: an ICE preceded by diagnostics that point out the problem that caused the ICE downstream is not P-high

pnkfelix (Feb 27 2020 at 14:38, on Zulip):

But an ICE with no relevant diagnostic (and whose panic message doesn’t make it blisteringly obvious to an average user what is wrong) is P-high

nikomatsakis (Feb 27 2020 at 14:38, on Zulip):

https://hackmd.io/dvgegmdgQVSMbC4rcjG6EA?view

nikomatsakis (Feb 27 2020 at 14:38, on Zulip):

there is your document I think :point_up:

nikomatsakis (Feb 27 2020 at 14:38, on Zulip):

pnkfelix said:

My intuition had been: is this a bug that a user gets no useful immediate feedback as to its existence and/or how to work around it

yes I tried to capture some of that

nikomatsakis (Feb 27 2020 at 14:38, on Zulip):

pnkfelix off-the-cuff proposal: Maybe instead of “P-high, P-medium, P-low”, we should directly encode the intended visit frequency for each bug. Hz-weekly, Hz-monthly, Hz-yearly… (and find some way to also slice up the sets that are visited, perhaps via modular arithmetic on issue#) so that we could make headway each week on a deterministcally-predictable subset of the Hz-monthly/Hz-yearly bugs)

nikomatsakis (Feb 27 2020 at 14:38, on Zulip):

ooh, I forgot about that

nikomatsakis (Feb 27 2020 at 14:38, on Zulip):

I love that

pnkfelix (Feb 27 2020 at 14:39, on Zulip):

Likewise: incorrect codegen and compiler seg faults, link errors, all P-high

nikomatsakis (Feb 27 2020 at 14:39, on Zulip):

there's other good stuff in here too

nikomatsakis (Feb 27 2020 at 14:39, on Zulip):

e.g., keeping up with future compatibility warnings and planned migrations

nikomatsakis (Feb 27 2020 at 14:39, on Zulip):

I think that would be a prime thing for a working group to tackle

nikomatsakis (Feb 27 2020 at 14:39, on Zulip):

some of the things in here we've made progress on, notably helping to do bisections (ICE-breaker group) etc

nikomatsakis (Feb 27 2020 at 14:40, on Zulip):

though I think we should probably do a kind of "periodic publicity blast" around that sort of thing

nagisa (Feb 27 2020 at 16:03, on Zulip):

I think we’re trying to get through too many things in 1 hour to be honest.

nagisa (Feb 27 2020 at 16:04, on Zulip):

on one hand its necessary to keep up with anything, on the other hand, we hardly have time to talk about any one individual thing to completion, really.

centril (Feb 27 2020 at 16:51, on Zulip):

I only use milestones for relnoted PRs, to have a full list of PRs with changes

centril (Feb 27 2020 at 16:51, on Zulip):

and those are used to e.g. write the blog post

nikomatsakis (Feb 27 2020 at 18:44, on Zulip):

nagisa said:

I think we’re trying to get through too many things in 1 hour to be honest.

you mean in the standard triage meeting, @nagisa ?

nikomatsakis (Feb 27 2020 at 18:45, on Zulip):

I have the same fear about upcoming design meeting :) --

DPC (Feb 27 2020 at 19:29, on Zulip):

catching up with the messages here.. yeah the list shared by niko (1400 utc) looks good to me

pnkfelix (Feb 27 2020 at 21:43, on Zulip):

Perhaps even more worrying is that the vast vast majority of the bugs (1,365) are not prioritized at all.

That list now has 2,213 open issues. That's more than 100 unprioritizd issues each month (since its been 8 months since I wrote that doc).

pnkfelix (Feb 27 2020 at 21:46, on Zulip):

(though to be fair it would probably be good to filter out issues with tags like { C-enhancement, C-feature-request, C-cleanup, C-tracking-issue }, probably...)

nagisa (Feb 28 2020 at 04:14, on Zulip):

nikomatsakis said:

you mean in the standard triage meeting, nagisa ?

yes.

Last update: Sep 28 2020 at 16:15UTC