Stream: t-compiler/rust-analyzer

Topic: new labels


matklad (Oct 14 2020 at 13:44, on Zulip):

Hey, triage team, I've added a couple of new "process" labels: https://github.com/rust-analyzer/rust-analyzer/pull/6228/commits/190011168dda6ce4836b105b892f0a26eaf07411

Jeremy Kolb (Oct 14 2020 at 16:58, on Zulip):

Hey @matklad what do the E- and S- prefixes stand for?

Laurențiu (Oct 14 2020 at 17:01, on Zulip):

Status and.. easiness?

matklad (Oct 14 2020 at 17:01, on Zulip):

status and experience(?)

matklad (Oct 14 2020 at 17:01, on Zulip):

copy-pasted from rust-lang/rust

Laurențiu (Oct 14 2020 at 17:02, on Zulip):

image.png

Laurențiu (Oct 15 2020 at 12:02, on Zulip):

If anyone is wondering, is:open is:issue -label:S-actionable -label:S-unactionable

matklad (Oct 15 2020 at 12:08, on Zulip):

Lol, I wondered exactly that like ten minutes ago, ended up wit

is:issue is:open sort:updated-desc -label:"S-actionable"  -label:"S-unactionable"
Laurențiu (Dec 21 2020 at 10:56, on Zulip):

Still 430 issues to tag :sweat:

matklad (Dec 21 2020 at 10:57, on Zulip):

@Laurențiu Nicola thanks for making a dent there :-)

I don't think "all issues are triaged & properly addressed" is a an achievable or worthwhile goal -- there are many more issue reporters than issue fixers/triagers

matklad (Dec 21 2020 at 10:58, on Zulip):

I guess, the primary goal is to surface some percentage of S-actionable issues

matklad (Dec 21 2020 at 10:59, on Zulip):

Though, tbh, I can's say I actively use our issue tracker at all. It feels like there's some process-level failure here. It indeed is not easy for me to answer "what is the primary value provided by the issue tracker"?

Laurențiu (Dec 21 2020 at 11:05, on Zulip):

I don't think "all issues are triaged & properly addressed" is a an achievable or worthwhile goal -- there are many more issue reporters than issue fixers/triagers

I know, but it's worth trying. We closed plenty of already-fixed issues in the last day.

Laurențiu (Dec 21 2020 at 11:06, on Zulip):

what is the primary value provided by the issue tracker

It's where people having a problem should look before filing a new issue :smiley:

Florian Diebold (Dec 21 2020 at 12:03, on Zulip):

yeah, I don't understand that question -- clearly if the issue tracker didn't exist, we'd a) be inundated with bug reports here, b) miss many problems because people wouldn't report them, or possibly c) both of those at the same time

Florian Diebold (Dec 21 2020 at 12:03, on Zulip):

I have some feeling that it'd be useful to distinguish between bugs and feature requests, but I don't know if it would actually be

Laurențiu (Dec 21 2020 at 12:08, on Zulip):

I read that as "I don't look on the issue tracker for stuff to implement"

Florian Diebold (Dec 21 2020 at 12:09, on Zulip):

possibly, but I think other people do, considering we regularly get new contributors picking up something from there

Laurențiu (Dec 21 2020 at 12:10, on Zulip):

I think issues are generally either fixed quickly, or maybe never. From time to time we've had PRs to implement some long-standing feature, but that doesn't happen too often.

matklad (Dec 21 2020 at 12:10, on Zulip):

The issue tracker is clearly useful for many things. It's just that I don't have a well-articulated idea of what exactly is it good for.

matklad (Dec 21 2020 at 12:11, on Zulip):

For example, it's both clearly a todo list (I often log various inspections/intentions I wish were implemented), and not a todo list (there's a whole bunch of unactionable issues there, which we will never reasonably address)

Florian Diebold (Dec 21 2020 at 12:13, on Zulip):

I think issues that we know we will never address should be closed. But it's reasonable to have issues about long-term design problems, both as a place to keep notes and as something to point people to when they ask about them

matklad (Dec 21 2020 at 12:15, on Zulip):

I think issues that we know we will never address should be closed.

Empirically, that doesn't work. I haven't seen a public issue tracker which is not a graveyard of issues. This is not a problem in itself, but also not something we can solve

Florian Diebold (Dec 21 2020 at 12:15, on Zulip):

it would be nice to be able to see at a glance only issues that can be addressed right now, but I don't know if that's realistic / if the effort for that would be worth it

matklad (Dec 21 2020 at 12:15, on Zulip):

@Florian Diebold in theory, that's S-actionable label

Florian Diebold (Dec 21 2020 at 12:16, on Zulip):

yeah

Laurențiu (Dec 21 2020 at 12:17, on Zulip):

Yeah, I'm not sure you'll be happy with how I tagged some of the issues. I used that as "this can be implemented, or it's blocked on something that can be implemented, or we're waiting for the reporter or there's a way to make some amount of progress, if not completely fix it". But there's still some amount of subjectivity involved.

Florian Diebold (Dec 21 2020 at 12:18, on Zulip):

hmm ok yeah I think we have differing conceptions of the label then :sweat_smile: I would not tag things that are waiting on information from the reporter as actionable

Laurențiu (Dec 21 2020 at 12:19, on Zulip):

You might have a point there :-)

Florian Diebold (Dec 21 2020 at 12:19, on Zulip):

like, that's even the primary example of non-actionable issues I'd think of

Laurențiu (Dec 21 2020 at 12:20, on Zulip):

But hopefully the reporters will answer soon and we can clean those up

matklad (Dec 21 2020 at 12:20, on Zulip):

or we're waiting for the reporter

Yup, this is unactionable

Laurențiu (Dec 21 2020 at 12:21, on Zulip):

So how do I find issues I've updated recently? :D

matklad (Dec 21 2020 at 12:21, on Zulip):

Also, seeing unactionable label is an impersonal prod for the reporter to come up with minimal reproduciable example .

matklad (Dec 21 2020 at 12:21, on Zulip):

@Laurențiu Nicola sort by updated?

Laurențiu (Dec 21 2020 at 12:26, on Zulip):

or we're waiting for the reporter

Actually, I don't think there's too many of those where I did that

matklad (Dec 21 2020 at 12:29, on Zulip):

:thumbs_up: . I'd also probably tag "blocked on something else" as unactionable.

Generally, I think it's better to classify actionable issues as unactionable, than to classify una as a. The goal is to have as many unactionable issues as possible => that means we can better focus on actionable ones!

Florian Diebold (Dec 21 2020 at 13:13, on Zulip):

maybe we could improve the descriptions of the labels? How about "Someone could pick this issue up and work on it right now" for actionable, and "Issue requires feedback, design decisions or is blocked on other work" for unactionable?

Florian Diebold (Dec 21 2020 at 13:19, on Zulip):

(the current description makes it sound a bit like E-has-instructions)

matklad (Dec 21 2020 at 13:20, on Zulip):

Yeah, thouse are better ones

matklad (Dec 21 2020 at 13:20, on Zulip):

Could you update them?

(trying to manually get release out of the door without breaking code 1.51 users :( )

Florian Diebold (Dec 21 2020 at 13:46, on Zulip):

will do

Joshua Nelson (Dec 21 2020 at 13:46, on Zulip):

what is the point of the S-unactionable label? If you don't plan to fix the issue, why not close it?

Florian Diebold (Dec 21 2020 at 13:46, on Zulip):

S-unactionable doesn't mean we're not planning to fix it, it means we can't fix it right now

Joshua Nelson (Dec 21 2020 at 13:46, on Zulip):

and personally I think 'unactionable' should be different from 'waiting-on-author', when I saw 'unactionable' I thought that was the same as 'wontfix'

Florian Diebold (Dec 21 2020 at 13:47, on Zulip):

well, the new descriptions should help with that

Florian Diebold (Dec 21 2020 at 13:48, on Zulip):

we could distinguish various kinds of unactionable, but it's again questionable whether that would bring any actual benefit

Joshua Nelson (Dec 21 2020 at 13:49, on Zulip):

Florian Diebold said:

S-unactionable doesn't mean we're not planning to fix it, it means we can't fix it right now

hmm ok - could you post why you can't fix it now when you add the label? like for https://github.com/rust-analyzer/rust-analyzer/issues/6547 I have no idea what you need from me

Florian Diebold (Dec 21 2020 at 13:50, on Zulip):

for that one, I think it's just unclear whether we should do it at all

Joshua Nelson (Dec 21 2020 at 13:50, on Zulip):

I see

Joshua Nelson (Dec 21 2020 at 13:51, on Zulip):

well, I still think it would be helpful to distinguish those different kinds of unactionable

Joshua Nelson (Dec 21 2020 at 13:51, on Zulip):

right now that issue sort of seems in limbo, I can't see anyone ever planning to fix it

Florian Diebold (Dec 21 2020 at 13:51, on Zulip):

the way I see it, the labels aren't directed at the issue author; they're a tool for us to organize the issues

Florian Diebold (Dec 21 2020 at 13:57, on Zulip):

maybe that one should be actionable though

Joshua Nelson (Dec 21 2020 at 14:02, on Zulip):

well I'm not trying to get special dispensation :P just understand the thought process

matklad (Dec 21 2020 at 14:11, on Zulip):

right now that issue sort of seems in limbo, I can't see anyone ever planning to fix it

That's exactly the meaning of unactionable, yeah.

matklad (Dec 21 2020 at 14:12, on Zulip):

In my experience, a significant fraction of issues end up in limbo, no matter what triage process is used.

matklad (Dec 21 2020 at 14:12, on Zulip):

S-unactionable is a way to make this "de facto slips through the cracks" explicit

Last update: Jul 24 2021 at 21:15UTC