Stream: rustdoc

Topic: A-rustdoc vs T-rustdoc


view this post on Zulip Joshua Nelson (May 07 2021 at 20:42):

@Manish Goregaokar have you had another chance to think about this? I do think it's really confusing to have both

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:43):

I'm not convinced

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:43):

it's confusing because we were using them wrong for ages

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:44):

And yes, there is no A-compiler label, because compiler is huge

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:45):

We should mostly be using A-rustdoc and use T- when we need rfcbot

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:46):

Otherwise it's going to be hard for us to integrate with tooling

view this post on Zulip Joshua Nelson (May 07 2021 at 20:46):

Manish Goregaokar said:

The core issue is that we've been using the tool wrong for ages, and if we ever want to autolabel our PRs, this will be a problem

why do you think it will be a problem? you mentioned early having both T-compiler and T-rustdoc could be confusing, but I think if it affects rustdoc maybe we should have input by default, or remove the label manually in special cases

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:47):

there's a difference between "should have input" and "should have blocking input on a particular decision"

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:47):

like

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:47):

Yes, we should have input if it touches rustdoc, but we _do_ have input

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:47):

Nobody stops us from commenting

view this post on Zulip Joshua Nelson (May 07 2021 at 20:47):

I just think "oops I didn't mean to include rustdoc in the FCP" is less confusing and easier to fix than "what's the difference between a-rustdoc and t-rustdoc"

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:48):

however if the compiler team wishes to MCP they're MCPing in their own team

view this post on Zulip Joshua Nelson (May 07 2021 at 20:48):

and it's difficult for me to explain to people not on the team as well

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:48):

I don't see why it's difficult

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:48):

"use A-rustdoc"

view this post on Zulip Joshua Nelson (May 07 2021 at 20:48):

maybe it's because I don't fully understand the difference - should prioritized issues use A-rustdoc or T-rustdoc?

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:48):

the rest of the project already works like this

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:49):

everything uses A-rustdoc

view this post on Zulip Joshua Nelson (May 07 2021 at 20:49):

Manish Goregaokar said:

the rest of the project already works like this

I don't agree with this :confused: the only teams using A-* instead have separate issue trackers

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:49):

T-rustdoc is only if we want to nominate things for meetings and fcps and such

view this post on Zulip Joshua Nelson (May 07 2021 at 20:49):

it's already hard enough to find "issues related to rustdoc" without having to OR them together

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:49):

We should not OR them

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:49):

There should be no situation where an issue is T but not A

view this post on Zulip Joshua Nelson (May 07 2021 at 20:50):

Manish Goregaokar said:

There should be no situation where an issue is T but not A

there's no way to enforce this, and it happens in practice because people don't know which label to use

view this post on Zulip Joshua Nelson (May 07 2021 at 20:50):

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

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:51):

They all have A tags? I guess this is more of a "we have multiple A tags that are somewhat nested" problem

view this post on Zulip Joshua Nelson (May 07 2021 at 20:52):

only 2 of the 5 have a-rustdoc tags

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:52):

Oh

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:52):

I meant that one had a-rustdoc-ui

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:52):

But this mostly seems to be team members being used to the old thing

view this post on Zulip Joshua Nelson (May 07 2021 at 20:52):

and the ratio gets worse when you look at PRs https://github.com/rust-lang/rust/labels/T-rustdoc

view this post on Zulip Joshua Nelson (May 07 2021 at 20:52):

Manish Goregaokar said:

But this mostly seems to be team members being used to the old thing

well, my concern is that the benefit doesn't seem to be worth the churn

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:53):

I'm pushing back because the old situation was terrible for everyone else, it was always confusing, and I'd keep getting asked

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:53):

the rest of the project used A- labels, and we would often not have an appropriate A- label

view this post on Zulip Joshua Nelson (May 07 2021 at 20:53):

Manish Goregaokar said:

I'm pushing back because the old situation was terrible for everyone else, it was always confusing, and I'd keep getting asked

hmm that was not my experience :confused: you can always just use T-rustdoc if you couldn't find the exact label

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:53):

right

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:53):

That was confusing

view this post on Zulip Joshua Nelson (May 07 2021 at 20:54):

I don't understand why.

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:54):

Because the rest of the project used T- as something team members applied exclusively

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:54):

So people didn't feel empowered to apply it

view this post on Zulip Joshua Nelson (May 07 2021 at 20:54):

oh I see, the issue is people not on the team don't have permissions to add T-rustdoc

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:54):

they do!

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:55):

they don't feel empowered to use it

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:55):

because _that's not what T-foo is for_

view this post on Zulip Joshua Nelson (May 07 2021 at 20:55):

Manish Goregaokar said:

because _that's not what T-foo is for_

that is definitely not true in practice though https://github.com/rust-lang/rust/labels/T-compiler

view this post on Zulip Joshua Nelson (May 07 2021 at 20:55):

almost 3/4 of the open issues are labeled with t-compiler

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:56):

like, the problem within the team we're seeing with this change? the rest of the project had the same problem, and it's easier for us to fall in line rather and teach _our team_ to do things differently

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:56):

sure

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:56):

this is something that perhaps should be gated by tooling

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:56):

but it's not a big deal if things are mislabeled

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:56):

as far as "having to type OR", we should be bookmarking searches or something if that's such a huge deal

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:56):

can ask for tooling to help with this

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:57):

e.g. get triage dashboards for this

view this post on Zulip Joshua Nelson (May 07 2021 at 20:57):

I still don't understand the problem I think. I haven't seen people be hesistant to add T-rustdoc, but if they are, maybe we can address that with a rustdoc issue template?

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:57):

i don't think that's going to be super effective

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:57):

nobody experienced with the project uses issue templates

view this post on Zulip Joshua Nelson (May 07 2021 at 20:58):

Manish Goregaokar said:

nobody experienced with the project uses issue templates

well, I do

view this post on Zulip Joshua Nelson (May 07 2021 at 20:58):

but maybe not most

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:58):

yes

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:58):

but if the problem is people new to the project not understanding A vs T, that we can fix with an issue template

view this post on Zulip Joshua Nelson (May 07 2021 at 20:58):

I still don't think that's a strong enough drawback to have two almost identical tags for :/

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:58):

they're not identical!

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:58):

they're _very_ different

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:59):

A is an area tag

view this post on Zulip Joshua Nelson (May 07 2021 at 20:59):

I don't think rustdoc needs to be the one to pioneer "only use T-* for FCPs"

view this post on Zulip Joshua Nelson (May 07 2021 at 20:59):

right yes you've made clear what you think the difference is

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:59):

that's not what I'm proposing

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:59):

I'm proposing we do what the other teams do: studiously apply A- tags to everything, not care too much about which things are T- tagged but T-tag things when we need FCPs

view this post on Zulip Manish Goregaokar (May 07 2021 at 20:59):

like

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:00):

there's a bit of postel's law going on here with T- tags

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:00):

we choose to apply it to certain specific cases but we don't care if they get applied more broadly

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:00):

which is what the other teams do

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:00):

and then we'll be able to have tooling auto apply A- tags

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:00):

to PRs

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:00):

i don't see us pioneering anything here

view this post on Zulip Joshua Nelson (May 07 2021 at 21:01):

Manish Goregaokar said:

and then we'll be able to have tooling auto apply A- tags

I still don't understand the issue with auto-applying T-rustdoc :confused: I feel like we're going in circles

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:01):

because lots of PRs touch rustdoc in inconsequential ways

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:01):

and these are almost always larger compiler PRs and i don't want to step on their MCPs

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:01):

but also i've had this issue with T-rustdoc for ages where other teams keep being confused about it

view this post on Zulip Joshua Nelson (May 07 2021 at 21:02):

Manish Goregaokar said:

but also i've had this issue with T-rustdoc for ages where other teams keep being confused about it

(here's a circle: https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/nominated.3A.20.2381070/near/237888419)

view this post on Zulip Joshua Nelson (May 07 2021 at 21:03):

anyway I don't have energy to keep arguing about this

view this post on Zulip Joshua Nelson (May 07 2021 at 21:03):

I'll add A-rustdoc to the issues missing it

view this post on Zulip triagebot (May 07 2021 at 21:04):

Issue #81070's nomination request has been removed.

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:04):

Joshua Nelson said:

Manish Goregaokar said:

but also i've had this issue with T-rustdoc for ages where other teams keep being confused about it

(here's a circle: https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/nominated.3A.20.2381070/near/237888419)

I already answered that :(

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:04):

like the problem is that because everyone else uses T- differently, the people _on a team_ will be reluctant to apply T-rustdoc

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:04):

people not on a team don't think abotu this

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:04):

but it's friction

view this post on Zulip Joshua Nelson (May 07 2021 at 21:05):

and I already responded to your answer :P https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/nominated.3A.20.2381070/near/237888649 and you responded to that https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/nominated.3A.20.2381070/near/237888776

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:05):

there have also been problems in the past with team members going around and applying T-foo overbroadly on things belonging to other teams and causing friction

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:05):

Joshua Nelson said:

and I already responded to your answer :P https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/nominated.3A.20.2381070/near/237888649 and you responded to that https://rust-lang.zulipchat.com/#narrow/stream/266220-rustdoc/topic/nominated.3A.20.2381070/near/237888776

no, that's not a response

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:05):

"people on a team"

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:05):

vs "random people"

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:05):

Manish Goregaokar said:

there have also been problems in the past with team members going around and applying T-foo overbroadly on things belonging to other teams and causing friction

but this is also a thing i'd like to avoid if we auto-apply

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:05):

T- labels have caused strife in the past because of this

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:05):

it sounds dumb but they have

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:06):

i have to go but i can tell you privately

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:06):

also when random people apply it it tends to be something folks are happy to undo

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:06):

like when i triage an issue i'll often redo the labels

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:06):

but if, say, oli-obk filed an issue, i won't remove labels because he probably had reasons

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:06):

this is why it's nuanced

view this post on Zulip Joshua Nelson (May 07 2021 at 21:07):

Manish Goregaokar said:

T- labels have caused strife in the past because of this

ok, I think this was the bit I was missing

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:07):

"people use T- labels wrong already" is not an answer to "people _on teams_ are reluctant to use T- labels wrong"

view this post on Zulip Joshua Nelson (May 07 2021 at 21:07):

not 100% convinced A-rustdoc is better than the alternative but willing to keep trying it for now and see if people pick it up :)

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:07):

this is like saying "let's not build a zebra crossing" because "i can see people jaywalking"

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:08):

sure

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:08):

i think we can bookmark searches and stuff to make this automatic

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:08):

and tell everyone on the team to use the label

view this post on Zulip Manish Goregaokar (May 07 2021 at 21:08):

i think this will be annoying for a couple months and after that fine

view this post on Zulip Noah Lev (May 07 2021 at 22:22):

Note: I moved this discussion into its own topic, because it's not related to the nominated issue.

view this post on Zulip GuillaumeGomez (May 08 2021 at 10:15):

Manish Goregaokar said:

I'm pushing back because the old situation was terrible for everyone else, it was always confusing, and I'd keep getting asked

Until you brought it up, no one had issue with it and we simply put T-rustdoc on every rustdoc issue alongside the appropriate A-*. I personally don't care much about whether we have A-rustdoc or T-rustdoc, but having both is weird

view this post on Zulip GuillaumeGomez (May 08 2021 at 10:40):

Also, I still wonder for example why there isn't both A-compiler and T-compiler for example.

view this post on Zulip Manish Goregaokar (Jun 03 2021 at 17:50):

That's because the compiler team has an A- tag for everything, and they're _big_ so it's not useful to filter by "all compiler bugs"

I'm fine with getting rid of A-rustdoc _if we have sufficient A- tags to cover all rustdoc things_, and people are using those A tags, not the T tag

The point is that we should not be using T-rustdoc unless decisions are involved.

view this post on Zulip Joshua Nelson (Jun 04 2021 at 17:55):

Manish Goregaokar said:

The point is that we should not be using T-rustdoc unless decisions are involved.

you cannot get people to stop using it though :/ it looks like the right thing unless you click through to the label description, which no one does. https://github.com/rust-lang/rust/issues/85995#issuecomment-854880669

view this post on Zulip Manish Goregaokar (Jun 04 2021 at 18:13):

@Joshua Nelson T-rustdoc-team then? Like, I would prefer to _remove_ that label in that case

view this post on Zulip Manish Goregaokar (Jun 04 2021 at 18:13):

T-rustdoc-team-discussion

view this post on Zulip Manish Goregaokar (Jun 04 2021 at 18:13):

and we get rfcbot to learn the new label

view this post on Zulip GuillaumeGomez (Jun 04 2021 at 19:38):

T-rustdoc-team-discussion would be better I tihnk. At least there is no confusion possible

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:28):

Hi, random person new to the project who has been categorizing issues as a way to help out here.

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:28):

This division is extremely confusing to me.

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:30):

I even read the label. When I looked at it, "decisions" didn't seem like it's an important meaning-carrying word in that description. I'm sorry I didn't pick that up.

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:33):

As far as I can tell, the way these are divided is different from the way everyone else does it. T-* labels are supposed to show which team is responsible for dealing with it. A-* labels are supposed to show topics that are involved in the problem.

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:33):

Or, at least, that's what I've surmised. This is not documented anywhere, which means you pretty much have to work it out for yourself.

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:35):

As a random helper person, I add T-compiler if something looks like it would be within the compiler team's scope. I figure that at some point they'll look through the issues and deal with the ones that they think need dealing with.

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:40):

If T-* labels were just for the relevant team to add, they wouldn't be on the unauthenticated list. The fact that they are implied to me that I should be adding them.

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:44):

It was also mentioned in #t-release/triage that triagers are encouraged to add them. (I'm not actually in that WG; I'm interested in joining someday perhaps, but I'm a bit scared of the responsibility).

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:46):

The final point I'll make is that there are more specific labels to be placed if something needs a decision. I-nominated is one; that one implies that it will be discussed at the next meeting of the team, and is restricted. There's also I-needs-decision, which currently anyone can add. There are a few others that may be applicable in some situations.

view this post on Zulip inquisitivecrystal (Jun 05 2021 at 03:46):

I apologize. I feel like I'm being way too verbose and also too assertive. Those are both bad habits of mine.

view this post on Zulip GuillaumeGomez (Jun 05 2021 at 12:21):

No, you provided good explanations. Thanks for providing your input, it's very appreciated! So I guess we should simply remove A-rustdoc and use I-nominated and I-needs-decision to make it work like other teams. Unification is a good thing.

view this post on Zulip Joshua Nelson (Jun 21 2021 at 00:44):

@Manish Goregaokar whatever conflict happened around this in the past, I really don't think the confusion is worth it. We can deal with that when and if it happens, but right now it just ends up with people applying wrong or duplicate labels:

view this post on Zulip Joshua Nelson (Jun 21 2021 at 00:44):

(everything currently labeled with T-rustdoc is wrong according to your criteria I think: https://github.com/rust-lang/rust/labels/T-rustdoc)

view this post on Zulip Joshua Nelson (Jun 21 2021 at 00:45):

I would like to get rid of A-rustdoc for now and switch everything using it back to T-rustdoc.

view this post on Zulip Joshua Nelson (Jun 21 2021 at 00:49):

I am ok with not automatically applying T-rustdoc if you have concerns (although I do think that without clear guidelines on what shouldn't have the label it doesn't actually help with preventing conflicts, just makes it less likely).

view this post on Zulip Manish Goregaokar (Jun 21 2021 at 02:26):

@Joshua Nelson okay, that's fair. go for it. I still think we should as much as possible avoid automatically applying T-rustdoc but also okay wit hthat if folks feel strongly

view this post on Zulip Joshua Nelson (Jun 21 2021 at 02:29):

we have some time to decide; https://github.com/rust-lang/triagebot/pull/1321 is a long way from landing. I'll make sure I talk with you before sending a PR to automatically apply anything :)

view this post on Zulip Joshua Nelson (Jun 21 2021 at 02:30):

@Manish Goregaokar how did you flip the labels the first time, did you run a script or something? I don't want to go through 21 pages of issues by hand haha

view this post on Zulip Manish Goregaokar (Jun 21 2021 at 17:57):

I renamed the existing label and added a new one

view this post on Zulip Manish Goregaokar (Jun 21 2021 at 17:58):

find the label that has less, do a search for having one label but not the other, bulk add the other label, and then rename

view this post on Zulip Joshua Nelson (Jun 22 2021 at 00:06):

ok, done

view this post on Zulip Joshua Nelson (Jun 22 2021 at 00:07):

thanks for being flexible about this :)


Last updated: Oct 11 2021 at 22:34 UTC