W3C home > Mailing lists > Public > public-test-infra@w3.org > April to June 2018

Re: Towards a multi-vendor WPT triage process?

From: Philip Jägenstedt <foolip@google.com>
Date: Thu, 05 Apr 2018 08:48:38 +0000
Message-ID: <CAARdPYcstpDvDONxMwXY7eo_c8bLJLR-ZMuopB=DcUhXE0hUoA@mail.gmail.com>
To: public-test-infra <public-test-infra@w3.org>
So, last week I tried using https://foolip.github.io/wpt-rotation/ myself
for a few days, made some of the circles green, but noticed some problems.

Neither "new issues (no comments)" nor "old issues (>90 days)" quite work I
think. Some new issues are quite clear (example
<https://github.com/w3c/web-platform-tests/issues/10021>) and not in need
of further commentary, and many old issues really aren't in need of further
triage but just fixing (example
<https://github.com/w3c/web-platform-tests/issues/5839>).

The fix for both of these problem for the infra label (
https://foolip.github.io/ecosystem-infra-rotation/) was to have three
priority labels, treat issues as untriaged if they have no label, and never
revisit the lowest priority. I'm sort of skeptical about extending this to
everything where nobody is yet being paid to do work in the area, so what
can one do?

I'm starting to think that a wider triage process (beyond the infra label)
by itself isn't that valuable, but that we'd actually need people working
to review PRs and fix known bugs in each area. Which is what James always
says, of course :)

I think we have a real problem that people send PRs and issues to wpt and
they just go nowhere or very slowly. Other ideas for making progress on
that?

On Wed, Mar 28, 2018 at 2:37 PM Philip Jägenstedt <foolip@google.com> wrote:

> Hi all,
>
> At the last TPAC we discussed the problem of WPT issue/PR triage:
> https://www.w3.org/2017/11/07-testing-minutes.html#item06
>
> I've put together something roughly in the shape of what I think a
> top-level or "fallback" triage process would look like:
> https://foolip.github.io/wpt-rotation/
>
> (Based on https://foolip.github.io/ecosystem-infra-rotation/)
>
> My rough idea is that we need something to push things into other labels,
> both for issues and PRs, and that if the triage of those labels isn't
> working well it all falls apart. That's a big if, because with more
> widespread 2-way sync I'd expect it to become *more* difficult to get
> review upstream, even if the project as a whole is more active.
>
> Anyway, does something like this look sensible as the top-level triage and
> the labels that have to do with the project itself as opposed to individual
> test suites? Does is seem plausible that we could get to a state where
> these circles are made green on a daily basis, and would that amount to any
> real value?
>
> To make this work, I think we'd need to define a multi-vendor rotation
> where we take turns doing this work, so this is also a call for volunteers
> if that makes sense.
>
> Thanks!
>
Received on Thursday, 5 April 2018 08:49:13 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 5 April 2018 08:49:14 UTC