Is there a way to make the test runner run a *single* test, specified in the URL?

(The online one at w3c-test.org/tools/runner/index.html)

Oh, if they're testharness I can just link straight to them, looks like.

yes (there's basically no reason to ever use than runner)

I mean I guess it's better than nothing for running multiple testharness.js tests

The docs currently point to that runner, and don't mention that you can just navigate directly to the test.

gsnedders: I'm seeing "POST w3c-test.org/tools/runner/update_manifest.py 500 (Internal Server Error)" is this normal?

TimothyGu: yeah, the runner has never really worked well on w3c-test.org :/

:(

honestly we should just have better tooling to run the tests; better UI for ./wpt run locally, some better way to run them in browser (and run more than just testharness.js tests), and make that less useful generally by having up-to-date results at wpt.fyi

gsnedders: is there a way to mute websocket connection notifications, like those shown in travis-ci.org/tmpvar/jsdom/jobs/329347954#L5163?

or, for more fun, travis-ci.org/tmpvar/jsdom/jobs/329347954#L5207-L5285

I assume you're the WPT guy but correct me if I'm wrong

TimothyGu: I… guess they come from the websocket server, so I don't know.

yeah, they come from pywebsocket

so no way to disable it?

Pretty sure it's coming directly from the third party code, so nothing we directly maintain. Might be configurable with logging levels and the like? I don't know. I've never touched the WebSockets stuff.

ok..

Sorry, I can't give a useful answer.

it's okay, just stopping by

annevk, spec question about valid floating point numbers

are whitespaces in front of the numbers allowed?

i.e. is " 123.345" considered a valid floating point number?

link to spec: html.spec.whatwg.org/multipage/comm…es.html#valid-floating-point-number

KiChjang: what would suggest it’s valid?

the parsing part of it suggest that it may be valid, since on step 6 it is allowed to skip whitespaces

KiChjang: we parse many invalid inputs

ah ok, that's the part i'm missing

my intuition was successful parsing == valid string

smaug____:FYI, I'm going to fix the WPT test, github.com/w3c/web-platform-tests/b…posed-path-with-related-target.html, as well as fixing Blink's implementation at chromium-review.googlesource.com/c/chromium/src/+/869693.

I wonder what GitHub uses to mark some of my comments as verified and some not

I do push from different machines, but I think I use identical setups...

And I'm not entirely sure it's related to that

Also, Travis being offline is kinda weird

annevk: it's GPG signing

if you merge on GitHub then it uses GitHub's PGP key and GH automatically recognizes it as verified

If you push from your own machine then depends on the setup

see jsdom/webidl2js bfb6ee1 (GitHub's key) and jsdom/webidl2js b26285a (my own key on my machine)

if you don't have signing enabled on your local machine, then only merge commits created on GitHub will be verified, I assume

TimothyGu: I merge everything through GitHub though; either rebase and merge or squash and merge

TimothyGu: and the difference is also not in the distinction between those, judging from recent whatwg/html changes

hmm, maybe fast forward vs. not fast forward? no clue

And locally I use GitHub Desktop, logged in to GitHub directly

It's probably some minor config difference somewhere, but I don't care enough to find out

annevk, TimothyGu: pretty sure the commit has to be created on GitHub, not merely rebased on GitHub (hence squash/rebase & merge won't do it)

Fact check for a talk: Microtasks were originally invented for mutation observers right?

JakeA: yeah

Ta!

<main> in <td> realto.ru/journal/articles/kogda-mo…taciya-yurista-po-zhilishnomu-pravu

because it’s using tables for layout

<main> in <section> mentecomunal.net/author/luis and artlist.io

annevk: foolip: stats on <main> are now included in validator.w3.org/nu/stats.html

but note that it caches very aggressively

so if you want to see fresh data, do something like validator.w3.org/nu/stats.html?a

then the next time do validator.w3.org/nu/stats.html?b

or whatever

etc

I am also logging URLs on the backend for the interesting cases

<main> in <td>, <main> in <section>, etc

so we will have real current cases to inspect and consider

another main in section case pticainfo.ru/article/?ELEMENT_ID=1780

multiple visible <main> ukv.de/content

anyway, with that I’m off soon for the night

hopefully my mostly untested code for those counters doesn’t cause some unexpected eff up in the checker while I’m sleeping

Thanks MikeSmith! ❤️

MikeSmith: thanks!

I also have stats about what the ancestors usually are now, but my draft comment is at work so it'll have to wait until tomorrow

But basically, form and section are the most common ones that we're contemplating making invalid

annevk: ^

foolip: do those pages make sense though or would a message from the checker do some good?

annevk: I looked at a few and quickly found one that seemed fine, where <main> was just inside a <section>, and inverting it seems like pointless work

foolip: I don't think that's the only consideration though

foolip: section is a sectioning element too and if it's used as such that seems rather wrong, at least if we ever want to make sections work properly

foolip: if it's a high number, maybe, but otherwise I'm not really convinced

are there any ways to generate a network request on a worker thread besides fetch(), xhr, improtScripts(), and es import()?

offline canvas or something?

^offline^offscreen

wanderview: new Worker in theory

annevk: oh yea... ok

wanderview: WebSockets/EventSource I guess

wanderview: pretty sure canvas has no networking btw

ok

annevk: thanks

annevk: seems like we need to look at some sample of the documents we have that use <main> in <section> and <form> and make a determination about which cases are misuses and which seem legit

foolip: ⬆

and if it turns out that the percentage of misuses are way higher than the percentage of legit uses, then that argues for making it non-conforming

and for the checker to emit an error for those

and regardless, unless the legit uses seem to greatly outweigh the misuses, then those still would seem to merit at least a “should not” in the spec

and for the checker to at least emit a warning

the stats at this point show so far about 800 cases of each that we could take a look at

to narrow things, I could also add some more-granular stats

like, I could get stats on how many of the <section> cases that contain <main> are <section> without a heading

which is arguably a misuse of <section>, in that it indicates the author is just using where they should be using <div>

and also because <section> without a heading can be bad for AT users, if the AT exposes each of those <section> even though they have no identifying heading (and so makes exposing them for navigation pretty worthless)

which is why the checker emits a warning for <section> without a heading

Afaict, the DOM methods that use selectors are: querySelector, querySelectorAll, closest, matches, webKitMatches

Are there any others?