@jgraham Hi James, what do you think of this change? Can you try it on Mozilla CI to see if it breaks anything?

@foolip Regarding the check failure, a documented assumption is that Ahem is installed on the device. I guess we'd need to break this assumption, or relax it to say that Ahem is available as a webfont (and fix the test generator here)?

see also #9105 for earlier discussion of this

@foolip Regarding the check failure, a documented assumption is that Ahem is installed on the device. I guess we'd need to break this assumption, or relax it to say that Ahem is available as a webfont (and fix the test generator here)?

If we're changing behaviour to always load Ahem as a web font, we have no need for that "assumption" and should drop it and the related test (don't even bother changing it: there's no point; it would just become a tautology that the web font is equal to the web font).

That said, more generally as I think was clear in #9105 I'm generally in favour of this but have never been able to justify anyone spending their time on it (so, uh, thanks!), but I do think we need an RFC for this change (which shouldn't be too hard to get through!).

I'd be interested in whether this made any notable difference to overall test execution time, but I'd hope the web font would get cached quite easily.

Don't you have to ensure to wait until the webfont is actually loaded?

Don't you have to ensure to wait until the webfont is actually loaded?

One of the guarantees of WPT is that the reftest is taken after the document.fonts.ready promise resolves (though see also #11476).

What's the intention of the change? To make the tests more robust due to issue with Ahem not being available from the system?

I don't have objections to this change in particular, but I think it'd be beneficial to improve our system font testing story, as outlined in #13203 (comment) or it would generally help if control over system fonts available to WPT became more reliable, or just .. available.

Regarding the motivation for this change: it's to allow us to run WPT on systems where it's unlikely or difficult to have a system font installed (specifically Chromium CI).

I've created an RFC for this here: web-platform-tests/rfcs#22

More generally, I think it's a big help for people running tests locally in their own browser, as that's a situation where it's much less likely they'll have Ahem installed versus (e.g.) fonts with decent coverage of characters.

Discarding this PR, will split as per RFC discussion