Closed eeeps closed 2 years ago
I might not have time for a deep code review, but in the meantime @eeeps could you look at some of the linter errors?
While running this through a suite of actual pages, I found an issue where if w
descriptors in the srcset
were wrong (e.g.—on netflix.com's hero right now!), this metric attributed that error to the sizes
attribute.
Addressing that now...
Ok, here are some WPT results, showing behavior in different contexts:
@tunetheweb I'm not sure if an HA admin needs to be around to merge PRs but if so could you keep an eye on this while I'm away this week?
Aye. Though travelling myself this week. But much shorter plane flights than you!
@eeeps if you could get the feedback addressed today that would be great!
@tunetheweb @yoavweiss I think I addressed all of Yoav's feedback but I would like to re-run a few validation tests to make sure that I didn't break anything in the process. I need to run at the moment, I'll try to get back to it in a few hours and give @tunetheweb a final thumbs up.
Thank you!!
Some updated test results, with @rviscomi and @yoavweiss 's changes:
Diffing vs the earlier results, the only changes I see are:
Which is as I'd expect. So.... I think we're good to merge, @tunetheweb
Progress on https://github.com/HTTPArchive/almanac.httparchive.org/issues/2884
As discussed in Slack starting around https://httparchive.slack.com/archives/C017HPKG614/p1652902287349589
I have successfully tested this on a couple of test pages, using WPT. Will run it against a handful of actual pages tomorrow; requesting some external code reviews from people familiar with how Chrome's selection algorithm works right now.