Yoav asked if I could do another HTTP Archive query to look for sites that hit the counters, contain "not all" and contain at least one media feature that's known to be unsupported in Chrome. The idea would be to then look at what those sites do, check if they break, etc.
Looking at a relevant table in MDN <https://developer.mozilla.org/en-US/docs/Web/CSS/@media#browser_compatibility>, I then queried for r'(inverted\-colors)|(overflow\-block)|(overflow\-inline)|(prefers\-reduced\-data)|(scripting)'. I found only a single interesting (non-Yandex-metrics) hit, which was https://marselshoes.by/assets/js/ym.js?hash=e427de0e11f74a8a57a8 . The code was obfuscated/minimized, so it was a bit hard to see if something would break from just reading the code. But at least nothing appeared to be broken if I tried to shop for shoes with the feature enabled. Looking a bit closer today though, I think this is actually the yandex metrics script as well, under a different name. (ym.js = "yandex metrics"?). So maybe there are actually zero interesting results. On Wed, Apr 13, 2022 at 11:59 AM Daniel Bratell <bratel...@gmail.com> wrote: > LGTM1, with a careful launch (as usual, but even more so). > > The usage counter indicates that the ceiling of breakage is quite high (in > the 0.1% magnitude range) but I believe that your analysis of is likely > accurate and the real amount of affected sites is likely to be much lower. > Furthermore, any breakage has a fair chance of not affecting a sites > functionality. > > /Daniel > On 2022-04-12 16:28, 'Joe Medley' via blink-dev wrote: > > Please ping me if it ends up being 103. I don't like missing things in the > beta release post. > > On Tuesday, April 12, 2022 at 12:23:37 AM UTC-7 and...@chromium.org wrote: > >> On Mon, Apr 11, 2022 at 7:45 PM Joe Medley <jme...@google.com> wrote: >> >>> Are you aiming for 102? >>> >> >> That's branching very soon, so definitely not. I'm not really aiming for >> any particular release. But if you want a prediction anyway, then likely >> 104. >> >> >>> >>> On Monday, April 11, 2022 at 7:17:45 AM UTC-7 Emilio Cobos Alvarez wrote: >>> >>>> >>>> >>>> On 4/11/22 15:02, Anders Hartvoll Ruud wrote: >>>> > Ah, I'm not familiar with that way of doing compat research. What >>>> would >>>> > we gain from doing that vs. regular use-counter + HTTP Archive? >>>> > >>>> > Do we expect those 0.1% to visibly break? (I guess that depends on >>>> > what they're feature testing for..) >>>> > >>>> > >>>> > I would not expect that at all based on the HTTP Archive query. If >>>> > testing against "not all" was commonplace, I'd expect more results >>>> > beyond the two Yandex scripts. Or, perhaps it is commonplace, but it >>>> > happens mostly on features that actually *are* supported at the >>>> moment. >>>> > >>>> > Just as an example (and to show that "not all" testing isn't a myth), >>>> > one of the few (non-Yandex-script) sites that did show up was >>>> > https://ww.sapo.pt <https://ww.sapo.pt>, which does the following: >>>> > >>>> > if(rule.mediaText.includes("not all") || ...) >>>> > >>>> > By the looks of it, it's an early-out related to the theme switching, >>>> > which prevents the code from amending the query if >>>> prefers-color-scheme >>>> > is not supported. Had we not supported prefers-color-scheme, then I >>>> > think the worst that could happen is that we end up with a more >>>> > complicated query that still ultimately evaluates to false. Testing >>>> the >>>> > page with the feature enabled (with and without dark mode >>>> preference), >>>> > their color switcher still works normally. >>>> > >>>> > That is just one site though, it's probably theoretically possible to >>>> > write *something* that breaks. I did try to look at the "sample URLs" >>>> > for the counters, but I couldn't actually reproduce the counters >>>> being hit. >>>> >>>> >>>> It's a thing that even Google DevRel people have recommended in the >>>> past, ftr :) >>>> >>>> https://web.dev/prefers-color-scheme/#supporting-dark-mode >>>> >>>> >> Ouch ... >> >> >>> But if you have data to indicate this is not a problem in the wild I'd >>>> be happy to implement it in Gecko after you ship this change. >>>> >>>> Cheers, >>>> >>>> -- Emilio >>>> >>> -- > You received this message because you are subscribed to the Google Groups > "blink-dev" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to blink-dev+unsubscr...@chromium.org. > To view this discussion on the web visit > https://groups.google.com/a/chromium.org/d/msgid/blink-dev/87a783d4-dbd7-4b0c-8ba1-7a118a87d759n%40chromium.org > <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/87a783d4-dbd7-4b0c-8ba1-7a118a87d759n%40chromium.org?utm_medium=email&utm_source=footer> > . > > -- You received this message because you are subscribed to the Google Groups "blink-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+unsubscr...@chromium.org. To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAKFBnUoLaQ6BCr2LZEz9KF3cmpj1F09RZGHrrOmrhr%3D4eov1Ww%40mail.gmail.com.
