Bug 295203 – object-fit: scale-down is ignored when srcset attribute and w descriptor is used

• Component: Layout and Rendering
• Status: REOPENED

Steps to reproduce the problem:

Open url https://codepen.io/zipper/pen/vdwVLb
Image inside first div is incorrectly scaled.
Similar behaviour is observed when sizes attribute is used https://codepen.io/zipper/pen/gvJBdp.
Actual results:

Image is up-scaled. Its size is either corresponding to sizes attribute or viewport width (when sizes is omitted). But source image is smaller and because of "object-fit: scale-down" it shouldn't be upscaled.

Expected results:

My understanding is, that sizes attribute should describe available space (and reserved space for the image), but then the image size is described using its descriptor in srcset. And then the object-fit comes to play, therefore the image should be 400px wide, since it has "object-fit" property set to "scale-down". In the second example, the first image should be 200px wide. Second image is correctly smaller regardless the sizes.

Related issues:
Reported bug in Firefox: https://bugzilla.mozilla.org/show_bug.cgi?id=1974747
Fixes issue in Chrome: https://issues.chromium.org/issues/41374573

[Radar WebKit Bug Importer]

<rdar://problem/154687132>

[Karl Dubost]

Created attachment 475738 [details]
rendering in safari, firefox, chrome

[Ahmad Saleem]

Does Firefox regressed? Since Chromium / Blink commit claims - 'Chrome's behavior now matches Firefox.' while image attached by Karl shows that Safari is matching Firefox.

[Ahmad Saleem]

Pull request: https://github.com/WebKit/WebKit/pull/47479

[Radek Šerý]

Yes, I think the Firefox regressed, I have already submitted an issue there https://bugzilla.mozilla.org/show_bug.cgi?id=1974747.

[Ahmad Saleem]

(In reply to Radek Šerý from comment #5)
> Yes, I think the Firefox regressed, I have already submitted an issue there
> https://bugzilla.mozilla.org/show_bug.cgi?id=1974747.

Thanks for confirming, I already did PR and will wait for review.

[EWS]

Committed 296914@main (ec4f65f46d45): <https://commits.webkit.org/296914@main>

Reviewed commits have been landed. Closing PR #47479 and removing active labels.

[WebKit Commit Bot]

Re-opened since this is blocked by bug 295344

[Radek Šerý]

FYI similar issue in FF has been closed as invalid. Which is technically correct - the reported bug is actually strictly speaking not a bug, although very unexpected behaviour from developer point of view. I have opened an issue on WHATWG to see if there is a will to change the spec https://github.com/whatwg/html/issues/11428. Feel free to express your opinion there, thank you.