229687 2021-08-30 14:05:02 -0700 Lazyload does not work properly with negative margins 2021-08-31 07:05:11 -0700 1 1 1 Unclassified WebKit Images Safari 14 Mac (Intel) Unspecified RESOLVED INVALID P2 Normal --- 1 domfarolino webkit-unassigned rbuis sabouhallawa simon.fraser oldest_to_newest 1788650 0 domfarolino 2021-08-30 14:05:02 -0700 I'm a Chromium engineer, and while looking into a lazyload bug report on Chromium (crbug.com/994329#c14) I found that WebKit's lazyload implementation did not behave properly on a site using `loading=lazy` images. Specifically, there is a discrepancy between whether the IntersectionObserver API sees an image as "intersecting" vs whether the lazyload implementation sees an image as "intersection. See https://lazyload-study.glitch.me/crbug-994329.html for an example. I think the IntersectionObserver API results should be the same as the lazyload implementation results, since LL is spec'd based off of IntersectionObserver: https://html.spec.whatwg.org/multipage/urls-and-fetching.html#start-intersection-observing-a-lazy-loading-element. This works correctly in Chrome and Firefox. 1788810 1 rbuis 2021-08-31 01:11:28 -0700 Hi Dom, did you assume here lazy load is enabled in Safari releases? It is not. I checked with a local build though, and using the mini browser the test case behaves as expected, i.e. the image does not load and IO status is false. Indeed we use IO internally so we should always pass this test case. 1788883 2 domfarolino 2021-08-31 06:27:59 -0700 Oh wow, I didn't realize it was not enabled yet, I just assumed all modern browsers supported it. Thanks, looks like this can be closed 1788888 3 rbuis 2021-08-31 07:05:11 -0700 (In reply to Dominic Farolino from comment #2) > Oh wow, I didn't realize it was not enabled yet, I just assumed all modern > browsers supported it. Thanks, looks like this can be closed Well, the test is still nice, I'll keep it in mind, thanks!