fix: root rects #1099
fix: root rects #1099
Conversation
| // Math.abs(win.innerWidth / visualViewport.scale - visualViewport.width) < | ||
| // 0.001 | ||
| // Fallback here: "Not Safari" userAgent | ||
| if (!/^((?!chrome|android).)*safari/i.test(navigator.userAgent)) { |
There was a problem hiding this comment.
If we end up using this check we may want to move it outside the function so that we don't run it more than once per session
There was a problem hiding this comment.
I think it's too impractical to use since it could change in the future, have you had a look at this? Maybe you can find a way to feature detect this that works for all edge cases? There might be something fundamentally wrong with the approach. Not being able to write a test for it doesn't exactly help either.
There was a problem hiding this comment.
Puppetter should allow us to enable the mobile emulation mode, if we want to write a test for this we can. I'll try to add one now.
There was a problem hiding this comment.
To summarize the issue:
Safari's getBoundingClientRect() returns "visual viewport" values by default (including pinch-zoom), but Chrome's returns "layout viewport" values (not including pinch-zoom). This means it already uses visual viewport by default, so the offsetLeft/Top values to determine the offset relative to getBoundingClientRect() always need to be 0, but in Chrome, we need to consider them.
So we need to be able to feature detect if getBoundingClientRect() is returning visual or layout viewport values, and if it's visual, don't add the offsets.
There was a problem hiding this comment.
I just pushed a commit to add the test for this bug, I haven't added the pinch to zoom test yet since I haven't found how to perform the action 😅
Change Log for @popperjs/core (2.4.0 → 2.4.1)Bug Fixes
|
Problems I encountered:
documentRectwidth was inaccurate (area outside of<html>/<body>bounds if it was horizontally scrollable) and possibly the height. I'm not sure what effects this has exactly since it seemed to work before withrootBoundary: "document"viewportRectin mobile emulation mode when pinch zooming broke the horizontal demo in Wrong tooltip positions when X axis has overflow (on mobile) #1097 - it looks like the feature detection doesn't work with when the browser has zoomed the viewport out. I cannot think of a way to fix it besides userAgent detection which can easily break in the future if Safari decides to start using "layout viewport"...