[cssom-view-1] offsetParent produces incorrect results for position:fixed elements that behave like absolute

[dbaron]

There are a number of cases (such as for transforms) where position:fixed elements behave like position:absolute, rather than attaching to the viewport.

The spec for offsetParent (and presumably offsetTop/offsetLeft as well) doesn't account for this, but instead returns null for position:fixed in step 1.

According to the relevant Chromium bug Firefox does this correctly -- and it seems like it would be best if Chromium changed to match Firefox since that behavior makes the most sense.

[tabatkins]

I'm not sure I understand. The spec states "If the computed value of the position property of ancestor is fixed, and no ancestor establishes a fixed position containing block," - that second condition is what catches and excludes the "fixed acting like absolute" cases, right?

[dbaron]

The problem is step 1 combined with the fourth bullet point within it:

1. If any of the following holds true return null and terminate this algorithm:
   • ...
   • The element’s computed value of the position property is fixed.

[dbaron]

(I think your quote is from the definition of scrollParent rather than offsetParent.)

[tabatkins]

Whoops, yup, I must have scrolled up and not realized I was reading from the wrong spot. Looks like a fix was applied to scrollParent and not copied to offsetParent.