Bugzilla

(In reply to Botond Ballo [:botond] from comment #13)
>  * The reason for _that_ early-exit is the intersection [here](https://searchfox.org/mozilla-central/rev/9b282b34b5aa0f836beb735656c55efb2cc4c617/layout/painting/nsDisplayList.cpp#501). The backdrop frame's visual overflow rect seems to include the toolbar height (and thus intersect the hit-test point) on the page in bug 1643212, but not the page in this bug. I don't know why yet.

Investigating why the backdrop frame's visual overflow rect is sized the way it is may not be a useful direction. I don't think we want to be relying on the presence of the backdrop frame to force us to descend into the scroll frame via the mark-for-display mechanism; for all we know, a future optimization may result in there not being a backdrop frame at all.

Rather, I think the problem is the **scroll frame's** visual overflow rect: it should include the toolbar height.

(In reply to Botond Ballo [:botond] from comment #13)
>  * The reason for _that_ early-exit is the intersection [here](https://searchfox.org/mozilla-central/rev/9b282b34b5aa0f836beb735656c55efb2cc4c617/layout/painting/nsDisplayList.cpp#501). The backdrop frame's visual overflow rect seems to include the toolbar height (and thus intersect the hit-test point) on the page in bug 1643212, but not the page in this bug. I don't know why yet.

Investigating why the backdrop frame's visual overflow rect is sized the way it is may not be a useful direction. I don't think we want to be relying on the presence of the backdrop frame to force us to descend into the scroll frame via the mark-for-display mechanism; for all we know, a future optimization may result in there not being a backdrop frame at all.

Rather, I think the problem is the **scroll frame's** visual overflow rect: it should include the toolbar height. (Then we would descend into the scroll frame regardless of whether we take the early-exit it `DescendIntoChild` or not.)