853833 - Changing computed box-shadow causes reflow

Status: RESOLVED DUPLICATE of bug 1194480

(Core :: Layout, defect)

Description

[Tim Taubert [:ttaubert] (inactive)]

Attachment: minimized test case to reproduce reflows

I'm currently trying to move about:newtab (the newtab page) from XUL flexbox to CSS3 flexbox. After re-writing most of the styles and opening the new page I noticed that the box-shadow style applied to .site:hover is quite unresponsive (esp. in the original layout, not really bad in the minimized example) when hovering the elements.

Looking at the profiler it seems that every time I'm entering and leaving one of the .site elements we do a reflow. Those reflows seem to go away when removing the .site:hover rule or removing the span.thumb elements in .link (which is quite weird because they have no styles at all).

http://people.mozilla.com/~bgirard/cleopatra/#report=2ae6b0f8335f3ba5379acfb0b9323916ed792371

I tried to minimize the test case as much as possible so that reflows are still there. The lag itself is not that noticeable on a fast machine.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Changes in box-shadow trigger a reflow.  See CalcShadowDifference in nsStyleStruct.cpp.

See also bug 719177 which attempted to fix this and got backed out...  Mats, is there a plan to reland it?

Comment 2

[Mats Palmgren (inactive)]

I don't plan to work on that, no.  I think it might even have been fixed
by others now.

Comment 3

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Not fixed yet, but bug 984226 did most of the prerequisite work needed.

Comment 4

[Gregory Pappas [:gregp]]

Hmm, the test case seems to perform well now. Do you think bug 1194480 fixed this case?

Comment 5

[Daniel Holbert [:dholbert]]

I think so, yeah. Thanks!