1463588 - Will-change:transform on a non-transformable element shouldn't create a stacking context

Status: NEW

(Core :: Layout, enhancement, P3)

Description

[Matt Woodrow (:mattwoodrow)]

See https://bugzilla.mozilla.org/show_bug.cgi?id=1462412#c12

I think we need to update the will-change bitflags so that NS_STYLE_WILL_CHANGE_{ABS|FIX}POS_CB and NS_STYLE_WILL_CHANGE_STACKING_CONTEXT can share that they only take effect on a transformable frame.

At the very least NS_STYLE_WILL_CHANGE_STACKING_CONTEXT added for will-change:opacity should still apply for non-transformable elements, and we need a way to distinguish the cases.

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I don't think this matches what the will-change spec says, though.  I think it says these effects happen unconditionally.  Maybe the spec should change, though?

What do other implementations do?

Comment 2

[Matt Woodrow (:mattwoodrow)]

"If any non-initial value of a property would create a stacking context on the element, specifying that property in will-change must create a stacking context on the element."

My interpretation of this was that a non-initial value of transform on a non-transformable element wouldn't create a stacking context, thus will-change:transform on the same element shouldn't either.

Comment 3

[Matt Woodrow (:mattwoodrow)]

Attachment: Testcase

Stealing the testcase from the other bug.

Looks like we create a stacking context here (as expected), but chrome doesn't.

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Right -- sounds like this change would make us conform to the spec.  (That's good, given that you need it.)

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I think this is higher priority given that I *think* it's equally possible as a cause of crashes like in 1462412 comment 1 (although probably less common given the properties involved).

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I misunderstood the conditions around the crash; this is just a bug but won't trigger crashes.