1198996 - Layer::GetLocalOpacity() can return values outside of 0..1 range triggering Skia assertion in DrawTargetSkia

Status: RESOLVED FIXED

(Core :: Graphics: Layers, defect)

Description

[Lee Salzman [:lsalzman]]

In SkColor.h:SkColorSetARBInline(U8CPU a, U8CPU r, U8CPU g, U8CPU b) we're triggering the following assertion:

SkASSERT(a <= 255 && r <= 255 && g <= 255 && b <= 255);

The a value is exceeding 255, because inside the SkPaint setup we're passing in an alpha value exceeding 1 before conversion, and this value is originating from Layer::GetEffectiveOpacity().

Note that since GetEffectiveOpacity() is multiplying the result of GetLocalOpacity() together, if these values are outside the 0..1 range the result can be even more generally bogus if any of the intermediate values is negative for some reason.

Comment 1

[Lee Salzman [:lsalzman]]

Attachment: ensure layer local opacity is clamped to 0..1 range

Comment 2

[Lee Salzman [:lsalzman]]

Try run: https://treeherder.mozilla.org/#/jobs?repo=try&revision=518a5a6f5eaa

Comment 3

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/c5f9c9319016

Comment 4

[Robert O'Callahan (:roc) (email my personal email if necessary)]

How can GetLocalOpacity() get outside the range 0..1? That should never happen.

Comment 5

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/c5f9c9319016

Comment 6

[Lee Salzman [:lsalzman]]

(In reply to Robert O'Callahan (:roc) (Mozilla Corporation) from comment #4)
> How can GetLocalOpacity() get outside the range 0..1? That should never
> happen.

It should never happen, except when it does... And then other stuff downwind blows up as a consequence.

I couldn't definitively track down the source, but it looks like there is some style animation of opacity that is sometimes, on the test machines, producing values slightly over the end of range.