855233 - Image with opacity, border-radius covered by a positioned element disappears after hover (BasicLayers only)

Status: VERIFIED FIXED

(Core :: Layout, defect)

Description

[Cork]

Attachment: testcase image

Steps to reproduce:
1. Load the testcase
2. hover the image
3. blur the image

Actual results:
The image disappears

Expected results:
The image should go back to 0.8 opacity

In the regression range the image is never rendered. In the improvement range the image is shown on first render, but dispersal after hover (current state)


Regression:
Last good nightly: 2012-09-17
First bad nightly: 2012-09-18

Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=3f0587ce1774&tochange=0d3b17a88d5f


Improvement:
Last bad nightly: 2012-09-24
First good nightly: 2012-09-25

Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?startdate=2012-09-25&enddate=2012-09-24

Comment 1

[Cork]

Attachment: testcase

Comment 2

[Boris Zbarsky [:bzbarsky]]

Worksforme on Mac...

Comment 3

[Matt Woodrow (:mattwoodrow)]

Attachment: Create a new mask if the offset changes

This only reproduces with BasicLayers.

When the layer becomes inactive, mParameters.mOffset becomes non-zero and we were re-using the existing mask (with the incorrect transform).

With accelerated layers we are switching backend type, so that would force the required mask recreation.

I guess we could just update the transform on the mask if this changes, rather than rebuilding. I believe the same applies for the existing mXScale/mYScale though, so I just followed suit.

Comment 4

[Nick Cameron [:nrc]]

Comment on attachment 733158 [details] [diff] [review]
Create a new mask if the offset changes

Review of attachment 733158 [details] [diff] [review]:
-----------------------------------------------------------------

::: layout/base/FrameLayerBuilder.cpp
@@ +682,5 @@
>    // remain unchanged.
>    nsTArray<FrameLayerBuilder::Clip::RoundedRect> mRoundedClipRects;
>    // scale from the masked layer which is applied to the mask
>    float mScaleX, mScaleY;
> +  nsIntPoint mOffset;

please add a comment here

Comment 5

[Matt Woodrow (:mattwoodrow)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/e9113b268416

Comment 6

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/e9113b268416

Comment 7

[Cork]

Retested the testcases in todays nightly http://hg.mozilla.org/mozilla-central/rev/55f9e3e3dae7