674003 - Radial gradient in <canvas> on glass (Aero Window Title extension) doesn't have the same opaqueness as a linear gradient

Status: VERIFIED FIXED

(Core :: Graphics, defect)

Description

[Benjamin Smedberg]

Attachment: Screenshot

This is a followup to bug 672646. In that bug, the radial gradient was opaque. This fixed the problems in web content, but when painting onto glass in the Firefox toolbar there is still a bug. The gradient is no longer opaque, but it is more opaque than the linear gradient it matches up with. Screenshot attached.

This is a Windows-only regression from Azure. Disabling azure via pref fixes the issue.

Comment 1

[Benjamin Smedberg]

Attachment: Testcase .xpi

Here is a testcase extension: install it and then run Firefox with -chrome chrome://azuretestcase/content/

Comment 2

[Jeff Muizelaar [:jrmuizel]]

I can't get this test case to work for me. When I launch with -chrome chrome://azuretestcase/content/ I get no windows at all.

Comment 3

[Benjamin Smedberg]

Whoops, it was chrome://azure-testcase/content/

Comment 4

[Jeff Muizelaar [:jrmuizel]]

This should be easy to fix again. We're accidentally squaring the alpha.

Comment 5

[Jeff Muizelaar [:jrmuizel]]

Attachment: Fix premultiplying to only premultiply the color components.

We were accidentally squaring alpha. This patch avoids doing so and fixes the problem. I have a reftest that I'll upload later.

Comment 6

[Benoit Jacob [:bjacob] (mostly away)]

Comment on attachment 548626 [details] [diff] [review]
Fix premultiplying to only premultiply the color components.

R+ on the shader sources, didn't check the generated part.

Comment 7

[Asa Dotzler [:asa]]

Comment on attachment 548626 [details] [diff] [review]
Fix premultiplying to only premultiply the color components.

I don't think we want to take changes that only impact our chrome when we don't even use this in our default chrome.

Comment 8

[Joe Drew (not getting mail)]

This doesn't only affect chrome; all canvases are transparent on the web, and radial gradients are currently wrong whenever the author uses transparency.

Comment 9

[Joe Drew (not getting mail)]

Comment on attachment 548626 [details] [diff] [review]
Fix premultiplying to only premultiply the color components.

Renomming based on comment 8.

Comment 10

[Marco Bonardo [:mak]]

http://hg.mozilla.org/mozilla-central/rev/d146b9c0f92e

Comment 11

[Bas Schouten (:bas.schouten)]

Fwiw, this is incredibly safe and as Joe points out, this does affect all canvases.

Comment 12

[LegNeato]

Comment on attachment 548626 [details] [diff] [review]
Fix premultiplying to only premultiply the color components.

Approved for releases/mozilla-aurora

Comment 13

[Jeff Muizelaar [:jrmuizel]]

http://hg.mozilla.org/releases/mozilla-aurora/rev/62b90a0b9d4d

Comment 14

[Marco Bonardo [:mak]]

reftest
http://hg.mozilla.org/mozilla-central/rev/bae3e43a5173

Comment 15

[Bas Schouten (:bas.schouten)]

I was thinking an interesting (and simpler) reftest here might also be to compare on a white background a gradient with stops rgba(0%, 0%, 0%, 100%) rgba(100%, 100%, 100%, 100%) and one with stops rgba(0%, 0%, 0%, 100%) rgba(0%, 0%, 0%, 0%).

Comment 16

[AndreiD[QA]]

Build ID: Mozilla/5.0 (Windows NT 6.1; rv:7.0) Gecko/20100101 Firefox/7.0
Setting this bug as Verified. The issue reported is fixed on Fx7 Beta. Tested following the steps in the description.