Bug 52766 – [chromium] Composited render surfaces should allow writes to alpha channel.

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Adrienne Walker]

[chromium] Composited render surfaces should allow writes to alpha channel.

[Adrienne Walker]

Created attachment 79518 [details]
Patch

[Adrienne Walker]

See: http://crbug.com/70140

[Adrienne Walker]

Because render surfaces weren't enabling writes to the alpha channel, it caused issues with blending the text in the search drop down here: http://bodybrowser.googlelabs.com

[James Robinson]

Comment on attachment 79518 [details]
Patch

Why doesn't the comment in http://trac.webkit.org/browser/trunk/Source/WebCore/platform/graphics/chromium/LayerRendererChromium.cpp#L260 apply to non-root layers?

[Adrienne Walker]

(In reply to comment #4)
> (From update of attachment 79518 [details])
> Why doesn't the comment in http://trac.webkit.org/browser/trunk/Source/WebCore/platform/graphics/chromium/LayerRendererChromium.cpp#L260 apply to non-root layers?

My hope would be that we are not turning on subpixel antialiasing when we're not rendering into the backbuffer.  Since layers can be repositioned, how would you know which subpixel the layer lines up with?

I don't know where I'd look to verify that this is the case.

[Vangelis Kokkevis]

Comment on attachment 79518 [details]
Patch

LGMT.  Thanks for taking care of it.  Still needs a reviewer to bless it.

[James Robinson]

(In reply to comment #5)
> (In reply to comment #4)
> > (From update of attachment 79518 [details] [details])
> > Why doesn't the comment in http://trac.webkit.org/browser/trunk/Source/WebCore/platform/graphics/chromium/LayerRendererChromium.cpp#L260 apply to non-root layers?
> 
> My hope would be that we are not turning on subpixel antialiasing when we're not rendering into the backbuffer.  Since layers can be repositioned, how would you know which subpixel the layer lines up with?
> 
> I don't know where I'd look to verify that this is the case.

Ah, of course!  The place to check is:

http://trac.webkit.org/browser/trunk/Source/WebCore/platform/graphics/chromium/ContentLayerChromium.cpp#L261

and we do turn subpixel AA off for sublayers (currently for all sublayers, but we could revise this in the future for opaque layers or whatever).

[James Robinson]

Comment on attachment 79518 [details]
Patch

You need to remove the (OOPS!) before landing this - preferably with an explanation of why there's no new test added.

[James Robinson]

(In reply to comment #8)
> (From update of attachment 79518 [details])
> You need to remove the (OOPS!) before landing this - preferably with an explanation of why there's no new test added.

Or even more preferably with a list of tests that cover the change :)

[W. James MacLean]

I tried this patch on a ToT Mac build, and it didn't seem to work for me. Is something also required on the Chromium side?

[W. James MacLean]

(In reply to comment #10)
> I tried this patch on a ToT Mac build, and it didn't seem to work for me. Is something also required on the Chromium side?

False alarm - apparently I did not run the right image on the Mac, it works for me now. Sorry about that.

[Adrienne Walker]

Created attachment 79657 [details]
Patch

[James Robinson]

Comment on attachment 79657 [details]
Patch

Looks good!

[Adrienne Walker]

Committed r76299: <http://trac.webkit.org/changeset/76299>