982480 - Add support for the alpha option to the canvas context options

Status: RESOLVED FIXED

(Core :: Graphics: Canvas2D, defect)

Description

[Rik Cabanier]

WhatWG proposal is here: http://wiki.whatwg.org/wiki/CanvasOpaque

Summary
The proposed API is to allow developers to request an opaque backing store for 2D Canvas. This mirrors existing functionality in WebGL. From the discussions on WhatWG and webkit-dev, Mozilla has expressed public support, and has provided feedback to guide the proposal into its current form.

Chrome has already implemented this feature and it is being used by some Google products
Chrome bug: https://code.google.com/p/chromium/issues/detail?id=234297

Comment 1

[Rik Cabanier]

Attachment: Added implementation + test caset

try build: https://tbpl.mozilla.org/?tree=Try&rev=58ca1bee505c

Comment 2

[Rik Cabanier]

Attachment: Added implementation + test case

- Fixed logic so size is set before the options.
- changed interface so context is queried for opaque-ness
- made sure that if context is opaque, it is always created (since it should display as black)

Comment 3

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

Comment on attachment 8390088 [details] [diff] [review]
Added implementation + test case

Review of attachment 8390088 [details] [diff] [review]:
-----------------------------------------------------------------

::: content/canvas/src/CanvasRenderingContext2D.cpp
@@ +1065,5 @@
>      // Use software when there is going to be a lot of readback
>      mForceSoftware = attributes.mWillReadFrequently;
>    }
>  
> +  SetIsOpaque(attributes.mAlpha == false);

!attributes.mAlpha

::: content/canvas/test/test_canvas.html
@@ +21582,5 @@
> +<script type="text/javascript">
> +
> +function test_opaque() {
> +  var c = document.getElementById("c688");
> +  var ctx = c.getContext("2d", {alpha: true});

Shouldn't this be "alpha:false"?

Comment 4

[Rik Cabanier]

Comment on attachment 8390088 [details] [diff] [review]
Added implementation + test case

Review of attachment 8390088 [details] [diff] [review]:
-----------------------------------------------------------------

::: content/canvas/test/test_canvas.html
@@ +21582,5 @@
> +<script type="text/javascript">
> +
> +function test_opaque() {
> +  var c = document.getElementById("c688");
> +  var ctx = c.getContext("2d", {alpha: true});

yes!
How did this pass the try bots? I will run them again.

Comment 5

[Rik Cabanier]

Attachment: Added implementation + test case

Fixed test case.
Also fixed getImageData as not all platforms set the alpha channel to 255.

Comment 6

[Rik Cabanier]

try run: https://tbpl.mozilla.org/?tree=Try&rev=843ce14f1be2

Comment 7

[Rik Cabanier]

Attachment: Added implementation + test case

removed obsolete comment

Comment 8

[Ryan VanderMeulen [:RyanVM]]

patching file content/canvas/src/CanvasRenderingContext2D.cpp
bad hunk #3 @@ -3802,16 +3808,41 @@ CanvasRenderingContext2D::GetImageDataAr
 (17 16 41 41)

Comment 9

[Rik Cabanier]

Attachment: Added implementation + test case

Comment 10

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/9a2aab98a8d0

Comment 11

[Ryan VanderMeulen [:RyanVM]]

Backed out for D2D reftest failures.
https://hg.mozilla.org/integration/mozilla-inbound/rev/e182de48f628

https://tbpl.mozilla.org/php/getParsedLog.php?id=36214643&tree=Mozilla-Inbound

Comment 12

[Ryan VanderMeulen [:RyanVM]]

And WinXP mochitest-1 failures.
https://tbpl.mozilla.org/php/getParsedLog.php?id=36214670&tree=Mozilla-Inbound

Comment 13

[Rik Cabanier]

Attachment: Added implementation + test case + fixed issue where mozopaque was broken

Comment 14

[Rik Cabanier]

try run: https://tbpl.mozilla.org/?tree=Try&rev=391307a4c8d3

Comment 15

[Rik Cabanier]

Attachment: Added implementation + test case + fixed issue where mozopaque was broken

successful try run: https://tbpl.mozilla.org/?tree=Try&rev=fc25121a3260
everything is green this time. Sorry about not noticing the failures earlier.

Comment 16

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/e25f1c48cf0c

Comment 17

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/e25f1c48cf0c

Comment 18

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Rik, thanks for implementing this!

Comment 20

[Rik Cabanier]

See http://blogs.adobe.com/webplatform/2014/04/01/new-canvas-features/ for a description of the feature.

Comment 21

[Sylvestre Ledru [:Sylvestre]]

Added in the release notes with the title "Support for the alpha option to the canvas context options"

Comment 22

[Steve Fink [:sfink] [:s:]]

Is the performance of the code following this comment important?:

      // XXX Is there some useful swizzle MMX we can use here?

http://dxr.mozilla.org/mozilla-central/source/content/canvas/src/CanvasRenderingContext2D.cpp#3918

I just happened to notice it when auditing code for an unrelated reason. The answer: yes. I came up with one back in grad school (late in the last millenium), though I've never done anything with it. There are probably better ways now with MMX2 or something, but I just thought I'd check.

Comment 23

[Steve Fink [:sfink] [:s:]]

Actually, I'm not sure my algorithm will apply here. This looks like a simpler problem. Well, with an extra conversion to unpremultiplied alpha.

Comment 24

[Rik Cabanier]

(In reply to Steve Fink [:sfink] from comment #23)
> Actually, I'm not sure my algorithm will apply here. This looks like a
> simpler problem. Well, with an extra conversion to unpremultiplied alpha.

Yes, it seems that there should be a routine somewhere in the codebase that can do this quickly.

Of course, storing premultiplied data is wrong to begin with so this shouldn't be necessary :-)

Comment 25

[Florian Scholz (Open Web Docs)]

Added "alpha" context attribute to the reference:
https://developer.mozilla.org/en-US/docs/Web/API/HTMLCanvasElement.getContext
Release notes:
https://developer.mozilla.org/en-US/Firefox/Releases/30#Interfaces.2FAPIs.2FDOM