645792 - Implement correct behavior for getContext() failures

Status: RESOLVED FIXED

(Core :: Graphics: CanvasWebGL, defect)

Description

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

We shoud:
 * return null and NOT generate an exception
 * generate a webglcontextcreationerror event

As per the spec:
http://www.khronos.org/registry/webgl/specs/latest/#5.14.3

Quote:

webglcontextcreationerror

    This event occurs in response to a call to getContext() on the HTMLCanvasElement receiving the event, when some error occurs during that call. When such an error occurs the getContext() returns with a null value. Later, at the normal event delivery time, this event is delivered with the details of the failure. If the HTMLCanvasElement is listening for this event, it will be delivered whenever getContext() returns null.

    The statusMessage attribute may contain a platform dependent string about the nature of the failure.

Comment 1

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

I don't remember what the current status is, on whether we should generate an exception or not. Checking that is the first step here. http://www.whatwg.org/specs/web-apps/current-work/multipage/the-canvas-element.html#dom-canvas-getcontext

Comment 2

[Everton da Silva Marques]

Similar issue:

https://bugzilla.mozilla.org/show_bug.cgi?id=844244

Comment 4

[Kelsey Gilbert [:jgilbert]]

This appears to still be the case. I'm amused that the 1.0.1 conformance suite doesn't test this. (Nor 1.0.2, I believe)

Comment 5

[Guillaume Abadie]

I didn't find any test either.
I'm looking at the code in question, it seams to be very old... So the question I'm asking is: should we get away from stuff like:
 nsCOMPtr<nsICanvasRenderingContextInternal> ctx =
    do_CreateInstance(ctxString.get(), &rv);

Should we re-write entirely HTMLCanvasElement::GetContextHelper and HTMLCanvasElement::GetContext?

Comment 6

[Guillaume Abadie]

Attachment: patch revision 1

Tested with a return false in WebGLContext::InitAndValidate and:

> function init() {
>   outputInit();
> 
>   var renderingCanvas = document.getElementById("renderingZone");
> 
>   try
>   {
>     gl = renderingCanvas.getContext("experimental-webgl");
>   }
>   catch (exp)
>   {
>     outputPrint("Test failed");
>     return;
>   }
> 
>   outputPrint("Test passed");
> }

https://tbpl.mozilla.org/?tree=Try&rev=a19e6557d5f5

Comment 7

[Guillaume Abadie]

Attachment: patch revision 2

This patch also fix unit tests using a try/catch in javascript.

Comment 8

[Guillaume Abadie]

https://tbpl.mozilla.org/?tree=Try&rev=bc88e4164604

Comment 9

[Guillaume Abadie]

https://hg.mozilla.org/integration/mozilla-inbound/rev/a58d895d8197

Comment 10

[Ryan VanderMeulen [:RyanVM]]

Backed out for OSX leaks.
https://hg.mozilla.org/integration/mozilla-inbound/rev/bf2568d0fec5

https://tbpl.mozilla.org/php/getParsedLog.php?id=27181714&tree=Mozilla-Inbound
https://tbpl.mozilla.org/php/getParsedLog.php?id=27180694&tree=Mozilla-Inbound

Comment 11

[Ryan VanderMeulen [:RyanVM]]

The leaks weren't from this patch. Re-landed.
https://hg.mozilla.org/integration/mozilla-inbound/rev/13fc79f9c13a

Comment 12

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/13fc79f9c13a