Workaround issue in Firefox 35

[mramato]

In Firefox 35 we need to force the context alpha to true because setting it to false affects all framebuffers. Fixes #2431

Also removed old workaround for a Chrome issue that was fixed in Feb 2014 as well as an unused glsl #ifdef that was left over from an older workaround for Firefox.

On a side note, we should look for all of the places in the code that we have workaround and determine if they are still relevant. We should probably also start writing up issues for new workarounds that we add, similar to how we handle deprecation. This way we revisit them from time to time until they are no longer an issue.

[pjcozzi]

we should look for all of the places in the code that we have workaround and determine if they are still relevant

+1

[pjcozzi]

Sandcastle looks good. Do you get any new test failures? I went from 4 to 20.

[mramato]

Looks like you're right (I could have sworn I ran the tests). Other than a separate hack to avoid activating the workaround during the tests, I'm not sure what we can do. With the fix all of the Sandcastle examples seem to work, so these failures may have to be something we live with until they patch Firefox. I'm up for any ideas if people have them.

[mramato]

Should we also check for ANGLE and only activate it then? This is a FF Angle only issue.

[mramato]

@bagnell pointed out that we can't detect ANGLE before we create the context, so I'm going to just add a check for Windows instead, which I hate doing but we don't have much of a choice.

[mramato]

@shunter any potential issues with my Windows detection?

This should be ready unless we have any ideas on how to avoid the test failures in Firefox (which are now Windows only).

[pjcozzi]

This is all good with me. I expect the failures will go away in FF 36. We can make a note in the issue to remove the workaround (perhaps even add a new label for those issues).

Tested on Windows and Mac.

@shunter please follow-up if we can improve the test.