I stupidly didn't make sure GfxInfo::Init was called on non-Windows platforms. This patch fixes that.

Comment on attachment 505522 [details] [diff] [review] call Init in the constructor function on Android and Cocoa Test for ogl layers on 10.6 please

This is a mochitest that makes sure that acceleration status is what we think it is on Windows and OS X. It would have prevented this bug!

Comment on attachment 505553 [details] [diff] [review] test acceleration status >+ <title>Test for Bug 627498</title> A better title

http://hg.mozilla.org/mozilla-central/rev/2736d8740c66 is the fix. The test is currently going through Try.

Jeff, this appears to detect if gl is enabled on windowsXP+, and if it isn't then it will fail. I am not sure what the relationship is between direct2D/webgl/canvas- maybe they are the same or have similar traits when it comes to support. I am testing running the windows 7 tests in the cloud. For machines that have a graphics card, this test case (test_acceleration.html) passes, but for the ones that do not have a dedicated graphics card this test fails. We are looking to run the webgl tests (in the 'gl' job) as well as reftests on higher end machines that have full graphics support. Most likely, I could put this test in the 'gl' job- I just don't understand how this related to webgl. Jeff, can you help me understand this more and determine if it makes sense to put this test case in the 'gl' job.

This is testing whether we have layers acceleration and is unrelated to webgl. It should run on a machine that has a GPU, but I'm not sure which is the best job for it.