We currently fail a number of times in the 'framebuffer-object-attachment' test. These fails appear to be caused by readPixels not properly generating an INVALID_FRAMEBUFFER_OPERATION when needed. It looks like this is because the optimization to skip reading when the width or height is 0 kicks in before we test for all error conditions.
Created attachment 570022 [details] [diff] [review] Fix WebGL conf. errors in 'framebuffer-object-attachment' Test should now pass cleanly on all platforms.
Created attachment 570032 [details] [diff] [review] Fix WebGL conf. errors in 'framebuffer-object-attachment' Unbitrot.
Removed [inbound] from whiteboard as per https://wiki.mozilla.org/Tree_Rules/Inbound