Disabled the tests temporarily: http://hg.mozilla.org/mozilla-central/rev/72e61f2a524f
Created attachment 413272 [details] [diff] [review] remove bogus NS_ABORT_IF_FALSE, re-enable tests Asserting that Get(Current)DrawableImgFrame() must always return non-null is bogus, and we shouldn't do that.
Comment on attachment 413272 [details] [diff] [review] remove bogus NS_ABORT_IF_FALSE, re-enable tests I don't really have any idea what the implications of this are, but it seems like this is better to have in the tree than not. Still, it would be nice to hear what went wrong here.
I answered this in real life, but for posterity: The whole point of GetDrawableImgFrame and friends is that it can return null in the case that we failed to create a drawable image frame from a paletted frame. Before bug 523528, we would just try to draw a paletted image frame and crash. Now, we notice when we fail to create an imgFrame because new returned null or imgFrame::Init() returned an error. (Init returns an error when we're low on virtual memory - see nsIMemory - or when the image's size is too large.) In those cases, GetDrawableImgFrame returns null, instead of returning a paletted image frame that crashes when we call Draw() on it.
Comment on attachment 413272 [details] [diff] [review] remove bogus NS_ABORT_IF_FALSE, re-enable tests Comment below the removed line is now slightly inaccurate, but w/e. r+