We should stop PresShell::GetReferenceRenderingContext from creating a Thebes backed gfxContext.

sorry had to backout this change for test failures like https://tbpl.mozilla.org/php/getParsedLog.php?id=37267409&tree=Mozilla-Inbound

It seems this patch to use a Moz2D reference DrawTarget instead of a gfxASurface reference surface screws up text metrics measurement. Specifically for the on Linux and Windows for the tests: https://mxr.mozilla.org/mozilla-central/source/layout/reftests/text/subpixel-glyphs-x-1a.html?force=1 https://mxr.mozilla.org/mozilla-central/source/layout/reftests/text/475092-pos.html?force=1 And on Linux for the test: https://mxr.mozilla.org/mozilla-central/source/toolkit/content/tests/chrome/test_arrowpanel.xul See: https://tbpl.mozilla.org/?tree=Try&rev=d3c1bc13e890

As an aside, I'd like to understand why some code gets away with calling gfxPlatform::ScreenReferenceDrawTarget() while others calls PresShell::GetReferenceRenderingContext() (the former not taking account of whether we're printing while the latter supposedly does).

I'm not sure where to start on these failures. Matt, any insight to help speed up the debugging? Or is anyone able to take this and finish it while I'm on PTO?

I had a look at this and haven't got very far. I've narrowed it down to this user of the function - http://mxr.mozilla.org/mozilla-central/source/layout/base/nsPresShell.cpp#8393 The failing tests are disabled on OSX. I can reproduce a change in rendering with 475092-pos.html if I remove the failing annotation.

This took *way* too long to track down :( Fixes the rendering changes on OSX, fingers crossed it does everywhere else too. https://tbpl.mozilla.org/?tree=Try&rev=b64b00c3e83a

Comment on attachment 8419210 [details] [diff] [review] Use a DT for the pres shell reference Sweet, than you, Matt!

(In reply to Matt Woodrow (:mattwoodrow) from comment #7) > The failing tests are disabled on OSX. I can reproduce a change in rendering > with 475092-pos.html if I remove the failing annotation. Were you going to remove the failing annotation too?

(In reply to Jonathan Watt [:jwatt] from comment #11) > Were you going to remove the failing annotation too? Nope, they still fail. The PresShell change made them fail in a *different* way, and adding the gfxContext change took them back to failing the normal way.