We need this so that we can use gfxQuartzSurface as the drawing backend on mac.

Comment on attachment 678210 [details] [diff] [review] Make gfxSharedImageSurface take a template parameter for the base surface type I'm not too enthused about the interface here, but the implementation looks fine. Would this alternative approach for the interface work? - create gfxBaseSharedMemorySurface.{h,cpp} with the code you factored out here - cut down gfxSharedImageSurface.h to something like class gfxSharedImageSurface : public gfxBaseSharedMemorySurface<gfxImageSurface> - create gfxSharedQuartzSurface similarly

Comment on attachment 678211 [details] [diff] [review] Use gfxQuartzSurface for OMTC drawing on mac >diff --git a/gfx/layers/ipc/ShadowLayerUtilsMac.cpp b/gfx/layers/ipc/ShadowLayerUtilsMac.cpp >+bool >+ShadowLayerForwarder::PlatformAllocBuffer(const gfxIntSize& aSize, >+ gfxASurface::gfxContentType aContent, >+ uint32_t aCaps, >+ SurfaceDescriptor* aBuffer) >+{ >+ return false; This seems OK for now, but any reason why we wouldn't want to use quartz from the content context? It probably be problematic with low-rights content processes, but shouldn't be an issue with omtc.

(In reply to Chris Jones [:cjones] [:warhammer] from comment #4) > This seems OK for now, but any reason why we wouldn't want to use > quartz from the content context? It probably be problematic with > low-rights content processes, but shouldn't be an issue with omtc. I'm not sure what you mean by this sorry. We should be using quartz for drawing content with my patch?

Your patch with return a gfxImageSurface to content, which have cairo use the cairo-image drawing backend right?

https://hg.mozilla.org/integration/mozilla-inbound/rev/877f3c82d44a https://hg.mozilla.org/integration/mozilla-inbound/rev/8083f8825153

https://hg.mozilla.org/mozilla-central/rev/877f3c82d44a https://hg.mozilla.org/mozilla-central/rev/8083f8825153