Patch that gets this somewhat working attached. There's still a lot of bugs and we hit asserts frequently.

A couple of things that I saw: DrawTargetD2D1 and SourceSurfaceD2D1 can both hold references to each other, causing cycles. It also causes us to fail this assertion http://mxr.mozilla.org/mozilla-central/source/gfx/layers/d3d11/TextureD3D11.cpp#248 Nothing appears to ever remove things from DrawTargetD2D1::mDependentTargets, so we hit this assertion: http://mxr.mozilla.org/mozilla-central/source/gfx/2d/DrawTargetD2D1.cpp#681

Comment on attachment 8465219 [details] [diff] [review] allow-d2d11 Review of attachment 8465219 [details] [diff] [review]: ----------------------------------------------------------------- ::: gfx/layers/d3d11/TextureD3D11.cpp @@ +165,4 @@ > MOZ_ASSERT(mDrawTarget->refCount() == 1); > + if (mTexture) { > + LockD3DTexture(mTexture.get()); > + } else { MOZ_ASSERT(mTexture10) etc. ::: gfx/thebes/gfxPrefs.h @@ +157,5 @@ > DECL_GFX_PREF(Live, "gfx.color_management.rendering_intent", CMSRenderingIntent, int32_t, 0); > > DECL_GFX_PREF(Once, "gfx.direct2d.disabled", Direct2DDisabled, bool, false); > DECL_GFX_PREF(Once, "gfx.direct2d.force-enabled", Direct2DForceEnabled, bool, false); > + DECL_GFX_PREF(Once, "gfx.direct2d.use1_1", Direct2DUse1_1, bool, false); We might as well make it live I suppose? ::: gfx/thebes/moz.build @@ +193,5 @@ > 'gfxDWriteTextAnalysis.cpp', > ] > > + if CONFIG['MOZ_ENABLE_DIRECT2D1_1']: > + DEFINES['USE_D2D1_1'] = True It's unclear to me what we need this for? :) I'm guessing in some intermediate stage you had something conditional inside gfxWindowsPlatform?

gfxWindowsPlatform includes 2D.h which checks USE_D2D1_1.

Updated patch without the racy device.

Slightly cleaned up and splitted out Content device additions.

This is basically Matt Woodrow's patch split up and rebased. Carrying my t+.

Last bit of Matt's patch, rebased. Carrying my r+.

Comment on attachment 8469227 [details] [diff] [review] Part 1: Add content device to gfxWindowsPlatform Review of attachment 8469227 [details] [diff] [review]: ----------------------------------------------------------------- ::: gfx/thebes/gfxWindowsPlatform.cpp @@ +1495,5 @@ > + if (FAILED(hr)) { > + return; > + } > + > + hr = d3d11CreateDevice(adapter, D3D_DRIVER_TYPE_UNKNOWN, nullptr, You might want to #ifdef this piece in USE_D3D1_1 to avoid allocating it if it won't ever be used. Not that it matters much.

This adds a bunch of code to assure temporary DTs created are also of the right backend.

The right patch this time.

Comment on attachment 8488165 [details] [diff] [review] Part 3: Use Direct2D 1.1 when preffed on in D3D11 layers v2 Review of attachment 8488165 [details] [diff] [review]: ----------------------------------------------------------------- ::: gfx/layers/moz.build @@ +99,5 @@ > 'd3d11/CompositorD3D11.cpp', > 'd3d11/ReadbackManagerD3D11.cpp', > + ] > + if CONFIG['MOZ_ENABLE_DIRECT2D1_1']: > + DEFINES['USE_D2D1_1'] = True Whitespace!

https://hg.mozilla.org/mozilla-central/rev/418e2846120f https://hg.mozilla.org/mozilla-central/rev/20e97c8496e4 https://hg.mozilla.org/mozilla-central/rev/7b9201731195

Comment on attachment 8488165 [details] [diff] [review] Part 3: Use Direct2D 1.1 when preffed on in D3D11 layers v2 >+ if (gfxPrefs::Direct2DUse1_1() && Factory::SupportsD2D1()) { >+ contentMask |= BackendTypeBit(BackendType::DIRECT2D1_1); >+ defaultBackend = BackendType::DIRECT2D1_1; >+ } else { >+ defaultBackend = BackendType::DIRECT2D; >+ } SupportsD2D1() is #ifdef in 2D.h so the compile fails if you don't have it. Perhaps you could write #ifdef USE_D2D1_1 if (gfxPrefs::Direct2DUse1_1() && Factory::SupportsD2D1()) { contentMask |= BackendTypeBit(BackendType::DIRECT2D1_1); defaultBackend = BackendType::DIRECT2D1_1; } else #endif defaultBackend = BackendType::DIRECT2D;

My bad, I now see bug 1068590 exists for that.