1278236 - Convert GetSurfaceForGdkDrawable to return a DrawTarget

Status: RESOLVED FIXED

(Core :: Widget: Gtk, defect)

Description

[Jonathan Watt [:jwatt]]

As part of getting rid of gfxXlibSurface (and ultimately gfxASurface) I have a patch that pushes the DrawTarget wrapping of the gfxXlibSurface created by GetSurfaceForGdkDrawable back to just after it is created (rather than passing it out to callers and having it wrapped later).

Once we get to a point where all gfxXlibSurface surfaces are immediately wrapped it becomes easy to replace them all with a DrawTarget that directly wraps a cairo_xlib_surface_t.

Comment 1

[Jonathan Watt [:jwatt]]

Attachment: patch

Comment 2

[Bas Schouten (:bas.schouten)]

Comment on attachment 8760256 [details] [diff] [review]
patch

Review of attachment 8760256 [details] [diff] [review]:
-----------------------------------------------------------------

::: widget/gtk/nsWindow.cpp
@@ +6462,5 @@
>      return keyBindings->Execute(aEvent, aCallback, aCallbackData);
>  }
>  
>  #if defined(MOZ_X11) && (MOZ_WIDGET_GTK == 2)
>  /* static */ already_AddRefed<gfxASurface>

I don't see how this compiles since the return value here is wrong.

Comment 3

[Jonathan Watt [:jwatt]]

Good catch. I guess that must mean all our linux builders on Try are using gtk3 now.

Comment 4

[Jonathan Watt [:jwatt]]

Attachment: patch

Comment 5

[Jonathan Watt [:jwatt]]

FWIW I need this patch to make progress on tier 1 work, but karlt considers this code tier 2 - we don't bother to build and test it ourselves, but we should try to write code that will work for it:

https://bugzilla.mozilla.org/show_bug.cgi?id=1278282#c1

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/f26550274a45
Convert GetSurfaceForGdkDrawable to return a DrawTarget. r=Bas

Comment 7

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/f26550274a45

Comment 8

[[e-mail]]

/media/data/develop/mozilla/central/widget/gtk/nsDragService.cpp:446:48: error: no matching constructor for initialization of 'IntRect'

Comment 9

[Jonathan Watt [:jwatt]]

Attachment: potential compilation fix

Comment 10

[Jonathan Watt [:jwatt]]

Does the patch is comment 9 fix things?

Comment 11

[[e-mail]]

I can compile it with the patch.

Comment 12

[Bill Gianopoulos [:WG9s]]

(In reply to Jonathan Watt [:jwatt] from comment #9)
> Created attachment 8762689 [details] [diff] [review]
> potential compilation fix

This patch does not fix the compilation issue for me.  I additionally need this as well:

diff --git a/widget/gtk/nsWindow.h b/widget/gtk/nsWindow.h
--- a/widget/gtk/nsWindow.h
+++ b/widget/gtk/nsWindow.h
@@ -316,17 +316,17 @@ public:
    nsresult            UpdateTranslucentWindowAlphaInternal(const nsIntRect& aRect,
                                                             uint8_t* aAlphas, int32_t aStride);
 
     already_AddRefed<mozilla::gfx::DrawTarget> GetDrawTarget(const LayoutDeviceIntRegion& aRegion,
                                                              mozilla::layers::BufferMode* aBufferMode);
 
 #if (MOZ_WIDGET_GTK == 2)
     static already_AddRefed<DrawTarget> GetDrawTargetForGdkDrawable(GdkDrawable* aDrawable,
-                                                                    const IntSize& aSize);
+                                                                    const nsIntSize& aSize);
 #endif
     NS_IMETHOD         ReparentNativeWidget(nsIWidget* aNewParent) override;
 
     virtual nsresult SynthesizeNativeMouseEvent(LayoutDeviceIntPoint aPoint,
                                                 uint32_t aNativeMessage,
                                                 uint32_t aModifierFlags,
                                                 nsIObserver* aObserver) override;