1015211 - Implement StartRemoteDrawing for GTK3

Status: RESOLVED DUPLICATE of bug 1013552

(Core :: Graphics: Layers, defect)

Description

[Frederic Plourde (:fred23)]

Attachment: Start_remote_drawing_implementation_for_GTK3.patch

We need GTK3 support in StartRemoteDrawing() since I'm making BasicCompositor kick in there days for GTK3 on OMTC linux.

The patch is simple, it basically gets a soft ref (not with cairo_reference, of course) on the current cairo_t when it's passed to OnExposeEvent(...) and use that when StartRemoteDrawing() is called to retrieve the ThebesSurface.

Comment 1

[Frederic Plourde (:fred23)]

Attachment: GTK3_StartRemoteDrawing.patch

Sorry, the previous patch was not the most up-to-date one on my system. Here's the one that works ;-)

Comment 2

[Chris Lord [:cwiiis]]

Comment on attachment 8428745 [details] [diff] [review]
GTK3_StartRemoteDrawing.patch

Review of attachment 8428745 [details] [diff] [review]:
-----------------------------------------------------------------

I think there are better people to review this, but the change is small and I think it looks fine with the second comment addressed (rename mCr to mCairoCtx, or something similar).

::: widget/gtk/nsWindow.h
@@ +332,5 @@
>      bool                mEnabled;
>      // has the native window for this been created yet?
>      bool                mCreated;
> +    // with GTK3 builds, we need to keep track of the associated cairo_t
> +    // for stuff like ::StartRemoteDrawing()

The cairo_t associated with the window I presume?

@@ +333,5 @@
>      // has the native window for this been created yet?
>      bool                mCreated;
> +    // with GTK3 builds, we need to keep track of the associated cairo_t
> +    // for stuff like ::StartRemoteDrawing()
> +    cairo_t*            mCr;

mCr isn't a good name for this, mCairoCtx would make the code a lot more readable.

Comment 3

[Karl Tomlinson (:karlt)]

Comment on attachment 8428745 [details] [diff] [review]
GTK3_StartRemoteDrawing.patch

I'm going to mark this r- sorry, because the cairo_t is only valid for the life of the expose event.  The pointer should at least be nulled before the expose event handler returns, but I think we'll need something after that point anyway, so I prefer the approach in bug 1013552.

Can you include function names and more context in future patches please?
Mercurial can be configured to do this automatically as described at
https://developer.mozilla.org/en-US/docs/Mercurial_FAQ#How_can_I_generate_a_patch_for_somebody_else_to_check-in_for_me.3F

Comment 4

[Frederic Plourde (:fred23)]

(In reply to Karl Tomlinson (needinfo?:karlt) from comment #3)
> Comment on attachment 8428745 [details] [diff] [review]
> GTK3_StartRemoteDrawing.patch
> 
> I'm going to mark this r- sorry, because the cairo_t is only valid for the
> life of the expose event.  The pointer should at least be nulled before the
> expose event handler returns, but I think we'll need something after that
> point anyway, so I prefer the approach in bug 1013552.

ah, cool !  Didn't know someone already got a nice fix for that, thanks !
I'll pull the change from 1013552, which seems to be doing the trick