Created attachment 221331 [details] [review]
MetaWindowGroup: Handle painting inside a ClutterClone

Review of attachment 221331 [details] [review]:

I'm curious if we can skip the whole origin calculation in the case that we know we're not in a clone paint.

Math seems correct, but I didn't investigate too much. Mostly cosmetic fixes. Welcome to land as-is.

::: src/compositor/meta-window-group.c
@@ +126,3 @@
+#define MTX_GL_SCALE_X(x,w,v1,v2) ((((((x) / (w)) + 1.0f) / 2.0f) * (v1)) + (v2))
+#define MTX_GL_SCALE_Y(y,w,v1,v2) ((v1) - (((((y) / (w)) + 1.0f) / 2.0f) * (v1)) + (v2))
+

Z is unused.

@@ +151,3 @@
+  meta_screen_get_size (window_group->screen, &width, &height);
+
+

If you don't care about ANSI C, this is probably a lot cleaner with an initializer.

@@ +164,3 @@
+  vertices[3].z = 0;
+
+painting_untransformed (MetaWindowGroup *window_group,

Not ANSI C.

@@ +169,3 @@
+  for (i = 0; i < 4; i++)
+    {
+{

Why the initializing?

Thanks for the review - I decided to not go the C99 route because I'm not
sure if we're doing it anywhere else - and adding it here felt like pulling
on a string (do we need to add AC_PROG_CC_C99 to get the right compiler
options? if we do that, do we need to change autogen.sh so that autoconf-2.60
is required? what is the GNOME minimum version?) 

It would conceptually be fine though, and does make that code block more readable

The initialization of w is there because clutter_matrix_transform_point is in-out.

Attachment 221331 [details] pushed as c27b4b7 - MetaWindowGroup: Handle painting inside a ClutterClone

We use C99-style initializers in meta-window-actor.c, which you OKed a little while ago.