shapes

1 files changed, 49 insertions, 0 deletions

diff --git a/docs/pixman shapes b/docs/pixman shapes
index 3d8c5d06..d9a19277 100644
--- a/docs/pixman shapes
+++ b/docs/pixman shapes
@@ -1,3 +1,52 @@
+In
+
+    http://www.mail-archive.com/pixman@lists.freedesktop.org/msg01344.html
+
+Taekyun Kim suggests:
+
+    pixman_image_composite_xrender(op, src, shape, clip, dst)
+    pixman_image_composite_bounded(op, src, shape, clip, dst)
+    pixman_image_composite_simple(op, src, shape, clip, dst)
+
+to directly support cairo's composite operations. Additional changes
+that would be convenient:
+
+    - Add PIXMAN_TYPE_W images that are like PIXMAN_TYPE_A except
+      the R, G, B channels are considered copies of the alpha channel.
+
+    - All compositing is component_alpha. With the W images, we can
+      emulate unified alpha. The component_alpha property is 
+      ignored.
+
+    - clip regions are given as direct arguments:
+
+        pixman_image_composite_xrender (op, src, shape, clip, region, dst)
+        pixman_image_composite_bounded (op, src, shape, clip, region, dst)
+        pixman_image_composite_simple (op, src, shape, clip, region, dst)
+
+      Or maybe there should be a region image.q
+
+    - Source images (src, shape, clip) that are NULL are considered
+      solid white.
+
+    - Overlapping compositing is explicitly allowed
+
+    - Ignore alpha maps?
+    - Ignore convolution filters?
+
+    - "clone" a pixman image? This is something we need for
+      overlapping support anyway. The clone
+      - has no transformations
+      - has the same alpha maps as the dest
+      - has a copy of the data
+
+Other convenient things: 
+    - When image_create() fails, return a fail image like the one
+      cairo has. We can make it do whatever.
+
+Original question: 
+
+
 There are several separate questions related to this:
 
 1. Can we support cairo's compositing operation better?