[webkit-reviews] review requested: [Bug 70634] Mark GraphicsLayers as opaque when possible : [Attachment 117907] Patch

[bugzilla-daemon at webkit.org]

Dana Jansens <danakj at chromium.org> has asked  for review:
Bug 70634: Mark GraphicsLayers as opaque when possible
https://bugs.webkit.org/show_bug.cgi?id=70634

Attachment 117907: Patch
https://bugs.webkit.org/attachment.cgi?id=117907&action=review

------- Additional Comments from Dana Jansens <danakj at chromium.org>
@smfr, thanks for the conversation on Friday. I'm submitting a new patch that
is more along the lines you described
on IRC. This considers only a single RenderObject for a backing in a simple
way. It updates contentsOpaque when bounds
or content changed.

I experimented with just doing this computation during the painting step, but
that made using DRT for testing it not
possible, as nothing showed up opaque there.

I've approached this with a better understanding of what "contents" means. This
patch will check the background and
box decorations for a backing to see if they are opaque. If the background is
not, then it will try the foreground, but
only if it will be painted (ie is part of the "contents"). If the foreground
will be composited then it is not tested
here.

There are 3 sets of layout tests. One to test combinations or box decorations
and backgrounds. One to test these same
things with composited image foregrounds. And one to test these together with
painted image foregrounds.