[blink] Support (semi-)transparent background colors in WebView/Frame.

third_party/WebKit/Source/web/WebLocalFrameImpl.cpp

Index: third_party/WebKit/Source/web/WebLocalFrameImpl.cpp
diff --git a/third_party/WebKit/Source/web/WebLocalFrameImpl.cpp b/third_party/WebKit/Source/web/WebLocalFrameImpl.cpp
index ebda68aa738d402ad920e1bc0c4419b893699343..1a389bb88c343a6e50b7df67167f62fee4040a24 100644
--- a/third_party/WebKit/Source/web/WebLocalFrameImpl.cpp
+++ b/third_party/WebKit/Source/web/WebLocalFrameImpl.cpp
@@ -166,6 +166,7 @@
 #include "platform/WebFrameScheduler.h"
 #include "platform/clipboard/ClipboardUtilities.h"
 #include "platform/fonts/FontCache.h"
+#include "platform/graphics/Color.h"
 #include "platform/graphics/GraphicsContext.h"
 #include "platform/graphics/GraphicsLayerClient.h"
 #include "platform/graphics/paint/ClipRecorder.h"
@@ -1715,12 +1716,11 @@ void WebLocalFrameImpl::createFrameView() {
   IntSize initialSize = (isMainFrame || !frameWidget())
                             ? webView->mainFrameSize()
                             : (IntSize)frameWidget()->size();
-  bool isTransparent = !isMainFrame && parent()->isWebRemoteFrame()
-                           ? true
-                           : webView->isTransparent();
+  Color baseBackgroundColor = webView->baseBackgroundColor();
+  if (!isMainFrame && parent()->isWebRemoteFrame())
+    baseBackgroundColor = Color::transparent;

[chrishtr, 2017/03/28 19:45:56]

I think the old code just used a conditional here to detect alpha and
pass it along, but still respected the non-alpha parts of
webView->baseBackgroundColor(). Also, tested?

[Eric Seckler, 2017/03/29 10:50:18]

My understanding is that the old code was setting FrameView::isTransparent to
true, which resulted in !paintsBackgroundColor in ViewPainter, thus no
background color would have been painted for the FrameView. Setting a completely
transparent baseBackgroundColor should have the same effect now.

Based on the fact that I forgot to actually use the local variable in createView
below, I think it's a safe bet that this isn't tested :P

lfg@, do you have a suggestion of how to test this, since you added this
condition in https://codereview.chromium.org/1565893004 ? :)

[lfg, 2017/03/29 15:19:43]

Yes, that's my understanding as well.
Yes, ideally, this should be a pixel test where we can check that the background
is being drawn properly. At the time I added that, there was no way to do pixel
tests on out-of-process iframes (which is what we are checking here -- remote
parent), I know that lukasza@ (OOO) did a lot of work around this area (see bug
587175), but I'm not sure if it's fully functional yet.

Maybe alternatively, we could write a unit test in WebFrameTest that just
creates a LocalFrame with a RemoteFrame parent, and asserts that the background
color is transparent. It's not as good as a layouttest, but at least gives us
some test coverage (you can file a bug and assign it to me, no need to do it in
this CL).

[Eric Seckler, 2017/03/30 10:46:38]

Added a unit test and filed https://crbug.com/706782 for the layout test.

 
-  frame()->createView(initialSize, webView->baseBackgroundColor(),
-                      isTransparent);
+  frame()->createView(initialSize, webView->baseBackgroundColor());
   if (isMainFrame) {
     frame()->view()->setInitialViewportSize(
         webView->pageScaleConstraintsSet().initialViewportSize());