Set CRWWebViewContentView's size before resetting scroll proxy.

ios/web/web_state/ui/crw_web_view_content_view.mm

Index: ios/web/web_state/ui/crw_web_view_content_view.mm
diff --git a/ios/web/web_state/ui/crw_web_view_content_view.mm b/ios/web/web_state/ui/crw_web_view_content_view.mm
index 36ff49cbc3c78ce80e4d4072c27ab8bff02b4642..82c32e1342b58c1a2a923943d14e049b4a565ef5 100644
--- a/ios/web/web_state/ui/crw_web_view_content_view.mm
+++ b/ios/web/web_state/ui/crw_web_view_content_view.mm
@@ -79,17 +79,6 @@ const CGFloat kBackgroundRGBComponents[] = {0.75f, 0.74f, 0.76f};
                                            green:kBackgroundRGBComponents[1]
                                             blue:kBackgroundRGBComponents[2]
                                            alpha:1.0];
-    // The frame needs to be set immediately after the web view is added
-    // as a subview. The change in the frame triggers drawing operations and
-    // if not done after it's added as a subview, the web view exhibits
-    // strange behavior where clicks from certain web sites are not triggered.
-    // The actual value of the frame doesn't matter as long as it's not
-    // CGRectZero.  The CRWWebViewContentView's frame will be reset to a correct
-    // value in a subsequent layout pass.
-    // TODO(crbug.com/577793): This is an undocumented and not-well-understood
-    // workaround for this issue.
-    const CGRect kDummyRect = CGRectMake(10, 20, 30, 50);
-    self.frame = kDummyRect;

[stuartmorgan, 2016/02/09 21:27:50]

Would it be worth adding a DCHECK here that the frame isn't zero/invalid, with a
brief comment explaining why, just to be sure we never regress this on accident?

[shreyasv1, 2016/02/09 21:58:29]

The fact that it's non zero is a coincidence from the embedder's side.
When a UIView is instantiated it's bounds is CGRectZero. 
Whenever a WebController is alloc, inited the CRWWebControllerContainerView has
a bounds of CGRectZero.

Given that, adding a DCHECK is the wrong thing here.

[stuartmorgan, 2016/02/09 22:15:19]

If having it be non-zero avoids critical bugs, then it's not a coincidence, it's
something embedders need to do.

[kkhorimoto, 2016/02/09 22:17:11]

I think it'd be better to add a DCHECK in
|-loadRequestForCurrentNavigationItem|, since all that really matters is that it
has a valid size before attempting to load/render anything.

   }
 }