Fix missing backing store in Linux.

Fix missing backing store in Linux.

Issue: In the renderer host, the RenderWidgetHostView gets created before the ViewMsg_New gets sent.  In the Linux code, gtk packing of the RenderWidgetHostView will cause a configure-event to happen, which leads to ViewHost_Resize getting sent before the ViewMsg_New gets sent.  This eventually leads to the backing store never getting created.  See the bug report for details.

Fix:
Add a SetChildSize() method to WebContentsView.
Windows and Mac keep their current flow of just calling SetSize() on the child.
Linux defers packing from the RenderWidgetHostView creation to the SetChildSize() call.

BUG=9830
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=13725

[willchan no longer on Chromium, 2009-04-10 18:25:48]

I don't really understand the whole design of the various classes here, so my
fix very likely is the wrong way to do it.  I haven't tested out on mac or
windows either.  I'm just sending this out for opinions.  Please add other
reviewers as necessary, since it wasn't clear from the logs who understands this
code best.

[brettw, 2009-04-10 19:31:16]

http://codereview.chromium.org/67044/diff/1010/11
File chrome/browser/tab_contents/web_contents_view.h (right):

http://codereview.chromium.org/67044/diff/1010/11#newcode125
Line 125: void SetChildSize(RenderWidgetHostView* rwh_view) {
Why can't this be the virtual interface? I don't see why this has to call the
"internal" version.

[willchan no longer on Chromium, 2009-04-10 19:47:43]

http://codereview.chromium.org/67044/diff/1010/11
File chrome/browser/tab_contents/web_contents_view.h (right):

http://codereview.chromium.org/67044/diff/1010/11#newcode125
Line 125: void SetChildSize(RenderWidgetHostView* rwh_view) {
tOn 2009/04/10 19:31:16, brettw wrote:
> Why can't this be the virtual interface? I don't see why this has to call the
> "internal" version.

I'm using Sutter's NVI idiom here (http://www.gotw.ca/publications/mill18.htm). 
It's definitely inconsistent with most of the code I've seen in Chrome, so I'm
happy to undo it, but I figured I'd put it in and see what people thought.  It's
my personal preference since it prevents subclasses from calling the parent
implementation, which I generally don't like.

[willchan no longer on Chromium, 2009-04-14 16:29:29]

ping

[Evan Martin, 2009-04-14 23:30:43]

lgtm

http://codereview.chromium.org/67044/diff/37/1019
File chrome/browser/tab_contents/web_contents_view_win.h (right):

http://codereview.chromium.org/67044/diff/37/1019#newcode104
Line 104: RenderWidgetHostView* rwh_view_;
comment on member variables?

[willchan no longer on Chromium, 2009-04-15 00:10:37]

Woops, that member variable wasn't supposed to be there.  I've deleted it. 
Thanks for catching that.