Making ExtensionService less dependent on Browser.

Making ExtensionService less dependent on Browser.

As part of the effort to get unit-tests linking on Android, making
ExtensionService depend less on Browser. More specifically abstracting the
use of the GlobalError class through a new ExtensionErrorUI class so that
ExtensionService does not need direct access to Browser.

BUG=None
TEST=Unit-tests should still pass.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=146870

[Yoyo Zhou, 2012-07-13 20:08:47]

https://chromiumcodereview.appspot.com/10696210/diff/2003/chrome/browser/exte...
File chrome/browser/extensions/extension_error_ui_default.cc (right):

https://chromiumcodereview.appspot.com/10696210/diff/2003/chrome/browser/exte...
chrome/browser/extensions/extension_error_ui_default.cc:30:
global_error_->ShowBubbleView(browser);
It certainly seems like browser_ should be able to be set here...

Looking at the GlobalErrorBubble implementations, the various Browser* arguments
to callback functions are always the same Browser that was passed in to
ShowBubbleView. This makes sense and simultaneously means that those arguments
are probably unnecessary. I would check with Sailesh (sail), but I think that's
worth trying first.

It would simplify (but probably not eliminate the need for) this CL.

[Jay Civelli, 2012-07-16 17:54:29]

https://chromiumcodereview.appspot.com/10696210/diff/2003/chrome/browser/exte...
File chrome/browser/extensions/extension_error_ui_default.cc (right):

https://chromiumcodereview.appspot.com/10696210/diff/2003/chrome/browser/exte...
chrome/browser/extensions/extension_error_ui_default.cc:30:
global_error_->ShowBubbleView(browser);
On 2012/07/13 20:08:47, Yoyo Zhou wrote:
> It certainly seems like browser_ should be able to be set here...
> 
> Looking at the GlobalErrorBubble implementations, the various Browser*
arguments
> to callback functions are always the same Browser that was passed in to
> ShowBubbleView. This makes sense and simultaneously means that those arguments
> are probably unnecessary. I would check with Sailesh (sail), but I think
that's
> worth trying first.
> 
> It would simplify (but probably not eliminate the need for) this CL.

As we discussed with IM, I kept the GlobalError class as is but changed browser_
to be set here.

[Yoyo Zhou, 2012-07-16 18:50:29]

LGTM, thanks for doing this.