[mac] Implement {frame: 'none'} app window on Mac.

[mac] Implement {frame: 'none'} app window on Mac.

Support for draggable regions hasn't landed yet, so we hard-code 20px of
draggable space at the top of the window.

BUG=130182
R=rsesek@chromium.org


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=150062

[jeremya, 2012-07-25 01:11:52]

rsesek- browser/ui/cocoa
jam- content/. Perhaps adding OSX-specific stuff here is bad? I'm not sure how
else to do it.

[jam, 2012-07-25 01:27:15]

http://codereview.chromium.org/10824004/diff/1/content/public/browser/web_con...
File content/public/browser/web_contents_view.h (right):

http://codereview.chromium.org/10824004/diff/1/content/public/browser/web_con...
content/public/browser/web_contents_view.h:122: virtual void
SetMouseDownCanMoveWindow(bool can_move) = 0;
the event tracking functions right above are only there for mac, and are empty
for other platforms. probably best to copy that pattern

[jeremya, 2012-07-25 03:15:59]

http://codereview.chromium.org/10824004/diff/1/content/public/browser/web_con...
File content/public/browser/web_contents_view.h (right):

http://codereview.chromium.org/10824004/diff/1/content/public/browser/web_con...
content/public/browser/web_contents_view.h:122: virtual void
SetMouseDownCanMoveWindow(bool can_move) = 0;
On 2012/07/25 01:27:15, John Abd-El-Malek wrote:
> the event tracking functions right above are only there for mac, and are empty
> for other platforms. probably best to copy that pattern

Hm, it seems misleading to have a method exist on platforms where it makes no
sense.

Since this is really just a hack to get the WebContentsView to respond in a
particular way to the -[NSView mouseDownCanMoveWindow] message, perhaps I should
just cast the NSView down to (id) and send it the message directly instead of
defining a new method on WebContentsView.

[jam, 2012-07-25 16:20:06]

+Avi for his thoughts, I don't understand the WCV workaround you mentioned.

[Avi (use Gerrit), 2012-07-25 16:48:34]

I'm a little lost here. You want to make the whole window a webview, yet make it
draggable anywhere? Won't that completely interfere with the functioning of the
webpage?

[jeremya, 2012-07-25 23:48:24]

On 2012/07/25 16:48:34, Avi wrote:
> I'm a little lost here. You want to make the whole window a webview, yet make
it
> draggable anywhere? Won't that completely interfere with the functioning of
the
> webpage?

Some context: this is for new-style packaged apps, which show wall-to-wall web
content and have complete control over their window frame. The drive-by web
doesn't have access to this sort of window.

Apps will be able to specify regions that are non-draggable within the window
(coming up in a CL near you this fall). The approach I'm planning to use to
implement that is to have the WebContentsView draggable everywhere by default,
and to put on top of it views which respond negatively to [NSView
mouseDownCanMoveWindow] (the ControlRegionView class in shell_window_cocoa.mm).

Another detail is that even when the WCV responds YES to [NSView
mouseDownCanMoveWindow], it still receives mouse move, mouse down and mouse up
messages (though I think it doesn't get mouse move messages while the mouse is
down, since the window is being dragged at that point).

The alternative implementation I talked about in my last message is this, which
requires no content interface changes (but is a bit hacky):
https://chromiumcodereview.appspot.com/10834016

[Avi (use Gerrit), 2012-07-26 02:58:54]

Can't you overlay views that respond YES to mouseDownCanMoveWindow? I'm not
thrilled with either of the two current approaches :(

[jeremya, 2012-07-26 04:37:39]

On 2012/07/26 02:58:54, Avi wrote:
> Can't you overlay views that respond YES to mouseDownCanMoveWindow? I'm not
> thrilled with either of the two current approaches :(

Unfortunately not, the docs for mouseDownCanMoveWindow say:

> Returns YES if the receiver does not need to handle a mouse down and can pass
it through to superviews; NO if it needs to handle the mouse down.

I tried that and it didn't work :(

[Avi (use Gerrit), 2012-07-26 14:32:03]

This approach throws a platform method into the public interface, and the other
approach relies on the client magically knowing about a
setMouseDownCanMoveWindow method, neither of which I'm fond of.

RenderWidgetHostViewMacDelegate seems to me to be the best home for this.

[jeremya, 2012-07-31 05:12:03]

On 2012/07/26 14:32:03, Avi wrote:
> This approach throws a platform method into the public interface, and the
other
> approach relies on the client magically knowing about a
> setMouseDownCanMoveWindow method, neither of which I'm fond of.
> 
> RenderWidgetHostViewMacDelegate seems to me to be the best home for this.

Hm, it's the WebContentsViewMac that needs to be customised, not the RWHV. Did
you mean WebContentsViewDelegate? That has a bunch of platform-specific methods
already.

[jeremya, 2012-07-31 05:51:12]

On 2012/07/31 05:12:03, jeremya wrote:
> On 2012/07/26 14:32:03, Avi wrote:
> > This approach throws a platform method into the public interface, and the
> other
> > approach relies on the client magically knowing about a
> > setMouseDownCanMoveWindow method, neither of which I'm fond of.
> > 
> > RenderWidgetHostViewMacDelegate seems to me to be the best home for this.
> 
> Hm, it's the WebContentsViewMac that needs to be customised, not the RWHV. Did
> you mean WebContentsViewDelegate? That has a bunch of platform-specific
methods
> already.

Well, it looks like WCVDelegate isn't really the right place either.

Perhaps it belongs in ui/base/cocoa/base_view.h with a tri-state
"YES/NO/DEFAULT"?

[Avi (use Gerrit), 2012-07-31 15:03:04]

Sorry about the suggestion for RenderWidgetHostViewMacDelegate; I had in my head
the idea that it was the RWHV.

I don't know. None of the options seem very palatable.
https://chromiumcodereview.appspot.com/10834016 seems the cleanest so far.

[jeremya, 2012-08-03 04:06:48]

I'll go with that (with added DCHECKs) and add a comment that it's a hack
for now, then.


On Wed, Aug 1, 2012 at 1:03 AM, <avi@chromium.org> wrote:

> Sorry about the suggestion for RenderWidgetHostViewMacDelegat**e; I had
> in my head
> the idea that it was the RWHV.
>
> I don't know. None of the options seem very palatable.
>
https://chromiumcodereview.**appspot.com/10834016<https://chromiumcodereview....
the cleanest so far.
>
>
http://codereview.chromium.**org/10824004/<http://codereview.chromium.org/108...
>

[jeremya, 2012-08-03 04:48:46]

avi- I've removed the changes to the WebContentsView interface.

rsesek- could you take a look at the shell_window_cocoa parts?

Thanks :)

[Avi (use Gerrit), 2012-08-03 05:19:49]

lgtm

https://chromiumcodereview.appspot.com/10824004/diff/9002/chrome/browser/ui/c...
File chrome/browser/ui/cocoa/extensions/shell_window_cocoa.mm (right):

https://chromiumcodereview.appspot.com/10824004/diff/9002/chrome/browser/ui/c...
chrome/browser/ui/cocoa/extensions/shell_window_cocoa.mm:167: //
WebContentsViewCocoa view.
Grrr. :(

[jeremya, 2012-08-05 22:52:42]

https://chromiumcodereview.appspot.com/10824004/diff/9002/chrome/browser/ui/c...
File chrome/browser/ui/cocoa/extensions/shell_window_cocoa.mm (right):

https://chromiumcodereview.appspot.com/10824004/diff/9002/chrome/browser/ui/c...
chrome/browser/ui/cocoa/extensions/shell_window_cocoa.mm:167: //
WebContentsViewCocoa view.
On 2012/08/03 05:19:49, Avi wrote:
> Grrr. :(

I know :( :(

I'm thinking that maybe the setMouseDownCanMoveWindow: method belongs on
BaseView, which is visible here. That seems like the cleanest way. I'll prepare
a CL to do that.

[commit-bot: I haz the power, 2012-08-06 04:02:17]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/jeremya@chromium.org/10824004/9002

[commit-bot: I haz the power, 2012-08-06 04:57:35]

Change committed as 150062

[sail, 2012-08-07 17:40:11]

Drive by comment.

https://chromiumcodereview.appspot.com/10824004/diff/9002/chrome/browser/ui/c...
File chrome/browser/ui/cocoa/extensions/shell_window_cocoa.mm (right):

https://chromiumcodereview.appspot.com/10824004/diff/9002/chrome/browser/ui/c...
chrome/browser/ui/cocoa/extensions/shell_window_cocoa.mm:145: NSView*
controlRegion = [[ControlRegionView alloc] init];
this should be scoped_nsobject