Set disabled style on GPU window and plugin intermediate window so mouse mess...

Set disabled style on GPU window and plugin intermediate window so mouse messages pass through to the browser window.
TEST=trybots
BUG=none


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=36311

[darin (slow to review), 2010-01-12 22:26:57]

http://codereview.chromium.org/549025/diff/1/4
File webkit/glue/webplugin_impl.cc (right):

http://codereview.chromium.org/549025/diff/1/4#newcode266
webkit/glue/webplugin_impl.cc:266: if (first_geometry_update_ ||
what does the callstack look like in the case where we get redundant
updateGeometry calls?  i wonder because it is unclear to me that adding this
filtering is best done at this layer and not in the caller.

i also remember that we would sometimes generate seemingly redundant
updateGeometry calls in some cases, which are actually crucial.

http://codereview.chromium.org/549025/diff/1/4#newcode270
webkit/glue/webplugin_impl.cc:270: move.cutout_rects != last_move_.cutout_rects)
{
what if there was only a change to the cutout_rects?  it seems like an equality
method on WebPluginGeometry would be ideal here.

[apatrick_chromium, 2010-01-12 23:42:32]

It seems to be working now. I found I couldn't disable the intermediate window
in all cases. It prevents the child window from receiving any input.

WebPluginImpl::updateGeometry seems to be on the normal painting code path. It
is called whenever the page is painted. That was happening because the test page
I was using output the events it was receiving via html.

Call stack:
>	chrome.dll!webkit_glue::WebPluginImpl::updateGeometry(const WebKit::WebRect &
window_rect={...}, const WebKit::WebRect & clip_rect={...}, const
WebKit::WebVector<WebKit::WebRect> & cutout_rects={...}, bool is_visible=true) 
Line 253	C++
 	chrome.dll!WebKit::WebPluginContainerImpl::reportGeometry()  Line 222 + 0x54
bytes	C++
 	chrome.dll!WebKit::WebPluginContainerImpl::setFrameRect(const WebCore::IntRect
& frameRect={...})  Line 77 + 0x13 bytes	C++
 	chrome.dll!WebCore::Widget::move(int x=11, int y=47)  Line 143 + 0x50
bytes	C++
 	chrome.dll!WebCore::RenderWidget::paint(WebCore::RenderObject::PaintInfo &
paintInfo={...}, int tx=11, int ty=47)  Line 262	C++

http://codereview.chromium.org/549025/diff/4001/5002
File webkit/glue/webplugin_impl.h (right):

http://codereview.chromium.org/549025/diff/4001/5002#newcode301
webkit/glue/webplugin_impl.h:301: 
already fixed

[cpu_(ooo_6.6-7.5), 2010-01-13 00:50:30]

I don't understand the geometry fu enough to have an opinion.
some comments:

http://codereview.chromium.org/549025/diff/4001/5004
File gpu/gpu_plugin/gpu_plugin.cc (right):

http://codereview.chromium.org/549025/diff/4001/5004#newcode33
gpu/gpu_plugin/gpu_plugin.cc:33: const LPCTSTR kPluginObject =
TEXT("GPUPluginObject");
why use LPCSTR? and TEXT(), we specifically stay away from these macros and we
never build ansi versions of chrome.

http://codereview.chromium.org/549025/diff/4001/5004#newcode54
gpu/gpu_plugin/gpu_plugin.cc:54: WS_EX_NOACTIVATE | GetWindowLong(hwnd_,
GWL_EXSTYLE));
uhh do we still need WS_EX_NOACTIVATE ?  my new understanding of this flag is
that does nothing for focus if you are a child window, just for painting..

[apatrick_chromium, 2010-01-13 01:19:34]

http://codereview.chromium.org/549025/diff/4001/5004
File gpu/gpu_plugin/gpu_plugin.cc (right):

http://codereview.chromium.org/549025/diff/4001/5004#newcode33
gpu/gpu_plugin/gpu_plugin.cc:33: const LPCTSTR kPluginObject =
TEXT("GPUPluginObject");
Done.

http://codereview.chromium.org/549025/diff/4001/5004#newcode54
gpu/gpu_plugin/gpu_plugin.cc:54: WS_EX_NOACTIVATE | GetWindowLong(hwnd_,
GWL_EXSTYLE));
It does seem to work without this flag now, maybe because disabling prevents the
child window from taking focus anyway.

[darin (slow to review), 2010-01-13 06:22:40]

http://codereview.chromium.org/549025/diff/8001/8005
File gpu/gpu_plugin/gpu_plugin.cc (right):

http://codereview.chromium.org/549025/diff/8001/8005#newcode4
gpu/gpu_plugin/gpu_plugin.cc:4: 
you need to include build/build_config.h before using OS_WIN

http://codereview.chromium.org/549025/diff/8001/8002
File webkit/glue/webplugin.h (right):

http://codereview.chromium.org/549025/diff/8001/8002#newcode37
webkit/glue/webplugin.h:37: bool operator==(const WebPluginGeometry& rhs) const;
the google style guide does not allow operator==.  google style would be to
create a method name Equals.

http://codereview.chromium.org/549025/diff/8001/8004
File webkit/glue/webplugin_impl.cc (right):

http://codereview.chromium.org/549025/diff/8001/8004#newcode269
webkit/glue/webplugin_impl.cc:269: // Only UpdateGeometry if either the window
or clip rects hace changed.
hace -> have

[apatrick_chromium, 2010-01-13 23:18:03]

http://codereview.chromium.org/549025/diff/8001/8005
File gpu/gpu_plugin/gpu_plugin.cc (right):

http://codereview.chromium.org/549025/diff/8001/8005#newcode4
gpu/gpu_plugin/gpu_plugin.cc:4: 
On 2010/01/13 06:22:40, darin wrote:
> you need to include build/build_config.h before using OS_WIN

Done.

http://codereview.chromium.org/549025/diff/8001/8002
File webkit/glue/webplugin.h (right):

http://codereview.chromium.org/549025/diff/8001/8002#newcode37
webkit/glue/webplugin.h:37: bool operator==(const WebPluginGeometry& rhs) const;
On 2010/01/13 06:22:40, darin wrote:
> the google style guide does not allow operator==.  google style would be to
> create a method name Equals.

Done.

http://codereview.chromium.org/549025/diff/8001/8004
File webkit/glue/webplugin_impl.cc (right):

http://codereview.chromium.org/549025/diff/8001/8004#newcode269
webkit/glue/webplugin_impl.cc:269: // Only UpdateGeometry if either the window
or clip rects hace changed.
On 2010/01/13 06:22:40, darin wrote:
> hace -> have

Done.

[cpu_(ooo_6.6-7.5), 2010-01-14 00:55:55]

lgtm.

Sorry it took this long.