We need to set the clipping region of the HDC passed in to a windowless flash...

We need to set the clipping region of the HDC passed in to a windowless flash plugin instance in NPP_HandleEvent for WM_PAINT. The windowless flash plugin instance queries the clipping region of the DC and on not finding it proceeds
to paint the plugin window rect, which causes unnecessary CPU spikes.

This fixes bug http://code.google.com/p/chromium/issues/detail?id=8835, where the plugin process would consume CPU even when we scrolled down to the static video images.

With this fix the CPU usage for windowless flash plugins is on par with Firefox.

Added a uitest which validates whether the hdc passed in to HandleEvent for WM_PAINT has a valid clipping region.

Bug=8835
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=12651

[jam, 2009-03-26 20:09:39]

I assume this is already done by WebKit, which is why --in-process-plugins case
doesn't need this?

Can a ui test be added to check that a plugin that's not visible has an empty
clipping region?

[ananta, 2009-03-26 22:17:43]

On 2009/03/26 20:09:39, John Abd-El-Malek wrote:
> I assume this is already done by WebKit, which is why --in-process-plugins
case
> doesn't need this?
> 
> Can a ui test be added to check that a plugin that's not visible has an empty
> clipping region?

Added a ui test

[jam, 2009-03-26 23:29:42]

On 2009/03/26 22:17:43, ananta wrote:
> On 2009/03/26 20:09:39, John Abd-El-Malek wrote:
> > I assume this is already done by WebKit, which is why --in-process-plugins
> case
> > doesn't need this?
> > 
> > Can a ui test be added to check that a plugin that's not visible has an
empty
> > clipping region?
> 
> Added a ui test

Can't you just reuse the paint function in ExecuteScriptDeleteTest::HandleEvent?

[ananta, 2009-03-26 23:32:17]

On 2009/03/26 23:29:42, John Abd-El-Malek wrote:
> On 2009/03/26 22:17:43, ananta wrote:
> > On 2009/03/26 20:09:39, John Abd-El-Malek wrote:
> > > I assume this is already done by WebKit, which is why --in-process-plugins
> > case
> > > doesn't need this?
> > > 
> > > Can a ui test be added to check that a plugin that's not visible has an
> empty
> > > clipping region?
> > 
> > Added a ui test
> 
> Can't you just reuse the paint function in
ExecuteScriptDeleteTest::HandleEvent?

I thought about doing that initially. Eventually decided that this was  much
cleaner and easier to debug.

[ananta, 2009-03-27 00:37:29]

On 2009/03/26 23:32:17, ananta wrote:
> On 2009/03/26 23:29:42, John Abd-El-Malek wrote:
> > On 2009/03/26 22:17:43, ananta wrote:
> > > On 2009/03/26 20:09:39, John Abd-El-Malek wrote:
> > > > I assume this is already done by WebKit, which is why
--in-process-plugins
> > > case
> > > > doesn't need this?
> > > > 
> > > > Can a ui test be added to check that a plugin that's not visible has an
> > empty
> > > > clipping region?
> > > 
> > > Added a ui test
> > 
> > Can't you just reuse the paint function in
> ExecuteScriptDeleteTest::HandleEvent?
> 
> I thought about doing that initially. Eventually decided that this was  much
> cleaner and easier to debug.

Based on our discussion added the new test to the plugin_windowless_test.cc
file, which was renamed from plugin_execute_script_delete_test.*

Thanks
Ananta

[jam, 2009-03-27 00:57:49]

lgtm

http://codereview.chromium.org/53106/diff/1014/1016
File webkit/glue/plugins/test/plugin_client.cc (right):

http://codereview.chromium.org/53106/diff/1014/1016#newcode143
Line 143: new_test = new NPAPIClient::WindowlessPluginTest(instance,
can these be put together (i.e. if (strcmp() == 0 || strcmp() == 0))