Move npapi_layout_test_plugin target to webkit_support to use it from...

Move npapi_layout_test_plugin target to webkit_support to use it from
WebKit DumpRenderTree.

The following code change affects only WebKit DumpRenderTree.

* webkit_support.cc
 - WebPluginImplWithPageDelegate passes a plugin path and MIME type to WebPluginImpl.
 - CreateWebPlugin checks a MIME type.
   This is the same logic as test_shell.

* test_webplugin_page_delegate.h
  Creates and returns a WebPluginDelegateImpl instance.

BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=49372

[tkent, 2010-06-10 01:09:02]

Note: I'd like to use the npapi_layout_test_plugin for now instead of
TestNetscapePlugIn i n WebKitTree because os some build issues of
TestNetscapePlugIn.

[tony, 2010-06-10 01:11:19]

On 2010/06/10 01:09:02, Kent Tamura wrote:
> Note: I'd like to use the npapi_layout_test_plugin for now instead of
> TestNetscapePlugIn i n WebKitTree because os some build issues of
> TestNetscapePlugIn.

What type of build issues?  It should work on mac at least, we could start with
that.  I think the sooner we switch to the upstream plugin, the better.

[tkent, 2010-06-10 01:31:31]

On 2010/06/10 01:11:19, tony wrote:
> What type of build issues?  It should work on mac at least, we could start
with
> that.  I think the sooner we switch to the upstream plugin, the better.

I don't remember all of them precisely. One of them was an NPEvent issue.
NPEvent is used in TestNetscapePlugIn and not used in npapi_layout_test_plugin,
and NPEvent is not defined in our Mac build.

I also found a behavior difference. npapi_layout_test_plugin has
NP_GetMIMEDescription() and TestNetscapePlugIn doesn't.

Of course, I'll resolve the build issues and try to upstream the behavior change
later.

[tony, 2010-06-10 02:01:41]

Isn't it better to just focus our efforts on using the upstream plugin?  I would
rather just mark the plugin tests as failing for DRT in test_expectations.txt
and focus on the long term solution.

[tkent, 2010-06-10 03:34:59]

On 2010/06/10 02:01:41, tony wrote:
> Isn't it better to just focus our efforts on using the upstream plugin?  I
would
> rather just mark the plugin tests as failing for DRT in test_expectations.txt
> and focus on the long term solution.

I think using npapi_layout_test_plugin is better:
 - It decreases the number of failures now
  It's very important for switching test_shell to DRT.
 - Easy to find remaining issues with plugin tests
 - When we merge npapi_layout_test_plugin into TestNetscapePlugIn in the future,
we can find regressions caused by the merger easily.

[tony, 2010-06-10 04:15:45]

You convinced me with the "finding regressions when changing to the upstream
plugin" argument.  LGTM.

http://codereview.chromium.org/2729004/diff/4001/1011
File webkit/support/webkit_support.cc (right):

http://codereview.chromium.org/2729004/diff/4001/1011#newcode151
webkit/support/webkit_support.cc:151: const bool allow_wildcard = true;
Nit: variable should be named kAllowWildcard since it's const.

http://codereview.chromium.org/2729004/diff/4001/1013
File webkit/support/webkit_support.gypi (right):

http://codereview.chromium.org/2729004/diff/4001/1013#newcode19
webkit/support/webkit_support.gypi:19: 'npapi_layout_test_plugin',
Do we really need this?  Shouldn't the final target depend on this (e.g,
test_shell and DumpRenderTree)?  In other words, every target that depends on
webkit_support does not need npapi_layout_test_plugin.

[tkent, 2010-06-10 04:37:43]

Thanks.  I'll commit after confirming try server results.

http://codereview.chromium.org/2729004/diff/4001/1011
File webkit/support/webkit_support.cc (right):

http://codereview.chromium.org/2729004/diff/4001/1011#newcode151
webkit/support/webkit_support.cc:151: const bool allow_wildcard = true;
On 2010/06/10 04:15:46, tony wrote:
> Nit: variable should be named kAllowWildcard since it's const.

Done.

http://codereview.chromium.org/2729004/diff/4001/1013
File webkit/support/webkit_support.gypi (right):

http://codereview.chromium.org/2729004/diff/4001/1013#newcode19
webkit/support/webkit_support.gypi:19: 'npapi_layout_test_plugin',
Oh, it isn't needed.  Removed.