It looks like the Chrome NPAPI plugin installer has been broken since we firs...

It looks like the Chrome NPAPI plugin installer has been broken since we first updated chrome webkit after 1.0 shipped. 

Basically in the 1.0 branch when the plugin was instantiated in its instantiation it would get the mime type along with the list of other arguments. If an object tag was specified with the classid, it would get mapped to the mime type. With the webkit merge the classs id is passed in along with the mime type.

The plugin installer thinks that this is an activex installation on receiving a valid class id and 
and ends up checking if it is a white listed classid, etc.

All this code will be taken out along with the activex shim in the near future. For now we take this code path only if we don't have a valid mime type.

This fixes http://code.google.com/p/chromium/issues/detail?id=8584

Added a plugin test for the argument parsing functionality in the default plugin. I changed the ParseInstantiationArguments function in the
plugin installer to a static function to be able to unit test this.


Bug=8584
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=12733

[jam, 2009-03-27 18:33:00]

I haven't heard about the activex shim going away.  I personally would like it
to, but some people feel we should continue using it for WMP because there's no
WMP plugin in Vista, and people claim that the ActiveX plugin is less buggy than
the WMP one.  Also we need admin access to set the registry entry that makes it
use the new skin.

Either way, since this is a pretty big regression, I think we need a test to
make sure it doesn't regress unnoticed again.

[ananta, 2009-03-27 22:46:04]

On 2009/03/27 18:33:00, John Abd-El-Malek wrote:
> I haven't heard about the activex shim going away.  I personally would like it
> to, but some people feel we should continue using it for WMP because there's
no
> WMP plugin in Vista, and people claim that the ActiveX plugin is less buggy
than
> the WMP one.  Also we need admin access to set the registry entry that makes
it
> use the new skin.
> 
> Either way, since this is a pretty big regression, I think we need a test to
> make sure it doesn't regress unnoticed again.

Added a plugin test which tests the argument parsing functionality in the
default plugin.

Thanks
Ananta

[jam, 2009-03-28 02:22:27]

lgtm

http://codereview.chromium.org/42684/diff/2020/2024
File chrome/chrome.sln (right):

http://codereview.chromium.org/42684/diff/2020/2024#newcode101
Line 101: ndProject
E is gone..

http://codereview.chromium.org/42684/diff/2020/2022
File webkit/default_plugin/plugin_impl_win.h (right):

http://codereview.chromium.org/42684/diff/2020/2022#newcode228
Line 228: static bool ParseInstantiationArguments(NPMIMEType mime_type,
36 lines is huge for a function comment!  can you trim this down a little? 
perhaps just one line for each parameter..