Invoke PluginImplementedObjectDestroyed for ppp_class deallocate

Invoke PluginImplementedObjectDestroyed for ppp_class deallocate

Deallocate on the plugin's GetInstanceObject ScriptableObject gets invoked
twice during plugin destruction, typically resulting in a crash, for proxied
plugins that expose a scriptable object. Once by PPP_ClassProxy::OnMsgDeallocate
and once by PluginVarTracker::DidDeleteInstance. This patch removes the object
from the var tracker when it is destroyed in PPP_Class_Proxy.

It also tweaks a unittest to verify Deallocate is invoked only once during the
test.

BUG=133950
TEST=Run gmail, reload page (ctrl-r) and observe for plugin crashes
TEST=Run ppapi_example out-of-process, reload page, observe for plugn crashes
TEST=ppapi_unittests --gtest_filter="PluginVar*"


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=144531

[Josh Horwich, 2012-06-27 00:38:25]

Please take a look.

Also, please verify that I don't need to wrap the entire OnMsgDeallocate call
with ProxyAutoLock - I freely admit I'm not familiar with it :(

[brettw, 2012-06-27 05:04:52]

That's definitely a bug! I'm actually kind of confused how it worked at all for
Flash. Nice!

LGTM

We should be sure to get this on the branch

[brettw, 2012-06-27 05:22:38]

https://chromiumcodereview.appspot.com/10678007/diff/1/ppapi/proxy/ppp_class_...
File ppapi/proxy/ppp_class_proxy.cc (right):

https://chromiumcodereview.appspot.com/10678007/diff/1/ppapi/proxy/ppp_class_...
ppapi/proxy/ppp_class_proxy.cc:345:
PluginGlobals::Get()->plugin_var_tracker()->PluginImplementedObjectDestroyed(
Actually, I'd feel slightly better if this was moved to before the Deallocate
call. Just in case we get some weird reentrancy in there.

[Josh Horwich, 2012-06-27 17:03:55]

Please take a look

https://chromiumcodereview.appspot.com/10678007/diff/1/ppapi/proxy/ppp_class_...
File ppapi/proxy/ppp_class_proxy.cc (right):

https://chromiumcodereview.appspot.com/10678007/diff/1/ppapi/proxy/ppp_class_...
ppapi/proxy/ppp_class_proxy.cc:345:
PluginGlobals::Get()->plugin_var_tracker()->PluginImplementedObjectDestroyed(
On 2012/06/27 05:22:38, brettw wrote:
> Actually, I'd feel slightly better if this was moved to before the Deallocate
> call. Just in case we get some weird reentrancy in there.

Done.

[commit-bot: I haz the power, 2012-06-27 19:07:43]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/jhorwich@chromium.org/10678007/6001

[commit-bot: I haz the power, 2012-06-27 20:15:28]

Change committed as 144531