Instantiate the WebIDBFactoryImpl lazily, and call setIDBFactory

Instantiate the WebIDBFactoryImpl lazily, and call setIDBFactory
again between tests.  Introduces a ResetTestEnvironment method
to webkit_support, which is intended to be called by DRT's
TestShell::resetTestController method.

ResetTestEnvironment subsumes the Android-specific ReleaseMediaResources
method.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=177629

[darin (slow to review), 2013-01-16 23:10:37]

Here's the corresponding WebKit patch, which I should really upload to bugzilla:

Index: chromium/TestShell.cpp
===================================================================
--- chromium/TestShell.cpp      (revision 139838)
+++ chromium/TestShell.cpp      (working copy)
@@ -332,9 +332,6 @@
 #if ENABLE(NOTIFICATIONS) || ENABLE(LEGACY_NOTIFICATIONS)
     m_notificationPresenter->reset();
 #endif
-#if OS(ANDROID)
-    webkit_support::ReleaseMediaResources();
-#endif
     m_drtDevToolsAgent->reset();
     if (m_drtDevToolsClient)
         m_drtDevToolsClient->reset();
@@ -344,6 +341,8 @@
     webView()->mainFrame()->clearOpener();
     WebTestingSupport::resetInternalsObject(webView()->mainFrame());
     WebCache::clear();
+
+    webkit_support::ResetTestEnvironment();
 }

 void TestShell::loadURL(const WebURL& url)

[jsbell, 2013-01-16 23:12:17]

+dgrogan, alecflett

[jsbell, 2013-01-16 23:19:28]

lgtm

https://codereview.chromium.org/11946027/diff/1/webkit/support/test_webkit_pl...
File webkit/support/test_webkit_platform_support.cc (right):

https://codereview.chromium.org/11946027/diff/1/webkit/support/test_webkit_pl...
webkit/support/test_webkit_platform_support.cc:399:
factory_.reset(WebKit::WebIDBFactory::create());
I don't know if this is plausible, but it would be nice to capture the creation
thread and DCHECK() that later calls to EnsureFactory() are occurring on the
same thread. Possibly overkill for test code, but would e.g. make mysterious DRT
failures less mysterious.

[darin (slow to review), 2013-01-17 00:01:54]

On 2013/01/16 23:19:28, jsbell wrote:
> lgtm
> 
>
https://codereview.chromium.org/11946027/diff/1/webkit/support/test_webkit_pl...
> File webkit/support/test_webkit_platform_support.cc (right):
> 
>
https://codereview.chromium.org/11946027/diff/1/webkit/support/test_webkit_pl...
> webkit/support/test_webkit_platform_support.cc:399:
> factory_.reset(WebKit::WebIDBFactory::create());
> I don't know if this is plausible, but it would be nice to capture the
creation
> thread and DCHECK() that later calls to EnsureFactory() are occurring on the
> same thread. Possibly overkill for test code, but would e.g. make mysterious
DRT
> failures less mysterious.

I can do that.

I also just realized that we are leaking the old TestWebIDBFactory instance.  We
could delete that in the Reset() method, but I'm just not sure if that will be
possible to do without tripping over the thread assertions in WebKit :-/

[jsbell, 2013-01-17 00:05:26]

I'd just put a TODO in about the leak, referencing a bug with more context.

[darin (slow to review), 2013-01-17 00:35:54]

Yeah... the existing code has the same leak.


On Wed, Jan 16, 2013 at 4:05 PM, <jsbell@chromium.org> wrote:

> I'd just put a TODO in about the leak, referencing a bug with more context.
>
>
https://codereview.chromium.**org/11946027/<https://codereview.chromium.org/1...
>

[pilgrim_google, 2013-01-17 02:03:52]

I think I understand this, kind of. If this lets us land the final IDB patch
upstream, then I'm happy.

[darin (slow to review), 2013-01-17 16:24:29]

@jsbell - PTAL

This is step 1 of landing the patch.  Note, the setIDBFactory call is disabled
and ReleaseMediaResources is not yet deleted.

[darin (slow to review), 2013-01-17 16:31:50]

WebKit side is here:
https://bugs.webkit.org/show_bug.cgi?id=107132

[jsbell, 2013-01-17 18:07:19]

lgtm

[commit-bot: I haz the power, 2013-01-18 06:39:41]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/darin@chromium.org/11946027/10004

[commit-bot: I haz the power, 2013-01-18 10:03:29]

Change committed as 177629