Force garbage collection after running any unit tests that initialize WebViews.

Force garbage collection after running any unit tests that initialize WebViews.

It's quite possible that we just plain shouldn't allow any unit tests to
initialize WebKit for philosophical and theoretical reasons.  But this does
seem to work, so I guess we should just go this route for now.

BUG=22971
TEST=none


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=27239

[darin (slow to review), 2009-09-25 06:34:55]

http://codereview.chromium.org/240001/diff/1/2
File webkit/api/src/ChromiumBridge.cpp (right):

http://codereview.chromium.org/240001/diff/1/2#newcode372
Line 372: webKitClient()->suddenTerminationChanged(enabled);
ugh... we should either universally null check webKitClient() or never null
check it.  i have thus far tried hard to make it so that we would never have to
null check it.

there is a way to force GC.  would that be another option?  just use
WebFrame::collectGarbage().

[jorlow, 2009-09-25 18:25:16]

New patch and a new approach.

[darin (slow to review), 2009-09-25 19:43:33]

LGTM

http://codereview.chromium.org/240001/diff/3001/4002
File chrome/test/render_view_test.cc (right):

http://codereview.chromium.org/240001/diff/3001/4002#newcode114
Line 114: ProcessPendingMessages();
i might be slightly better to call GC after this line.  that way any work
performed by those tasks have a chance to release object references that might
be intertwined with GC'able objects.  i'm thinking about JS timers, etc.

[jorlow, 2009-09-25 19:56:11]

http://codereview.chromium.org/240001/diff/3001/4002
File chrome/test/render_view_test.cc (right):

http://codereview.chromium.org/240001/diff/3001/4002#newcode114
Line 114: ProcessPendingMessages();
On 2009/09/25 19:43:33, darin wrote:
> i might be slightly better to call GC after this line.  that way any work
> performed by those tasks have a chance to release object references that might
> be intertwined with GC'able objects.  i'm thinking about JS timers, etc.

Not possible.  GetMainThread() won't work at this point.

[darin (slow to review), 2009-09-25 20:36:36]

On 2009/09/25 19:56:11, Jeremy Orlow wrote:
> http://codereview.chromium.org/240001/diff/3001/4002
> File chrome/test/render_view_test.cc (right):
> 
> http://codereview.chromium.org/240001/diff/3001/4002#newcode114
> Line 114: ProcessPendingMessages();
> On 2009/09/25 19:43:33, darin wrote:
> > i might be slightly better to call GC after this line.  that way any work
> > performed by those tasks have a chance to release object references that
might
> > be intertwined with GC'able objects.  i'm thinking about JS timers, etc.
> 
> Not possible.  GetMainThread() won't work at this point.

Ah, because of SendCloseMessage.  OK, LGTM