Suppresses leaks in tracked_objects.

tools/valgrind/memcheck/suppressions.txt

Index: tools/valgrind/memcheck/suppressions.txt
===================================================================
--- tools/valgrind/memcheck/suppressions.txt	(revision 108066)
+++ tools/valgrind/memcheck/suppressions.txt	(working copy)
@@ -5198,6 +5198,14 @@
    fun:_ZN2v88internalL6InvokeEbNS0_6HandleINS0_10JSFunctionEEENS1_INS0_6ObjectEEEiPS5_Pb
    fun:_ZN2v88internal9Execution4CallENS0_6HandleINS0_6ObjectEEES4_iPS4_Pbb
 }
+{
+   bug_102327
+   Memcheck:Leak
+   fun:_Znw*
+   fun:_ZN15tracked_objects10ThreadData10InitializeEv
+   fun:_ZN15tracked_objects10ThreadData30InitializeAndSetTrackingStatusEb
+   fun:_ZN15tracked_objects10ThreadData29ShutdownSingleThreadedCleanupEb

[jar (doing other things), 2011/11/01 03:55:15]

This is an excellent example of a race to supress.

This Shutdown... function is *ONLY* called in unit tests, and hence it is
possible that other threads, lingering from previous tests, are racing to see
status as this unit test runs.  As a result, this is an artifact of the unit
test only, and will *NEVER* take place in Chrome (the browser). 

In addition, the examination of status (raced against the setting of status) is
inconsequential, as the code handles any of the status values, in any order,
over time.  Hence any read is old or new values is handled gracefully.

Bottom line: Totally benign race.

Thanks for adding this.

+}
 
 #-----------------------------------------------------------------------
 # 4. These only occur on our Google workstations