[tracing] fix a data race in MemoryDumpManager

base/trace_event/memory_dump_manager_unittest.cc

Index: base/trace_event/memory_dump_manager_unittest.cc
diff --git a/base/trace_event/memory_dump_manager_unittest.cc b/base/trace_event/memory_dump_manager_unittest.cc
index 3112615c79546d040e1ddc5df445c4a90de31b1f..c15748c9ab71e10d13b4850f667cc2f17e98772f 100644
--- a/base/trace_event/memory_dump_manager_unittest.cc
+++ b/base/trace_event/memory_dump_manager_unittest.cc
@@ -191,18 +191,11 @@ TEST_F(MemoryDumpManagerTest, MultipleDumpers) {
   DisableTracing();
 }
 
-// Fails on Linux TSan. http://crbug.com/499983
-#if defined(OS_LINUX)
-#define MAYBE_RespectTaskRunnerAffinity DISABLED_RespectTaskRunnerAffinity
-#else
-#define MAYBE_RespectTaskRunnerAffinity RespectTaskRunnerAffinity
-#endif
-
 // Checks that the MemoryDumpManager respects the thread affinity when a
 // MemoryDumpProvider specifies a task_runner(). The test starts creating 8
 // threads and registering a MemoryDumpProvider on each of them. At each
 // iteration, one thread is removed, to check the live unregistration logic.
-TEST_F(MemoryDumpManagerTest, MAYBE_RespectTaskRunnerAffinity) {
+TEST_F(MemoryDumpManagerTest, RespectTaskRunnerAffinity) {
   const uint32 kNumInitialThreads = 8;
 
   ScopedVector<Thread> threads;