Fix race condition on utility thread shutdown in unit tests. The problem was that the main thread could be accessing the MessageLoop of the in-process utility thread, but that used to also kill itself in UtilityThreadImpl::ReleaseProcessIfNeeded().

BUG=260754
R=scottmg@chromium.org

Review URL: https://codereview.chromium.org/19437002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@211891 0039d316-1c4b-4281-b951-d872f2087c98

1 files changed, 4 insertions, 6 deletions

diff --git a/content/utility/utility_thread_impl.cc b/content/utility/utility_thread_impl.cc
index af6db9d..16d87ef 100644
--- a/content/utility/utility_thread_impl.cc
+++ b/content/utility/utility_thread_impl.cc
@@ -65,13 +65,11 @@ void UtilityThreadImpl::ReleaseProcessIfNeeded() {
     return;
 
   if (single_process_) {
-    // Just quit the message loop directly so that unit tests don't need to
-    // pump the client message loop again. In normal multi-process mode, need
-    // normal shutdown as the IO thread could still be sending the result IPC.
-    // Also close the IPC channel manually as normally that's done by the child
-    // process exiting, which doesn't happen here.
+    // Close the channel to cause UtilityProcessHostImpl to be deleted. We need
+    // to take a different code path than the multi-process case because that
+    // depends on the child process going away to close the channel, but that
+    // can't happen when we're in single process mode.
     channel()->Close();
-    base::MessageLoop::current()->Quit();
   } else {
     ChildProcess::current()->ReleaseProcess();
   }