[PATCH] [Tsan] Fix the signal_recursive.cc test to pass on systems with high loads

[Viktor Kutuzov]

================
Comment at: test/tsan/signal_recursive.cc:45
@@ -42,3 +44,3 @@
   do {
     g_busy_thread_received_restart = false;
     if (sigsuspend(&g_suspend_handler_mask) != -1 || errno != EINTR)
----------------
dvyukov wrote:
> Please remove this line as well. It not only pointless, but also wrong. If the signal comes before this line, we lose the signal and deadlock.
> Maybe it will fix the test w/o sleep...
It didn't worked. In all cases when the signal comes before sigsuspend(), it deadlocks. Sigh... :-)

http://reviews.llvm.org/D6504