SafeStack: Fix flaky test (PR39001)

[vlad.tsyrklevich]

Summary

pthread_join() can return before a thread finishes exit()ing in the
kernel and a subsequent tgkill() can report the thread still alive.
Update the pthread-cleanup.c test to sleep and retry if it hits this
possible flake.

Thanks to Jeremy Morse for reporting.

[vitalybuka]

test/safestack/pthread-cleanup.c

38–52

Why just 3?

[jmorse]

LGTM -- obviously this isn't an ideal situation, but the pthreads/OS-threads abstraction isn't supposed to be un-peeled anyway.

This doesn't trigger the fault in my test setup after ~30 mins, alas I don't have time for a longer soak test.

[vlad.tsyrklevich]

test/safestack/pthread-cleanup.c

38–52

This test seems to only be flaky under high load when one thread runs immediately after the other. 3 seconds of wall time seemed like a reasonably high enough amount of time that the first thread would have been able to finish exiting in the kernel (without making this test very laggy under circumstances where it would actually fail for some reason.)

test/safestack/pthread-cleanup.c

int main(int argc, char **argv)
   pthread_t t1, t2;
   char *t1_buffer = NULL;
 
   if (pthread_create(&t1, NULL, start, NULL))
     abort();
   if (pthread_join(t1, &t1_buffer))
     abort();
 
+  // Stack has not yet been deallocated
   memset(t1_buffer, 0, kBufferSize);
 
   if (arg == 0)
     return 0;
 
+  for (int i = 0; i < 3; i++) {
     if (pthread_create(&t2, NULL, start, NULL))
       abort();
     // Second thread destructor cleans up the first thread's stack.
     if (pthread_join(t2, NULL))
       abort();
 
-  // should segfault here
+    // Should segfault here
     memset(t1_buffer, 0, kBufferSize);
+
+    // PR39001: Re-try in the rare case that pthread_join() returns before the
+    // thread finishes exiting in the kernel--hence the tgkill() check for t1
+    // returns that it's alive despite pthread_join() returning.
+    sleep(1);
+  }
   return 0;
 }