Fixed intermittent hang on tests with "target teams if(0)" construct with no parallel inside.

Differential Revision: https://reviews.llvm.org/D29597

llvm-svn: 298373
diff --git a/openmp/runtime/src/kmp_runtime.cpp b/openmp/runtime/src/kmp_runtime.cpp
index b62d98e..2210239 100644
--- a/openmp/runtime/src/kmp_runtime.cpp
+++ b/openmp/runtime/src/kmp_runtime.cpp
@@ -5267,18 +5267,22 @@
             // Wait for threads to reach reapable state
             for (f = 1; f < team->t.t_nproc; ++f) {
                 KMP_DEBUG_ASSERT(team->t.t_threads[f]);
-                volatile kmp_uint32 *state = &team->t.t_threads[f]->th.th_reap_state;
+                kmp_info_t *th = team->t.t_threads[f];
+                volatile kmp_uint32 *state = &th->th.th_reap_state;
                 while (*state != KMP_SAFE_TO_REAP) {
 #if KMP_OS_WINDOWS
                     // On Windows a thread can be killed at any time, check this
                     DWORD ecode;
-                    if (__kmp_is_thread_alive(team->t.t_threads[f], &ecode))
-                        KMP_CPU_PAUSE();
-                    else
+                    if (!__kmp_is_thread_alive(th, &ecode)) {
                         *state = KMP_SAFE_TO_REAP; // reset the flag for dead thread
-#else
-                    KMP_CPU_PAUSE();
+                        break;
+                    }
 #endif
+                    // first check if thread is sleeping
+                    kmp_flag_64 fl(&th->th.th_bar[bs_forkjoin_barrier].bb.b_go, th);
+                    if (fl.is_sleeping())
+                        fl.resume(__kmp_gtid_from_thread(th));
+                    KMP_CPU_PAUSE();
                 }
             }
 
diff --git a/openmp/runtime/test/misc_bugs/teams-no-par.c b/openmp/runtime/test/misc_bugs/teams-no-par.c
new file mode 100644
index 0000000..0ef8d9a
--- /dev/null
+++ b/openmp/runtime/test/misc_bugs/teams-no-par.c
@@ -0,0 +1,64 @@
+// RUN: %libomp-compile-and-run
+//
+// The test checks the teams construct pseudocode executed on host
+//
+
+#include <stdio.h>
+#include <omp.h>
+
+#ifndef N_TEAMS
+#define N_TEAMS 4
+#endif
+#ifndef N_THR
+#define N_THR 3
+#endif
+
+static int err = 0;
+
+// Internal library staff to emulate compiler's code generation:
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+  int reserved_1;
+  int flags;
+  int reserved_2;
+  int reserved_3;
+  char *psource;
+} ident_t;
+
+static ident_t dummy_loc = {0, 2, 0, 0, ";dummyFile;dummyFunc;0;0;;"};
+
+int __kmpc_global_thread_num(void*);
+void __kmpc_push_num_teams(ident_t const*, int, int, int);
+void __kmpc_fork_teams(ident_t const*, int argc, void *microtask, ...);
+
+#ifdef __cplusplus
+}
+#endif
+
+// Outlined entry point:
+void foo(int *gtid, int *tid, int *nt)
+{ // start "serial" execution by master threads of each team
+  if ( nt ) {
+    printf(" team %d, param %d\n", omp_get_team_num(), *nt);
+  } else {
+    printf("ERROR: teams before parallel: gtid, tid: %d %d, bad pointer: %p\n", *gtid, *tid, nt);
+    err++;
+    return;
+  }
+}
+
+int main()
+{
+  int nt = 4;
+  int th = __kmpc_global_thread_num(NULL); // registers initial thread
+  __kmpc_push_num_teams(&dummy_loc, th, N_TEAMS, N_THR);
+  __kmpc_fork_teams(&dummy_loc, 1, &foo, &nt); // pass 1 shared parameter "nt"
+  if (err)
+    printf("failed with %d errors\n",err);
+  else
+    printf("passed\n");
+  return err;
+}