Fix a race condition - ensure we don't call terminate in orterun more than once, even if the timeout fires while we are doing so

This commit was SVN r17766.

orte/runtime/orte_globals.c

@@ -147,8 +147,8 @@ int orte_register_params(void)
     orte_max_timeout = 1000000.0 * value;  /* convert to usec */
 
     mca_base_param_reg_int_name("orte", "timeout_step",
-                                "Time to wait [in usecs/proc] before aborting an ORTE operation (default: 10 usec/proc)",
+                                "Time to wait [in usecs/proc] before aborting an ORTE operation (default: 100 usec/proc)",
-                                false, false, 10, &orte_timeout_usec_per_proc);
+                                false, false, 100, &orte_timeout_usec_per_proc);
     
     /* default hostfile */
     mca_base_param_reg_string_name("default", "hostfile",

orte/tools/orterun/orterun.c

@@ -112,14 +112,15 @@ static orte_std_cntr_t total_num_apps = 0;
 static bool want_prefix_by_default = (bool) ORTE_WANT_ORTERUN_PREFIX_BY_DEFAULT;
 static opal_event_t *orterun_event, *orteds_exit_event;
 static char *ompi_server=NULL;
+static bool terminating=false;
 
 /*
  * Globals
  */
-struct globals_t orterun_globals;
+struct orterun_globals_t orterun_globals;
-bool globals_init = false;
+static bool globals_init = false;
 
-opal_cmd_line_init_t cmd_line_init[] = {
+static opal_cmd_line_init_t cmd_line_init[] = {
     /* Various "obvious" options */
     { NULL, NULL, NULL, 'h', NULL, "help", 0,
       &orterun_globals.help, OPAL_CMD_LINE_TYPE_BOOL,
@@ -183,13 +184,6 @@ opal_cmd_line_init_t cmd_line_init[] = {
       NULL, OPAL_CMD_LINE_TYPE_STRING,
       "Provide a cartography file" },
 
-    /* Don't wait for the process to finish before exiting */
-#if 0
-    { NULL, NULL, NULL, '\0', "nw", "nw", 0,
-      &orterun_globals.no_wait_for_job_completion, OPAL_CMD_LINE_TYPE_BOOL,
-      "Launch the processes and do not wait for their completion (i.e., let orterun complete as soon a successful launch occurs)" },
-#endif
-    
     /* Export environment variables; potentially used multiple times,
        so it does not make sense to set into a variable */
     { NULL, NULL, NULL, 'x', NULL, NULL, 1,
@@ -644,6 +638,12 @@ static void terminated(int trigpipe, short event, void *arg)
     orte_proc_t **procs;
     orte_vpid_t i;
     
+    /* flag that we are here to avoid doing it twice */
+    if (terminating) {
+        return;
+    }
+    terminating = true;
+    
     /* close the trigger pipe so it cannot be called again */
     if (0 <= trigpipe) {
         close(trigpipe);

orte/tools/orterun/orterun.h

@@ -37,13 +37,12 @@ int orterun(int argc, char *argv[]);
 /**
  * Global struct for catching orterun command line options.
  */
-struct globals_t {
+struct orterun_globals_t {
     bool help;
     bool version;
     bool verbose;
     bool quiet;
     bool exit;
-    bool no_wait_for_job_completion;
     bool by_node;
     bool by_slot;
     bool do_not_launch;
@@ -62,19 +61,10 @@ struct globals_t {
 };
 
 /**
- * Struct holding values gleaned from the orterun command line
+ * Struct holding values gleaned from the orterun command line -
+ * needed by debugger init
  */
-ORTE_DECLSPEC extern struct globals_t orterun_globals;
+ORTE_DECLSPEC extern struct orterun_globals_t orterun_globals;
-
-/**
- * Whether orterun_globals has been initialized yet or not
- */
-ORTE_DECLSPEC extern bool globals_init;
-
-/**
- * Struct holding list of allowable command line parameters
- */
-ORTE_DECLSPEC extern opal_cmd_line_init_t cmd_line_init[];
 
 END_C_DECLS