Merge pull request #5458 from rhc54/cmr40/px

Default to internal PMIx if newer than external

config/opal_check_pmi.m4

@@ -375,7 +375,7 @@ AC_DEFUN([OPAL_CHECK_PMIX],[
                                                       #endif
                                                       ], [])],
                                            [AC_MSG_RESULT([found])
-                                            opal_external_pmix_version=1.2.x
+                                            opal_external_pmix_version=1x
                                             opal_external_pmix_version_found=1
                                             opal_external_have_pmix1=1
                                             opal_external_pmix_happy=yes],
@@ -402,6 +402,15 @@ AC_DEFUN([OPAL_CHECK_PMIX],[
            AC_MSG_WARN([header/library files were not found])
            AC_MSG_ERROR([Cannot continue])])
 
+    # Final check - if they didn't point us explicitly at an external version
+    # but we found one anyway, use the internal version if it is higher
+    AS_IF([test "$opal_external_pmix_version" != "internal" && (test -z "$with_pmix" || test "$with_pmix" = "yes")],
+          [AS_IF([test "$opal_external_pmix_version" != "3x"],
+                 [AC_MSG_WARN([discovered external PMIx version is less than internal version 3.x])
+                  AC_MSG_WARN([using internal PMIx])
+                  opal_external_pmix_version=internal
+                  opal_external_pmix_happy=no])])
+
     AC_MSG_CHECKING([PMIx version to be used])
     AS_IF([test "$opal_external_pmix_happy" = "yes"],
           [AC_MSG_RESULT([external($opal_external_pmix_version)])
@@ -415,8 +424,8 @@ AC_DEFUN([OPAL_CHECK_PMIX],[
                        [Whether the external PMIx library is v1])
     AM_CONDITIONAL([OPAL_WANT_PRUN], [test "$opal_prun_happy" = "yes"])
 
-    AS_IF([test "$opal_external_pmix_version" = "1.2.x"],
-          [OPAL_SUMMARY_ADD([[Miscellaneous]],[[PMIx support]], [opal_pmix], [$opal_external_pmix_version: WARNING - DYNAMIC OPS NOT SUPPORTED])],
+    AS_IF([test "$opal_external_pmix_version" = "1x"],
+          [OPAL_SUMMARY_ADD([[Miscellaneous]],[[PMIx support]], [opal_pmix], [1.2.x: WARNING - DYNAMIC OPS NOT SUPPORTED])],
           [OPAL_SUMMARY_ADD([[Miscellaneous]],[[PMIx support]], [opal_pmix], [$opal_external_pmix_version])])
 
     OPAL_VAR_SCOPE_POP

opal/mca/pmix/ext1x/configure.m4

@@ -33,9 +33,7 @@ AC_DEFUN([MCA_opal_pmix_ext1x_CONFIG],[
     AS_IF([test "$opal_external_pmix_happy" = "yes"],
           [ # check for the 1.x version ( >= 1.1.4 ?)
            AC_MSG_CHECKING([if external component is version 1.x])
-           AS_IF([test "$opal_external_pmix_version" = "11" ||
-                  test "$opal_external_pmix_version" = "12" ||
-                  test "$opal_external_pmix_version" = "1x"],
+           AS_IF([test "$opal_external_pmix_version" = "1x"],
                  [AC_MSG_RESULT([yes])
                   AS_IF([test "$opal_event_external_support" != "yes"],
                         [AC_MSG_WARN([EXTERNAL PMIX SUPPORT REQUIRES USE OF EXTERNAL LIBEVENT])

orte/orted/pmix/pmix_server_gen.c

@@ -356,6 +356,17 @@ void pmix_server_notify(int status, orte_process_name_t* sender,
         }
     }
 
+    /* protect against infinite loops by marking that this notification was
+     * passed down to the server by me */
+    if (NULL == cd->info) {
+        cd->info = OBJ_NEW(opal_list_t);
+    }
+    val = OBJ_NEW(opal_value_t);
+    val->key = strdup("orte.notify.donotloop");
+    val->type = OPAL_BOOL;
+    val->data.flag = true;
+    opal_list_append(cd->info, &val->super);
+
     opal_output_verbose(2, orte_pmix_server_globals.output,
                         "%s NOTIFYING PMIX SERVER OF STATUS %d",
                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), ret);
@@ -382,6 +393,14 @@ int pmix_server_notify_event(int code, opal_process_name_t *source,
                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                         ORTE_NAME_PRINT(source), code);
 
+    /* check to see if this is one we sent down */
+    OPAL_LIST_FOREACH(val, info, opal_value_t) {
+        if (0 == strcmp(val->key, "orte.notify.donotloop")) {
+            /* yep - do not process */
+            goto done;
+        }
+    }
+
     /* a local process has generated an event - we need to xcast it
      * to all the daemons so it can be passed down to their local
      * procs */
@@ -448,6 +467,7 @@ int pmix_server_notify_event(int code, opal_process_name_t *source,
     /* maintain accounting */
     OBJ_RELEASE(sig);
 
+  done:
     /* execute the callback */
     if (NULL != cbfunc) {
         cbfunc(ORTE_SUCCESS, cbdata);