In accordance with prior releases, we are supposed to default to looking at the openmpi-default-hostfile as a default hostfile. Restore that behavior, but ignore the file if it is empty. Allow the user to ignore any MCA param setting pointing to a default hostfile by setting the param to "none" (via cmd line or whatever) - this allows them to override a setting in the system default MCA param file.

This commit was SVN r25851.

orte/runtime/orte_mca_params.c

@@ -28,6 +28,7 @@
 #include <stdio.h>
 
 #include "opal/mca/base/mca_base_param.h"
+#include "opal/mca/installdirs/installdirs.h"
 #include "opal/util/output.h"
 #include "opal/util/argv.h"
 
@@ -250,9 +251,15 @@ int orte_register_params(void)
                                 false, false, 1000, &orte_timeout_usec_per_proc);
     
     /* default hostfile */
+    asprintf(&orte_default_hostfile, "%s/etc/openmpi-default-hostfile", opal_install_dirs.prefix);
     mca_base_param_reg_string_name("orte", "default_hostfile",
-                                   "Name of the default hostfile (relative or absolute path)",
-                                   false, false, NULL, &orte_default_hostfile);
+                                   "Name of the default hostfile (relative or absolute path, \"none\" to ignore environmental or default MCA param setting)",
+                                   false, false, orte_default_hostfile, &orte_default_hostfile);
+    if (0 == strcmp(orte_default_hostfile, "none")) {
+        free(orte_default_hostfile);
+        orte_default_hostfile = NULL;
+    }
+
     
 #ifdef __WINDOWS__
     mca_base_param_reg_string_name("orte", "ccp_headnode",

orte/util/hostfile/hostfile.c

@@ -556,9 +556,17 @@ int orte_util_filter_hostfile_nodes(opal_list_t *nodes,
     OBJ_CONSTRUCT(&exclude, opal_list_t);
     if (ORTE_SUCCESS != (rc = hostfile_parse(hostfile, &newnodes, &exclude, false))) {
         OBJ_DESTRUCT(&newnodes);
+        OBJ_DESTRUCT(&exclude);
         return rc;
     }
     
+    /* if the hostfile was empty, then treat it as a no-op filter */
+    if (0 == opal_list_get_size(&newnodes)) {
+        OBJ_DESTRUCT(&newnodes);
+        OBJ_DESTRUCT(&exclude);
+        return ORTE_SUCCESS;
+    }
+
     /* remove from the list of newnodes those that are in the exclude list
      * since we could have added duplicate names above due to the */
     while (NULL != (item1 = opal_list_remove_first(&exclude))) {