master: mpirun does not emit error message with non-existent executable

[jsquyres]

On master HEAD, mpirun (i.e., prun), does not seem to emit an error message if there is no executable found:

$ mpirun -n 1 uptime
 15:03:51 up 200 days, 10:36, 80 users,  load average: 0.59, 1.66, 1.29
$ mpirun -n 1 uptimefasdfgaskfasdlkfsadf
$ echo $?
78

[awlauria]

I can take a look today.

[rhc54]

@awlauria I have done a little bit on it, but am happy to hand it over to you. The problem stems in part from trying to distinguish between prte running as a persistent DVM vs running as a one-shot mpirun. In the former, it is actually required that there be no app present on the cmd line. In the latter, you must have one. I think some of the "show help" support got lost there.

We also need to ensure that prte leaves via the prrte_quit function so we output the final error messages. We might be missing that step while in mpirun mode?

Anyway, here is what I have so far:

diff --git a/src/tools/prte/prte.c b/src/tools/prte/prte.c
index 9150b94ff0..e3dd8fe91c 100644
--- a/src/tools/prte/prte.c
+++ b/src/tools/prte/prte.c
@@ -519,6 +519,7 @@ int main(int argc, char *argv[])
     char **tmp;
     prrte_job_t *jdata;
     prrte_app_context_t *dapp;
+    bool proxyrun = false;
 
     /* init the globals */
     PRRTE_CONSTRUCT(&job_info, prrte_list_t);
@@ -580,6 +581,8 @@ int main(int argc, char *argv[])
             PRRTE_ERROR_LOG(rc);
             return rc;
         }
+    } else {
+        proxyrun = true;
     }
 
     /* get our session directory */
@@ -928,6 +931,14 @@ int main(int argc, char *argv[])
 
     /* did they provide an app? */
     if (PMIX_SUCCESS != rc || 0 == prrte_list_get_size(&apps)) {
+        if (proxyrun) {
+            if (PRRTE_ERR_NOT_FOUND != rc) {
+                /* this is an error */
+                prrte_show_help("help-prun.txt", "prun:nothing-to-do",
+                               true, prrte_tool_basename);
+            }
+            goto DONE;
+        }
         /* nope - just need to wait for instructions */
         goto proceed;
     }
@@ -1413,6 +1424,8 @@ static int create_app(int argc, char* argv[],
     /* See if we have anything left */
     if (0 == count) {
         rc = PRRTE_ERR_NOT_FOUND;
+        prrte_show_help("help-prun.txt", "prun:executable-not-specified",
+                       true, prrte_tool_basename, prrte_tool_basename);
         goto cleanup;
     }

Holler if you hit a snag

[awlauria]

Thanks for the info - makes sense. I'll apply your patch above and see what else needs to be done. Thanks!

[rhc54]

Will be fixed by openpmix/prrte#473 when it comes to OMPI