From 1e93437cd4b3b5d220e430d7c9b090c67e77aee7 Mon Sep 17 00:00:00 2001
From: Ralph Castain <rhc@open-mpi.org>
Date: Sat, 16 Oct 2010 03:29:47 +0000
Subject: [PATCH] To help with debugging, add a new mca param that instructs
 ORTE_ERROR_LOG to output "silent" errors. Helps to track down silent errors
 that don't have an associated error message (e.g., via show_help).

This commit was SVN r23893.
---
 orte/mca/errmgr/base/errmgr_base_fns.c |  8 ++++++--
 orte/runtime/orte_globals.c            |  1 +
 orte/runtime/orte_globals.h            |  1 +
 orte/runtime/orte_mca_params.c         |  6 +++++-
 orte/util/error_strings.c              | 13 +++++++++++--
 5 files changed, 24 insertions(+), 5 deletions(-)

diff --git a/orte/mca/errmgr/base/errmgr_base_fns.c b/orte/mca/errmgr/base/errmgr_base_fns.c
index bd85d0acd4..3058d7e1e9 100644
--- a/orte/mca/errmgr/base/errmgr_base_fns.c
+++ b/orte/mca/errmgr/base/errmgr_base_fns.c
@@ -177,16 +177,20 @@ void orte_errmgr_predicted_map_destruct( orte_errmgr_predicted_map_t *item)
  */
 void orte_errmgr_base_log(int error_code, char *filename, int line)
 {
+    char *errstring = NULL;
+
     OPAL_TRACE(1);
     
-    if (ORTE_ERR_SILENT == OPAL_SOS_GET_ERROR_CODE(error_code)) {
+    errstring = (char*)ORTE_ERROR_NAME(error_code);
+
+    if (NULL == errstring) {
         /* if the error is silent, say nothing */
         return;
     }
     
     opal_output(0, "%s ORTE_ERROR_LOG: %s in file %s at line %d",
                 ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
-                ORTE_ERROR_NAME(error_code), filename, line);
+                errstring, filename, line);
 }
 
 void orte_errmgr_base_abort(int error_code, char *fmt, ...)
diff --git a/orte/runtime/orte_globals.c b/orte/runtime/orte_globals.c
index 1d29a094cf..11d7b56f29 100644
--- a/orte/runtime/orte_globals.c
+++ b/orte/runtime/orte_globals.c
@@ -194,6 +194,7 @@ bool orte_xml_output = false;
 FILE *orte_xml_fp = NULL;
 char *orte_job_ident = NULL;
 bool orte_execute_quiet = false;
+bool orte_report_silent_errors = false;
 
 /* See comment in orte/tools/orterun/debuggers.c about this MCA
    param */
diff --git a/orte/runtime/orte_globals.h b/orte/runtime/orte_globals.h
index bb2d11d582..a70beeb612 100644
--- a/orte/runtime/orte_globals.h
+++ b/orte/runtime/orte_globals.h
@@ -56,6 +56,7 @@ ORTE_DECLSPEC extern bool orte_help_want_aggregate;  /* instantiated in orte/uti
 ORTE_DECLSPEC extern char *orte_job_ident;  /* instantiated in orte/runtime/orte_globals.c */
 ORTE_DECLSPEC extern bool orte_create_session_dirs;  /* instantiated in orte/runtime/orte_init.c */
 ORTE_DECLSPEC extern bool orte_execute_quiet;  /* instantiated in orte/runtime/orte_globals.c */
+ORTE_DECLSPEC extern bool orte_report_silent_errors;  /* instantiated in orte/runtime/orte_globals.c */
 
 /* Shortcut for some commonly used names */
 #define ORTE_NAME_WILDCARD      (&orte_name_wildcard)
diff --git a/orte/runtime/orte_mca_params.c b/orte/runtime/orte_mca_params.c
index 63b52855c8..a1b14ad942 100644
--- a/orte/runtime/orte_mca_params.c
+++ b/orte/runtime/orte_mca_params.c
@@ -93,7 +93,11 @@ int orte_register_params(void)
                                 false, false, (int) false, &value);
     orte_execute_quiet = OPAL_INT_TO_BOOL(value);
     
-    
+    mca_base_param_reg_int_name("orte", "report_silent_errors",
+                                "Report all errors, including silent ones",
+                                false, false, (int) false, &value);
+    orte_report_silent_errors = OPAL_INT_TO_BOOL(value);
+
 #if !ORTE_DISABLE_FULL_SUPPORT
     
     mca_base_param_reg_int_name("orte", "send_profile",
diff --git a/orte/util/error_strings.c b/orte/util/error_strings.c
index 4b9b690157..2e11b59e4a 100644
--- a/orte/util/error_strings.c
+++ b/orte/util/error_strings.c
@@ -26,6 +26,7 @@
 
 #include "opal/util/opal_sos.h"
 #include "orte/util/error_strings.h"
+#include "orte/runtime/orte_globals.h"
 
 const char *orte_err2str(int errnum)
 {
@@ -80,7 +81,11 @@ const char *orte_err2str(int errnum)
         retval = "Multiple applications were specified, but at least one failed to specify the number of processes to run";
         break;
     case ORTE_ERR_SILENT:
-        retval = NULL;
+        if (orte_report_silent_errors) {
+            retval = "Silent error";
+        } else {
+            retval = NULL;
+        }
         break;
     case ORTE_ERR_ADDRESSEE_UNKNOWN:
         retval = "A message is attempting to be sent to a process whose contact information is unknown";
@@ -136,7 +141,11 @@ const char *orte_err2str(int errnum)
 
             
     default:
-        retval = NULL;
+        if (orte_report_silent_errors) {
+            retval = "Unknown error";
+        } else {
+            retval = NULL;
+        }
     }
 
     return retval;