Move the error name macro to the errmgr framework. Add a second level of tracing. Remove an obsolete file.

This commit was SVN r7445.

opal/util/trace.h

@@ -38,16 +38,23 @@ extern "C" {
             __func__, __FILE__, __LINE__);                                     \
        } while (0)
 
-#define OPAL_TRACE_ARG(verbose, foo) \
+#define OPAL_TRACE_ARG1(verbose, foo) \
     do {                                                                       \
         opal_output_verbose(verbose, opal_trace_handle, "TRACE: %s @ %s:%d arg: %lu",   \
             __func__, __FILE__, __LINE__, (unsigned long)foo);                                     \
        } while (0)
-        
+
+#define OPAL_TRACE_ARG2(verbose, foo, foo2) \
+    do {                                                                       \
+        opal_output_verbose(verbose, opal_trace_handle, "TRACE: %s @ %s:%d arg: %lu\n\t0x%x",   \
+            __func__, __FILE__, __LINE__, (unsigned long)foo, (unsigned long)foo2);        \
+       } while (0)
+
 #else
 
 #define OPAL_TRACE(verbose)
-#define OPAL_TRACE_ARG(verbose, foo)
+#define OPAL_TRACE_ARG1(verbose, foo)
+#define OPAL_TRACE_ARG2(verbose, foo, foo2)
 
 #endif /* ENABLE_TRACE */
 

orte/mca/errmgr/base/errmgr_base_log.c

@@ -1,32 +0,0 @@
-/*
- * Copyright (c) 2004-2005 The Trustees of Indiana University.
- *                         All rights reserved.
- * Copyright (c) 2004-2005 The Trustees of the University of Tennessee.
- *                         All rights reserved.
- * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart, 
- *                         University of Stuttgart.  All rights reserved.
- * Copyright (c) 2004-2005 The Regents of the University of California.
- *                         All rights reserved.
- * $COPYRIGHT$
- * 
- * Additional copyrights may follow
- * 
- * $HEADER$
- */
-
-
-#include "ompi_config.h"
-
-#include "opal/util/output.h"
-#include "runtime/runtime.h"
-#include "mca/ns/ns.h"
-
-#include "mca/errmgr/base/base.h"
-
-
-void mca_errmgr_base_log(char *msg, char *filename, int line)
-{
-    opal_output(0, "[%lu,%lu,%lu] ORTE_ERROR_LOG: %s in file %s at line %d",
-                ORTE_NAME_ARGS(orte_process_info.my_name), 
-                msg, filename, line);
-}

orte/mca/errmgr/errmgr.h

@@ -3,14 +3,14 @@
  *                         All rights reserved.
  * Copyright (c) 2004-2005 The Trustees of the University of Tennessee.
  *                         All rights reserved.
- * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart, 
+ * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
  *                         University of Stuttgart.  All rights reserved.
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
  * $COPYRIGHT$
- * 
+ *
  * Additional copyrights may follow
- * 
+ *
  * $HEADER$
  */
 /** @file:
@@ -39,6 +39,13 @@ extern "C" {
 /*
  * Macro definitions
  */
+/*
+ * Thess macros and associated error name array are used to output intelligible error
+ * messages.
+ */
+
+#define ORTE_ERROR_NAME(n)  opal_strerror(n)
+
 #define ORTE_ERROR_LOG(n) \
     orte_errmgr.log((n), __FILE__, __LINE__)
 
@@ -51,7 +58,7 @@ extern "C" {
  * Log an error
  * Log an error that occurred in the runtime environment, and call the "error_detected"
  * interface to see if further action is required.
- * 
+ *
  * @code
  * orte_errmgr.log("this is an error", __FILE__, __LINE__);
  * @endcode
@@ -65,7 +72,7 @@ typedef void (*orte_errmgr_base_module_log_fn_t)(int error_code, char *filename,
  * process errors should always be reported through the error_detected interface and
  * NOT here. The function is called when a message is received from the universe daemon
  * indicating that another process in the job failed. For now, this function will
- * simply cause the local process to gracefully finalize and terminate. 
+ * simply cause the local process to gracefully finalize and terminate.
  */
 typedef void (*orte_errmgr_base_module_proc_aborted_fn_t)(orte_process_name_t *proc);
 
@@ -75,7 +82,7 @@ typedef void (*orte_errmgr_base_module_proc_aborted_fn_t)(orte_process_name_t *p
  * one or more processes to start. The function decides on the strategy for dealing
  * with this "incomplete start" situation - for now, it simply orders the resource
  * manager to terminate the entire job.
- * 
+ *
  * This function is only called by the respective process launcher, which is responsible
  * for detecting incomplete starts.
  */
@@ -96,7 +103,7 @@ typedef void (*orte_errmgr_base_module_error_detected_fn_t)(int error_code);
  * problems occur - i.e., when process status entries change to abnormal termination
  * values. Process status entries are changed by the appropriate state-of-health monitor
  * and/or the process launcher, depending upon the stage at which the problem occurs.
- * 
+ *
  * Monitoring of the job begins once the job has reached the "executing" stage. Prior
  * to that time, failure of processes to start are the responsibility of the respective
  * process launcher - which is expected to call the error manager via the "incomplete
@@ -136,7 +143,7 @@ typedef orte_errmgr_base_module_t* (*orte_errmgr_base_component_init_fn_t)(
     int *priority);
 
 typedef int (*orte_errmgr_base_component_finalize_fn_t)(void);
- 
+
 /*
  * the standard component data structure
  */

orte/mca/schema/schema_types.h

@@ -21,13 +21,6 @@
 #include "mca/ns/ns_types.h"
 #include "opal/util/error.h"
 
-/*
- * This macro and array are used to output intelligible error
- * messages.
- */
-
-#define ORTE_ERROR_NAME(n)  opal_strerror(n)
-
 /*
  * Standard names used across the system
  */

orte/tools/orterun/orterun.c

@@ -4,14 +4,14 @@
  *                         All rights reserved.
  * Copyright (c) 2004-2005 The Trustees of the University of Tennessee.
  *                         All rights reserved.
- * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart, 
+ * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
  *                         University of Stuttgart.  All rights reserved.
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
  * $COPYRIGHT$
- * 
+ *
  * Additional copyrights may follow
- * 
+ *
  * $HEADER$
  */
 
@@ -354,10 +354,10 @@ int orterun(int argc, char *argv[])
 
     /* Prep to start the application */
 
-    opal_signal_set(&term_handler, SIGTERM, 
+    opal_signal_set(&term_handler, SIGTERM,
                    signal_callback, NULL);
     opal_signal_add(&term_handler, NULL);
-    opal_signal_set(&int_handler, SIGINT, 
+    opal_signal_set(&int_handler, SIGINT,
                    signal_callback, NULL);
     opal_signal_add(&int_handler, NULL);
 
@@ -436,8 +436,8 @@ static void dump_aborted_procs(orte_jobid_t jobid)
         NULL
     };
 
-    OPAL_TRACE_ARG(1, jobid);
+    OPAL_TRACE_ARG1(1, jobid);
-    
+
     /* query the job segment on the registry */
     if(ORTE_SUCCESS != (rc = orte_schema.get_job_segment_name(&segment, jobid))) {
         ORTE_ERROR_LOG(rc);
@@ -538,8 +538,8 @@ static void dump_aborted_procs(orte_jobid_t jobid)
 
 static void job_state_callback(orte_jobid_t jobid, orte_proc_state_t state)
 {
-    OPAL_TRACE_ARG(1, jobid);
+    OPAL_TRACE_ARG2(1, jobid, state);
-    
+
     OPAL_THREAD_LOCK(&orterun_globals.lock);
 
     /* Note that there's only two states that we're interested in
@@ -588,7 +588,7 @@ static void job_state_callback(orte_jobid_t jobid, orte_proc_state_t state)
 static void exit_callback(int fd, short event, void *arg)
 {
     OPAL_TRACE(1);
-    
+
     opal_show_help("help-orterun.txt", "orterun:abnormal-exit",
                    true, orterun_basename, orterun_basename);
 
@@ -616,9 +616,9 @@ static void signal_callback(int fd, short flags, void *arg)
     opal_event_t* event;
 
     static int signalled = 0;
-    
+
     OPAL_TRACE(1);
-    
+
     if (0 != signalled++) {
          return;
     }