Make lsf support compile

This commit was SVN r18889.

orte/mca/ess/lsf/ess_lsf_module.c

@@ -35,6 +35,7 @@
 #include "opal/util/opal_environ.h"
 #include "opal/class/opal_pointer_array.h"
 
+#include "orte/util/show_help.h"
 #include "orte/util/name_fns.h"
 #include "orte/runtime/orte_globals.h"
 #include "opal/mca/base/mca_base_param.h"
@@ -309,55 +310,38 @@ static int lsf_set_name(void)
 {
     int rc;
     int id;
-    char* name_string = NULL;
     int lsf_nodeid;
-
+    orte_jobid_t jobid;
-    /* start by getting our jobid, and vpid (which is the
+    orte_vpid_t vpid;
-       starting vpid for the list of daemons) */
+    char* jobid_string;
-    id = mca_base_param_register_string("orte", "ess", "name", NULL, NULL);
+    char* vpid_string;
-    mca_base_param_lookup_string(id, &name_string);
-
-    if (name_string != NULL) {
-        if (ORTE_SUCCESS != 
-            (rc = orte_util_convert_string_to_process_name(&ORTE_PROC_MY_NAME, name_string))) {
-            ORTE_ERROR_LOG(rc);
-            free(name_string);
-            return rc;
-        }
-        free(name_string);
-    } else {
-        orte_jobid_t jobid;
-        orte_vpid_t vpid;
-        char* jobid_string;
-        char* vpid_string;
       
-        id = mca_base_param_register_string("orte", "ess", "jobid", NULL, NULL);
+    id = mca_base_param_register_string("orte", "ess", "jobid", NULL, NULL);
-        mca_base_param_lookup_string(id, &jobid_string);
+    mca_base_param_lookup_string(id, &jobid_string);
-        if (NULL == jobid_string) {
+    if (NULL == jobid_string) {
-            ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
+        ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
-            return ORTE_ERR_NOT_FOUND;
+        return ORTE_ERR_NOT_FOUND;
-        }
-        if (ORTE_SUCCESS != 
-            (rc = orte_util_convert_string_to_jobid(&jobid, jobid_string))) {
-            ORTE_ERROR_LOG(rc);
-            return(rc);
-        }
-        
-        id = mca_base_param_register_string("orte", "ess", "vpid", NULL, NULL);
-        mca_base_param_lookup_string(id, &vpid_string);
-        if (NULL == vpid_string) {
-            ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
-            return ORTE_ERR_NOT_FOUND;
-        }
-        if (ORTE_SUCCESS !=
-            (rc = orte_util_convert_string_to_vpid(&vpid, vpid_string))) {
-            ORTE_ERROR_LOG(rc);
-            return(rc);
-        }
-
-        ORTE_PROC_MY_NAME->jobid;
-        ORTE_PROC_MY_NAME->vpid = vpid;
     }
+    if (ORTE_SUCCESS != 
+        (rc = orte_util_convert_string_to_jobid(&jobid, jobid_string))) {
+        ORTE_ERROR_LOG(rc);
+        return(rc);
+    }
+    
+    id = mca_base_param_register_string("orte", "ess", "vpid", NULL, NULL);
+    mca_base_param_lookup_string(id, &vpid_string);
+    if (NULL == vpid_string) {
+        ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
+        return ORTE_ERR_NOT_FOUND;
+    }
+    if (ORTE_SUCCESS !=
+        (rc = orte_util_convert_string_to_vpid(&vpid, vpid_string))) {
+        ORTE_ERROR_LOG(rc);
+        return(rc);
+    }
+    
+    ORTE_PROC_MY_NAME->jobid = jobid;
+    ORTE_PROC_MY_NAME->vpid = vpid;
     
     /* fix up the base name and make it the "real" name */
     lsf_nodeid = atoi(getenv("LSF_PM_TASKID"));

orte/mca/plm/lsf/plm_lsf.h

@@ -28,17 +28,17 @@
 
 BEGIN_C_DECLS
 
-    struct orte_plm_lsf_component_t {
+struct orte_plm_lsf_component_t {
-        orte_plm_base_component_t super;
+    orte_plm_base_component_t super;
-        int priority;
+    int priority;
-        bool timing;
+    bool timing;
-        char *orted;
+    char *orted;
-    };
+};
-    typedef struct orte_plm_lsf_component_t orte_plm_lsf_component_t;
+typedef struct orte_plm_lsf_component_t orte_plm_lsf_component_t;
 
-    /* Globally exported variables */
+/* Globally exported variables */
-    ORTE_DECLSPEC extern orte_plm_lsf_component_t mca_plm_lsf_component;
+ORTE_DECLSPEC extern orte_plm_lsf_component_t mca_plm_lsf_component;
-    extern orte_plm_base_module_t orte_plm_lsf_module;
+extern orte_plm_base_module_t orte_plm_lsf_module;
 
 END_C_DECLS
 

orte/mca/plm/lsf/plm_lsf_component.c

@@ -29,9 +29,9 @@
 #include <lsf/lsbatch.h>
 
 #include "opal/mca/base/mca_base_param.h"
-#include "orte/util/show_help.h"
 #include "opal/util/argv.h"
 
+#include "orte/util/show_help.h"
 #include "orte/util/proc_info.h"
 #include "orte/mca/errmgr/errmgr.h"
 
@@ -95,7 +95,6 @@ orte_plm_lsf_component_t mca_plm_lsf_component = {
 
 static int plm_lsf_open(void)
 {
-    int tmp, value;
     mca_base_component_t *comp = &mca_plm_lsf_component.super.base_version;
 
     mca_base_param_reg_int(comp, "priority", "Default selection priority",
@@ -106,15 +105,6 @@ static int plm_lsf_open(void)
                               false, false, "orted",
                               &mca_plm_lsf_component.orted);
 
-    tmp = mca_base_param_reg_int_name("orte", "timing",
-                                      "Request that critical timing loops be measured",
-                                      false, false, 0, &value);
-    if (value != 0) {
-        mca_plm_lsf_component.timing = true;
-    } else {
-        mca_plm_lsf_component.timing = false;
-    }
-    
     return ORTE_SUCCESS;
 }
 
@@ -131,7 +121,7 @@ static int orte_plm_lsf_component_query(mca_base_module_t **module, int *priorit
     /* check if lsf is running here */
     if (NULL == getenv("LSB_JOBID") || lsb_init("ORTE launcher") < 0) {
         /* nope, not here */
-        opal_output_verbose(10, orte_plm_base.plm_output,
+        opal_output_verbose(10, orte_plm_globals.output,
                             "plm:lsf: NOT available for selection");
         *module = NULL;
         return ORTE_ERROR:

orte/mca/plm/lsf/plm_lsf_module.c

@@ -55,12 +55,12 @@
 
 #include "opal/mca/installdirs/installdirs.h"
 #include "opal/util/argv.h"
-#include "orte/util/show_help.h"
 #include "opal/util/opal_environ.h"
 #include "opal/util/path.h"
 #include "opal/util/basename.h"
 #include "opal/mca/base/mca_base_param.h"
 
+#include "orte/util/show_help.h"
 #include "orte/runtime/runtime.h"
 #include "orte/runtime/orte_wakeup.h"
 #include "orte/runtime/orte_wait.h"
@@ -88,7 +88,7 @@ static int plm_lsf_finalize(void);
 /*
  * Global variable
  */
-orte_plm_base_module_1_3_0_t orte_plm_lsf_module = {
+orte_plm_base_module_t orte_plm_lsf_module = {
     plm_lsf_init,
     orte_plm_base_set_hnp_name,
     plm_lsf_launch_job,
@@ -130,7 +130,6 @@ static int plm_lsf_launch_job(orte_job_t *jdata)
     int argc;
     int rc;
     char** env = NULL;
-    char* var;
     char **nodelist_argv;
     int nodelist_argc;
     char *vpid_string;
@@ -147,7 +146,7 @@ static int plm_lsf_launch_job(orte_job_t *jdata)
     /* default to declaring the daemons failed*/
     failed_job = ORTE_PROC_MY_NAME->jobid;
 
-    if (mca_plm_lsf_component.timing) {
+    if (orte_timing) {
         if (0 != gettimeofday(&joblaunchstart, NULL)) {
             opal_output(0, "plm_lsf: could not obtain job start time");
         }        
@@ -257,8 +256,8 @@ static int plm_lsf_launch_job(orte_job_t *jdata)
        don't support different --prefix'es for different nodes in
        the SLURM plm) */
     cur_prefix = NULL;
-    for (i=0; i < map->num_apps; i++) {
+    for (i=0; i < jdata->num_apps; i++) {
-        char * app_prefix_dir = map->apps[i]->prefix_dir;
+        char * app_prefix_dir = apps[i]->prefix_dir;
          /* Check for already set cur_prefix_dir -- if different,
            complain */
         if (NULL != app_prefix_dir) {
@@ -284,7 +283,7 @@ static int plm_lsf_launch_job(orte_job_t *jdata)
     /* setup environment */
     env = opal_argv_copy(orte_launch_environ);
 
-    if (mca_plm_lsf_component.timing) {
+    if (orte_timing) {
         if (0 != gettimeofday(&launchstart, NULL)) {
             opal_output(0, "plm_lsf: could not obtain start time");
         }        
@@ -306,7 +305,10 @@ static int plm_lsf_launch_job(orte_job_t *jdata)
     /* wait for daemons to callback */
     if (ORTE_SUCCESS != 
         (rc = orte_plm_base_daemon_callback(map->num_new_daemons))) {
-        ORTE_ERROR_LOG(rc);
+        OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
+                             "%s plm:lsf: daemon launch failed for job %s on error %s",
+                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
+                             ORTE_JOBID_PRINT(active_job), ORTE_ERROR_NAME(rc)));
         goto cleanup;
     }
 
@@ -314,14 +316,17 @@ launch_apps:
     /* daemons succeeded - any failure now would be from apps */
     failed_job = active_job;
     if (ORTE_SUCCESS != (rc = orte_plm_base_launch_apps(active_job))) {
-        ORTE_ERROR_LOG(rc);
+        OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
+                             "%s plm:lsf: launch of apps failed for job %s on error %s",
+                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
+                             ORTE_JOBID_PRINT(active_job), ORTE_ERROR_NAME(rc)));
         goto cleanup;
     }
     
     /* declare the launch a success */
     failed_launch = false;
     
-    if (mca_plm_lsf_component.timing) {
+    if (orte_timing) {
         if (0 != gettimeofday(&launchstop, NULL)) {
              opal_output(0, "plm_lsf: could not obtain stop time");
          } else {
@@ -340,9 +345,6 @@ launch_apps:
     }
 
 cleanup:
-    if (NULL != map) {
-        OBJ_RELEASE(map);
-    }
     if (NULL != argv) {
         opal_argv_free(argv);
     }