This commit fixes runs when there is no available CRS component (BLCR is unavailable, and SELF is deactivated). Previously the run would fail out of MPI_INIT since the OPAL CRS framework could not select a component. This is because the framework did not recognize the 'none' component as a full component because it was part of crs/base.

I promoted the ''none'' component to a full component, and updated the other components to reflect this code movement. The ''none'' component is the default component unless the user requests '''-am ft-enable-cr''' to auto-select a component. There is an MCA parameter to show a warning if the application requested an FT enabled job, but the ''none'' component was selected ({{{crs_none_select_warning}}}).

This temporarily fixes the problem mentioned in r18739. The full fix will entail working on ticket #1291.

Thanks to Ethan from Sun for finding this bug.

This commit was SVN r18840.

The following SVN revision numbers were found above:
  r18739 --> open-mpi/ompi@a003fa7a50

opal/mca/crs/base/base.h

@@ -81,33 +81,6 @@ extern "C" {
     OPAL_DECLSPEC extern opal_crs_base_module_t opal_crs;
     OPAL_DECLSPEC extern char * opal_crs_base_snapshot_dir;
 
-    /**
-     * 'None' component functions
-     * These are to be used when no component is selected.
-     * They just return success, and empty strings as necessary.
-     */
-    int opal_crs_base_none_open(void);
-    int opal_crs_base_none_close(void);
-    int opal_crs_base_none_query(mca_base_module_t **module, int *priority);
-
-    int opal_crs_base_none_module_init(void);
-    int opal_crs_base_none_module_finalize(void);
-
-    int opal_crs_base_none_checkpoint(    pid_t pid, opal_crs_base_snapshot_t *sanpshot, opal_crs_state_type_t *state);
-
-    int opal_crs_base_none_restart(    opal_crs_base_snapshot_t *snapshot, bool spawn_child, pid_t *child_pid);
-
-    int opal_crs_base_none_disable_checkpoint(void);
-    int opal_crs_base_none_enable_checkpoint(void);
-    
-    OPAL_DECLSPEC int opal_crs_base_none_prelaunch(int32_t rank,
-                                                   char *base_snapshot_dir,
-                                                   char **app,
-                                                   char **cwd,
-                                                   char ***argv,
-                                                   char ***env);
-    OPAL_DECLSPEC int opal_crs_base_none_reg_thread(void);
-
     /**
      * Some utility functions
      */

opal/mca/crs/base/crs_base_fns.c

@@ -88,99 +88,7 @@ OBJ_CLASS_INSTANCE(opal_crs_base_snapshot_t,
                    opal_crs_base_construct,
                    opal_crs_base_destruct);
 
-int opal_crs_base_none_open(void)
-{
-    return OPAL_SUCCESS;
-}
 
-int opal_crs_base_none_close(void)
-{
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_query(mca_base_module_t **module, int *priority)
-{
-    *module = NULL;
-    *priority = 0;
-
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_module_init(void)
-{
-
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_module_finalize(void)
-{
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_checkpoint(pid_t pid, opal_crs_base_snapshot_t *snapshot, opal_crs_state_type_t *state)
-{
-    int ret;
-
-    *state = OPAL_CRS_CONTINUE;
-    
-    snapshot->component_name  = strdup("none");
-    snapshot->reference_name  = strdup("none");
-    snapshot->local_location  = strdup("");
-    snapshot->remote_location = strdup("");
-    snapshot->cold_start      = false;
-
-    /*
-     * Update the snapshot metadata
-     */
-    if( OPAL_SUCCESS != (ret = opal_crs_base_metadata_write_token(NULL, CRS_METADATA_COMP, "none") ) ) {
-        opal_output(0,
-                    "crs:none: checkpoint(): Error: Unable to write component name to the directory for (%s).",
-                    snapshot->reference_name);
-        return ret;
-    }
-
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_restart(opal_crs_base_snapshot_t *snapshot, bool spawn_child, pid_t *child_pid)
-{
-    *child_pid = getpid();
-
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_disable_checkpoint(void)
-{
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_enable_checkpoint(void)
-{
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_prelaunch(int32_t rank,
-                                 char *base_snapshot_dir,
-                                 char **app,
-                                 char **cwd,
-                                 char ***argv,
-                                 char ***env)
-{
-    char * tmp_env_var = NULL;
-
-    tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
-    opal_setenv(tmp_env_var,
-                "0", true, env);
-    free(tmp_env_var);
-    tmp_env_var = NULL;
-
-    return OPAL_SUCCESS;
-}
-
-int opal_crs_base_none_reg_thread(void)
-{
-    return OPAL_SUCCESS;
-}
 
 /*
  * Utility functions

opal/mca/crs/base/crs_base_open.c

@@ -85,7 +85,7 @@ int opal_crs_base_open(void)
     mca_base_param_reg_string_name("crs", NULL,
                                    "Which CRS component to use (empty = auto-select)",
                                    false, false,
-                                   NULL, &str_value);
+                                   "none", &str_value);
     
     /* Open up all available components */
     if (OPAL_SUCCESS != (ret = mca_base_components_open("crs", 

opal/mca/crs/base/crs_base_select.c

@@ -31,72 +31,11 @@
 #include "opal/mca/crs/base/base.h"
 #include "opal/runtime/opal_cr.h"
 
-/*
- * Instantiate the public struct with all of our public information
- * and pointer to our public functions in it
- */
-static opal_crs_base_component_t none_component = {
-    /* Handle the general mca_component_t struct containing 
-     *  meta information about the component itself
-     */
-    {
-        OPAL_CRS_BASE_VERSION_1_0_0,
-        /* Component name and version */
-        "none",
-        OMPI_MAJOR_VERSION,
-        OMPI_MINOR_VERSION,
-        OMPI_RELEASE_VERSION,
-            
-        /* Component open and close functions */
-        opal_crs_base_none_open,
-        opal_crs_base_none_close,
-        opal_crs_base_none_query
-    },
-
-    /* Next the MCA v1.0.0 component meta data */
-    {
-        /* The component is checkpoint ready */
-        MCA_BASE_METADATA_PARAM_CHECKPOINT
-    },
-        
-    /* Verbosity level */
-    0,
-    /* opal_output handler */
-    -1,
-    /* Default priority */
-    0
-};
-
-static opal_crs_base_module_t none_module = {
-    /** Initialization Function */
-    opal_crs_base_none_module_init,
-    /** Finalization Function */
-    opal_crs_base_none_module_finalize,
-
-    /** Checkpoint interface */
-    opal_crs_base_none_checkpoint,
-
-    /** Restart Command Access */
-    opal_crs_base_none_restart,
-
-    /** Disable checkpoints */
-    opal_crs_base_none_disable_checkpoint,
-    /** Enable checkpoints */
-    opal_crs_base_none_enable_checkpoint,
-
-    /** Prelaunch */
-    opal_crs_base_none_prelaunch,
-
-    /** Register thread */
-    opal_crs_base_none_reg_thread
-};
-
 int opal_crs_base_select(void)
 {
     int ret, exit_status = OPAL_SUCCESS;
     opal_crs_base_component_t *best_component = NULL;
     opal_crs_base_module_t *best_module = NULL;
-    char *include_list = NULL;
     int int_value = 0;
 
     /*
@@ -122,28 +61,6 @@ int opal_crs_base_select(void)
         return OPAL_SUCCESS;
     }
 
-    /*
-     * Register the framework MCA param and look up include list
-     */
-    mca_base_param_reg_string_name("crs", NULL,
-                                   "Which CRS component to use (empty = auto-select)",
-                                   false, false,
-                                   NULL, &include_list);
-
-    if(NULL != include_list && 0 == strncmp(include_list, "none", strlen("none")) ){ 
-        opal_output_verbose(10, opal_crs_base_output,
-                            "crs:select: Using %s component",
-                            include_list);
-        best_component = &none_component;
-        best_module    = &none_module;
-        /* JJH: Todo: Check if none is in the list */
-        /* Close all components since none will be used */
-        mca_base_components_close(0, /* Pass 0 to keep this from closing the output handle */
-                                  &opal_crs_base_components_available,
-                                  NULL);
-        goto skip_select;
-    }
-
     /*
      * Select the best component
      */
@@ -156,7 +73,6 @@ int opal_crs_base_select(void)
         goto cleanup;
     }
 
- skip_select:
     /* Save the winner */
     opal_crs_base_selected_component = *best_component;
     opal_crs = *best_module;
@@ -170,10 +86,5 @@ int opal_crs_base_select(void)
     }
 
  cleanup:
-    if( NULL != include_list ) {
-        free(include_list);
-        include_list = NULL;
-    }
-
     return exit_status;
 }

opal/mca/crs/blcr/crs_blcr.h

@@ -71,6 +71,13 @@ extern "C" {
     int opal_crs_blcr_disable_checkpoint(void);
     int opal_crs_blcr_enable_checkpoint(void);
 
+    int opal_crs_blcr_prelaunch(int32_t rank,
+                                char *base_snapshot_dir,
+                                char **app,
+                                char **cwd,
+                                char ***argv,
+                                char ***env);
+
     int opal_crs_blcr_reg_thread(void);
 
 #if defined(c_plusplus) || defined(__cplusplus)

opal/mca/crs/blcr/crs_blcr_module.c

@@ -64,7 +64,7 @@ static opal_crs_base_module_t blcr_module = {
     opal_crs_blcr_enable_checkpoint,
 
     /** Prelaunch */
-    opal_crs_base_none_prelaunch,
+    opal_crs_blcr_prelaunch,
 
     /** Register Thread */
     opal_crs_blcr_reg_thread
@@ -203,6 +203,25 @@ int opal_crs_blcr_module_init(void)
     
     return OPAL_SUCCESS;
 }
+
+int opal_crs_blcr_prelaunch(int32_t rank,
+                            char *base_snapshot_dir,
+                            char **app,
+                            char **cwd,
+                            char ***argv,
+                            char ***env)
+{
+    char * tmp_env_var = NULL;
+
+    tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
+    opal_setenv(tmp_env_var,
+                "0", true, env);
+    free(tmp_env_var);
+    tmp_env_var = NULL;
+
+    return OPAL_SUCCESS;
+}
+
 int opal_crs_blcr_reg_thread(void)
 {
     cr_client_id_t loc_client_id;

opal/mca/crs/none/Makefile.am

@@ -0,0 +1,40 @@
+#
+# Copyright (c) 2004-2008 The Trustees of Indiana University.
+#                         All rights reserved.
+# $COPYRIGHT$
+# 
+# Additional copyrights may follow
+# 
+# $HEADER$
+#
+
+AM_CPPFLAGS = \
+    $(LTDLINCL)
+
+dist_pkgdata_DATA = help-opal-crs-none.txt
+
+sources = \
+        crs_none.h \
+        crs_none_component.c \
+        crs_none_module.c
+
+# Make the output library in this directory, and name it either
+# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
+# (for static builds).
+
+if OMPI_BUILD_crs_none_DSO
+component_noinst =
+component_install = mca_crs_none.la
+else
+component_noinst = libmca_crs_none.la
+component_install =
+endif
+
+mcacomponentdir = $(pkglibdir)
+mcacomponent_LTLIBRARIES = $(component_install)
+mca_crs_none_la_SOURCES = $(sources)
+mca_crs_none_la_LDFLAGS = -module -avoid-version
+
+noinst_LTLIBRARIES = $(component_noinst)
+libmca_crs_none_la_SOURCES = $(sources)
+libmca_crs_none_la_LDFLAGS = -module -avoid-version

opal/mca/crs/none/configure.params

@@ -0,0 +1,13 @@
+# -*- shell-script -*-
+#
+# Copyright (c) 2004-2008 The Trustees of Indiana University.
+#                         All rights reserved.
+# $COPYRIGHT$
+# 
+# Additional copyrights may follow
+# 
+# $HEADER$
+#
+
+PARAM_INIT_FILE=crs_none_component.c
+PARAM_CONFIG_FILES="Makefile"

opal/mca/crs/none/crs_none.h

@@ -0,0 +1,76 @@
+/*
+ * Copyright (c) 2004-2008 The Trustees of Indiana University.
+ *                         All rights reserved.
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+/**
+ * @file
+ * 
+ * NONE CRS component
+ *
+ * Simple, braindead implementation.
+ */
+
+#ifndef MCA_CRS_NONE_EXPORT_H
+#define MCA_CRS_NONE_EXPORT_H
+
+#include "opal_config.h"
+
+#include "opal/util/output.h"
+
+#include "opal/mca/mca.h"
+#include "opal/mca/crs/crs.h"
+
+#if defined(c_plusplus) || defined(__cplusplus)
+extern "C" {
+#endif
+
+    /*
+     * Local Component structures
+     */
+    struct opal_crs_none_component_t {
+        opal_crs_base_component_t super;  /** Base CRS component */
+
+    };
+    typedef struct opal_crs_none_component_t opal_crs_none_component_t;
+    OPAL_MODULE_DECLSPEC extern opal_crs_none_component_t mca_crs_none_component;
+
+    int opal_crs_none_component_query(mca_base_module_t **module, int *priority);
+
+    /*
+     * Module functions
+     */
+    int opal_crs_none_module_init(void);
+    int opal_crs_none_module_finalize(void);
+
+    /*
+     * Actual funcationality
+     */
+    int opal_crs_none_checkpoint( pid_t pid, opal_crs_base_snapshot_t *snapshot, opal_crs_state_type_t *state);
+
+    int opal_crs_none_restart(    opal_crs_base_snapshot_t *snapshot, bool spawn_child, pid_t *child_pid);
+
+    int opal_crs_none_disable_checkpoint(void);
+    int opal_crs_none_enable_checkpoint(void);
+
+    int opal_crs_none_prelaunch(int32_t rank,
+                                char *base_snapshot_dir,
+                                char **app,
+                                char **cwd,
+                                char ***argv,
+                                char ***env);
+
+    int opal_crs_none_reg_thread(void);
+
+    extern bool opal_crs_none_select_warning;
+
+#if defined(c_plusplus) || defined(__cplusplus)
+}
+#endif
+
+#endif /* MCA_CRS_NONE_EXPORT_H */

opal/mca/crs/none/crs_none_component.c

@@ -0,0 +1,126 @@
+/*
+ * Copyright (c) 2004-2008 The Trustees of Indiana University.
+ *                         All rights reserved.
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "opal_config.h"
+
+#include "opal/constants.h"
+#include "opal/mca/crs/crs.h"
+#include "opal/mca/crs/base/base.h"
+#include "crs_none.h"
+
+/*
+ * Public string for version number
+ */
+const char *opal_crs_none_component_version_string = 
+"OPAL CRS none MCA component version " OMPI_VERSION;
+
+/*
+ * Local functionality
+ */
+static int crs_none_open(void);
+static int crs_none_close(void);
+
+/*
+ * Instantiate the public struct with all of our public information
+ * and pointer to our public functions in it
+ */
+opal_crs_none_component_t mca_crs_none_component = {
+    /* First do the base component stuff */
+    {
+        /* Handle the general mca_component_t struct containing 
+         *  meta information about the component itnone
+         */
+        {
+            OPAL_CRS_BASE_VERSION_1_0_0,
+            /* Component name and version */
+            "none",
+            OMPI_MAJOR_VERSION,
+            OMPI_MINOR_VERSION,
+            OMPI_RELEASE_VERSION,
+            
+            /* Component open and close functions */
+            crs_none_open,
+            crs_none_close,
+            opal_crs_none_component_query
+        },
+
+        /* Next the MCA v1.0.0 component meta data */
+        {
+            /* The component is checkpoint ready */
+            MCA_BASE_METADATA_PARAM_CHECKPOINT
+        },
+        
+        /* Verbosity level */
+        0,
+        /* opal_output handler */
+        -1,
+        /* Default priority */
+        0
+    }
+};
+
+/*
+ * None module
+ */
+static opal_crs_base_module_t loc_module = {
+    /** Initialization Function */
+    opal_crs_none_module_init,
+    /** Finalization Function */
+    opal_crs_none_module_finalize,
+
+    /** Checkpoint interface */
+    opal_crs_none_checkpoint,
+
+    /** Restart Command Access */
+    opal_crs_none_restart,
+
+    /** Disable checkpoints */
+    opal_crs_none_disable_checkpoint,
+    /** Enable checkpoints */
+    opal_crs_none_enable_checkpoint,
+
+    /** Prelaunch */
+    opal_crs_none_prelaunch,
+
+    /** Register Thread */
+    opal_crs_none_reg_thread
+};
+
+bool opal_crs_none_select_warning = false;
+
+static int crs_none_open(void) 
+{
+    int value = 0;
+
+    mca_base_param_reg_int(&mca_crs_none_component.super.base_version,
+                           "select_warning",
+                           "Enable warning when the 'none' component is selected when checkpoint/restart functionality is requested."
+                           "[Default = disabled/no-warning]",
+                           false, false,
+                           0, /* Disabled */
+                           &value);
+    opal_crs_none_select_warning = OPAL_INT_TO_BOOL(value);
+
+    return OPAL_SUCCESS;
+}
+
+static int crs_none_close(void)
+{
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_none_component_query(mca_base_module_t **module, int *priority)
+{
+    *module   = (mca_base_module_t *)&loc_module;
+    *priority = mca_crs_none_component.super.priority;
+
+    return OPAL_SUCCESS;
+}
+

opal/mca/crs/none/crs_none_module.c

@@ -0,0 +1,116 @@
+/*
+ * Copyright (c) 2004-2008 The Trustees of Indiana University.
+ *                         All rights reserved.
+ *
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "opal_config.h"
+
+#include <sys/types.h>
+#if HAVE_UNISTD_H
+#include <unistd.h>
+#endif  /* HAVE_UNISTD_H */
+
+#include "opal/util/opal_environ.h"
+#include "opal/util/output.h"
+#include "opal/util/show_help.h"
+#include "opal/util/argv.h"
+#include "opal/util/opal_environ.h"
+
+#include "opal/constants.h"
+#include "opal/mca/base/mca_base_param.h"
+
+#include "opal/mca/crs/crs.h"
+#include "opal/mca/crs/base/base.h"
+#include "opal/runtime/opal_cr.h"
+
+#include "crs_none.h"
+
+int opal_crs_none_module_init(void)
+{
+    /*
+     * If not a tool, and requesting C/R support print a warning.
+     */
+    if( opal_crs_none_select_warning &&
+        !opal_cr_is_tool && opal_cr_is_enabled ) {
+        opal_show_help("help-opal-crs-none.txt", "none:select-warning",
+                       true);
+    }
+
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_none_module_finalize(void)
+{
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_none_checkpoint(pid_t pid, opal_crs_base_snapshot_t *snapshot, opal_crs_state_type_t *state)
+{
+    int ret;
+
+    *state = OPAL_CRS_CONTINUE;
+    
+    snapshot->component_name  = strdup("none");
+    snapshot->reference_name  = strdup("none");
+    snapshot->local_location  = strdup("");
+    snapshot->remote_location = strdup("");
+    snapshot->cold_start      = false;
+
+    /*
+     * Update the snapshot metadata
+     */
+    if( OPAL_SUCCESS != (ret = opal_crs_base_metadata_write_token(NULL, CRS_METADATA_COMP, "none") ) ) {
+        opal_output(0,
+                    "crs:none: checkpoint(): Error: Unable to write component name to the directory for (%s).",
+                    snapshot->reference_name);
+        return ret;
+    }
+
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_none_restart(opal_crs_base_snapshot_t *base_snapshot, bool spawn_child, pid_t *child_pid)
+{
+    *child_pid = getpid();
+
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_none_disable_checkpoint(void)
+{
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_none_enable_checkpoint(void)
+{
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_none_prelaunch(int32_t rank,
+                            char *base_snapshot_dir,
+                            char **app,
+                            char **cwd,
+                            char ***argv,
+                            char ***env)
+{
+    char * tmp_env_var = NULL;
+
+    tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
+    opal_setenv(tmp_env_var,
+                "0", true, env);
+    free(tmp_env_var);
+    tmp_env_var = NULL;
+
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_none_reg_thread(void)
+{
+    return OPAL_SUCCESS;
+}

opal/mca/crs/none/help-opal-crs-none.txt

@@ -0,0 +1,20 @@
+ -*- text -*-
+#
+# Copyright (c) 2004-2008 The Trustees of Indiana University and Indiana
+#                         University Research and Technology
+#                         Corporation.  All rights reserved.
+# $COPYRIGHT$
+#
+# Additional copyrights may follow
+#
+# $HEADER$
+#
+# This is the US/English general help file for Open PAL CRS framework.
+#
+[none:select-warning]
+Error: 'none' component selected.
+       Checkpoint/Restart functionality may not work properly. 
+       Make sure that you have configured with and are using a fully functional
+       CRS component.
+       To disable this warning set the following MCA parmeter:
+           --mca crs_none_select_warning 0

opal/mca/crs/self/crs_self.h

@@ -81,6 +81,15 @@ extern "C" {
     int opal_crs_self_disable_checkpoint(void);
     int opal_crs_self_enable_checkpoint(void);
 
+    int opal_crs_self_prelaunch(int32_t rank,
+                                char *base_snapshot_dir,
+                                char **app,
+                                char **cwd,
+                                char ***argv,
+                                char ***env);
+
+    int opal_crs_self_reg_thread(void);
+
 
 #if defined(c_plusplus) || defined(__cplusplus)
 }

opal/mca/crs/self/crs_self_module.c

@@ -63,10 +63,10 @@ static opal_crs_base_module_t loc_module = {
     opal_crs_self_enable_checkpoint,
 
     /** Prelaunch */
-    opal_crs_base_none_prelaunch,
+    opal_crs_self_prelaunch,
 
     /** Register Thread */
-    opal_crs_base_none_reg_thread
+    opal_crs_self_reg_thread
 };
 
 /*
@@ -479,6 +479,29 @@ int opal_crs_self_enable_checkpoint(void)
     return OPAL_SUCCESS;
 }
 
+int opal_crs_self_prelaunch(int32_t rank,
+                            char *base_snapshot_dir,
+                            char **app,
+                            char **cwd,
+                            char ***argv,
+                            char ***env)
+{
+    char * tmp_env_var = NULL;
+
+    tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
+    opal_setenv(tmp_env_var,
+                "0", true, env);
+    free(tmp_env_var);
+    tmp_env_var = NULL;
+
+    return OPAL_SUCCESS;
+}
+
+int opal_crs_self_reg_thread(void)
+{
+    return OPAL_SUCCESS;
+}
+
 /******************
  * Local functions
  ******************/