More warnings clean up in the collectives code on behalf of Joshua Ladd. Refs trac:3243.

This commit was SVN r27090.

The following Trac tickets were found above:
  Ticket 3243 --> https://svn.open-mpi.org/trac/ompi/ticket/3243

ompi/mca/bcol/basesmuma/bcol_basesmuma_progress.c

@@ -23,7 +23,7 @@
 int bcol_basesmuma_progress(void)
 {
     /* local variables */
-   volatile uint64_t *cntr;
+   volatile int64_t *cntr;
     mca_bcol_basesmuma_component_t *cs = &mca_bcol_basesmuma_component;
 
     /* check to see if release of memory blocks needs to be done */
@@ -54,7 +54,7 @@ int bcol_basesmuma_progress(void)
                        ctl_buffs_mgmt[index].bank_gen_counter);
                      */
 
-                    cntr= (volatile uint64_t *) &(item_ptr->coll_buff->
+                    cntr= (volatile int64_t *) &(item_ptr->coll_buff->
                                 ctl_buffs_mgmt[index].bank_gen_counter);
                     item_ptr=(sm_nbbar_desc_t*)opal_list_remove_item((opal_list_t *)list,
                             ( opal_list_item_t *)item_ptr);

ompi/mca/bcol/basesmuma/bcol_basesmuma_rk_barrier.c

@@ -52,7 +52,7 @@ int bcol_basesmuma_k_nomial_barrier_init(bcol_function_args_t *input_args,
                 struct coll_ml_function_t *const_args)
 {
     /* local variables */
-    int flag_offset;
+    int flag_offset = 0;
     volatile int8_t ready_flag;
     mca_bcol_basesmuma_module_t *bcol_module = (mca_bcol_basesmuma_module_t *) const_args->bcol_module;
     mca_common_netpatterns_k_exchange_node_t *exchange_node = &bcol_module->knomial_allgather_tree;

ompi/mca/sbgp/basesmsocket/sbgp_basesmsocket_component.c

@@ -190,7 +190,7 @@ static int mca_sbgp_map_to_logical_socket_id(int *socket)
           */
           for( obj = first_pu_object; obj != NULL; obj = obj->next_cousin ) {/* WTF is a "next_cousin" ? */ 
               /* is this PU the same as the bit I pulled off the mask? */
-              if( obj->os_index == pu_os_index) {
+              if( obj->os_index == (unsigned int) pu_os_index) {
                   /* Then I found it, break out of for loop */
                   break;
               }
@@ -217,7 +217,14 @@ static int mca_sbgp_map_to_logical_socket_id(int *socket)
                         * Seems I'm bound to more than a single PU. Question
                         * is, am I bound to the same socket?? 
                         */
-                        if( this_pus_logical_socket_id != obj->logical_index ){
+                       /* in order to get rid of the compiler warning, I had to cast 
+                        * "this_pus_logical_socket_id", at a glance this seems ok, 
+                        * but if subgrouping problems arise, maybe look here. I shall 
+                        * tag this line with the "mark of the beast" for grepability
+                        * 666
+                        */  
+                        if( (unsigned int) this_pus_logical_socket_id != obj->logical_index ){
+                            /* 666 */
                             /* Then we're bound to more than one socket...fail */
                             this_pus_logical_socket_id = -1;
                             my_logical_socket_id = -1;