[svn-r24479] Description:

    Fix ph5diff worker command handling so it can correctly respond to
situations where there's no work to perform.  Also, a few minor code cleanups.

Tested on:
    Mac OSX/64 10.9.0 (amazon) w/parallel

3 files changed, 153 insertions, 148 deletions

diff --git a/tools/h5diff/ph5diff_main.c b/tools/h5diff/ph5diff_main.c
index 1fc563b..62e1e73 100644
--- a/tools/h5diff/ph5diff_main.c
+++ b/tools/h5diff/ph5diff_main.c
@@ -13,11 +13,12 @@
  * access to either file, you may request a copy from help@hdfgroup.org.     *
  * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
 
-#include "h5diff.h"
-#include "ph5diff.h"
 #include <stdlib.h>
 #include <string.h>
 #include <assert.h>
+#include "H5private.h"
+#include "h5diff.h"
+#include "ph5diff.h"
 #include "h5diff_common.h"
 #include "h5tools.h"
 #include "h5tools_utils.h"
@@ -30,9 +31,9 @@ static void ph5diff_worker(int );
 /*-------------------------------------------------------------------------
  * Function: main
  *
- * Purpose: h5diff/ph5diff main program
+ * Purpose: ph5diff main program
  *
- * Return: An  exit status of 0 means no differences were found, 1 means some
+ * Return: An exit status of 0 means no differences were found, 1 means some
  *   differences were found.
  *
  * Programmer: Pedro Vicente, pvn@ncsa.uiuc.edu
@@ -41,16 +42,6 @@ static void ph5diff_worker(int );
  *
  * Comments:
  *
- * Modifications: July 2004
- *  Introduced the four modes:
- *   Normal mode: print the number of differences found and where they occured
- *   Report mode: print the above plus the differences
- *   Verbose mode: print the above plus a list of objects and warnings
- *   Quiet mode: do not print output
- *
- * November 2004: Leon Arber (larber@uiuc.edu)
- *        Additions that allow h5diff to be run in parallel
- *
  * This function drives the diff process and will do a serial or parallel diff depending
  * on the value of the global variable g_Parallel (default is 0), set to 1 when the program
  * is run as "ph5diff"
@@ -140,112 +131,165 @@ int main(int argc, const char *argv[])
 static void
 ph5diff_worker(int nID)
 {
-    struct diff_mpi_args args;
-    hid_t file1_id, file2_id;
-    char    filenames[2][MAX_FILENAME];
-    char    out_data[PRINT_DATA_MAX_SIZE] = {0};
-    struct diffs_found  diffs;
-    int i;
-    MPI_Status Status;
-
-    outBuffOffset = 0;
+    hid_t file1_id = -1, file2_id = -1;
 
-    MPI_Recv(filenames, MAX_FILENAME*2, MPI_CHAR, 0, MPI_ANY_TAG, MPI_COMM_WORLD, &Status);
-    if(Status.MPI_TAG == MPI_TAG_PARALLEL)
+    while(1)
     {
-        /* disable error reporting */
-        H5E_BEGIN_TRY
+        MPI_Status Status;
+
+        MPI_Probe(0, MPI_ANY_TAG, MPI_COMM_WORLD, &Status);
+
+        /* Check for filenames */
+        if(Status.MPI_TAG == MPI_TAG_PARALLEL)
         {
-            /* Open the files */
-            if ((file1_id = H5Fopen (filenames[0], H5F_ACC_RDONLY, H5P_DEFAULT)) < 0)
+            char    filenames[2][MAX_FILENAME];
+
+            /* Retrieve filenames */
+            MPI_Recv(filenames, MAX_FILENAME*2, MPI_CHAR, 0, MPI_ANY_TAG, MPI_COMM_WORLD, &Status);
+
+            /* disable error reporting */
+            H5E_BEGIN_TRY
             {
-            printf ("h5diff Task [%d]: <%s>: unable to open file\n", nID, filenames[0]);
-            MPI_Abort(MPI_COMM_WORLD, 0);
+                /* Open the files */
+                if ((file1_id = H5Fopen (filenames[0], H5F_ACC_RDONLY, H5P_DEFAULT)) < 0)
+                {
+                    printf ("h5diff Task [%d]: <%s>: unable to open file\n", nID, filenames[0]);
+                    MPI_Abort(MPI_COMM_WORLD, 0);
+                }
+                if ((file2_id = H5Fopen (filenames[1], H5F_ACC_RDONLY, H5P_DEFAULT)) < 0)
+                {
+                    printf ("h5diff Task [%d]: <%s>: unable to open file\n", nID, filenames[1]);
+                    MPI_Abort(MPI_COMM_WORLD, 0);
+                }
+                /* enable error reporting */
             }
-            if ((file2_id = H5Fopen (filenames[1], H5F_ACC_RDONLY, H5P_DEFAULT)) < 0)
+            H5E_END_TRY;
+        }
+        /* Check for work */
+        else if(Status.MPI_TAG == MPI_TAG_ARGS)
+        {
+            struct diff_mpi_args args;
+            struct diffs_found  diffs;
+            int i;
+
+            /* Make certain we've received the filenames and opened the files already */
+            if(file1_id < 0 || file2_id < 0)
             {
-            printf ("h5diff Task [%d]: <%s>: unable to open file\n", nID, filenames[1]);
-            MPI_Abort(MPI_COMM_WORLD, 0);
+                printf("ph5diff_worker: ERROR: work received before/without filenames\n");
+                break;
             }
-            /* enable error reporting */
-        }
-        H5E_END_TRY;
 
+            /* Recv parameters for diff from manager task */
+            MPI_Recv(&args, sizeof(args), MPI_BYTE, 0, MPI_TAG_ARGS, MPI_COMM_WORLD, &Status);
 
-        while(1)
-        {
-            MPI_Probe(0, MPI_ANY_TAG, MPI_COMM_WORLD, &Status);
+            /* Do the diff */
+            diffs.nfound = diff(file1_id, args.name1, file2_id, args.name2, &(args.options), &(args.argdata));
+            diffs.not_cmp = args.options.not_cmp;
 
-            if(Status.MPI_TAG == MPI_TAG_ARGS)
+            /* If print buffer has something in it, request print token.*/
+            if(outBuffOffset>0)
             {
-                /*Recv parameters for diff from manager task */
-                MPI_Recv(&args, sizeof(args), MPI_BYTE, 0, MPI_TAG_ARGS, MPI_COMM_WORLD, &Status);
-                /*Do the diff */
-                diffs.nfound = diff(file1_id, args.name1, file2_id, args.name2, &(args.options), &(args.argdata));
-                diffs.not_cmp = args.options.not_cmp;
-
-                /*If print buffer has something in it, request print token.*/
-                if(outBuffOffset>0)
-                {
-                    MPI_Send(NULL, 0, MPI_BYTE, 0, MPI_TAG_TOK_REQUEST, MPI_COMM_WORLD);
-                    /*Wait for print token. */
-                    MPI_Recv(NULL, 0, MPI_BYTE, 0, MPI_TAG_PRINT_TOK, MPI_COMM_WORLD, &Status);
+                MPI_Send(NULL, 0, MPI_BYTE, 0, MPI_TAG_TOK_REQUEST, MPI_COMM_WORLD);
 
-                    /*When get token, send all of our output to the manager task and then return the token */
-                    for(i=0; i<outBuffOffset; i+=PRINT_DATA_MAX_SIZE)
-                        MPI_Send(outBuff+i, PRINT_DATA_MAX_SIZE, MPI_BYTE, 0, MPI_TAG_PRINT_DATA, MPI_COMM_WORLD);
+                /* Wait for print token. */
+                MPI_Recv(NULL, 0, MPI_BYTE, 0, MPI_TAG_PRINT_TOK, MPI_COMM_WORLD, &Status);
 
+                /* When get token, send all of our output to the manager task and then return the token */
+                for(i=0; i<outBuffOffset; i+=PRINT_DATA_MAX_SIZE)
+                    MPI_Send(outBuff+i, PRINT_DATA_MAX_SIZE, MPI_BYTE, 0, MPI_TAG_PRINT_DATA, MPI_COMM_WORLD);
 
-                    /* An overflow file exists, so we send it's output to the manager too and then delete it */
-                    if(overflow_file)
-                    {
-                        int tmp;
+                /* An overflow file exists, so we send it's output to the manager too and then delete it */
+                if(overflow_file)
+                {
+                    char out_data[PRINT_DATA_MAX_SIZE];
+                    int tmp;
 
-                        memset(out_data, 0, PRINT_DATA_MAX_SIZE);
-                        i=0;
+                    memset(out_data, 0, PRINT_DATA_MAX_SIZE);
+                    i=0;
 
-                        rewind(overflow_file);
-                        while((tmp = getc(overflow_file)) >= 0)
+                    rewind(overflow_file);
+                    while((tmp = getc(overflow_file)) >= 0)
+                    {
+                        *(out_data + i++) = (char)tmp;
+                        if(i==PRINT_DATA_MAX_SIZE)
                         {
-                            *(out_data + i++) = (char)tmp;
-                            if(i==PRINT_DATA_MAX_SIZE)
-                            {
-                                MPI_Send(out_data, PRINT_DATA_MAX_SIZE, MPI_BYTE, 0, MPI_TAG_PRINT_DATA, MPI_COMM_WORLD);
-                                i=0;
-                                memset(out_data, 0, PRINT_DATA_MAX_SIZE);
-                            }
-                        }
-
-                        if(i>0)
                             MPI_Send(out_data, PRINT_DATA_MAX_SIZE, MPI_BYTE, 0, MPI_TAG_PRINT_DATA, MPI_COMM_WORLD);
-
-                        fclose(overflow_file);
-                        overflow_file = NULL;
+                            i=0;
+                            memset(out_data, 0, PRINT_DATA_MAX_SIZE);
+                        }
                     }
 
-                    fflush(stdout);
-                    memset(outBuff, 0, OUTBUFF_SIZE);
-                    outBuffOffset = 0;
+                    if(i>0)
+                        MPI_Send(out_data, PRINT_DATA_MAX_SIZE, MPI_BYTE, 0, MPI_TAG_PRINT_DATA, MPI_COMM_WORLD);
 
-                    MPI_Send(&diffs, sizeof(diffs), MPI_BYTE, 0, MPI_TAG_TOK_RETURN, MPI_COMM_WORLD);
+                    fclose(overflow_file);
+                    overflow_file = NULL;
                 }
-                else
-                    MPI_Send(&diffs, sizeof(diffs), MPI_BYTE, 0, MPI_TAG_DONE, MPI_COMM_WORLD);
-            }
-            else if(Status.MPI_TAG == MPI_TAG_END)
-            {
-                MPI_Recv(NULL, 0, MPI_BYTE, 0, MPI_TAG_END, MPI_COMM_WORLD, &Status);
-            /*  printf("exiting..., task: %d\n", nID);
-                fflush(stdout);*/
-                break;
+
+                fflush(stdout);
+                memset(outBuff, 0, OUTBUFF_SIZE);
+                outBuffOffset = 0;
+
+                MPI_Send(&diffs, sizeof(diffs), MPI_BYTE, 0, MPI_TAG_TOK_RETURN, MPI_COMM_WORLD);
             }
             else
-            {
-                printf("ph5diff_worker: ERROR: invalid tag (%d) received\n", Status.MPI_TAG);
-                MPI_Abort(MPI_COMM_WORLD, 0);
-            }
+                MPI_Send(&diffs, sizeof(diffs), MPI_BYTE, 0, MPI_TAG_DONE, MPI_COMM_WORLD);
+        }
+        /* Check for leaving */
+        else if(Status.MPI_TAG == MPI_TAG_END)
+        {
+            MPI_Recv(NULL, 0, MPI_BYTE, 0, MPI_TAG_END, MPI_COMM_WORLD, &Status);
+            break;
+        }
+        else
+        {
+            printf("ph5diff_worker: ERROR: invalid tag (%d) received\n", Status.MPI_TAG);
+            break;
+        }
 
+    }
+
+    return;
+}
+
+/*-------------------------------------------------------------------------
+ * Function: print_manager_output
+ *
+ * Purpose: special function that prints any output accumulated by the
+ *      manager task.
+ *
+ * Return: none
+ *
+ * Programmer: Leon Arber
+ *
+ * Date: Feb 7, 2005
+ *
+ *-------------------------------------------------------------------------
+ */
+void print_manager_output(void)
+{
+    /* If there was something we buffered, let's print it now */
+    if( (outBuffOffset>0) && g_Parallel)
+    {
+        printf("%s", outBuff);
+
+        if(overflow_file)
+        {
+            int     tmp;
+            rewind(overflow_file);
+            while((tmp = getc(overflow_file)) >= 0)
+                putchar(tmp);
+            fclose(overflow_file);
+            overflow_file = NULL;
         }
+
+        HDfflush(stdout);
+        HDmemset(outBuff, 0, OUTBUFF_SIZE);
+        outBuffOffset = 0;
+    }
+    else if( (outBuffOffset>0) && !g_Parallel)
+    {
+        HDfprintf(stderr, "h5diff error: outBuffOffset>0, but we're not in parallel!\n");
     }
 }
 
@@ -268,12 +312,14 @@ ph5diff_worker(int nID)
 void h5diff_exit(int status)
 {
     /* if in parallel mode, dismiss workers, close down MPI, then exit */
-    if((g_nTasks > 1) && g_Parallel) {
-        phdiff_dismiss_workers();
-        MPI_Barrier(MPI_COMM_WORLD);
-    }
-    if(g_Parallel)
+    if(g_Parallel) {
+        if(g_nTasks > 1) {
+            phdiff_dismiss_workers();
+            MPI_Barrier(MPI_COMM_WORLD);
+        }
         MPI_Finalize();
+        status = EXIT_SUCCESS;  /* Reset exit status, since some mpiexec commands generate output on failure status */
+    }
 
     exit(status);
 }
diff --git a/tools/lib/h5diff.c b/tools/lib/h5diff.c
index 73bbd07..4ff3440 100644
--- a/tools/lib/h5diff.c
+++ b/tools/lib/h5diff.c
@@ -106,47 +106,6 @@ void phdiff_dismiss_workers(void)
 
 
 /*-------------------------------------------------------------------------
- * Function: print_manager_output
- *
- * Purpose: special function that prints any output accumulated by the
- *      manager task.
- *
- * Return: none
- *
- * Programmer: Leon Arber
- *
- * Date: Feb 7, 2005
- *
- *-------------------------------------------------------------------------
- */
-void print_manager_output(void)
-{
-    /* If there was something we buffered, let's print it now */
-    if( (outBuffOffset>0) && g_Parallel)
-    {
-        printf("%s", outBuff);
-
-        if(overflow_file)
-        {
-            int     tmp;
-            rewind(overflow_file);
-            while((tmp = getc(overflow_file)) >= 0)
-                putchar(tmp);
-            fclose(overflow_file);
-            overflow_file = NULL;
-        }
-
-        HDfflush(stdout);
-        HDmemset(outBuff, 0, OUTBUFF_SIZE);
-        outBuffOffset = 0;
-    }
-    else if( (outBuffOffset>0) && !g_Parallel)
-    {
-        HDfprintf(stderr, "h5diff error: outBuffOffset>0, but we're not in parallel!\n");
-    }
-}
-
-/*-------------------------------------------------------------------------
  * Function: print_incoming_data
  *
  * Purpose: special function that prints any output that has been sent to the manager
diff --git a/tools/lib/ph5diff.h b/tools/lib/ph5diff.h
index becbd6c..2a75228 100644
--- a/tools/lib/ph5diff.h
+++ b/tools/lib/ph5diff.h
@@ -17,18 +17,18 @@
 #define _PH5DIFF_H__
 
 /* Send from manager to workers */
-#define MPI_TAG_ARGS    1
-#define MPI_TAG_PRINT_TOK  2
+#define MPI_TAG_ARGS            1
+#define MPI_TAG_PRINT_TOK       2
 
 /*Sent from workers to manager */
-#define MPI_TAG_TOK_REQUEST  3
-#define MPI_TAG_DONE    4
-#define MPI_TAG_TOK_RETURN  5
-#define MPI_TAG_PRINT_DATA  6
+#define MPI_TAG_TOK_REQUEST     3
+#define MPI_TAG_DONE            4
+#define MPI_TAG_TOK_RETURN      5
+#define MPI_TAG_PRINT_DATA      6
 
 /* Operational tags used to init and complete diff */
-#define MPI_TAG_END    7
-#define MPI_TAG_PARALLEL  8
+#define MPI_TAG_END             7
+#define MPI_TAG_PARALLEL        8
 
 struct diff_mpi_args
 {