[AD] Type-pun warnings with GCC 4.1

[Chris <chris.kcat@xxxxxxxxxx>]

When building the latest SVN from SF using GCC 4.1.1, I get some type-pun 
warnings when compiling dat.c and grabber.c (which would invariably cause 
failure with --enable-strictwarn). The attached patch fixes dat.c, but 
grabber.c may prove to be a little trickier. For reference, the lines are 
516, 1743, and 3271. The first of which is below:

static void get_selection_info(DATAFILE **dat, DATAFILE ***parent)
{
   if ((SELECTED_ITEM <= 0) || (SELECTED_ITEM >= data_count)) {
      *dat = NULL;
      *parent = &datafile;
   }
   else {
      *dat = data[SELECTED_ITEM].dat;
      if ((*dat)->type == DAT_FILE)
     *parent = (DATAFILE **)&(*dat)->dat; // error here!
      else
     *parent = data[SELECTED_ITEM].parent;
   }
}

Unfortunately it's not so easy to fix, because you can't put it into a temp 
DATAFILE* var and pass the address to that since the function returns, making 
the pointer invalid. As well, putting the pointer in a DATAFILE** var would 
probably cause the same warning message. Ideas?

Index: tools/dat.c
===================================================================
--- tools/dat.c	(revision 7316)
+++ tools/dat.c	(working copy)
@@ -432,8 +432,10 @@
       }
       else {
 	 if ((*dat)[i].type == DAT_FILE) {
+	    DATAFILE *dattmp = (*dat)[i].dat;
 	    strcat(tmp, "/");
-	    do_delete((DATAFILE **)(&((*dat)[i].dat)), tmp);
+	    do_delete(&dattmp, tmp);
+	    (*dat)[i].dat = dattmp;
 	 }
       }
    }
@@ -681,8 +683,10 @@
       }
       else {
 	 if ((*dat)[i].type == DAT_FILE) {
+	    DATAFILE *dattmp = (*dat)[i].dat;
 	    strcat(tmp, "/");
-	    do_setpal((DATAFILE **)(&((*dat)[i].dat)), tmp);
+	    do_setpal(&dattmp, tmp);
+	    (*dat)[i].dat = dattmp;
 	 }
       }
    }