Re: [patch #PR 25933] memory leaks

[Uttam Pawar <uttamp at us dot ibm dot com>]

Thanks Ian for reviewing this patch and being patient with me. 

> Well, I was trying to suggest that you also move the sbitmap_alloc
> call until after the call to tpa_init.  Then you don't have to do
> anything special if tpa_init fails--you can just return NULL as the
> code currently does.
> 
> The patch is OK with that change.

Here is the updated complete patch.

------------------------------------------------------------------------------
2006-01-23  Uttam Pawar  <uttamp@us.ibm.com>

        PR middle-end/25933
        * loop-unroll.c (unroll_loop_runtime_iterations): Fix for
        a memory leak.
        * tree-ssa-live.c (type_var_init): Fix for a memory leak.

diff -urpN trunk.orig/gcc/loop-unroll.c trunk/gcc/loop-unroll.c
--- trunk.orig/gcc/loop-unroll.c        2006-01-18 10:03:20.000000000 -0800
+++ trunk/gcc/loop-unroll.c     2006-01-24 10:00:16.000000000 -0800
@@ -1168,6 +1168,9 @@ unroll_loop_runtime_iterations (struct l
             ";; Unrolled loop %d times, counting # of iterations "
             "in runtime, %i insns\n",
             max_unroll, num_loop_insns (loop));
+
+  if (dom_bbs)
+    free (dom_bbs);
 }

 /* Decide whether to simply peel LOOP and how much.  */
diff -urpN trunk.orig/gcc/tree-ssa-live.c trunk/gcc/tree-ssa-live.c
--- trunk.orig/gcc/tree-ssa-live.c      2006-01-18 10:03:20.000000000 -0800
+++ trunk/gcc/tree-ssa-live.c   2006-01-27 09:39:37.000000000 -0800
@@ -1078,13 +1078,13 @@ type_var_init (var_map map)
   tree t;
   sbitmap seen;

-  seen = sbitmap_alloc (num_partitions);
-  sbitmap_zero (seen);
-
   tv = tpa_init (map);
   if (!tv)
     return NULL;

+  seen = sbitmap_alloc (num_partitions);
+  sbitmap_zero (seen);
+
   for (x = num_partitions - 1; x >= 0; x--)
     {
       t = partition_to_var (map, x);