Hi!
This is something I didn't notice in my testing because my ld doesn't
support plugins and thus -flto didn't act as -fwhole-file.
The bug got fixed on the trunk by changing build_int_cst, which is
too risky to backport, so this patch instead just
adjusts the single place relevant to this patch to use build_int_cst_type
instead. Although it is very likely type is always non-NULL, I haven't
proved it and I'm trying to play safe on the release branch and thus kept
build_int_cst in the case type would be NULL. The testcase doesn't use
-flto, instead if makes the vars static which results in the same thing
as -fwhole-file or -flto with linker plugin.
Bootstrapped/regtested on the 4.6 branch on x86_64-linux and i686-linux,
ok for 4.6 and the testcase also for trunk?
2011-05-23 Jakub Jelinek <ja...@redhat.com>
PR lto/49123
* fold-const.c (constant_boolean_node): If type is non-NULL,
use build_int_cst_type instead of build_int_cst.
* gcc.c-torture/execute/pr49123.c: New test.
--- gcc/fold-const.c.jj 2011-05-04 10:46:52.000000000 +0200
+++ gcc/fold-const.c 2011-05-23 16:19:13.000000000 +0200
@@ -5953,8 +5953,10 @@ constant_boolean_node (int value, tree t
return value ? integer_one_node : integer_zero_node;
else if (type == boolean_type_node)
return value ? boolean_true_node : boolean_false_node;
+ else if (type)
+ return build_int_cst_type (type, value);
else
- return build_int_cst (type, value);
+ return build_int_cst (NULL_TREE, value);
}
--- gcc/testsuite/gcc.c-torture/execute/pr49123.c.jj 2011-05-23
16:15:25.000000000 +0200
+++ gcc/testsuite/gcc.c-torture/execute/pr49123.c 2011-05-23
16:12:59.000000000 +0200
@@ -0,0 +1,14 @@
+/* PR lto/49123 */
+
+extern void abort (void);
+static struct S { int f : 1; } s;
+static int v = -1;
+
+int
+main ()
+{
+ s.f = v < 0;
+ if ((unsigned int) s.f != -1U)
+ abort ();
+ return 0;
+}
Jakub
