Hi,
When I was developing a patch 2 years ago, I ran across us
mis-compiling libcpp/expr.c's num_equality_op. I noticed today I
never added the testcase upstream for this mis-compiling. I noticed
because I was porting the patch I had developed to the trunk and I had
missed the fix and I was getting a bootstrap failure but there was no
testsuite failures that was testing the problem.
Committed after a quick test to make sure the testcase is still a
valid executable testcase now.
Thanks,
Andrew Pinski
* gcc.c-torture/execute/20141107-1.c: New testcase.
Index: gcc.c-torture/execute/20141107-1.c
===================================================================
--- gcc.c-torture/execute/20141107-1.c (revision 0)
+++ gcc.c-torture/execute/20141107-1.c (revision 217242)
@@ -0,0 +1,27 @@
+#define bool _Bool
+
+bool f(int a, bool c) __attribute__((noinline));
+bool f(int a, bool c)
+{
+ if (!a)
+ c = !c;
+ return c;
+}
+
+void checkf(int a, bool b)
+{
+ bool c = f(a, b);
+ char d;
+ __builtin_memcpy (&d, &c, 1);
+ if ( d != (a==0)^b)
+ __builtin_abort();
+}
+
+int main(void)
+{
+ checkf(0, 0);
+ checkf(0, 1);
+ checkf(1, 1);
+ checkf(1, 0);
+ return 0;
+}
Index: ChangeLog
===================================================================
--- ChangeLog (revision 217241)
+++ ChangeLog (revision 217242)
@@ -1,3 +1,7 @@
+2014-11-07 Andrew Pinski <apin...@cavium.com>
+
+ * gcc.c-torture/execute/20141107-1.c: New testcase.
+
2014-11-07 Richard Biener <rguent...@suse.de>
PR tree-optimization/63605
