Hi!
This wrong-code issue has been fixed with r15-7249.
We still emit warnings which are questionable and perhaps we'd
get better generated code if niters determined the loop has only a single
iteration without UB and we'd punt on vectorizing it (or unrolling).
Tested on x86_64-linux -m32/-m64, committed to trunk as obvious.
2025-01-31 Jakub Jelinek <ja...@redhat.com>
PR middle-end/117498
* gcc.c-torture/execute/pr117498.c: New test.
--- gcc/testsuite/gcc.c-torture/execute/pr117498.c.jj
+++ gcc/testsuite/gcc.c-torture/execute/pr117498.c
@@ -0,0 +1,35 @@
+/* PR middle-end/117498 */
+
+int a, d, f;
+char g;
+volatile int c = 1;
+
+int
+foo ()
+{
+ if (c == 0)
+ return -1;
+ return 1;
+}
+
+void
+bar (int h, int i, char *k, char *m)
+{
+ for (; d < i; d += 2)
+ for (int j = 0; j < h; j++)
+ m[j] = k[4 * j];
+}
+
+void
+baz (long h)
+{
+ char n = 0;
+ bar (h, 4, &n, &g);
+}
+
+int
+main ()
+{
+ f = foo ();
+ baz ((unsigned char) f - 4);
+}
Jakub
