On 5/10/2026 11:50 PM, Lino Hsing-Yu Peng wrote:
diff --git
a/gcc/testsuite/gcc.target/riscv/rvv/zvfofp8min/non-policy/non-overloaded/riscv_vector.h
b/gcc/testsuite/gcc.target/riscv/rvv/zvfofp8min/non-policy/non-overloaded/riscv_vector.h
new file mode 100644
index 00000000000..fbb4858fc86
--- /dev/null
+++
b/gcc/testsuite/gcc.target/riscv/rvv/zvfofp8min/non-policy/non-overloaded/riscv_vector.h
@@ -0,0 +1,11 @@
+/* Wrapper of riscv_vector.h, prevent riscv_vector.h including stdint.h from
+ C library, that might cause problem on testing RV32 related testcase when
+ we disable multilib. */
+#ifndef _RISCV_VECTOR_WRAP_H
+
+#define _GCC_WRAP_STDINT_H
+#include "stdint-gcc.h"
+#include_next <riscv_vector.h>
+#define _RISCV_VECTOR_WRAP_H
+
+#endif
This looks wrong to me. Why do we need it now when things have worked
just fine on rv32 without it for a few years at this point?
I note that these are only compile tests. Any plans to add execution
tests? They're not strictly necessary, but obviously runtime coverage
is a good thing.
jeff
