On October 5, 2018 2:48:24 PM GMT+02:00, Richard Sandiford
<richard.sandif...@arm.com> wrote:
>We've traditionally allowed vectors of enums (not sure if that's
>deliberate) but vector_types_compatible_elements_p checked for
>INTEGER_TYPE rather than INTEGRAL_TYPE_P.
>
>Tested on aarch64-linux-gnu. OK to install?
OK.
Richard.
>Richard
>
>
>2018-10-05 Richard Sandiford <richard.sandif...@arm.com>
>
>gcc/c-family/
> PR c/87286
> * c-common.c (vector_types_compatible_elements_p): Use
> INTEGRAL_TYPE_P instead of checking only for INTEGER_TYPE.
>
>gcc/testsuite/
> PR c/87286
> * gcc.dg/pr87286.c: New test.
>
>Index: gcc/c-family/c-common.c
>===================================================================
>--- gcc/c-family/c-common.c 2018-10-05 13:46:08.287811117 +0100
>+++ gcc/c-family/c-common.c 2018-10-05 13:47:08.291325001 +0100
>@@ -7465,8 +7465,11 @@ vector_types_compatible_elements_p (tree
>
> enum tree_code c1 = TREE_CODE (t1), c2 = TREE_CODE (t2);
>
>- gcc_assert ((c1 == INTEGER_TYPE || c1 == REAL_TYPE || c1 ==
>FIXED_POINT_TYPE)
>- && (c2 == INTEGER_TYPE || c2 == REAL_TYPE
>+ gcc_assert ((INTEGRAL_TYPE_P (t1)
>+ || c1 == REAL_TYPE
>+ || c1 == FIXED_POINT_TYPE)
>+ && (INTEGRAL_TYPE_P (t2)
>+ || c2 == REAL_TYPE
> || c2 == FIXED_POINT_TYPE));
>
> t1 = c_common_signed_type (t1);
>@@ -7476,7 +7479,7 @@ vector_types_compatible_elements_p (tree
> if (t1 == t2)
> return true;
> if (opaque && c1 == c2
>- && (c1 == INTEGER_TYPE || c1 == REAL_TYPE)
>+ && (INTEGRAL_TYPE_P (t1) || c1 == REAL_TYPE)
> && TYPE_PRECISION (t1) == TYPE_PRECISION (t2))
> return true;
> return false;
>Index: gcc/testsuite/gcc.dg/pr87286.c
>===================================================================
>--- /dev/null 2018-09-14 11:16:31.122530289 +0100
>+++ gcc/testsuite/gcc.dg/pr87286.c 2018-10-05 13:47:08.291325001 +0100
>@@ -0,0 +1,3 @@
>+enum foo { F };
>+typedef enum foo vec_foo __attribute__((vector_size (16)));
>+vec_foo add (vec_foo x, vec_foo y) { return x + y; }
