On Sat, 28 Mar 2026, Martin Uecker wrote:
>
> Bootstrapped and regression tested on x86_64.
LGTM, but please give Joseph the chance to chime in.
Richard.
>
> We get a checking error in verify_type, because I did not copy
> TYPE_REVERSE_STORAGE_ORDER over from the original when reconstructing
> derived types during the contruction of canonical types.
>
> PR c/124635
>
> gcc/c/ChangeLog:
> * c-typeck.cc (c_reconstruct_complex_type): Copy
> TYPE_REVERSE_STORAGE_ORDER.
>
> gcc/testsuite/ChangeLog:
> * gcc.dg/pr124635.c: New test.
>
> diff --git a/gcc/c/c-typeck.cc b/gcc/c/c-typeck.cc
> index 4452bb32728..043e49794ca 100644
> --- a/gcc/c/c-typeck.cc
> +++ b/gcc/c/c-typeck.cc
> @@ -548,8 +548,14 @@ c_reconstruct_complex_type (tree type, tree bottom)
> }
> else if (TREE_CODE (type) == ARRAY_TYPE)
> {
> + bool rso = TYPE_REVERSE_STORAGE_ORDER (type);
> inner = c_reconstruct_complex_type (TREE_TYPE (type), bottom);
> outer = c_build_array_type (inner, TYPE_DOMAIN (type));
> + if (rso)
> + {
> + outer = build_distinct_type_copy (outer);
> + TYPE_REVERSE_STORAGE_ORDER (outer) = 1;
> + }
>
> gcc_checking_assert (C_TYPE_VARIABLE_SIZE (type)
> == C_TYPE_VARIABLE_SIZE (outer));
> diff --git a/gcc/testsuite/gcc.dg/pr124635.c b/gcc/testsuite/gcc.dg/pr124635.c
> new file mode 100644
> index 00000000000..59883ba2365
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/pr124635.c
> @@ -0,0 +1,10 @@
> +/* { dg-do compile } */
> +/* { dg-options "-std=c23 -g" } */
> +
> +#pragma scalar_storage_order big-endian
> +
> +struct S {
> + struct S *s[1];
> + char c;
> +};
> +
>
--
Richard Biener <[email protected]>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Jochen Jaser, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)
