https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126017
--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #7)
> For the streaming itself it could be something like:
> --- a/gcc/lto-streamer-out.cc 2026-07-20 09:57:50.693934604 +0200
> +++ b/gcc/lto-streamer-out.cc 2026-09-14 18:51:29.227337823 +0200
> @@ -560,6 +560,12 @@ lto_is_streamable (tree expr)
> {
> enum tree_code code = TREE_CODE (expr);
>
> + /* Allow streaming C++ std::meta::info. */
> + if (code == LANG_TYPE
> + && TYPE_NAME (expr)
> + && TREE_CODE (TYPE_NAME (expr)) == IDENTIFIER_NODE
> + && id_equal (TYPE_NAME (expr), "decltype(^^int)"))
> + return true;
> /* Notice that we reject SSA_NAMEs as well. We only emit the SSA
> name version in lto_output_tree_ref (see output_ssa_names). */
> return !is_lang_specific (expr)
>
> I wish we haven't abused LANG_TYPE here and really used META_TYPE in the
> middle-end.
> Because having is_lang_specific return true for LANG_TYPE is natural.
Or the above can go into is_lang_specific and we can just nuke LANG_TYPE line
from lto_is_streamable because it is already duplicated there.
Doing it for arbitrary LANG_TYPE where we know nothing about it is dangerous.
