https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122785
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org,
| |mpolacek at gcc dot gnu.org,
| |nshead at gcc dot gnu.org
--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Now on the trunk.
I think we have multiple problems for -freflection vs. modules (even when I've
tried to export all the <meta> metafns in the std module).
One is this bug, another one is that for members_of sorting we rely on DECL_UID
order so that class members are ordered as required by the standard - in
TYPE_FIELDS that is not the case, some elements of that chain are forcibly
moved later. I'm worried that module streaming out and streaming back in will
not preserve that order.
And yet another one is that members_of on namespaces don't even try to handle
BINDING_VECTOR trees at all. I have no idea what those are even about, yet how
they should be handled for enumerating namespace members. Though, in that case
the standard doesn't require any particular sorting (but GCC sorts based on
DECL_UID anyway, because we don't want code generation to depend on hash table
traversal ordering).
