interop: check the registry spelled on an any virtual parameter

[jll63]

(Written by Claude Code, on behalf of @jll63.)

Closes #123.

detail::validate_method_parameter for the any and Boost.TypeErasure interop headers took a ParamRegistry and never compared it with the method's. Every other virtual parameter shape rejects a mismatch; this one accepted it silently — and the parameter is not inert, because it decides the registry of a method that names none. Overriders registered against another registry then land in a different dispatch table, and the call reports no_overrider after initialize().

One missing route, not twelve missing checks

The specializations were unconditional std::true_type, which waved the registry check through as a side effect. They existed only because the generic rule did not count a virtual_traits::vptr as a way to reach a v-table: it knew two of the three routes method::vptr tries — the ADL boost_openmethod_vptr hook, and a class the rtti policy deems polymorphic — and a type-erased value has neither.

detail::has_vptr_source now names all three, in the order they are tried. The generic rule then covers the interop shapes and applies its registry check and its affinity recheck to them. The eleven blanket specializations are gone, from interop/std_any.hpp, interop/boost_any.hpp and interop/boost_type_erasure.hpp; the last keeps the one specialization that is a rejection rather than a blanket accept, an owning any<C, _self> passed by value.

The spelling matters. has_vptr<virtual_traits<T, R>, decltype(peek(declval<T>()))> is what method::vptr itself uses; the shorter const virtual_type& holds for ten of the eleven shapes and fails for any<C, _self&>, whose vptr takes the reference-bound any. A comment says so, because the shorter form is the obvious later "simplification".

virtual_any is unaffected: its specializations match a different shape, virtual_any<Any, Registry>&, and already compared.

Behaviour

parameter spelling R	decides an unannotated method's registry	checked against a method naming another
virtual_<Animal&, R> — ordinary carrier	yes	yes
virtual_<const std::any&, R> — interop	yes	yes, now
virtual_any<A, R>&	no, abstains	yes

An interop parameter is an ordinary carrier, so deciding the registry is the designed behaviour and stays. Only the missing check was a bug.

Tests

• compile_fail_std_any_registry_mismatch.cpp — the shape none of the five existing registry-mismatch tests used. The diagnostic is the ordinary "registry mismatch: the parameter belongs to another registry", confirmed on GCC and Clang.
• test_dispatch_std_any.cpp gains static_asserts for the three routes, and for the parameter still deciding the registry of a method that names none.
• The static_assert that names the routes gains the third; it keeps the substring compile_fail_non_polymorphic_virtual_parameter.cpp matches.
• CLAUDE.md's affinity table recorded this as an open bug, and is corrected.

Control: git stash push -- include/ brings the issue's reproducer back. It compiles on develop and is rejected here.

Verified locally:

Build	Result
CMake, clang-22	198/198
CMake, g++-16	198/198
b2, gcc (-Wextra -Werror)	128 passed, no warnings

Net 19 insertions against 56 deletions in include/.

🤖 Generated with Claude Code

[cppalliance-bot]

An automated preview of the documentation is available at https://130.openmethod.prtest3.cppalliance.org/libs/openmethod/doc/html/index.html

If more commits are pushed to the pull request, the docs will rebuild at the same URL.

2026-10-13 21:51:13 UTC