overrider: report a registry mismatch instead of "cannot find method"

[jll63]

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

Diagnostics only — nothing that compiled before stops compiling.

The problem

Since #96 a class can declare a registry of its own through boost_openmethod_registry. An overrider whose class resolves to a different registry than the method's therefore has parameters that simply do not convert — virtual_ptr<Poodle, kennel_registry> against virtual_ptr<Animal, zoo_registry>. The guide declared by BOOST_OPENMETHOD fails to match, and BOOST_OPENMETHOD_OVERRIDE reported only:

cannot find 'poke' method that accepts the same arguments as the overrider

which points nowhere near the cause. With this change:

static assertion failed: registry mismatch

and the instantiation trace names both registries — clang spells it std::is_same_v<zoo_registry, kennel_registry>.

How

BOOST_OPENMETHOD declares a second, relaxed guide beside the real one, <id>_guide_any_registry, driven by enable_guide_ignoring_registry: it rewrites every virtual_ptr parameter into the method's own registry (rebind_parameter_registry, plus its reference forms) before trying the call. An overrider that matches that guide and not the strict one differs from the method in nothing but a registry. The guide never finds a method to call; it exists only to be asked.

BOOST_OPENMETHOD_DETAIL_LOCATE_METHOD grows a rung under its existing fallback. The old generic static_assert moves into ..._explain_method_aux, whose specialization fires when the relaxed guide succeeds and inherits explain_overrider_mismatch. That pairs the method's parameters with the overrider's through validate_overrider_parameter — the same instantiation a call through the thunk would perform — so the diagnosis is the one the user would have got had the guide not failed first. The chain is now: strict guide → relaxed guide (registry mismatch) → "cannot find".

Tests

test/compile_fail_overrider_registry_mismatch.cpp, expecting registry mismatch.

compile_fail_override_method_not_found.cpp still produces the old generic message, deliberately: that is the proof a genuinely absent method is unaffected.

Verified on the tree as committed:

• gcc 13, Debug — 186/186
• g++-16, -DBOOST_OPENMETHOD_ENABLE_REFLECTION=ON -DBUILD_SHARED_LIBS=ON — 192/192
• compile-fail diagnostics also matched under clang 18

Note for anyone building this locally: the compile_fail_*.cpp glob has no CONFIGURE_DEPENDS, so the new file needs a manual cmake re-run before ctest sees it.

🤖 Generated with Claude Code

[cppalliance-bot]

An automated preview of the documentation is available at https://112.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.

2026-09-15 22:12:09 UTC

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.25%. Comparing base (d44aa59) to head (e89fe7c).
⚠️ Report is 3 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #112      +/-   ##
===========================================
- Coverage    93.51%   93.25%   -0.26%     
===========================================
  Files           22       22              
  Lines         1695     1706      +11     
  Branches       504      509       +5     
===========================================
+ Hits          1585     1591       +6     
- Misses          66       71       +5     
  Partials        44       44              

Files with missing lines	Coverage Δ
include/boost/openmethod/core.hpp	91.60% <ø> (-1.49%)	⬇️

---

Continue to review full report in Codecov by Harness.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8db9d0c...e89fe7c. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.