ci: fix gcc compiler warnings

[rhaschke]

Description

Removed redundant declarations causing the following warnings (including pybind11.h):

In file included from pybind11/include/pybind11/cast.h:15,
                 from pybind11/include/pybind11/attr.h:14,
                 from pybind11/include/pybind11/detail/class.h:12,
                 from pybind11/include/pybind11/pybind11.h:12,
                 from include/py_binding_tools/ros_msg_typecasters.h:39,
                 from src/ros_msg_typecasters.cpp:37:
pybind11/include/pybind11/detail/type_caster_base.h:521:13: warning: redundant redeclaration of ‘bool pybind11::detail::deregister_instance(pybind11::detail::instance*, void*, const pybind11::detail::type_info*)’ in same scope [-Wredundant-decls]
  521 | inline bool deregister_instance(instance *self, void *valptr, const type_info *tinfo);
      |             ^~~~~~~~~~~~~~~~~~~
In file included from pybind11/include/pybind11/detail/type_caster_base.h:14,
                 from pybind11/include/pybind11/cast.h:15,
                 from pybind11/include/pybind11/attr.h:14,
                 from pybind11/include/pybind11/detail/class.h:12,
                 from pybind11/include/pybind11/pybind11.h:12,
                 from include/py_binding_tools/ros_msg_typecasters.h:39,
                 from src/ros_msg_typecasters.cpp:37:
pybind11/include/pybind11/trampoline_self_life_support.h:19:13: note: previous declaration of ‘bool pybind11::detail::deregister_instance(pybind11::detail::instance*, void*, const pybind11::detail::type_info*)’
   19 | inline bool deregister_instance(instance *self, void *valptr, const type_info *tinfo);
      |             ^~~~~~~~~~~~~~~~~~~
In file included from pybind11/include/pybind11/pybind11.h:12,
                 from include/py_binding_tools/ros_msg_typecasters.h:39,
                 from src/ros_msg_typecasters.cpp:37:
pybind11/include/pybind11/detail/class.h:502:13: warning: redundant redeclaration of ‘std::string pybind11::detail::error_string()’ in same scope [-Wredundant-decls]
  502 | std::string error_string();
      |             ^~~~~~~~~~~~
In file included from pybind11/include/pybind11/detail/internals.h:19,
                 from pybind11/include/pybind11/gil.h:17,
                 from pybind11/include/pybind11/detail/type_caster_base.h:12,
                 from pybind11/include/pybind11/cast.h:15,
                 from pybind11/include/pybind11/attr.h:14,
                 from pybind11/include/pybind11/detail/class.h:12,
                 from pybind11/include/pybind11/pybind11.h:12,
                 from include/py_binding_tools/ros_msg_typecasters.h:39,
                 from src/ros_msg_typecasters.cpp:37:
pybind11/include/pybind11/pytypes.h:743:20: note: previous declaration of ‘std::string pybind11::detail::error_string()’
  743 | inline std::string error_string() {
      |                    ^~~~~~~~~~~~

Suggested changelog entry:

CI testing now includes ``-Wwrite-strings -Wunreachable-code -Wpointer-arith -Wredundant-decls`` in some jobs.

[rhaschke]

This affects master branch too, not only smart_holder.

[rhaschke]

Looks like some tests sparsely use includes. Will check later.

[rhaschke]

Let's start over and push commits one-by-one.

[rhaschke]

The pypy-3.8 build seems to be flaky.

[rwgk]

The pypy-3.8 build seems to be flaky.

Yes, very often.

[henryiii]

We need to move to pypy-3.10. Older pypy's aren't supported by upstream.

[rwgk]

Sorry I think this is going in the wrong direction:

It brings in a ton of dependencies unrelated to the gil_safe_call_once functionality.

It's seems counter-productive to turn on compiler warnings just to appease them.

A great step forward would be to refactor the code organization. I'd be very happy if you did that, even if several pieces of code need to be moved around.

If you think that's too much work: Simply disable the warnings for the troublesome existing forward declaration. That'd be a net gain: Having the warnings enabled will nudge us in the right direction when working on new code. The warning suppressions will remind us that we have a code organization cleanup job waiting for an intrepid volunteer.

[rhaschke]

Fair enough. For restructuring the code, I'm not an expert enough. I will disable the warnings. Switching to draft state for now.

[rhaschke]

This should be ready for review. The failing test seems to be unrelated to my changes.

[rwgk]

Nice, thanks!

Waiting a bit for @henryiii to take a look.

[rwgk]

Thanks for catching this!

[rwgk]

@rhaschke I made a quick change to the Suggested changelog entry. Could you please review, and maybe make it more precise and correct?