[BUG]: -Wmaybe-uninitialized warnings in O3 optimization

Required prerequisites

• Make sure you've read the documentation. Your issue may be addressed there.
• Search the issue tracker and Discussions to verify that this hasn't already been reported. +1 or comment there if it has.
• Consider asking first in the Gitter chat room or in a Discussion.

What version (or hash if on master) of pybind11 are you using?

3.0.4

Problem description

I get a lot of warnings, when compiling the example.cpp file found in the docs. I also put it below. Other than the warnings, it just compiles fine.
I am on Linux (fedora 44) using the pybind11 release 3.0.4
I first asked this in a Discussion, where i got asked to file a bug report. The Discussion is here.
I tried a few variations of the following command to compile:
gcc -O3 -Wall -shared -std=c++11 -fPIC $(python3-config --includes) -I "$PYBIND11_BASE/include" example.cpp -o "example$(python3-config --extension-suffix)"
I tried using -isystem instead of -I or using c++ or g++ instead of gcc, but the same warnings were always there.
The output i get from the compiler is the following:

In file included from /usr/include/c++/16/vector:69,
                 from /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/common.h:240,
                 from /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/attr.h:13,
                 from /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/class.h:12,
                 from /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h:12,
                 from example.cpp:4:
In member function ‘void std::_Bvector_base<_Alloc>::_M_deallocate() [with _Alloc = std::allocator<bool>]’,
    inlined from ‘std::_Bvector_base<_Alloc>::~_Bvector_base() [with _Alloc = std::allocator<bool>]’ at /usr/include/c++/16/bits/stl_bvector.h:676:28,
    inlined from ‘std::vector<bool, _Alloc>::~vector() [with _Alloc = std::allocator<bool>]’ at /usr/include/c++/16/bits/stl_bvector.h:926:37,
    inlined from ‘pybind11::detail::inline_array_or_vector<long unsigned int, 1, bool>::heap_vector::~heap_vector()’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/argument_vector.h:58:12,
    inlined from ‘pybind11::detail::inline_array_or_vector<ArrayT, InlineSize, VectorT>::~inline_array_or_vector() [with ArrayT = long unsigned int; long unsigned int InlineSize = 1; VectorT = bool]’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/argument_vector.h:75:33,
    inlined from ‘pybind11::detail::small_vector<bool, kRequestedInlineSize>::small_vector(std::size_t, bool) [with long unsigned int kRequestedInlineSize = 6]’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/argument_vector.h:241:5,
    inlined from ‘static PyObject* pybind11::cpp_function::dispatcher(PyObject*, PyObject* const*, pybind11::size_t, PyObject*)’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h:1203:87:
/usr/include/c++/16/bits/stl_bvector.h:701:30: warning: ‘*(std::_Bvector_base<std::allocator<bool> >*)((char*)&<unnamed> + offsetof(pybind11::detail::args_convert_vector<6>,pybind11::detail::small_vector<bool, 6>::m_repr) + 8).std::_Bvector_base<std::allocator<bool> >::_M_impl.std::_Bvector_base<std::allocator<bool> >::_Bvector_impl::std::_Bvector_base<std::allocator<bool> >::_Bvector_impl_data.std::_Bvector_base<std::allocator<bool> >::_Bvector_impl_data::_M_start.std::_Bit_iterator::std::_Bit_iterator_base.std::_Bit_iterator_base::_M_p’ may be used uninitialized [-Wmaybe-uninitialized]
  701 |         if (_M_impl._M_start._M_p)
      |             ~~~~~~~~~~~~~~~~~^~~~
/home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h: In static member function ‘static PyObject* pybind11::cpp_function::dispatcher(PyObject*, PyObject* const*, pybind11::size_t, PyObject*)’:
/home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h:1203:87: note: ‘<anonymous>’ declared here
 1203 |                         = args_convert_vector<arg_vector_small_size>(func.nargs, false);
      |                                                                                       ^
In function ‘std::ptrdiff_t std::operator-(const _Bit_iterator_base&, const _Bit_iterator_base&)’,
    inlined from ‘std::vector<bool, _Alloc>::size_type std::vector<bool, _Alloc>::size() const [with _Alloc = std::allocator<bool>]’ at /usr/include/c++/16/bits/stl_bvector.h:1128:32,
    inlined from ‘std::size_t pybind11::detail::small_vector<bool, kRequestedInlineSize>::size() const [with long unsigned int kRequestedInlineSize = 6]’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/argument_vector.h:247:39,
    inlined from ‘std::size_t pybind11::detail::small_vector<bool, kRequestedInlineSize>::size() const [with long unsigned int kRequestedInlineSize = 6]’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/argument_vector.h:243:17,
    inlined from ‘static PyObject* pybind11::cpp_function::dispatcher(PyObject*, PyObject* const*, pybind11::size_t, PyObject*)’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h:1145:77:
/usr/include/c++/16/bits/stl_bvector.h:302:44: warning: ‘((const std::_Bit_iterator*)((char*)&used_kwargs + offsetof(pybind11::detail::small_vector<bool, 6>,pybind11::detail::small_vector<bool, 6>::m_repr) + 8))[1].std::_Bit_iterator::std::_Bit_iterator_base.std::_Bit_iterator_base::_M_p’ may be used uninitialized [-Wmaybe-uninitialized]
  302 |       return (int(_S_word_bit) * (__x._M_p - __y._M_p)
      |                                   ~~~~~~~~~^~~~~~~~~~
/home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h: In static member function ‘static PyObject* pybind11::cpp_function::dispatcher(PyObject*, PyObject* const*, pybind11::size_t, PyObject*)’:
/home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h:1100:59: note: ‘((const std::_Bit_iterator*)((char*)&used_kwargs + offsetof(pybind11::detail::small_vector<bool, 6>,pybind11::detail::small_vector<bool, 6>::m_repr) + 8))[1].std::_Bit_iterator::std::_Bit_iterator_base.std::_Bit_iterator_base::_M_p’ was declared here
 1100 |                 small_vector<bool, arg_vector_small_size> used_kwargs(
      |                                                           ^~~~~~~~~~~
In function ‘std::ptrdiff_t std::operator-(const _Bit_iterator_base&, const _Bit_iterator_base&)’,
    inlined from ‘std::vector<bool, _Alloc>::size_type std::vector<bool, _Alloc>::size() const [with _Alloc = std::allocator<bool>]’ at /usr/include/c++/16/bits/stl_bvector.h:1128:32,
    inlined from ‘std::size_t pybind11::detail::small_vector<bool, kRequestedInlineSize>::size() const [with long unsigned int kRequestedInlineSize = 6]’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/argument_vector.h:247:39,
    inlined from ‘std::size_t pybind11::detail::small_vector<bool, kRequestedInlineSize>::size() const [with long unsigned int kRequestedInlineSize = 6]’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/detail/argument_vector.h:243:17,
    inlined from ‘static PyObject* pybind11::cpp_function::dispatcher(PyObject*, PyObject* const*, pybind11::size_t, PyObject*)’ at /home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h:1145:77:
/usr/include/c++/16/bits/stl_bvector.h:303:21: warning: ‘((const std::_Bit_iterator*)((char*)&used_kwargs + offsetof(pybind11::detail::small_vector<bool, 6>,pybind11::detail::small_vector<bool, 6>::m_repr) + 8))[1].std::_Bit_iterator::std::_Bit_iterator_base.std::_Bit_iterator_base::_M_offset’ may be used uninitialized [-Wmaybe-uninitialized]
  303 |               + __x._M_offset - __y._M_offset);
      |                 ~~~~^~~~~~~~~
/home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h: In static member function ‘static PyObject* pybind11::cpp_function::dispatcher(PyObject*, PyObject* const*, pybind11::size_t, PyObject*)’:
/home/sytoplis/pythonModuleTest/Dependencies/pybind11/pybind11-3.0.4/include/pybind11/pybind11.h:1100:59: note: ‘((const std::_Bit_iterator*)((char*)&used_kwargs + offsetof(pybind11::detail::small_vector<bool, 6>,pybind11::detail::small_vector<bool, 6>::m_repr) + 8))[1].std::_Bit_iterator::std::_Bit_iterator_base.std::_Bit_iterator_base::_M_offset’ was declared here
 1100 |                 small_vector<bool, arg_vector_small_size> used_kwargs(
      |                                                           ^~~~~~~~~~~

From the Discussion i read that this should have been working in 3.0.1, but I didnt test that myself.
In the Discussion there was also a possible solution hypothesized, but i know the codebase too little to judge if that would actually work.
My g++ version is g++ (GCC) 16.1.1 20260515 (Red Hat 16.1.1-2)

Reproducible example code

#include <pybind11/pybind11.h>

namespace py = pybind11;

int add(int i, int j) {
    return i + j;
}

PYBIND11_MODULE(example, m, py::mod_gil_not_used()) {
    m.doc() = "pybind11 example plugin"; // optional module docstring

    m.def("add", &add, "A function that adds two numbers");
}

Is this a regression? Put the last known working version here if it is.

3.0.1

Start with include/pybind11/detail/argument_vector.h and the dispatcher paths in include/pybind11/pybind11.h, then reproduce the report using the docs example.cpp and its GCC 16 -O3 command. Done means the example still compiles and works while the reported -Wmaybe-uninitialized warnings no longer appear.