itlib's CMakeList.txt requires CMake 3.10, however I think it does not use any newer commands and thus would be compatible with older versions.

https://github.com/iboB/itlib/blob/1dc9abf235c64c02c7d0482ef0c81e7c3b759ab8/CMakeLists.txt#L1

shrink_at () returns wrong iterator

T* shrink_at(const T* cp, size_type num) { const auto s = size(); if (s == num) { clear(); return m_end; }

    auto position = const_cast<T*>(cp);

    std::memmove(position, position + num, size_t(m_end - position - num) * sizeof(T));

    m_end -= num;

    return ++position
}

Should be:

T* shrink_at(const T* cp, size_type num) { const auto s = size(); if (s == num) { clear(); return m_end; }

    auto position = const_cast<T*>(cp);

    std::memmove(position, position + num, size_t(m_end - position - num) * sizeof(T));

    m_end -= num;

    return position;
}

https://github.com/iboB/itlib/blob/master/include/itlib/make_ptr.hpp#L59 https://github.com/iboB/itlib/blob/master/include/itlib/make_ptr.hpp#L60

Typo make_shaerd => make_shared

This works with msvc, but with clang and gcc it produces:

[build] itlib/include/itlib/ufunction.hpp:106:63: error: invalid use of ‘const_cast’ with type ‘int (&)(int, int)’, which is a pointer or reference to a function type
[build]   106 |         copy_wrapper(const copy_wrapper& other) : func_object(const_cast<FO&&>(other.func_object)) { throw 0; } // should never get to here
[build]       |                                                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Apparently the const_cast fails. However we need this const cast for the fake copy constructor.

Potential solutions would be:

1. Instead of a const cast add a function which works with const_cast for non-free functions and just copies free functions
2. Devise some other way of obtaining a FO instance... something like a code-level declval

The key is that this only needs to compile. It's not important to do something which does sensible stuff when executed.

Not sure how to approach this yet.

Corner case in resize if going from dynamic memory to static , causing corrupt memory.

Reproduce: itlib::small_vector<int, 4,3> v(5); v.resize(2);

but just checking new_buf in resize() , I think is enough:

if (m_begin != static_begin_ptr()) { // we've moved from dyn to dyn memory, so deallocate the old one if (new_buf != static_begin_ptr()) atraits::deallocate(get_alloc(), m_begin, m_capacity);

 }

Currently pod_vector doesn't handle the alignment of T at all. This is bad (though due to the fact that malloc returns aligned pages, it's mitigated, at least until one tries to create a custom allocator)

The question here is should pod_vector handle the alignment (by modifying and tweaking the memory returned by malloc and realloc) or should the allocator handle it by providing an additional argument - alignment - to these functions