Next lib version #3
Conversation
| constexpr void operator()(object_t & aObject, acceptor_t & aAcceptor) const { | ||
| (*m_invokerPtr)(aObject, m_tag, aAcceptor); | ||
| } | ||
| constexpr void operator()(object_t* aObject, acceptor_t & aAcceptor) const { |
There was a problem hiding this comment.
Why don't just call the 1st overload?
Like
constexpr void operator()(object_t * aObject, acceptor_t & aAcceptor) const {
assert(aObject && "No object");
(*this)(*aObject, aAcceptor);
}There was a problem hiding this comment.
Some time we get ref, some time get ref to pointer. So overload is simplest solve problem. In another cases necessary change code in many places as i think. maybe i'm wrong, but i have question. You try checkout last version and use your suggestion? Code complile and work?
There was a problem hiding this comment.
Hello sorry for this delay:)
The ref to a pointer has another meaning, ref to a pointer means that you'd like to "sync" changes to the address stored inside the pointer inside and outside the function/unit.
Here we're interesing only in accessing object's method.
And having an overload, which accepts pointer, means that we MUST think of it, since pointers may be NULL. So you MUST write "if (p != nullptr)" every time to avoid UB.
From these points it's safer just to not have public API, which accepts pointer at all.
Just use a reference.
| using compare_obj_t = std::remove_const_t<std::remove_pointer_t<Obj>>; | ||
| static_assert(std::is_same_v<std::remove_const_t<class_t>, compare_obj_t>, "must be one type"); | ||
|
|
||
| (obj->*aFx)(conditionalAddressOf<std::tuple_element_t<Idx, qalified_t>>(std::get<Idx>(tuple))...); |
There was a problem hiding this comment.
See not any check that obj is not nullptr.
There was a problem hiding this comment.
Ok. I will fix. But better exception as i think. If all ok exception will be cheap.
I think, I'll add gtest into the library. |
|
Maybe it will be easier to exchange Skype or somehow cooperate? Russian is our native language, we will understand each other faster |
2 participants
This branch is almost end of common lib work.
Now need better error messages.
Think about better test organization.
Also some functions can be called more times then need.
Also i checked with valgrind. No memory leaks now.