Lightning Talk: Const Mayhem in C++ - Ofek Shilon - CppCon 2022

---

Lightning Talk: Const Mayhem in Cpp - Ofek Shilon - CppCon 2022

"Did you know that:
(1) const methods can legally modify members?
(2) A const object cannot be default-initialized? (this makes sense after a minute)
(2)(b) Unless... a default constructor is declared non-inline?
(3) The standard does not properly account for volatile members of const objects?
All this and (probably not much) more, at this lightning talk."
---

Ofek Shilon
__

#cppcon #programming #cpp

[genx]

5:43
The thing is that int* m_p in context of constMethod() is a "const pointer" not a "pointer to const". So we can't change its value (the address) while we can dereference (and change the value pointed to).

[Roibarkan]

Very good talk. People interested in this topic might be interested in the existence of which is a smart pointer that can’t modify its pointed-to memory when it’s in const context. Part of the “library fundamentals TS v2”

[User-cvee]

Were the "slides" out of sync? Also is the talk available somewhere else, I would love to hear the complete version.

[Raspredval]

5:40 but shouldn't that constness of the variable turn the int into a const int and then it would trigger at least a warning about making a mutable pointer from a const variable?

[ohwow]

Why does the standard allow this? It seems like a purely harmful UB case. I guess it should be detected by the compiler and result in an error.