[clang] [clang][Sema] Track trivial-relocatability as a type trait (PR #84621)

[via cfe-commits]

cor3ntin wrote:

> But this isn't useful at all; every library has already invented its own constraints and type traits. The point of having a language-based solution is that the "is TR" property gets automatically propagated when you compose types. No library solution can achieve that, especially lacking reflection.

I am not sure i understand, assignability is already a transitive property for defaults operators? so the scenario you are concerned about is triviality copyable types with a user defined assignment operators?

This would for example not impact reference wrappers which are not assignable.

The larger point is that there are 2 concerns here, one is whether it is safe to relocate a type at the language level, which this type trait is intended to provide, and one is how a library uses relocatability for optimization.

A library may decide it doesn't want to relocate types that are not assignable for some operations (which btw doesn't need to apply recursively to subobjects) , and it can do so today.

I would in fact imagine that the logic would be 1/checking for assignability in all cases 2/checking for relocatability for optimization.


It would be useful to illustrate with code examples scenarios in which that wouldn't work.

And again, https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/p2959r0.html goes into some of these questions and might be worth checking out.

https://github.com/llvm/llvm-project/pull/84621