Potential Design Issues to re-evaluate from TR

[JeffGarland]

constexpr all the things

Probably all the methods should be constexpr (see #37)

is scope_guard exposition only or not

Maybe this should actually be part of the user api instead of hidden?

boost divergence

There's some differences in implementation there (cant remember details without looking at review) - is there something new and valuable there?

[robert-andrzejuk]

Don't pay, for what you don't use - should the release method be by default in the scope_guard class? Maybe it should be a mixin?

[robert-andrzejuk]

scope_fail and scope_success both require to use the header <exception> (for std::uncaught_exceptions()).
Should <scope> be reliant on this header?

[JeffGarland]

• exception header

Since scope if fundamentally about dealing with exceptions I think that is a requirement.

[robert-andrzejuk]

A released variable is required, to guard against multiple calls of the exit function when moved from the object.

Publicly available release function is a by-product of this variable.

Maybe turn this around - a scope_guard is moveable if it is releasable (it contains release functionality)?

[JeffGarland]

Don't pay, for what you don't use - should the release method be by default in the scope_guard class? Maybe it should be a mixin?

This is documented in #27