Why does the C++ standard specifies a destructor for std::bitset::reference?

Question

I am wondering why std::bitset::reference and std::vector<bool>::reference specifies an explicit destructor (not compiler generated one). Because, for example, boost::dynamic_bitset::reference does not seem to specify such a destructor.

Answer 1

Just because the Standard mentions ~reference() as a destructor, does not mean it has to be user-provided as a no-op {} (which is how libstdc++ and SGI/STL do it). It could also be user-declared as =default, or even left defined implicitly (which is how libc++ does it). Regardless, the Standard could be updated to have this explicit mention of the destructor removed. You could file an editorial change (I don't think it warrants a real proposal).

As noted by @BoPersson in the comments, std::bitset is a very old component of the Standard Library. Many of its features (implicit constructor from unsigned integer, member instead of non-member operator==) pre-date the language standardization in 1998. Shameless plugs: see e.g. this Q&A for more discussion how this might have arisen, and this Q&A for why it might break code when this would be fixed.

<rant mode>

The best way out of the legacy of std::bitset would be a clean break in namespace experimental. Preferably, this would also solve the mixed abstraction of std::bitset, which at the same time tries to be a space optimized version of array<bool> as well as a set<int>. Ideally, there would be a proposal for a bool_array<N> and a bounded_int_set<N> that provided these abstractions. Similarly, bool_vector<Alloc> (currently known as vector<bool, Alloc>) and int_set<Alloc> (currently a mixture of boost::dynamic_bitset and boost::container::flat_set<int, Alloc>) could be defined.

</rant mode>