Re: [committed] libstdc++: Fix constexpr functions in <experimental/internet>

[Jonathan Wakely <jwakely@redhat.com>]

On Thu, 30 Mar 2023 at 15:44, Jonathan Wakely wrote:
>
> On Thu, 30 Mar 2023 at 00:40, Jonathan Wakely via Libstdc++
> <libstdc++@gcc.gnu.org> wrote:
> >
> > Tested powerpc64le-linux. Pushed to trunk.
> >
> > -- >8 --
> >
> > Change ip::basic_endpoint to work in constant expressions, but only for
> > C++20 and later (due to the use of a union, which cannot change active
> > member in constexpr evaluation until C++20).
> >
> > During constant evaluation we cannot inspect the common initial sequence
> > of basic_endpoint's union members to check whether sin_family == AF_INET
> > or AF_INET6.  This means we need to store an additional boolean member
> > that remembers whether we have a v4 or v6 address. The address type can
> > change behind our backs if a user copies an address to the data()
> > pointer and then calls resize(n), so we need to inspect the sa_family_t
> > member in the union after a resize and update the boolean. POSIX only
> > guarantees that the sa_family_t member of each protocol-specific address
> > structure is at the same offset and of the same type, not that there is
> > a common initial sequence. The check in resize is done using memcmp, so
> > that we avoid accessing an inactive member of the union if the
> > sockaddr_in and sockaddr_in6 structures do not have a common initial
> > sequence that includes the sa_family_t member.
>
> If we had https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/p2641r2.html
> then we wouldn't need the extra boolean. During constant evaluation we
> could use the new magic function to check which union member is
> active, and during runtime we could inspect the sa_family_t member
> (still using memcmp, because POSIX doesn't guarantee a common initial
> sequence, even though I think everybody does give it one in practice).

In fact, thinking about P2641 some more, I might revert this change.
Instead of adding an extra bool member to support constexpr, I think
I'll just remove the 'constexpr' keywords from basic_endpoint for now,
and implement it in terms of just inspecting the sa_family_t member of
the union members. And then later, once we have something like P2641,
we can re-add the constexpr keywords and use is_within_lifetime during
constant evaluation. That way we don't add a bool then need to take it
away again, changing the ABI each time.