[Bug libstdc++/70238] [5/6 Regression] std::future_category ABI change

["matz at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70238

--- Comment #5 from Michael Matz <matz at gcc dot gnu.org> ---
(In reply to Michael Matz from comment #4)
> 
> Basically our shared library (not just the object files) has a change in
> one symbol, and we didn't bump either the symversion or the soname,
> which I'd say is a bug (we can't fix anymore :-/ ).

Now having read the backlog of the parallel IRC conversation, I want to amend 
the above with saying: I do think it's a bug of us, and in the future we should
not put experimental stuff that we don't claim can be held stable into
the shared library.  Alternatively, once we do put something into the shared
lib, those symbols must remain available with the same ABI forever, i.e. we
need compat hacks when we do changes even on the experimental parts.

Otherwise our whole nicely architected sym-version infrastructure and us
not bumping the soname is all just a lie.