Matt Borland wrote:
> > On May 10, 2023, at 5:42 PM, Dominique Devienne
> <ddevienne(a)gmail.com> wrote:
> >
> > On Wed, May 10, 2023 at 4:01 PM Matt Borland via Boost
> <boost(a)lists.boost.org> wrote:
> >
> >> [...] We use errno rather than std::error_code as specified in the standard.
> >
> > Hi. Sounded like a good candidate for Boost.Compat (or
> > Boost.Polyfill), discussed recently, until that sentence. I'm lucky
> > enough to be able to use
> > C++17 at work, but if I was on C++11, I'd tend to use such a library
> > C++until
> > I could use the std one, so deviations like this would complicate
> > adoption because moving to std would then be harder. My $0.02. —DD
>
> Dominique,
>
> Thanks for the response. We discussed Boost.Compat but decided it is outside
> of the scope. For instance <latch> is what set off the discussion and it can be
> implemented in ~100 LOC. This implementation of charconv will be around
> 10k LOC so it should be in its own library. If the demand signal is to use
> std::error_code instead of errno we can make the change.
boost/charconv.hpp was initially supposed to be a very light header (with only
a few declarations), but since from/to_chars for integral types were moved
to the headers because of constexpr, that's no longer the case, so we can
probably afford the inclusion of <system_error> without anyone noticing.
(Note that std::from/to_chars use std::errc, not std::error_code.)
