[PATCH 1/5] libstdc++: Import the fast_float library

Jonathan Wakely jwakely@redhat.com
Tue Nov 16 11:34:48 GMT 2021


On Tue, 16 Nov 2021 at 09:46, Florian Weimer via Libstdc++ <
libstdc++@gcc.gnu.org> wrote:

> * Jonathan Wakely:
>
> > On Tue, 16 Nov 2021 at 08:01, Florian Weimer wrote:
> >>
> >> * Patrick Palka via Libstdc:
> >>
> >> > This copies the fast_float library[1] into the compiled-in library
> >> > sources.  We're going to use this library in our floating-point
> >> > std::from_chars implementation for faster and more portable parsing of
> >> > binary32/64 decimal strings.
> >> >
> >> > [1]: https://github.com/fastfloat/fast_float
> >> >
> >> > Series tested on x86_64, i686, ppc64, ppc64le and aarch64, does it
> >> > look OK for trunk?
> >>
> >> Missing Signed-off-by:?
> >
> > That's not needed if Patrick is still covered by an FSF assignment.
>
> But the submission is not covered by the FSF assignment.
>

Good point.


> > I think we could use Apache as well, because this code isn't going to
> > appear in public headers so the problematic clause doesn't apply. But
> > MIT is simpler.
>
> Okay, so you consider dynamic linking only?  I think the historic
> libstdc++ license is more permissive than Apache or MIT when used with
> GCC.  There aren't any notification or other requirements.
>
>
Another good point - the Apache license is (once again) problematic here.
So it's good we can choose the MIT one.


More information about the Gcc-patches mailing list