Re: scons check breaks with latest version of gpsd

[Gary E. Miller]

Yo Greg!

On Tue, 13 Jul 2021 18:26:29 -0400
Greg Troxel <gdt@lexort.com> wrote:

> "Gary E. Miller" <gem@rellim.com> writes:
> 
> > I just updated arm64 to glibc 2.33.  They changed how printf()
> > rounds.
> >
> > Now Instead of this: 
> >
> >     "climb":0.000
> >
> > I get this: 
> >
> >     "climb":-0.000
> >
> > Works on amd64.
> >
> > I've long feared that gpsd would have to do its own rounding...  
> 
> I would expect that all the rounding is specified between IEE754 and
> POSIX.  So was this a glibc bug fix?

C99 and ISO/IEC 9899:TC3 (Draft C 2007) say of printf("%f"):

    The value is rounded to the appropriate number of digits.

That does not explicitly explain 0.00 and -0.00.

POSIX 2007 says it include C99 definition of pritnf() by reference.

There has been at least one other RasPi report of rounding issues:

https://gitlab.com/gpsd/gpsd/-/issues/65

    "lat":41.575030049

vs:

    "lat":41.575030048

But that was on Debian Buster Pi that no one else could duplicate.

> (I need to test on NetBSD/earmv7hf-el, and to install NetBSD/aarch64
> as we call it on something.)

That would be helpful.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        gem@rellim.com  Tel:+1 541 382 8588

            Veritas liberabit vos. -- Quid est veritas?
    "If you can't measure it, you can't improve it." - Lord Kelvin

pgpjYVCqZAwKn.pgp
Description: OpenPGP digital signature