Re: libc implementations supporting C99 formats

["Joseph S. Myers" <jsm28 at cam dot ac dot uk>]

On Tue, 17 Oct 2000, Robert Lipe wrote:

> The new SCO UDK for UnixWare and OpenServer purports to support all
> library but "return value for snprintf(3S) when the destination array is
> not long enough" and all language features other than complex arithmetic
> and VLA's.
> 
> 
> Though I'm not personally involved in those tools, I can try tests if you
> need a sanity check on something as I have them installed locally.

An obvious sanity check that the printf/scanf may be working C99
implementations would be to printf an infinity with all eight floating
point formats (%a %e %f %g %A %E %F %G) and check that it comes out as
"inf" or "infinity" with the lowercase ones and "INF" or "INFINITY" with
the uppercase ones (this being the only difference between %f and the new
%F; the choice of inf versus infinity should be documented in your
conformance document).  Then check that scanf will accept the same eight
formats and scan infinities back in (printf %f etc. taking a double
argument; scanf %f etc. taking a float * argument).

Essentially, when producing an announcement of format checking work for
the News section (which Gerald Pfeifer asked for some time ago), it seems
appropriate to include a note that you need library support to use the new
formats as well as a compiler which understands them with -Wall, and to
indicate something about the availability of such support; subject to any
limitations on not appearing to promote non-free software it seems it
would be useful to give some indication of where, apart from glibc 2.2 for
GNU/Linux and HURD, such formats can be used.

-- 
Joseph S. Myers
jsm28@cam.ac.uk