httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Trawick <trawi...@bellsouth.net>
Subject Re: cvs commit: apache-1.3/src/ap ap_snprintf.c
Date Thu, 10 May 2001 12:32:20 GMT
"Ralf S. Engelschall" <rse@engelschall.com> writes:

> In article <20010427175209.A901@waka.ebuilt.net> you wrote:
> 
> > It also isn't portable, which means that 1.3 is now broken on all but a
> > few platforms.  1.3 doesn't have autoconf, so this change will have to
> > disabled by default and enabled on a per-platform basis.
> > 
> >> > [...]
> >> >  Make ap_snprintf() more robust against border situations with
> >> >  floating point numbers.
> >> 
> >> The patch adds handling for "nan" and "inf" only for %f, %e and %E. What
> >> about %g abd %G? I think the same checks should be applied to them, too.
> 
> Ops, yes, Roy is right: isnan(3) and isinf(3) are non-standard
> functions. For 2.0 this should be apr_isnan and apr_isinf and handled
> inside APR with Autoconf. For 1.3 it should be checked with TestCompile
> (recommended) or at least wrapped with a few platform-specific #ifdef's
> (alternatively).

We need to agree ASAP on how to solve this, or people will be wasting
their time as they play with 1.3 in prep for a release of 1.3.20.

TestCompile sounds right but then I've never played with it :)
Anybody want to be a hero?

Simpler approach: If your platform has isinf() then add -DHAVE_ISINF
in src/Configure, fix up ap_snprintf.c, and the rest of us can relax.

(I never saw any isnan() breakage with APR, though perhaps there are
some ancient systems without it.)

(Tru64 broken, suspect at least two of AIX, Solaris, and OS/390 will
be broken as well)
-- 
Jeff Trawick | trawickj@bellsouth.net | PGP public key at web site:
       http://www.geocities.com/SiliconValley/Park/9289/
             Born in Roswell... married an alien...


Mime
View raw message