From: ville.mattila@...
Date: 2006-06-30T22:37:38+09:00
Subject: Re: 1.8.5p1 build failure on Solaris 10

--=_alternative 0048F5C1C225719D_=
Content-Type: text/plain; charset="US-ASCII"

Yukihiro Matsumoto <matz@ruby-lang.org> wrote on 30.06.2006 15:21:11:

> Hi,
> 
> In message "Re: 1.8.5p1 build failure on Solaris 10"
>     on Fri, 30 Jun 2006 04:57:10 +0900, "Daniel Berger" <Daniel.
> Berger@qwest.com> writes:
> |> -solaris*2.10)  ac_cv_func_isinf=yes
> |> +solaris*2.10)  if test -z "$GCC"; then
> |> +                      ac_cv_func_isinf=yes
> |> +              fi
> 
> |Actually, shouldn't we test for the feature instead of the 
> platform?  I don't 
> |remember if older versions of Solaris have this.
> 
> Actually this hack is added to resolve usual test failure for isinf
> on  Solaris , maybe it was due to the forte compiler, so that just
> restoring test may not work well.

    This does not restore the test if gcc is not used. I tested this patch 
with gcc
   and the forte compiler. All tests passed OK.


> 
> |So, AC_CHECK_LIB(sunmath, isinf) && AC_CHECK_HEADER(sunmath.h, isinf), 
with 
> |#ifdef's in the source?
> 
> I have no knowledge to tell that this is appropriate or not.  Any idea?

        I don't like it too much, as it would require additional ifdefs to 
*.[ch] files.


--=_alternative 0048F5C1C225719D_=--