This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: finite() question
- From: Steven Bosscher <stevenb at suse dot de>
- To: sje at cup dot hp dot com, fortran at gcc dot gnu dot org
- Date: Sat, 20 Nov 2004 14:06:12 +0100
- Subject: Re: finite() question
- Organization: SUSE Labs
- References: <200411151731.JAA15352@hpsje.cup.hp.com>
On Monday 15 November 2004 18:31, Steve Ellcey wrote:
> I recently built libgfortran on an IA64 HP-UX platform and ran into a
> problem with 'finite()'. The write_float routine in io/write.c calls
> finite and HP-UX does not have the routine under this name. It does
> have 'isfinite()' and 'isinf()', but not 'finite()'. I would like to
> fix this so things work on HP-UX but would like some suggestions as to
> the best way to do the fix. HP-UX also has 'fpclassify()' if we wanted
> to use that.
This is PR15960.
> My first thought for a fix would be to have the configure script check
> for finite, isfinite, and isinf, and use which ever one it found
> (probably in the order listed). I am not sure what write_float should
> do if it finds none of them.
When no such function is available and fpclassify is also not available,
libgfortran configure should just fail IMO.
> I also noticed that there does not seem to be a check for fpclassify in
> configure, even though intrinsics/c99_functions.c checks for
> HAVE_FPCLASSIFY. How does HAVE_FPCLASSIFY get set if it is not set in
> configure?
Looks like a bug.
Gr.
Steven