Re: ARGH. Who wrote intrinsic_set_exponent.f90?

[Steve Kargl <sgk at troutmask dot apl dot washington dot edu>]

On Tue, Aug 29, 2006 at 10:50:14PM +0800, Feng Wang wrote:
> --- Steve Kargl <sgk@troutmask.apl.washington.edu>????:
> > On Tue, Aug 29, 2006 at 04:06:22PM +0800, Feng Wang wrote:
> > > Steve,
> > > 
> > > > With my patch?
> > > 
> > > Did you mean the patch in
> > http://gcc.gnu.org/ml/fortran/2006-08/msg00371.html?
> > 
> > Yes.  That patch fixes the range checking as explained in the email.
> > 
> > > No. I didn't apply any patch. 
> > > 
> > > After applying your patch, I got an error message on i686-pc-linux-gnu.
> > > 
> > > [wf@WFPC bug]$ gfortran intrinsic_set_exponent.f90
> > >  In file intrinsic_set_exponent.f90:39
> > > 
> > >   i = o'20037777777'
> > >      1
> > > Error: Arithmetic overflow converting INTEGER(8) to INTEGER(4) at (1)
> > > 
> > 
> > And, this is the problem that caused the start of this thread.
> > 
> 
> OK. I understand what you mean. I will try another way to construct these
> special floating-point numbers. Maybe TRANSFER is good choice.
> 

I have not converted o'20037777777' to a decimal integer value,
so can't say if it indeed overflow.  However, you can probably
use IBSET to get the bit pattern you.  Note, I have not looked
into the portability or conformance of this approach.

    i = 0
    i = ibset(i, 23)  !  Yet, another test of a standard intrinsic,
    i = ...
    x = transfer(i,x) !  Added testing of transfer.

Thanks for looking into this issue.

-- 
Steve