Peter,
On 2005 Apr 1 , at 12.01, Peter W. Draper wrote:
> On Fri, 1 Apr 2005, Norman Gray wrote:
>
>> Sorry, I wasn't clear. I added 'f77 -old_f77' to (the end of) the
>> list
>> of compiler+flag combinations that the Fortran support attempts, in a
>> push toward making setting FC/F77 unnecessary.
>
> OK, then that doesn't work. If I don't set FC/F77 then the f95 command
> is
> picked up, with much the same problems as plain f77.
I see what I've done. The check that's done is only whether one of the
commands `lf95 f95 fort xlf95...' is in the path, so any flags
specified there are just ignored at this stage (so I've taken out this
redundant addition). The remaining checks determine whether this
compiler is suitable, but there's no mechanism in autoconf for
backtracking and finding a better compiler if the first one found is
deficient. Which is a pain.
>>> cpp: Severe: No such file or directory
>>> ... file is '`-DLANGUAGE_FORTRAN_95'
>>> f90: Severe: The input stream is empty
>>
> As I sais I think the compiler is
> substantially broken in this area, not us.
I think you're right -- I can't see any way we/I could have provoked
that!
> Clearly when gfortran comes along (GCC4 is standard in FC4), this will
> become more of an issue (%VAL etc. aside), so we should be thinking of
> looking at this as part of any future plan.
Doomed we are! We're all doooooomed!
...as usual.
Norman
--
----------------------------------------------------------------------
Norman Gray : Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk
|