[SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio

Robert Kern robert.kern@gmail....
Mon Jul 14 18:35:24 CDT 2008


On Mon, Jul 14, 2008 at 18:09, Michael Abshoff
<michael.abshoff@googlemail.com> wrote:
> Robert Kern wrote:
>
> <SNIP>
>
>>>> Please provide the output of "f90 -V".
>>> This is SunStudio12-200709:
>>>
>>> -bash-3.00$ ./f77  -V
>>> NOTICE: Invoking ./f90 -f77 -ftrap=%none -V
>>> f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>>> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
>>> -bash-3.00$ ./f90  -V
>>> f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>>> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
>>> -bash-3.00$ ./f95  -V
>>> f95: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>>> Usage: f95 [ options ] files.  Use 'f95 -flags' for details
>>>
>>> I currently do not have access to older Sun Studio releases personally,
>>> but I could ask some people if they could check. This is unlikely to
>>> work, but I will try.
>>
>> Manuals sometimes give a little history. If it got dropped in version
>> 8.0, it's probably documented in there, at least as a bit of
>> advertising. Are the manuals publicly accessible online anywhere? I'm
>> happy to do some searching myself.
>>
>
> Hi Robert,
>
> caught a lucky break and found a box with Sun Studio 11: It ships a
> libf77compat.so.
>
> To identify:
>
> mabshoff@cerberus: /usr/local/SUNWspro/bin>./f77 -V
> NOTICE: Invoking ./f90 -f77 -ftrap=%none -V
> f90: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
> mabshoff@cerberus: /usr/local/SUNWspro/bin>./f90 -V
> f90: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
> mabshoff@cerberus: /usr/local/SUNWspro/bin>./f95 -V
> f95: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
> Usage: f95 [ options ] files.  Use 'f95 -flags' for details
>
> Hope this helps.

Yup. We can do a version check. For >= 8.3, we don't put in f77compat.
Can you and Phil check that if you modify that line of code in sun.py
to omit f77compat on your Sun Fortran 95 8.3 boxes, that it works?
(Ugh, that sentence got away from me, but too late now!)

> It seems that Phil is not poking around. I have some odd numpy 1.0.4cvs
> failure with gcc in 32 bit mode on Solaris 10 Intel when computing roots
> of polynomials. Since Sage will finally upgrade to numpy 1.1 soon I will
> try to see if I can reproduce the problem there and for fun I can also
> build numpy 1.1 + scipy in 64 bit mode with Sun Forte. So if there are
> any patches let me know and I can try them out.

I'm not sure if any are relevant, but we have a branch for the
upcoming 1.1.1 bugfix release:

  http://svn.scipy.org/svn/numpy/branches/1.1.x/

> Otherwise I will likely
> do some crude patching to get the flags right and let you guys work out
> the fines details :)

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


More information about the SciPy-user mailing list