[SciPy-user] Portability problems (on Solaris)

[Robert Kern]

On Mon, Jul 7, 2008 at 19:36, James Turner <jturner at gemini.edu> wrote:
> Following my last post on numpy-discussion (Re: Ctypes required?
> Fails to build.), I have found some additional portability
> issues -- this time in SciPy (0.6.0) -- that cause my build to
> fail on Solaris 9 with the Sun WorkShop 6 compiler.
>
> Is this a good place to report these, or should I make a ticket
> for them? They don't really need any troubleshooting -- just
> changing (or not).
>
> 1. In the ndimage module, there is a C++ style comment
>    at lines 144-145 of ni_interpolation.c.

I can fix that. Unfortunately, there are a lot of new such comments in SVN.

> 2. The Sun ForTran compiler doesn't like the "END" syntax
>    used in linalg/iterative -- for example
>    "END SUBROUTINE sSTOPTEST2" fails but "END" on its
>    own works (a shame, since the first syntax is nicer).
>    This applies to all the ForTran files in that directory.

Isn't that standard Fortran-77? If it isn't, I'll change it. If it is,
I am going to have to insist on not supporting such broken compilers.

> 3. The main INSTALL.txt file says this: "Currently the SciPy
>    build process does not use a C++ compiler, but the SciPy
>    module Weave uses a C++ compiler at run time, so it is
>    good to have a C++ compiler around as well". However,
>    when I try to build SciPy with no C++ compiler available,
>    the build stops at "scipy/sparese/sparsetools/
>    sparsetools_wrap.cxx" with the error "c++: not found" and
>    SciPy doesn't get installed. Do I need to tell it not to
>    build certain modules?

The docs are old and wrong. scipy.sparse does require C++.

-- 
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