[Numpy-discussion] Ctypes required? Fails to build.

[Robert Kern]

On Fri, Jul 4, 2008 at 19:47, James Turner <jturner at gemini.edu> wrote:
> Thanks, Robert and Stefan for your helpful replies. It makes
> a big difference to know which problem I need to solve and which
> I don't :-).
>
> Unfortunately I'm still getting those undefined symbol errors
> for certain maths functions. I tried "python setup.py build_ext
> -lm build" and don't have $LDFLAGS defined, though I am setting
> LD_LIBRARY_PATH and LD_RUN_PATH. But I think the compiler is
> actually finding libm because "-lm" is included on some lines
> without any complaint. Oddly enough, when I do numpy.test() in
> Python, everything passes except for ctypes, so maybe it's OK?
> I'll include the relevant output below in case anyone has ideas.

Oh, don't worry about these. These are configuration tests to look for
various math functions which vary in their availability, like those
for long-double floats. numpy itself will build fine once it has this
information. Your installation is fine.

> I also found a second problem that I was able to solve but I'm
> wondering whether it's a bug? The C compiler failed with a
> syntax error on line 1518 of numpy/core/src/umathmodule.c.src
> because of a C++ style comment ("//"). When I remove the comment
> it works. Should NumPy coding style cater for my eccentric C
> compiler or is this all above board?

Yes, we should only use /**/. I have fixed this. Thank you.

-- 
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