Re^6: Install Math::GSL fail

[syphilis]

return (gsl_nan());

That looks wrong to me, too - I think that the value returned by gsl_nan() needs instead to be pushed onto the stack and returned using XSRETURN() ... but I don't have time to test, tonight.

Cheers,
Rob

[aixtal]

Wonderful! Your reply opened my eyes. I changed the two occurrences of return to XSRETURN, and the build succeeded. I'll file a pull request on Github. Many thanks! --j

[syphilis]

I changed the two occurrences of return to XSRETURN, and the build succeeded

Whilst that will enable the build to succeed, it's unfortunately not the right solution.
XSRETURN() takes, as its argument, the number of elements to return from the stack. According to perldoc perlapi:

 XSRETURN
  Return from XSUB, indicating number of items on the stack. This is u
+sually handled by "xsubpp".

  void    XSRETURN(int nitems)

I'm not sure what will happen if your version of that block of code gets executed and an attempt is made to return gsl_nan() items off the stack ... but I fully expect that it won't be pretty.
If that block of code gets executed, then it needs to XSRETURN(1), and the one item that it needs to return from the stack is whatever-it-is-that-is-returned-by gsl_nan(). (Presumably that's either a NaN, or a Math::GSL object that holds the value NaN.)

So long as that block of code never gets executed I think your fix should, however, be fine.

Cheers,
Rob