[Numpy-discussion] vectorize problem with f2py and gfortran 4.3

Hi all, A user on the pymc user list has reported a problem with f2py wrapped fortran functions compiled with gfortran 4.3, which is the standard Ubuntu Jaunty fortran compiler. I noticed the same bug in some of my own routines. The problem, as far as I can understand, is that vectorize tries to find the number of arguments by calling the function with no arguments and parsing the error message. With numpy 1.3, python 2.6 and gfortran 4.3, the error message is not what numpy expects, and does not contain the expected number of arguments. So I am wondering if there is a reliable way to introspect compiled extensions to provide the number of arguments needed by vectorize ? Thanks, David