> However, the error handling in that routine (smu_init) needs much more > work. Indeed, the smu = NULL at the end of the function seems a bit brutal - a lot of allocated data ends up not getting freed. I can try to do something about it. Did you see something else? julia