link@stew.ssl.berkeley.edu (Richard Link) (03/21/91)
Has anyone else had a similar problem: I have a program in which the main routine passes two REAL*8 scalars to a subroutine. Immediately before the CALL, and immediately upon entering the subroutine, the passed values differ by over 20 orders of magnitude in f77 V1.3.1, SunOs 4.1.1 The program worked correctly under an older version of f77 (1988 or so - the one with f77cvt). First running the old f77cvt on the *.for file to produce a *.f file, and then compiling under f77 V1.3.1, produced correct results. I traced the problem to the following: Replacing the IMPLICIT REAL*8 (A-H,O-Z) with IMPLICIT DOUBLE PRECISION (A-H,O-Z) in this one subroutine solved the problem. (There should be no difference between the two). f77cvt automatically does this conversion, producing the correct results on the newer compliler. How do I notify Sun about this bug? Dr. Richard Link Computational Physics, Inc. P.O. Box 788 Annandale, VA 22003