MOPAC 6 bug

Manish Sud writes:
 > Using keyword ESP in conjunction with MULLIK in MOPAC 6.0 generates
 > excessively large values for the charges - scaled and unscaled - for
 > any molecular system.
 I have traced this to the use of the array C in COMMON/VECTOR/.
 In routine MULLIK, C is used as scratch, so the eigenvectors are
 destroyed.  Unfortunately, the esp routines expect to find the eigenvectors
 in that array.  I solved this problem in my own version by creating a scratch
 array of size MORB2 in routine WRITMO (which calls MULLIK).  I saved a copy
 of C at the start of WRITMO, and copy it back into C at the end of WRITMO.
 I realize that there are more elegant solutions, but this was the quickest one
 I could code up.
 Does anybody know of a formal way to submit this fix to QCPE?
 George Fitzgerald
 Cray Research, Inc.