Gaussian03 and bug in IntelFortran 8.1, w/ REPAIR

 In order to switch to the latest LINUX version, I had to re-compile all my
 programs with Intel Fortran 8.1 (instead of version 7 I used before).
 I observed that many Gaussian test jobs bombed thereafter (starting with
 test043.com).
 It could be traced back to wrong compilation of the subroutine ryspol.f.
 The C matrix is not initialized as programmed in the loop with label 60/70.
 The optimizer lost track that matrix C is actually used later on.
 1.) Is this a known bug?
 2.) Although it is definitely a compiler problem (Intel, do you hear?)
     the problem can be solved by cleaning the code from the unnecessary
     EQUIVALENCE statment (Gaussian, do you hear?). So do the following:
     a) comment out the EQUIVALENCE of C00, C01 and C11 (one single line)
        in subroutine ryspol (which is located in utilnz.F)
     b) replace (a few lines below), after label 70 in this subr.
        "C00" --> "C(1,1)"
        "C01" --> "C(1,2)"
        "C11" --> "C(2,2)"
 After recompilation, the affected test jobs are OK. I report on the
 complete test suite later.
 +---------------------------------+----------------------------------+
 | Prof. Christoph van W|llen      | Tele-Phone (+49) (0)30 314 27870 |
 | Technische Universitdt Sekr. C3 | Tele-Fax   (+49) (0)30 314 23727 |
 | Stra_e des 17. Juni 135         | eMail                            |
 | D-10623 Berlin, Germany         | Christoph.vanWullen*at*TU-Berlin.De |
 +---------------------------------+----------------------------------+
 --