> program tmp1 > real :: a(2,2),b(3,3) > > a=1.0 > b=2.0 > > b = b + a > print *, b > end program tmp1 Many commercial compilers do indeed issue an error (Intel, Portland Group, Sun). gfortran does not (g95 neither), which is a bug, since this is not standard-conforming (and not a common extension either, AFAIK). I can't find it in bugzilla. FX