LAPACK failures

[Thomas Koenig]

> Seems to be in the complex routines.  Should I try to pull previous
> snapshots out of cvs to see when this may have happened?

That would be great.  There were some changes to complex routines
recently, and I have a sneaking feeling that old Fortran codes
are the only ones that really strain the complex routines.

A Lapack regression tester would be ideal for this.  Does anybody
have code to run this?

	Thomas