It is a compiler bug of some sort, but I also think that piece of code is
not completely safe. My version (which has an extra trap for
TENS(1)=TENS(2)=TENS(3) and I think is safe) is:
!...DEFINE VXX,VYY,VZZ ACCORDING ABS( TENS)
JMAX=1
ABSMAX=abs(TENS(1))
JMIN=1
I don't understand that.
It can only be a compiler bug (due to optimization), where it
executes the code AFTER the do-loop BEFORE it finishes the do-loop ???
(or you have values for TENS() which are gt. 100 or not defined,
since the diagonalization in EIGEN3 went wrong and the test of "RES"
I am getting a sigsev in efg.f with the version on the web (today). I think
it is the trap for max/min values, so after
!...DEFINE VXX,VYY,VZZ ACCORDING ABS( TENS)
ABSMAX=-1.
ABSMIN=100.
I added:
!
JMAX=1
ABSMAX=abs(TENS(1))
JMIN=1
3 matches
Mail list logo
