https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85868
--- Comment #3 from Harald Anlauf <anlauf at gmx dot de> --- Something has changed recently in 9-trunk, I now get with rev.268162 for the testcase in comment #0: 2.00000000 Printing the array bounds in subroutine te gives the expected values, so it must be the offset that is off by one.
