The results now produced by flang(-new) for Fujitsu Fortran test 0676_0160 are better than the desired results of the test.
The test uses the SUM intrinsic function to compute the sum of 100 complex numbers. These values are all identical. The new results from the SUM intrinsic function match the value of 100 times this value exactly. The expected value used by the test does not, and perhaps it reflects the result of a compiler using a less accurate summation method than flang is now using.