Comment # 2 on bug 97338 from Nicolai Hähnle

Hi Daniel, thank you for the detailed report and the initial investigation! I
can indeed reproduce this (including the NaN values) on Polaris.