Before, I have made several test where white layer wouldn't change anything for all modulo mode to verify that they all work as expected. Then over time, I realized that at this point, modulo shift continuous just doesn't make sense from the functional point of view and the former test just don't take into account of that.
The problem with older version: If you set 0 on the blend layer using modulo shift continuous, it assume that there's a shift because of inversion. However, 0 shift is 0 shift as in there shouldn't be any changes at all.
This patch fix that issue from the functional point of view.