Local densities calculation using 800x800 [um] mislocated window

[KrzysztofHerman]

Minimal DRC run reports AFill.g2 and M1-5Fill.h violations. All related to areas of 800x800 um where the density is calculated.
The issue is that the windows of observation are mislocated as seen on the following figure.

GDS to reproduce this case can be found here

[akrinke]

Currently, this is the expected (although surprising) behaviour. The tile size is 800x800 µm². The processing starts in the lower left corner and the tile is then shifted by 400 µm in x/y direction, so that the individual tiles overlap. This process stops as soon as the complete layout has been covered. For this concrete layout, there are tiles that only cover a small piece of the layout (as shown in the screenshot). The area outside the chip area is ignored when calculating the density. The small chip area in these tiles fails the density check, because there is a little bit of empty room at the border.

Long story short, I agree that this can be considered a bug. We will look into changing the behavior so that each tile is within the chip boundary.