Hi all see https://www.mkgmap.org.uk/pipermail/mkgmap-dev/2024q2/034252.html I think the experiments by Felix Herwegh showed a bug in the blocksize calculation. If the user combines a huge number of tiles (e.g. 6000) in one gmapsupp the code may not find the best blocksize due to an integer overflow. Not sure, maybe this also happens with < 2048 tiles in some cases. The attached patch fixes the calculation. It will result in smaller gmapsupp sizes where the error occured. Gerd