Hi all, right after the merge to trunk I noticed that the split algo is not fully predictable (and it never was). It is possible that a file is split very well with a max-nodes=x and much worser with max-nodes=x-1000 or max-nodes=x+1000 This also means that the new --num-tiles option may not work with e.g. --num-tiles=1000 for Europe. I may also disable the support for polygons in combination with --num-tiles. I think nobody needs this and the calculation time can be extreme. I try to find a better algo, in the meantime you may try to increase or decrease max-nodes a little bit if splitter doesn't find a good split. Gerd -- View this message in context: http://gis.19327.n5.nabble.com/splitter-r387-tp5806869.html Sent from the Mkgmap Development mailing list archive at Nabble.com.