On Thu, Aug 14, 2008 at 09:11:01AM +1200, Rob Reid wrote: > I think a recent patch, probably 9702 > <http://trac.openstreetmap.org/changeset/9702> may have gone a bit too > far with blank tile handling.
While there is definitely something that needs fixing as your tileset indicates (thanks for the test case, by the way, helps tremendously), this was not cause by the checkin you indicated. The new server was not using the 0-byte files at all and indeed they wouldn't have helped here. (It used to produce a 0-byte tile when the parent tile (13,6994,4270) was of the same blankness as the subtile (14,13988) which wouldn't have applied here. It would have copied an empty land tile for zoom levels 12 and 15 only in the previous incarnation, so that would clearly not have helped either. I don't know when this broke, but I don't believe this checkin did. I have simply done away with that skip blank tile creation now. It was actually a minor optimization that we could easily drop. The server has to recreate those empty tiles recursively and I am not sure whether it's more costly to the server to process those 1500 tile files or whether to recreate the blankness state by recursing down. We will still not create blank tiles if the whole parent tile was empty, ------------------------------------------------------------------------ r9818 | spaetz | 2008-08-14 11:56:17 +0200 (Thu, 14 Aug 2008) | 1 line Do away with the whole "do we need to create a blank tile here" optimization. The server needs those anyway, it's less work for the client. Tileset uploads are only slightly bigger, and we still skip empty sub-regions. ------------------------------------------------------------------------ I have rerendered your test case and it looks good now. spaetz _______________________________________________ Tilesathome mailing list [email protected] http://lists.openstreetmap.org/listinfo/tilesathome
