On Thu, Dec 03, 2009 at 11:07:27PM +0000, svn commit wrote:

Version 1415 was commited by markb on 2009-12-03 23:07:27 +0000 (Thu, 03 Dec 2009)

Set mkgmap:dead-end-check=false when end points have FIXME=* Based on patch from Marko M?\195?\164kel?\195?\164.

Thanks, Mark! I see that you added a check for currentWay.isBoolTag("oneway") when adding the first Node to the way. I am not sure if it is going to work with an event-based parser, because the OSM files I checked (saved from JOSM or output by splitter.jar) had all the <nd ref=...> before any <tag> for the <way>. Thus, currentWay would never carry any tags at the time an <nd ref=...> is added, and the condition you added would never hold. If you want to avoid setting mkgmap:dead-end-check=false for non-oneways whose first node has FIXME=*, this could probably be best achieved by defining private boolean currentWayStartsWithFIXME in Osm5XmlHandler, which would be initialized when adding the first node to the way and examined when closing the way. Sure enough, my worries seem to be warranted. My original patch did not complain about http://www.openstreetmap.org/browse/way/39701441 coming from nowhere even with --report-dead-ends=2, but yours does. A corrected patch is attached. The patch also adds some initialization to addWay(), for additional safety against parsing errors or such. My apologies for the line splits or indentation; I am not sure which tab stop size and indentation rules you are following. Best regards, Marko PS: I got only 116 additional oneway warnings for --report-dead-ends=2 compared to the 0 reported by --report-dead-ends=1 for Finland (113 after the attached patch). I think that I will enable --report-dead-ends=2 permanently and fix any warnings that pop up (or "fix" them by adding fixme=survey if there is no obvious error, such as an unconnected junction).