This bug was highlighted in r3334 when the test public_transport=platform was added to these four rules in the default points style file (lines 45-63): (public_transport=platform | highway=bus_stop | railway=tram_stop | railway=halt | railway=station) & (ref=* | route_ref=*) { set ref='${ref|def:}(${route_ref})'; } (public_transport=platform | highway=bus_stop | railway=tram_stop | railway=halt | railway=station) & lit=yes & (shelter=yes | covered=yes) { set ref='${ref|def:}*'; } (public_transport=platform | highway=bus_stop | railway=tram_stop | railway=halt | railway=station) & lit!=yes & (shelter=yes | covered=yes) { set ref='${ref|def:}+'; } (public_transport=platform | highway=bus_stop | railway=tram_stop | railway=halt | railway=station) & (shelter=no | covered=no) { set ref='${ref|def:}-'; } What happens is that if a node has two of the tags in the list public_transport=platform | highway=bus_stop | railway=tram_stop | railway=halt | railway=station then the node is processed twice by each rule, so the route_ref and the shelter marker (*, + or -) are doubled-up. Looking at https://www.openstreetmap.org/node/1424432523 - this node is tagged public_transport=platform and highway=bus_stop. Prior to r3334, the node had only one tag that matched and there was no doubling-up. From r3334 onwards, there are two tags that match and the result is Agglo Hérault Méditerranée: (1;2;3;4;5,4,4,3,3)(1;2;3;4;5,4,4,3,3)++ Gare SNCF. I believe it is not intended that style rules should work in this way, therefore it is a bug in the code. In which case it is a question for the Java experts. Adrian