Relations with oneway=yes are wrongly converting roads to one-way

[RedAuburn]

There are around 2000 relations with oneway=yes tagging.
https://overpass-turbo.eu/s/1LQf

as reported on telegram, this road:
https://www.openstreetmap.org/way/6363768
is part of this relation:
https://www.openstreetmap.org/relation/1473170

the road incorrectly inherits oneway=yes from the relation.

[RedAuburn]

ah ok, 1911 uses are on relations with type=route (and only 2 on 'type=restriction'). The fix would be to not apply surface/oneway etc tags from route relations.

[muralito]

Which is the semantic of tag oneway=yes on a relation element? It's not specified in the wiki https://wiki.openstreetmap.org/wiki/Key:oneway

[pastk]

Probably we shouldn't infer all tags from relations, but should apply whitelisted ones only.
@vng

[vng]

Good catch. We don't take all tags, but the current logic is very complicated and aims to patch OSM mapping traits.
See RelationTagsWay::Process

[vng]

I will put a crutch, but I still don't have a clear vision of OSM mapping here.

1. For example, this https://www.openstreetmap.org/relation/12126834 I suspect that oneway=yes should be taken from this kind of relation (route=ferry). I see a contradiction here with route=road.
2. Should we take "surface" from route=road relations?

[pastk]

1. For example, this https://www.openstreetmap.org/relation/12126834 I suspect that oneway=yes should be taken from this kind of relation (route=ferry).

Nope, here is a "backward" route that uses the same way: https://www.openstreetmap.org/relation/12234840

2. Should we take "surface" from route=road relations?

I think its mistagging.

[mvgorcum]

This is still an issue for me, since I get weird routes when using organicmaps for navigation in this area.

Note that openstreetmap itself and also osmand don't inherit oneway on these streets.