You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
TrailView default view: a see-through wedge with lane lines at the Axenstrasse tunnel portal after features revision 24 #3582
Part of #3096. Found on production after the update of central-switzerland to features revision 24 (release 16b107f9, 2026-11-17).
Fault
At the Axenstrasse tunnel portal north of Flüelen, the default view (no ground=plan) shows a pale, partly see-through wedge with faint lane lines between the two roads, in front of the portal. With the release before (features revision 12) a solid slope divides the two roads there. A rider sees the wedge at close range.
Picture, before | after, same camera: /media/timo/OSM1/tmp/explore-3506/after/pair-close-3-axen-portal.png (on the owner's PC; view 3 "Axenstrasse tunnel portal from the road" of the Explore session's production check).
What is known
The viewer code is the same in both columns (one still page built from develop b8f3268). Only the map data differs: features revision 12 against 24.
The other 37 production views of that check show no such fault. The Beckenried motorway portal looks the same in both.
The cause is not found. Candidates: the levels of tunnel ways that revision 23 added (the stretches near a tunnel's tag end are OPEN, that is a ground way); a road profile near the portal from one of the revisions 13 to 22.
To do
Find the feature that draws the wedge (decode the tile at the place in both releases; compare the ways near the portal).
Find the revision that made it, with a small box bake.
Fix the cause with a general rule, and check every tunnel portal of the region for the same fault in the default view.
Cause found by an investigation (notes: /media/timo/OSM1/tmp/ground-plan/tunnels/axen/STATE.md on the owner's PC). No fix is built.
The geometry of the ways is the same in revision 12 and 24. The heights changed. The junction of the primary road and the tertiary road 31 m north of the portal (8.614451, 46.933864) went from 501.75 m to 498.25 m. The primary road now sags to that point and climbs 13 % to the portal. The tertiary road stays at the primary road's height for about 25 m, where it ran 3 to 6 m higher before.
The tunnel levels of revision 23 are not the cause. A box bake at revision 21 gives the same heights as revision 24. The change came with the road-height solver (revisions 13 to 21).
Mechanism: the rule "ways on the ground stay on the terrain" holds a way within a band of 5 m about the terrain across its width (GROUND_M in host/obc-trailview-features/src/roads/mod.rs). The terrain grid has a pit of 10 to 15 m under the junction, where the road ledge meets the cliff to the lake. The band's top there is 498.0 m, and the junction lands at 498.25 m.
Which drawn element is the pale wedge is not confirmed (a guess: retaining walls or cut faces over the cycleway's trench between the two roads).
Other places: of 3,754 tunnel portals in the region, 22 have a road junction within 50 m that sank 2 to 10 m against revision 12 (list: tunnels/axen/logs/census-sag.txt). Whether each shows a fault is not checked.
Options: (a) the band's upper bound gives way over a short, narrow terrain pit, the way the solver's weights already ignore a gully, and the census gate follows; (b) the band measures against terrain smoothed along the way. Both need a full-fixture proof and a region rebake.
#3479 (closed as a duplicate) has the same cause at the Axenstrasse: the 5 m GROUND_M band holds a way to a pit in the stored terrain. Its sidewalk drops 27 m while its road stays level. It adds a second rule for this fix: in roads/mod.rs, the PARALLEL_RISE_M (0.5 m) exception cancels the side read "a sidewalk takes its road's height". That exception should depend on whether the two ways part, not on the terrain under them. Prove it on test box B with the count "paths more than 2 m off their parallel road" (100 at the time). One fix and one region rebake prove both issues.
Activity
timohueser commentedon Nov 17, 2026
Cause found by an investigation (notes:
/media/timo/OSM1/tmp/ground-plan/tunnels/axen/STATE.mdon the owner's PC). No fix is built.GROUND_Minhost/obc-trailview-features/src/roads/mod.rs). The terrain grid has a pit of 10 to 15 m under the junction, where the road ledge meets the cliff to the lake. The band's top there is 498.0 m, and the junction lands at 498.25 m.tunnels/axen/logs/census-sag.txt). Whether each shows a fault is not checked.Options: (a) the band's upper bound gives way over a short, narrow terrain pit, the way the solver's weights already ignore a gully, and the census gate follows; (b) the band measures against terrain smoothed along the way. Both need a full-fixture proof and a region rebake.
Generated by Claude Code
timohueser commentedon Nov 20, 2026
#3479 (closed as a duplicate) has the same cause at the Axenstrasse: the 5 m
GROUND_Mband holds a way to a pit in the stored terrain. Its sidewalk drops 27 m while its road stays level. It adds a second rule for this fix: inroads/mod.rs, thePARALLEL_RISE_M(0.5 m) exception cancels the side read "a sidewalk takes its road's height". That exception should depend on whether the two ways part, not on the terrain under them. Prove it on test box B with the count "paths more than 2 m off their parallel road" (100 at the time). One fix and one region rebake prove both issues.Generated by Claude Code