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 ground plan: a deck and its ground stub meet on one line (shards at deck ends) #3484
Part of #3096, follow-up of #3482 (GP12.3b part B, #3353). The owner said on 2026-11-14 "Merge as it is", with these faults known.
Fault
Where a bridge way is split into a ground stub and a deck, the two do not meet on one line. Small shards lie on the road: wedges, an arc-shaped end piece, a free parapet block (Acheregg east deck end, in the owner's own Acheregg view; Attinghausen), and a seam crack across a motorway at one joint (R09).
Causes (from the joint round, crossings/gp123b/STATE-JOINTS.md on timosPC)
A turn at the split: the deck and the stub each end square to their own chord.
Staggered end lines of side-by-side decks of one sheet: a deck's band, edge face and parapet lie on the neighbour's stub.
The storey's outline rounds a concave corner at an end line.
A green line across lanes at R09: most likely the viewer's height composition between staggered end lines. Not traced.
State
Two tries are saved as patches on 0925917 (crossings/gp123b/joints-try1.patch, joints-try2.patch). Try 2: the deck's open end lies on storey 0's end line of its stub; no storey covers the paved stroke of a stub of its own decks; no outline round at a storey's open end. A strict check (no corner of an edge face, parapet, railing or post in a storey-0 paved face within 1 m of the deck) goes from 86 to 1 on the fixture and from 59, 50, 63, 10 to 1, 2, 3, 1 at four places of test box B. It was not pushed: 64 seams on the fixture (T-junctions and steps of a few millimetres), because the deck's end edge does not hold all points of storey 0.
Wanted
The deck's end edge takes storey 0's points by construction (exact, from the ground edges; no tolerance test). Holes for the paved ground arms and junction faces at a deck's end. The strict check with no budget; the seam count at 0. Pictures of T09, the owner's Acheregg view, R09, Attinghausen and the Lucerne interchange, checkpoint 5 | new.
Also open from #3482: the last fix round (672422a..0925917: the junction rule ground_junctions, the two shell probes removed) had no review of its own; the reviewer of this follow-up reads that delta too. The owner decides whether a deck beside a ground road keeps its parapet (the "beside" counts are in the notes).
More places of the same kind, seen on 2026-11-15 in views of the browser fixture with ground=plan (build 8a399ce, sheets on the owner PC in /media/timo/OSM1/tmp/ground-plan/develop-doc/notes/sheets/, crops in sheets/crops/):
b315-bridge-junction and b315-bridge-junction-top: a thin dark spike runs from the upper parapet into the road; thin dark lines leave the west parapet.
b45-bridge-junction and b45-bridge-junction-top (the same junction from two cameras): the near parapet has a torn blue-grey shard that points into the grass, and a thin green spike off its other end over the road.
d-bridge-top: the deck shows faint seams across the road at both ends, and the edge line breaks at the deck.
With ground=plan, no sidewalk or railing part crosses the road at these three places. On the default path, sidewalk slabs and parapet blocks cross the road at b315 and b45; the owner named that as "still bad" on the merge document.
One more cause: a concrete fin beside the approach at a deck end (found on #3585)
Fault. At a deck's end line, a concrete triangle stands beside the approach: about 0.9 m to 1.2 m tall on the end line, and it falls to 0 m over 0.5 m to 1.9 m. Examples in cell 34300/23097 of the fixture: slot 3699 (0.914 m over 1.85 m), slot 3711 (1.193 m over 1.83 m), slots 3803 to 3817 with end face piece 4377 (0.970 m over 0.45 m).
Cause (host/obc-ground-plan/src/faces/heights/mod.rs, decks and ceiling). The ceiling holds the ground of storey 0 on a deck's flat end line. The slope beside the approach is one label on the two sides of that line past the shoulder. Thus the slope's corner on the line reads the held height (deck minus drop), and the paved shoulder stays at deck level. The step between them is drawn as concrete.
Why it shows now. The crate makes these steps on develop too. The viewer on develop reads no deck ceiling in bins with trailing storey items, and this hides most of them. #3585 ports the heights correctly (GPU = crate to 0.1 mm at 48 corners), and then the fins show in the plan view.
Options.
(a) Corners on the end line read as free. Heights only. The end faces go, and the ground ramps up below the slab end.
(b) The ground below a deck gets a label of its own. Then the end line splits the ground across the full stroke, and the abutment face spans the band. Bytes and off digests change.
(c) The fins stay as wing walls and get a shape.
Decision of the orchestrator. (b), as a slice of this issue, with bridge pictures for the owner before the merge. #3585 stays a draft until that slice lands, because it must not make the owner's bridge views worse.
Evidence on timosPC: tunnels/s1/STATE.md ("Fix round 2"), tunnels/s1/shard2/INDEX.md, tunnels/s1/shard2/readback/table.txt.
Activity
timohueser commentedon Nov 15, 2026
More places of the same kind, seen on 2026-11-15 in views of the browser fixture with
ground=plan(build 8a399ce, sheets on the owner PC in/media/timo/OSM1/tmp/ground-plan/develop-doc/notes/sheets/, crops insheets/crops/):b315-bridge-junctionandb315-bridge-junction-top: a thin dark spike runs from the upper parapet into the road; thin dark lines leave the west parapet.b45-bridge-junctionandb45-bridge-junction-top(the same junction from two cameras): the near parapet has a torn blue-grey shard that points into the grass, and a thin green spike off its other end over the road.d-bridge-top: the deck shows faint seams across the road at both ends, and the edge line breaks at the deck.With
ground=plan, no sidewalk or railing part crosses the road at these three places. On the default path, sidewalk slabs and parapet blocks cross the road atb315andb45; the owner named that as "still bad" on the merge document.Generated by Claude Code
timohueser commentedon Nov 17, 2026
One more cause: a concrete fin beside the approach at a deck end (found on #3585)
Fault. At a deck's end line, a concrete triangle stands beside the approach: about 0.9 m to 1.2 m tall on the end line, and it falls to 0 m over 0.5 m to 1.9 m. Examples in cell 34300/23097 of the fixture: slot 3699 (0.914 m over 1.85 m), slot 3711 (1.193 m over 1.83 m), slots 3803 to 3817 with end face piece 4377 (0.970 m over 0.45 m).
Cause (
host/obc-ground-plan/src/faces/heights/mod.rs,decksandceiling). The ceiling holds the ground of storey 0 on a deck's flat end line. The slope beside the approach is one label on the two sides of that line past the shoulder. Thus the slope's corner on the line reads the held height (deck minus drop), and the paved shoulder stays at deck level. The step between them is drawn as concrete.Why it shows now. The crate makes these steps on
developtoo. The viewer ondevelopreads no deck ceiling in bins with trailing storey items, and this hides most of them. #3585 ports the heights correctly (GPU = crate to 0.1 mm at 48 corners), and then the fins show in the plan view.Options.
Decision of the orchestrator. (b), as a slice of this issue, with bridge pictures for the owner before the merge. #3585 stays a draft until that slice lands, because it must not make the owner's bridge views worse.
Evidence on timosPC:
tunnels/s1/STATE.md("Fix round 2"),tunnels/s1/shard2/INDEX.md,tunnels/s1/shard2/readback/table.txt.Generated by Claude Code