[Bug]: lonlat_to_merc does not fold longitudes outside -180..180: unwrapped antimeridian longitudes project past the world edge #940

What happened?

lonlat_to_merc (crates/geom/src/lib.rs) clamps latitude to ±85.05° but passes any longitude through as-is:

Coord::new(MERCATOR_R * c.x.to_radians(), …)

so a longitude outside −180..180 produces Web Mercator X metres past the ±20,037,508.34 m world edge. I verified against main (342b2b2) that:

Where such longitudes can arise in the current code: unwrapped antimeridian geometry is kept deliberately by geodesic_interpolate (it keeps its longitude next to a's) and by Aeqd::inverse (it keeps longitudes within 180° of the centre's, not folded into −180..180), which the buffer and Tissot tools use. A buffer of a point near 179.9°E that crosses the antimeridian comes back with vertices past 180, and extent_to_merc in turn produces an out-of-world extent when the result is zoomed to.

Steps to reproduce

A library test against rustgis-geom with the values in the table, or a geodesic buffer of a point at 179.9°E large enough to cross the antimeridian, then zoom-to-layer.

What I would like checked

Whether every caller either (a) never sees such longitudes, or (b) wraps them itself (rendering/mod-tile fetch handles it, so this may only affect extents, the camera and measuring). If callers do not, folding the longitude into −180..180 here — or documenting loudly that it does not — would make the extent case above stop producing past-world X metres. I am not sure which is intended, so treat this as a question first.

How are you running RustGIS?

Desktop app built from source, on Linux (WSL).

Version

main at 342b2b2 (after #928).