Skip to content

mesh-compile: a tangent contact is refused more often at UTM-scale placements #504

Open

Description

@GeneralPawz

The mesh boolean refuses a designed tangent contact ("meet tangentially") more often the farther the model sits from the origin. An I-beam with a round hole tangent to a flat face, turned 12 ways about z (0.6 + 0.05·k rad), is refused:

placement refused of 12
(12.5, -4, 3.2) 1
(250, -400, 3.2) 0
(1000, 2000, 10) 1
(3.6e5, 5.8e6, 120) (UTM scale) 5

Tolerance 1e-3, chord error 1e-3, BoolmeshBoolean. The rounding of the placement turns the exact tangency into a graze of a few thousandths of a radian, and whether the cut accepts it then depends on the coordinates' magnitude. The fixture and these counts are in crates/execution/compile/tests/far_placed_beam.rs, which #380 adds as a "no worse than before" guard.

This is older than #380, which does not change it. A far-placed model (georeferenced IFC at UTM coordinates) should mesh or refuse the same way at any placement. The local-frame cut (#356) might be the place to fix it: cut in a frame near the operands so the coordinates' magnitude stops mattering.

Done when the twelve turns get the same outcome at all four placements, with the test tightened to say so.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

needs-triageFiled but not yet accepted, sized, or scheduled

Type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions