Roadmap
Everything still open, in one place: what the workbenches lack, what
waits on the kernel, and the larger work ahead. What is built is in the
release notes (crates/app_shell/RELEASE_NOTES.md). Sizes
are rough effort (S: a day or less, M: a few days, L: a week or
more).
A kernel capability that is missing is filed on the kernel's
repository, noted on its item here, and the rest of the item is
built around it, with a test marked
#[ignore = "kernel: … (ogeom-rs#N)"] that passes once the
kernel has it.
Waiting on the kernel
Gaps in the kernel printCAD runs into. When a fix is released: raise
the ogeom requirement in the workspace
Cargo.toml, take the #[ignore] off its tests,
run the suite. The open ones, each with an ignored test naming it
(grep -rn "kernel:" crates/*/tests):
- Fillets and chamfers: chamfer on any edge (curved faces, a sphere's rim, spline outlines; ogeom-rs#711), a convex and a concave round meeting at one vertex (#712), a radius per edge in one call (#713), refusals naming the failing edge (#714), largest-fit queries (#715), overlapping rounds on a padded thin plate (#721) and a slow tangent loop of arcs (#723). Variable radius on any edge (#716), chamfer corners (#717), setback corners (#718), a rim's round taking a narrow strip off (#719) and a round over a staircase of narrow faces (#720) are filed and wait for an app side.
- Offset solid on free-form faces (#710).
Not yet filed:
Faster rebuilds
The rebuild skips the history an edit did not change, and the kernel's operations work near what an edit touches: a boolean passes the faces far from its tool through as they are (ogeom-rs#134), and a fillet reads only what is near each edge (ogeom-rs#135). What is left in printCAD:
A kernel operation found slow goes to the kernel's repository as an
issue with its API, a repro, rebuild_bench's numbers and an
acceptance test (the time follows the touched region, the result the
same as now).