- tests/kernel-cases/sources.json gives each looked-at number's kind: an open or closed issue, a
merged PR, one of the frozen truck-geometry PRs, a kernel-paused PR. The report's coverage by
source counts how many of each were looked at, became cases or were skipped.
- crashes/2357: one join of a ring of 90 overlapping cylinder segments. The right answer is one
closed ring body. The old kernel still panics in its intersection curves and runs past the
timeout.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Runs are compared with the golden's times, so a run sharing the machine between cases flags
slowdowns that are only contention (seen with --jobs 4: four cases two to ten times slower).
--jobs still overrides it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
The ladder orders kernel operations by complexity instead of by random goldens: a manifest
(tests/kernel-cases/ladder/ladder.json) lists tiers in order and each tier's cases in order;
every case is graded against its own answer like any rig case, and a golden
(tests/kernel-cases/ladder/golden.json) records per case and kernel the status, the commands'
time and what the kernel made. A run compares with the golden and flags slowdowns (over twice the
time and 0.25 s more), status changes and changed counts. `--record REASON` re-measures one case
at a time, rewrites the golden with its reason and sorts each tier by the first kernel's time.
The report, docs/kernel-ladder.md, gives per tier and kernel the counts per status and the
frontier: the first case in ladder order the kernel fails. It reuses the rig's process-per-case
runner, timeout and cells; results now carry their path under tests/kernel-cases/ so cases from
other areas can be listed in a tier. The README documents `base`, `missing` and the ladder.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
`"outcome": "solid_or_refused"` takes a clean refusal, or else grades the result as usual. Two
bodies that touch along a circle only (#168, #3841) have no one-solid union: a refusal or the two
bodies left apart are both right; fusing them is wrong. The two cases said opposite things
(two bodies, a refusal) and now agree.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- `"shells": N` in expected.json grades the total number of shells: a body with a sealed void
holds two, and some joins keep several lumps in one body. Those cases (#3175, #3455, #3526,
#3730) are graded on validity again instead of being marked not valid.
- When SMLib's record conversion says "not supported yet" (revolved surfaces, traced intersection
curves), its check gives no verdict. That's shown but not graded, since it isn't a fault of the
body. solidcraft_kernel::SmlibCheck carries the three outcomes.
- `$OUT` in a recipe is a scratch directory outside the repository, for files a case writes. The
two STEP-export crash cases write there instead of /tmp.
- cargo xtask kernel-cases --results FILE keeps every result as JSON.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- solidcraft-cli and solidcraft-engine forward an `smlib` feature to solidcraft-kernel/smlib, so
`cargo build -p solidcraft-cli --features smlib` gives the rig SMLib's check.
- solidcraft_kernel::boolean_kernel() says which kernel this process runs booleans on.
`default` until the SMLib switch (SOLIDCRAFT_BOOLEANS=smlib) lands and answers there. An smlib
run on a build without the switch reports "not supported yet" instead of grading the default
kernel under SMLib's name.
- An area may keep its skipped list in parts (skipped-<part>.json).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every geometry case the old kernel got wrong becomes tests/kernel-cases/<area>/<id>/: an
engine script or oracle recipe, and expected.json with the right answer, from analysis or a
hand calculation, or else the invariants. Never a kernel's output.
- solidcraft-cli kernel-cases <dir>...: runs a case and grades it. The checks are the outcome
(succeeds, or refuses cleanly), the body count, and one closed shell per body that passes the
record's structure and geometry checks and, with SMLib built, SMLib's check(All). Counts,
volume, area, centroid and box are graded where the case gives them. Each run is pass, fail,
not supported yet or crash.
- cargo xtask kernel-cases [--kernel default|smlib|both]: one process per case, so a native crash
or a hang (--timeout) is that case's result. It writes docs/kernel-cases.md: coverage per area
and source, results per kernel, and the sources looked at and skipped (<area>/skipped.json).
- solidcraft_kernel::validity and smlib_built, which the rig's checks use.
tests/kernel-cases/README.md gives the format.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>