Repository navigation
Description
What happens
A use-model floor that asks for Predicated with predicate isQuack is reported p-sufficient when the edge kept a different predicate, isLoud. The checker compares the fate with the predicate label erased, so the two cases cannot be told apart.
Reproduction
Run on 2026-11-17 with tropecheck-rs built from src/rust (last changed at 7f5aaa5, on main).
for p in isQuack isLoud; do
jq --arg p "$p" '
.edges[0].grade.fate.quality.predicate = $p
| .use_model.floor.fate.quality = {k: "Predicated", predicate: "isQuack"}' \
tests/conformance/fixtures/duck-collapse.ir.json > edge-$p.json
src/rust/target/release/tropecheck-rs edge-$p.json
doneedge predicate isQuack, floor predicate isQuack -> p-sufficient
edge predicate isLoud, floor predicate isQuack -> p-sufficient
The second line is the defect.
What was and was not tested
- Rust checker: run, result above.
- Idris2 checker: not run (no Idris2 toolchain was available).
Trope/Coords.idrstates in its header that the carrier erases the predicate label, so the same result is expected. That is a reading of the source, not a run.
Why it matters
This breaks no proved law. L1–L6 and the soundness results are stated over the erased carrier and say nothing about label equality; spec/calculus.adoc (Composition) says the verified developments erase predicate and fusion labels.
The gap is between the verdict's name and its content. The calculus's first firewall allows a false refusal ("The checker MAY be conservative: it MAY report p-insufficient where the true residual would in fact have sufficed"). It does not allow a false acceptance, and this is one.
The same question probably applies to fusion labels, which are erased in the same place. That was not tested.
Candidate remedies
No remedy is chosen here. Each one changes normative content (the floor rule or the carrier), so the choice is deferred to the owner.
- Compare predicate labels in the floor check: a
Predicated(P)floor is met only byPredicated(P)or something stronger. Smallest change; needs a decision on whether label equality is syntactic. - Refuse a
Predicatedfloor: report it as not decidable from the grade. Conservative and sound, and it makes such use-models unusable until a finer layer exists. - Keep the grade check as it is and move the question to a use-sensitive layer above the grade algebra, with a third outcome (indeterminate) for cases the grade cannot settle.
Suggested acceptance test
Whatever the remedy, add the isLoud/isQuack pair above as a conformance fixture whose expected verdict is not p-sufficient, and show that it fails against the current checker.
Related
- Write-up with the paper counterexample and three open theorems: Research note: a grade abstracts a transformation (one reproduced counterexample, three open theorems) trope-particularity-workbench#82 (
docs/theory/formalisms/grades-as-abstractions-of-transformations.adoc). - Bears on O2 (lowering correctness) and on Hermeneia's licensed band, which relies on this verdict.
Activity
coderabbitai commentedon Nov 17, 2026
🔗 Related PRs
#107 - Phase 0 + CI repair: make
#102's status surfaces true AND its checks green (supersedes#102) [closed]📝 Issue Planner
Check the box below or use the
@coderabbitai plancommand to generate an implementation plan and prompts that you can use with your favorite coding assistant.🧪 Issue enrichment is currently in open beta.
You can configure auto-planning by selecting labels in the issue_enrichment configuration.
To disable automatic issue enrichment, add the following to your
.coderabbit.yaml:💬 Have feedback or questions? Drop into our discord!