Repository navigation
Add the missing class and instance checks (M5 codes without a checker kind) #97
Description
Part of #77 (Phase 4). The M5 codes that have no corresponding checker kind at all — new checks, not new mappings. 30 corpus cases.
| Code | Cases | Needs |
|---|---|---|
InvalidInstanceHead |
7 | Validate the instance head against the class: constructor head vs. type variable head, the newtype-constructor special case. |
OrphanInstance |
7 | Reject an instance whose class is not in scope, or whose type is defined elsewhere. Needs a rule that composes with OrphanTypeDeclaration, which is already reported. |
InvalidNewtypeInstance |
6 | Coercible/Coerce instance validity — adjacent to FE-16's role checking, which is already in place. |
ClassInstanceArityMismatch |
4 | Compare the instance head's argument count against the class's. |
CycleInTypeClassDeclaration |
2 | Class-member and superclass cycles. |
MissingClassMember |
2 | Maps from the existing MissingInstanceMethod kind. |
PossiblyInfiniteInstance |
1 | Head-recursive use of the instance being solved. |
DuplicateTypeClass |
1 | Two classes with one name. |
DuplicateInstance |
1 | Two instances for the same head. |
Prerequisites
DuplicateInstance and DuplicateTypeClass are resolver work, not typechecker work — they belong with #90, which takes instances into the resolver. Do them there or coordinate explicitly, or they will be solved twice.
Acceptance
All 30 cases agree on the L5 scoreboard from #82.
Activity
Record the Prim.Row owner and what the kind work still needs
Record the Prim.Row owner and what the kind work still needs
biuld commentedon Oct 16, 2026
The 2026-10-17 L5 board is 77/97. Of the codes this issue lists, these now agree:
InvalidNewtypeInstance6/6,ClassInstanceArityMismatch4/4,DuplicateInstance1/1,MissingClassMember2/2.Still open:
InvalidInstanceHead6 (1/7),OrphanInstance7 (0/7),PossiblyInfiniteInstance1 (0/1),DuplicateTypeClass1 (0/1).Corpus casesmoves from 30 to 15.CycleInTypeClassDeclarationis not in the current L5 per-code table, so it is not part of the 15.Evidence: the L5 row in
docs/design/D-04-suite-roadmap.md(2026-10-17,886cd82c, annotations oracle).biuld commentedon Oct 18, 2026
All codes this issue owns agree on the L5 board, remeasured 2026-10-18 on master (annotations oracle, vendored corpus, package
e556bf2f): 98/99, denominator 97 → 99 after the two superclass-cycle cases left the blocked rows.InvalidInstanceHead7/7 — the determined-head rule at the common instance recorder (4a078b95), one rule covering the wildcard, free-variable-through-synonym (TypeSynonyms7), derived-head (3510,InvalidDerivedInstance2), and fundep-undetermined (RowInInstanceNotDetermined0/1/2) triggers; official covering-set minimalization (9a4c2e04).OrphanInstance7/7 — the official orphan rule (298cc585);3405.purs's produced set is exactly[OrphanInstance].PossiblyInfiniteInstance1/1 (438.purs),CycleInTypeClassDeclaration2/2 (Superclasses2,SelfCycleInTypeClassDeclaration) —4a078b95.DuplicateTypeClass1/1 — resolver-side distinction fromDeclConflict(fd7bf88a), withDeclConflictstaying 11/11 andDuplicateInstance1/1.InvalidNewtypeInstance6/6,ClassInstanceArityMismatch4/4,MissingClassMember2/2,DuplicateInstance1/1,CannotDeriveInvalidConstructorArg7/7 unchanged.The only remaining L5 mismatch is
ClassHeadNoVTA3.purs, which is #123's visible-type-application limit, not a code this issue owns. Boards: L2 72/72 + 403/413 passing, L3 44/48, L4 41/50, L6/M7 287/340 (serial rerun, same failure set; recorded in D-04 and the runtime-program scoreboard). TheProjectboard field updates for this item still need theprojectscope on theghtoken.