fix: ADR-0039 Phase-2 — resolve validation-layer + runtime own-reads the independent review found#140
Merged
Merged
Conversation
…ted values → resolving An independent review found 11 Python sites reading OWN attrs/children where the TS reference RESOLVES (own + inherited via extends), each carrying a comment that FALSELY claimed "TS uses ownAttr". Every such site silently drops a value inherited from an abstract base. Flip each to the resolving accessor (get_meta_attr / children) to match the TS oracle exactly, and correct the false comments. Sites (Python → TS oracle, all RESOLVING in TS): - runtime/n2m_resolver.py: own_children() → children() (n2m-resolver.ts:63). MOST SEVERE — an entity inheriting an M:N relationship via extends dropped it at runtime. - loader/validation_passes.py: @objectref via-path + single-hop inference (:522, :709); validateRelationships M:N attrs + root.children() (:1313-1324); @defaultSortField (:120); @filter (:1260); identity @fields (:291); and the full template-validation cluster — @payloadRef/@textRef/@kind/@subjectRef/ @htmlBodyRef/@requiredSlots + subtype-only-attrs + root.children() (:169-253; templates CAN extend). - loader/registered_validation.py: reference-attr value read (:66); recursion stays own. - loader/validate_source_parameter_ref.py: @parameterRef, the param VO's fields, and the referenced entity's field (:76, :122, :139); source/origin reads stay own. - loader/validate_field_readonly.py: identity @generation / @fields (:142,:144). Adds a Python unit test proving the runtime n2m fix: an entity inheriting an M:N relationship from an abstract base resolves the relation (fails pre-fix at own_children, passes at children). Full suite green (1287 passed) except the two pre-existing test_cli_staleness_nudge failures (agent-docs command string, unrelated). No golden/fixture changes. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GF9xLEQZaPus5Y6opk398n
…9 TS reference The C# validation layer read several inheritable attrs via OwnAttr while the TS oracle (server/typescript/packages/metadata) reads them RESOLVING — a concrete node can pick the attr up from an abstract base via `extends`, so own-only reads mis-validate inherited declarations. This flips every such site to the resolving `Attr()` / `Children()` accessor, matching TS 1:1, and leaves the deliberate own reads (origin.* which never inherits, @default declared-here coercibility, declaration-structure OwnChildren tree walks, @discriminator/@discriminatorValue root-vs-subtype semantics, @readonly downgrade own-vs-super comparison, physical-name-alias per-source constraint) untouched. Sites flipped (each vs its TS validation-passes.ts / validation-registry.ts / validate-source-parameter.ts / validate-field-readonly.ts counterpart): - ValidateDataGridSortFields @defaultSortField (TS:120) - ValidateFilterableHasIndex identity @fields, @filterable, @db.indexed (TS:291,301,304) - ValidateFilterableHasSupportedOps @filterable (TS:332) - ValidateViaPath / InferViaSingleHop relationship @objectref (TS:522,709) - ValidateDataGridFilterValues @filterable, @filter (TS:1252,1260) - ValidateRelationships @through/@sourceRefField/@symmetric/@cardinality/@objectref (TS:1320-1324) - JunctionReferenceFkFields / CountJunctionReferences → effective ReferenceIdentities() (TS:1293-1294) - ValidateTemplatePayloadRefs @kind/@textRef/@subjectRef/@htmlBodyRef/@payloadRef/@requiredSlots (TS:169-253) - PrimaryAssignedFieldNames identity @generation (TS:142) - ValidateSourceParameterRef @parameterRef, param fields, referenced field (TS:76,122,139) - RegisteredValidation.Walk reference @attr (@objectRef/@through) (validation-registry.ts:66) The M:N RUNTIME resolver (M2MResolver + M2MDerivation) was already correct — it resolves via Relationships()/Fields()/ReferenceIdentities()/Children() and the resolving MetaRelationship/MetaIdentity attr getters. But the M:N VALIDATOR read junction references via OwnChildren, so a junction inheriting its two identity.reference children via extends was mis-counted (found 0 → spurious ERR_INVALID_RELATIONSHIP), disagreeing with the runtime derivation. Fixed + gated by a new test. Corrects the misleading "matches TS 1:1 via ownAttr" class doc-comments on ValidationPasses and RegisteredValidation (the claim was false at these sites). New tests (ValidationTests): M2N_junction_references_inherited_via_extends_are_counted (fails against pre-fix source — genuine regression guard) + M2N_relationship_inherited_via_extends_validates_cleanly (coherence guard). MetaObjects.Conformance.Tests 676/676 green; Codegen.Tests 258 pass /1 skip with one PRE-EXISTING, unrelated failure (ApiDocsCrossPortConformanceTests .CsharpApiDocsSurface_MatchesTheSharedManifest — fails identically on the pristine pre-change tree). Zero golden changes. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GF9xLEQZaPus5Y6opk398n
…herited M:N/dataGrid/template/param-ref Align the JVM (Java) validation passes to the TS reference (ADR-0039): where the TS oracle RESOLVES an effective property to decide correctness, the Java port must resolve too. Own-only reads on inheritable attrs/children wrongly skip or reject metadata inherited via `extends`. Sites flipped to resolving (each mirrors its TS counterpart): - ValidationPhase.validateRelationshipsM2M: @through/@sourceRefField now read via the resolving getters (getThrough/getSourceRefField); the junction's identity.reference view (countJunctionReferences / junctionReferenceFkFields) now iterates getIdentities() (effective), mirroring TS referenceIdentities() and M2MFields.derive. Own-only would count 0 references on a junction that inherits its two identity.reference children via extends and falsely reject the M:N (ERR_INVALID_RELATIONSHIP) at load. (getObjectRef/getCardinality/ isSymmetric already resolve.) - ValidationPhase.validateDataGridLayouts: the layout.dataGrid lookup + @defaultSortField/@filter now resolve (an inherited grid is now validated), mirroring TS validateDataGridSortFields/FilterValues. - ValidationPhase.validateTemplates: @format/@promptStyle/@kind + the email/ document part-refs (@subjectRef/@htmlBodyRef/@textRef) now resolve — templates CAN extend (consistent with the merged codegen template decision), mirroring TS validateTemplatePayloadRefs. (@payloadRef/@requiredSlots getters + the payload field-name collection already resolved.) - ValidationPhase.validateSourceParameterRef: @parameterRef, the parameter value-object's fields, and the referenced entity's field now resolve, mirroring TS validate-source-parameter-ref.ts (origin.* + @from stay own — origin never inherits, ADR-0029). - RegisteredValidation: the registry-driven reference-attr read (@objectref on relationships, @references on identity.reference) now resolves, mirroring TS validation-registry.ts — own-only would let a dangling INHERITED reference slip through the resolution check. Runtime/codegen M:N discovery was already resolving (getRelationships() no-arg, M2MFields.derive, the OMDB/Spring/Kotlin resolvers) — no runtime-drop bug there. Kotlin needs no source change: it shares the JVM metadata loader for validation and its M:N discovery resolves. Genuinely-own declaration checks (required-attr PRESENCE, @isAbstract, TPH markers, the junction's OWN identity.reference declarations, explicit projection-identity @fields, origin.*) are left own. Adds M2MSlimVocabularyTest regression coverage: an M:N whose junction inherits both identity.reference children via extends loads cleanly AND derives its FK direction (fails against the own-only iteration; confirmed). Java metadata 1067 + codegen-spring 158 + codegen-kotlin 266 green; Java integration 30 + Kotlin integration M2M 6 green (Testcontainers PG). No golden changes. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GF9xLEQZaPus5Y6opk398n
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The independent post-sweep review (Phase 2) confirmed TS/JVM/C# codegen clean, but caught a straggler class in the VALIDATION layer + one runtime path that the sweep missed: own-only reads where the TS reference resolves, dropping extends-inherited values.
Genuine bugs fixed (each with a regression test that fails pre-fix): M:N junction-reference validation counted own children, so a junction inheriting its two identity.reference children via extends counted 0 → false ERR_INVALID_RELATIONSHIP (all 3 non-TS ports); Python runtime n2m_resolver iterated own_children, dropping an inherited M:N relationship at runtime. Plus consistency-aligned to the TS oracle: dataGrid @defaultSortField/@filter, relationship @objectRef/@through/@Cardinality, identity @fields/@generation, source @parameterRef, registered-validation reference-attr, and the template-validation cluster (TS resolved; JVM+Python read own — same template question, validation layer). Left genuinely-own: origin.* (never inherits), @Discriminator, @dbColumnType, required-attr presence, identity-passthrough author-disagreement checks.
Python b6013c2 · C# 327cce8 · JVM a0fa1a9. TS needed no change (it is the oracle). Zero golden changes; per-port suites + Testcontainers green.
🤖 Generated with Claude Code