ReviewOS

home-lang/home

TypeScript checker parity: eliminate the remaining 30 failures

#69
Closed chris opened this 22 days ago · 717 comments
22 days ago

Parent: #67

Scope

Make ts_checker fully green by porting the remaining TypeScript semantics and exact diagnostic behavior. Do not weaken assertions or encode fixture-specific outcomes.

Current verified state

After fixing distributive-conditional union member ownership:

  • 2,976 / 3,006 pass
  • 30 fail
  • 0 crash (down from 2 crashes)

The crash fix copies union member IDs before recursive conditional evaluation so interner reallocations cannot invalidate a borrowed slice.

Remaining failure clusters

  • Enum/member literal assignment compatibility.
  • Global-script this indexed access diagnostics.
  • Generic defaults, conditional distribution, NonNullable, mapped and template types.
  • Variadic tuple inference and callback arity.
  • Overload diagnostic anchoring and CheckJS/JSDoc arity.
  • Accessor read/write and spread typing.
  • Alias/generic-instance diagnostic rendering.
  • Recursive type depth/circular mapped diagnostics.

Completion gate

  • ./pantry/.bin/zig build test -Dfilter=ts_checker --summary failures reports 3,006 / 3,006.
  • No crash, timeout, or skipped parity assertion replaces a failure.
  • Each semantic cluster lands in a focused conventional commit with regression coverage.
22 days ago

Type-only import-equals recovery parity

Landed fix(ts-parser): recover type-only import assignments (ff2fe5586).

What changed:

  • Ported TypeScript's parseImportDeclarationOrImportEqualsDeclaration disambiguation for a type-only imported identifier followed by *.
  • Malformed import type defer * as ns1 from "./a" now enters import-equals recovery and reports TS1005 ('=' expected) at *.
  • The recovery leaves as, ns1, and from available as identifier statements, preserving the upstream TS1005, TS1434, and TS2304 sequence.
  • The former import-clause TS1005 ('from' expected) and TS1141 path is no longer taken.
  • Added parser coverage for the missing-equals branch and its downstream recovery diagnostics.

Parity movement:

  • Full corpus: 5,043 -> 5,044 passing out of 5,907.
  • Failures: 864 -> 863.
  • Sole removed failure: importDeferTypeConflict1.
  • Added failures: 0.

Audit:

  • First half: 2,701/3,000, all 299 failure identities unchanged.
  • Second half: 2,342 -> 2,343 passing out of 2,907; failures 565 -> 564.
  • Combined failure list now has 863 entries.
  • Focused import-defer exact set: 5/5 passing.
  • ts_parser suite: passing.
  • ts_checker suite: passing.
  • ts_conformance suite: passing.

Diff size:

  • packages/ts_parser/src/ts_parser.zig: +57/-3.

Tracks #69.

22 days ago

Parity progress: typeofImportDefer now exact

Pushed e2115aecd (fix(ts-parser): recover deferred import type queries) to main.

TypeScript behavior ported

TypeScript's recoverable parseExpected flow does not stop after the missing ( in:

export type X = typeof import.defer("./a").Foo;

It reports the malformed import type, ends that type query at defer, and leaves ("./a").Foo to parse as a value expression. Home now mirrors that behavior:

  • TS1005 "(" expected at .defer
  • TS1141 String literal expected at .defer
  • TS1005 ")" expected at the following (
  • normal semantic member checking of ("./a").Foo, producing TS2339 against the literal type "./a"

The parser recovery preserves the trailing expression in HIR instead of hard-coding the semantic error. The checker also now retains a string literal's literal type while resolving a direct property access, so the diagnostic is produced through the normal member-lookup path with TypeScript's exact target text.

Coverage and verification

  • ts_parser suite: pass
  • ts_checker suite: pass
  • ts_conformance package suite: pass
  • focused exact import-defer run: 3/3
  • full exact identity audit: 5,045/5,907
  • failures: 863 -> 862
  • first half: 2,701/3,000 (unchanged)
  • second half: 2,343 -> 2,344/2,907
  • failure-set delta: removed only typeofImportDefer; zero additions or changed identities
  • change size: 2 files, +85/-0, including parser and checker regression tests

This follows the preceding import-defer chunks for dynamicImportDefer, importDeferComments, importDeferTypeConflict2, and importDeferTypeConflict1, completing the currently isolated typeof import.defer recovery gap while keeping the broader exact corpus stable.

22 days ago

Pushed 9f72dff (fix(ts-checker): stabilize recursive member lookup). Recursive union/intersection property lookup now snapshots constituent TypeIds before interning can reallocate the shared member pool, fixing the tsgo nestedSpreadsAndWidening poisoned-TypeId crash. Added an 11-way nested-spread regression. Verification: ts_checker passes; tsgo-native compiler testdata passes 299/299 exact; default ts_conformance passes smoke 16/16, category 86/86, and baseline-aware 586/586; inherited TypeScript parity remains 5,045/5,907 with all 862 failure identities unchanged across the exact 5,907-case audit. Delta: 1 file, +34/-2.

22 days ago

Pushed e154609 (chore(ts-parity): pin typescript-go upstream). Home now owns a reproducible recursive microsoft/typescript-go source pin at b8276f35cd288aa163fad0516b60ddaacec87ee7; tsgo pins the inherited TypeScript corpus at 4d4f005c8541e0255a9d8791205fdce326e462bc. Conformance and tsgo-derived generators default to _submodules/typescript-go, while explicit environment overrides remain available. This advances 107 tsgo commits and 23 nested TypeScript commits; the expanded corpus and error baselines remain unchanged at 5,907 cases. Generator audit: 139-option CLI table and diagnostic reachability (459 = 0 active + 4 effectively dead + 455 dead) are byte-identical. Verification: native compiler testdata 299/299 exact; smoke 16/16; category 86/86; baseline-aware 586/586; inherited exact 5,045/5,907 with all 862 failure identities unchanged. Delta: 9 files, +69/-53, including the new gitlink.

22 days ago

Parity wave: finite template-union inference

Pushed 8a369c718 (fix(ts-checker): preserve finite template unions) to main.

What changed

  • Stop treating aggregate literal flags on union/intersection containers as one scalar literal payload.
  • Enumerate finite template substitution unions under the existing 32-result bound.
  • Relate a finite template-valued argument as its complete source set, so every generated string must fit the complete target union.
  • Preserve all branches through generic inference while continuing to reject narrower subset targets.
  • Added a focused regression covering a three-member inferred template union plus a two-member rejection.

Exact parity movement

  • Flipped: templateLiteralTypes2
  • Full tsgo exact: 5,045/5,907 -> 5,046/5,907 (+1)
  • Failures: 862 -> 861
  • First half: 2,701/3,000 -> 2,702/3,000
  • Second half: unchanged at 2,344/2,907
  • Failure identity audit: templateLiteralTypes2 was the only removed identity; no failures were added, and all 563 second-half identities were unchanged.

Verification

  • ts_checker: 3,118 passing
  • Focused exact fixture: 1/1
  • Default smoke: 16/16
  • Category survey: 86/86
  • Baseline-aware survey: 586/586
  • Full exact identity audit: 5,907 fixtures

Change size: 1 file, +67/-4, including one focused regression test.

22 days ago

Parity wave: generic class constructor overloads

Pushed e1648d3ce (fix(ts-checker): retain generic constructor metadata) to main.

What changed

  • Preserve the owned class type-parameter list when bodyless constructor overload signatures are rebuilt with the completed instance return type.
  • Let overload applicability and strict argument checking infer enclosing class parameters instead of treating them as fixed unresolved targets.
  • Added a strict-null-check regression for duplicate generic class constructor overloads.

Exact parity movement

  • Flipped: constructSignaturesWithIdenticalOverloads
  • Flipped: constructSignaturesWithOverloads
  • Flipped: constructSignaturesWithOverloads2
  • Full tsgo exact: 5,046/5,907 -> 5,049/5,907 (+3)
  • Failures: 861 -> 858
  • First half: 2,702/3,000 -> 2,705/3,000
  • Second half: unchanged at 2,344/2,907
  • Failure identity audit: those three construct-signature fixtures were the only removed identities; no failures were added, and all 563 second-half identities were unchanged.

Verification

  • ts_checker: 3,119 passing
  • Focused exact construct-signature set: 4/4
  • Default smoke: 16/16
  • Category survey: 86/86
  • Baseline-aware survey: 586/586
  • Full exact identity audit: 5,907 fixtures

Change size: 1 file, +21/-0, including one strict generic-constructor regression test.

22 days ago

Parity wave: named class static TS2741 sources

Pushed de21f5e50 (fix(ts-checker): name class static missing-property types) to main.

What changed

  • Render registered class static source types as typeof ClassName in missing-property diagnostics before anonymous object-shape fallback.
  • Preserve class identity for constructor values assigned where an instance type is required.
  • Added focused coverage requiring typeof A and typeof B in TS2741 prose.

Exact parity movement

  • Flipped: stringIndexerConstrainsPropertyDeclarations2
  • Full tsgo exact: 5,049/5,907 -> 5,050/5,907 (+1)
  • Failures: 858 -> 857
  • First half: 2,705/3,000 -> 2,706/3,000
  • Second half: unchanged at 2,344/2,907
  • Failure identity audit: stringIndexerConstrainsPropertyDeclarations2 was the only removed identity; no failures were added, and all 563 second-half identities were unchanged.

Verification

  • ts_checker: 3,120 passing
  • Focused exact fixture: 1/1
  • Default smoke: 16/16
  • Category survey: 86/86
  • Baseline-aware survey: 586/586
  • Full exact identity audit: 5,907 fixtures

Change size: 1 file, +27/-0, including one focused TS2741 rendering regression.

22 days ago

Parity wave: contextual generic interface heritage

Pushed f98355a8c (fix(ts-checker): refine generic heritage inference) to main.

What changed

  • Treat an incomplete contextual-signature result as inconclusive only when both child and parent interface member signatures are generic.
  • Detect child signatures that capture outer interface type parameters before applying generic fallback.
  • Preserve free-child to concrete-parent return mismatches as real TS2430 errors.
  • Expanded positive and negative generic call/construct heritage coverage.

Exact parity movement

  • Flipped: callSignatureAssignabilityInInheritance4
  • Flipped: constructSignatureAssignabilityInInheritance4
  • Full tsgo exact: 5,050/5,907 -> 5,052/5,907 (+2)
  • Failures: 857 -> 855
  • First half: 2,706/3,000 -> 2,708/3,000
  • Second half: unchanged at 2,344/2,907
  • Failure identity audit: those two inheritance fixtures were the only removed identities; no failures were added, and all 563 second-half identities were unchanged.
  • During refinement, the full identity audit caught five over-suppressed TS2430 fixtures; the final focused set and audit restored all five before commit.

Verification

  • ts_checker: 3,120 passing
  • Focused exact inheritance set: 7/7
  • Default smoke: 16/16
  • Category survey: 86/86
  • Baseline-aware survey: 586/586
  • Full exact identity audit: 5,907 fixtures

Change size: 1 file, +19/-5, with positive and negative generic heritage coverage.

22 days ago

Parity wave: nested type-parameter assignment details

Pushed 1fb06c633 (fix(ts-checker): nest type parameter assignment details) to main.

What changed

  • Ported typescript-go's assignment relation-error chain shape: TS2322 remains the primary assignment diagnostic while TS5075/TS5082 are nested explanatory details.
  • Applied the same reporting path to concrete and type-parameter sources.
  • Preserved unconstrained source type-parameter related hints.
  • Added positive and negative coverage for literal, constrained-object, unrelated-generic, and captured primitive assignments.

Exact parity movement

  • Flipped: invalidNumberAssignments
  • Flipped: invalidVoidAssignments
  • Flipped: invalidBooleanAssignments
  • Flipped: invalidStringAssignments
  • Flipped: keyofAndIndexedAccess2
  • Flipped: derivedClassWithoutExplicitConstructor
  • Flipped: classWithoutExplicitConstructor
  • Full tsgo exact: 5,052/5,907 -> 5,059/5,907 (+7)
  • Failures: 855 -> 848
  • First half: 2,708/3,000 -> 2,715/3,000
  • Second half: unchanged at 2,344/2,907
  • Failure identity audit: exactly those seven identities were removed; no failures were added, and all 563 second-half identities were unchanged.
  • typeParameterAssignability2 now matches all TS5075/TS5082 primary-code positions; it remains open only for separate assignability extras.

Verification

  • ts_checker: 3,121 passing
  • Focused invalid exact sweep: all four primitive assignment fixtures pass
  • Default smoke: 16/16
  • Category survey: 86/86
  • Baseline-aware survey: 586/586
  • Full exact identity audit: 5,907 fixtures

Change size: 1 file, +89/-25, including positive and negative diagnostic-chain coverage.

22 days ago

Parity wave: computed TS2564 bracket anchors

Pushed 553b9afff (fix(ts-checker): anchor computed field diagnostics) to main.

What changed

  • Routed strict-property-initialization diagnostics for computed class fields through the shared bracket-position lookup.
  • Matched typescript-go by anchoring TS2564 at the opening [ for resolved and dynamic computed keys.
  • Preserved ordinary field anchors and existing same-position diagnostic ordering.
  • Added a focused unique-symbol computed-field position regression.

Exact parity movement

  • Flipped: uniqueSymbols
  • Flipped: uniqueSymbolsDeclarations
  • Flipped: instanceMemberWithComputedPropertyName2
  • Flipped: parserComputedPropertyName28
  • Flipped: parserComputedPropertyName29
  • Flipped: parserComputedPropertyName31
  • Flipped: parserComputedPropertyName36
  • Flipped: parserSymbolProperty5
  • Flipped: parserES5SymbolProperty5
  • Flipped: symbolDeclarationEmit1
  • Flipped: symbolProperty6
  • Flipped: symbolProperty7
  • Full tsgo exact: 5,059/5,907 -> 5,071/5,907 (+12)
  • Failures: 848 -> 836
  • First half: 2,715/3,000 -> 2,724/3,000
  • Second half: 2,344/2,907 -> 2,347/2,907
  • Failure identity audit: exactly those twelve identities were removed; no failures were added.

Verification

  • ts_checker: 3,122 passing
  • Focused uniqueSymbols family: 6/7 exact; only the unrelated uniqueSymbolsPropertyNames duplicate-property fixture remains
  • Default smoke: 16/16
  • Category survey: 86/86
  • Baseline-aware survey: 586/586
  • Full exact identity audit: 5,907 fixtures

Change size: 1 file, +28/-3, with an explicit opening-bracket position regression.

22 days ago

Parity wave: stable unique-symbol object keys

Pushed e49bd1e63 (fix(ts-checker): detect stable symbol key duplicates) to main.

What changed

  • Preserved late-bound unique-symbol identity for object-literal keys reached through interface-typed member access.
  • Reused the existing canonical computed-symbol identity for calls returning typeof a unique-symbol binding.
  • Reported TS1117 only for repeated data properties with the same proven stable key.
  • Kept broad symbol variables and calls returning plain symbol dynamic and duplicate-free.

Exact parity movement

  • Flipped: uniqueSymbolsPropertyNames
  • Focused uniqueSymbols family: 6/7 -> 7/7 exact
  • Full tsgo exact: 5,071/5,907 -> 5,072/5,907 (+1)
  • Failures: 836 -> 835
  • First half: 2,724/3,000 -> 2,725/3,000
  • Second half: unchanged at 2,347/2,907
  • Failure identity audit: uniqueSymbolsPropertyNames was the only removed identity; no failures were added.

Verification

  • ts_checker: 3,123 passing
  • Focused uniqueSymbols family: 7/7 exact
  • Default smoke: 16/16
  • Category survey: 86/86
  • Baseline-aware survey: 586/586
  • Full exact identity audit: 5,907 fixtures

Change size: 1 file, +54/-1, with HIR-backed member identity and broad-symbol negative coverage.

22 days ago

Parity batch: computed object index signatures

Pushed 7ed99c02e (fix(ts-checker): relate computed object indexers) to main.

This ports tsgo's structural treatment of fresh object-literal indexers into the checker-owned object-literal assignment path. Source string, number, and symbol indexers are now compared directly with target index signatures before named-member checks. The scope deliberately stays out of the general relation engine, preserving specialized array and class behavior.

Exact fixture movement

  • Flipped 6 fixtures: computedPropertyNamesContextualType8_ES5, computedPropertyNamesContextualType8_ES6, computedPropertyNamesContextualType9_ES5, computedPropertyNamesContextualType9_ES6, computedPropertyNamesContextualType10_ES5, and computedPropertyNamesContextualType10_ES6.
  • First-half failure identity set remained unchanged.
  • Second-half failure identity set removed exactly those six fixtures with no additions.
  • Exact parity: 5,072/5,907 -> 5,078/5,907.
  • Remaining exact mismatches: 829.

Verification

  • ts_checker: 3,124/3,124
  • focused exact contextual computed-property family: 20/20
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • full exact split audit: 2,725/3,000 + 2,353/2,907 = 5,078/5,907

Change size: +39/-0 in packages/ts_checker/src/check.zig.

22 days ago

Parity batch: computed class-key TDZ

Pushed 5e829ad48 (fix(ts-checker): report computed class key tdz) to main.

Direct computed class member keys now treat the class's own binding as uninitialized, matching tsgo's TS2449 behavior. The reference walk visits bracketed getter, setter, and method names but still stops before deferred function bodies. Ordinary class self-references in decorators and static initializers remain legal, and nested object keys cannot be mistaken for direct class members.

Exact fixture movement

  • Flipped computedPropertyNamesWithStaticProperty.
  • Restored all 6 expected TS2449 diagnostics across the named class declaration and named class expression.
  • First-half failure identity set remained unchanged.
  • Second-half failure identity set removed exactly this fixture with no additions.
  • Exact parity: 5,078/5,907 -> 5,079/5,907.
  • Remaining exact mismatches: 828.

Verification

  • ts_checker: 3,125/3,125
  • focused exact fixture: 1/1
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • full exact split audit: 2,725/3,000 + 2,354/2,907 = 5,079/5,907

Change size: +74/-2 in packages/ts_checker/src/check.zig.

22 days ago

Parity batch: inferred computed class indexers

Pushed 38702f9e4 (fix(ts-checker): infer computed class indexers) to main.

Simple broad computed class field keys now establish instance or static index signatures as the class body is checked. Later dynamic fields are compared against those accumulated constraints, while invalid complex computed keys consume existing constraints without redefining the class shape. Numeric-key conflicts report number-index compatibility before string-index compatibility, matching tsgo ordering. TS2564 also renders computed field names from source text, preserving brackets around numeric keys such as [0].

Exact fixture movement

  • Flipped computedPropertyNames12_ES5 and computedPropertyNames12_ES6.
  • Restored both expected TS2411 diagnostics on [+s] in number-then-string order.
  • Restored Property '[0]' TS2564 rendering.
  • First-half failure identity set remained unchanged.
  • Second-half failure identity set removed exactly these two fixtures with no additions.
  • Exact parity: 5,079/5,907 -> 5,081/5,907.
  • Remaining exact mismatches: 826.

Verification

  • ts_checker: 3,126/3,126
  • focused exact family: 2/2
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • full exact split audit: 2,725/3,000 + 2,356/2,907 = 5,081/5,907

Change size: +77/-15 in packages/ts_checker/src/check.zig.

22 days ago

Parity checkpoint: callable assertion contextual typing

Pushed 58ec543d7 (fix(ts-checker): contextualize asserted callables) directly to main.

This ports TypeScript's target-first contextual typing behavior for arrow/function operands of type assertions:

  • lowers the asserted callable target before checking the operand;
  • supplies target parameter types to unannotated source parameters;
  • preserves the operand body's inferred return type and source parameter names in the source signature;
  • routes incompatible source/target return signatures through TS2352 overlap instead of emitting TS7006 or TS2322;
  • narrows the global Function overlap exemption to the exact builtin object;
  • anchors parenthesized function-like assertion diagnostics at the opening parenthesis, matching the upstream baseline without changing primitive assertion anchors.

Parity movement:

  • asOperatorContextualType: failing -> passing
  • exact full corpus: 5,081 -> 5,082 / 5,907 (+1)
  • failures: 826 -> 825
  • regression identity audit: zero regressions; the only removed failure identity was asOperatorContextualType
  • change size: +75/-4 in packages/ts_checker/src/check.zig

Verification:

  • ts_checker: 3,127/3,127
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • exact split audit: 2,725/3,000 and 2,357/2,907

The canonical conformance source remains the pinned microsoft/typescript-go (tsgo) main submodule.

22 days ago

Parity checkpoint: nested namespace assignment and value merging

Pushed 4389ddeea (fix(ts-checker): merge nested namespace values) directly to main.

This closes the four missing diagnostics in assignmentToParenthesizedIdentifiers by porting the relevant nested namespace value behavior:

  • resolves bare nested namespace assignment targets through the enclosing namespace scope, producing TS2631 for M3 = ...;
  • preserves exported nested namespaces as precise value objects instead of collapsing them to any;
  • carries { x: number } through M2.M3, (M2).M3, and (M2.M3) assignment targets, producing the three upstream TS2322 diagnostics;
  • merges nested namespace value contributions across virtual files by intersecting each matching object shape, so declarations such as A.Utils.mirror and A.Utils.Plane remain jointly visible.

The first full audit caught two temporary regressions in cross-file namespace-merging fixtures because the initial precise implementation returned only the first A.Utils contribution. The final implementation adds a focused virtual-file regression and restores both identities before landing.

Parity movement:

  • assignmentToParenthesizedIdentifiers: failing -> passing
  • exact full corpus: 5,082 -> 5,083 / 5,907 (+1)
  • failures: 825 -> 824
  • final identity audit: zero regressions; the only removed failure identity was assignmentToParenthesizedIdentifiers
  • change size: +99/-1 in packages/ts_checker/src/check.zig

Verification:

  • ts_checker: 3,129/3,129
  • focused target: 1/1 exact
  • focused TwoInternalModules* cluster: both audit regressions restored; pre-change failure identities preserved
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • exact split audit: 2,725/3,000 and 2,358/2,907

The canonical conformance source remains the pinned microsoft/typescript-go (tsgo) main submodule.

22 days ago

Parity update: paired accessor contextual returns

Pushed 3fb6f3167 (fix(ts-checker): contextualize getter returns).

What changed

  • Read paired setter parameter types from the class accessor pre-scan.
  • Use that type as the contextual return contract for an unannotated getter.
  • Carry an explicit paired-setter context flag through getter body checking so incompatible primitive returns produce TypeScript's TS2322 diagnostic.
  • Preserve nested function state, explicitly annotated getters, untyped setter pairs, and compatible getter returns.

Exact parity movement

  • getSetAccessorContextualTyping: failing -> passing
  • Passing fixtures: 5,083 -> 5,084 / 5,907 (+1)
  • Remaining failures: 824 -> 823
  • Regressions: 0
  • Source delta: 1 checker file, +51/-1

Verification

  • ts_checker: 3,130/3,130
  • Focused exact fixture: 1/1
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Full 5,907-fixture identity audit: 2,725/3,000 + 2,359/2,907 = 5,084/5,907

The canonical conformance source remains the microsoft/typescript-go (tsgo) main submodule.

22 days ago

Parity update: contextual fixed tuple spreads

Pushed 5efce8bfa (fix(ts-checker): expand fixed tuple spreads).

What changed

  • Reconstruct contextual array literal diagnostics by expanding a genuine fixed tuple spread into its known positional element types.
  • Compute tuple width diagnostics from the expanded arity rather than the array literal's raw syntax element count.
  • Require a recorded tuple origin before flattening, so custom array-like objects with literal length members are not mistaken for tuples.
  • Preserve the existing widened T[] diagnostic path for open-ended array spreads.
  • Added a two-path checker regression covering both fixed tuple expansion and unchanged open-array widening.

Exact parity movement

  • arrayLiteralExpressionContextualTyping: failing -> passing
  • Passing fixtures: 5,084 -> 5,085 / 5,907 (+1)
  • Remaining failures: 823 -> 822
  • Regressions: 0
  • Source delta: 1 checker file, +66/-13

Verification

  • ts_checker: 3,131/3,131
  • Focused exact fixture: 1/1
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Full 5,907-fixture identity audit: 2,725/3,000 + 2,360/2,907 = 5,085/5,907
  • Failure-name comparison: the sole removal was arrayLiteralExpressionContextualTyping

The canonical conformance source remains the microsoft/typescript-go (tsgo) main submodule.

22 days ago

Parity update: binding-pattern tuple origins

Pushed 3657b14f0 (fix(ts-checker): preserve binding tuple origins).

What changed

  • Register synthesized array binding-pattern parameter types as tuple origins at construction time.
  • Nested structural diagnostics now render those types with tuple syntax such as [any, any] instead of leaking { length: 2; 0: any; 1: any; }.
  • Added an exact nested TS2345 regression for an inferred destructured parameter with a tuple-valued property.

Exact parity movement

  • argumentExpressionContextualTyping: failing -> passing
  • destructuringParameterDeclaration3ES6: failing -> passing
  • destructuringParameterDeclaration3ES5iterable: failing -> passing
  • destructuringParameterDeclaration3ES5: failing -> passing
  • Passing fixtures: 5,085 -> 5,089 / 5,907 (+4)
  • Remaining failures: 822 -> 818
  • Regressions: 0
  • Source delta: 1 checker file, +20/-1

Verification

  • ts_checker: 3,132/3,132
  • Focused exact fixture: 1/1
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Full 5,907-fixture identity audit: 2,725/3,000 + 2,364/2,907 = 5,089/5,907
  • Failure-name comparison: exactly the four fixtures above were removed; no names were added

The canonical conformance source remains the microsoft/typescript-go (tsgo) main submodule.

22 days ago

Parity update: shorthand excess-property precedence

Pushed c8336be89 (fix(ts-checker): prioritize shorthand excess properties).

What changed

  • Run fresh-object excess-property checking before the generic missing-property fallback for failed variable initializers.
  • Suppress the competing TS2741 only when the direct excess property uses shorthand syntax.
  • Preserve missing outer-property diagnostics when an excess property is nested or explicitly declared.
  • Added a shorthand data-property regression while retaining the existing callable-excess and generic namespace regressions.

Exact parity movement

  • objectLiteralShorthandPropertiesAssignmentError: failing -> passing
  • objectLiteralShorthandPropertiesAssignmentErrorFromMissingIdentifier: failing -> passing
  • Passing fixtures: 5,089 -> 5,091 / 5,907 (+2)
  • Remaining failures: 818 -> 816
  • Regressions: 0
  • Source delta: 1 checker file, +30/-1

Verification

  • ts_checker: 3,133/3,133
  • Focused exact fixtures: 2/2
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Full 5,907-fixture identity audit: 2,725/3,000 + 2,366/2,907 = 5,091/5,907
  • Failure-name comparison: exactly the two fixtures above were removed; no names were added

The canonical conformance source remains the microsoft/typescript-go (tsgo) main submodule.

22 days ago

Parity checkpoint: computed enum accessor keys

Pushed 67e41823d (fix(ts-checker): canonicalize enum accessor keys) to main.

This ports tsgo's canonical computed-property behavior into paired accessor contextual typing:

  • paired getter lookup now uses the same canonical class member-name resolver as the accessor pre-scan;
  • recorded numeric enum values are converted to their runtime string property keys;
  • recorded string enum values are used directly as runtime property keys;
  • computed enum setter/getter pairs therefore share the setter's contextual return type during getter checking.

Exact parity movement

  • Flipped constEnumPropertyAccess1 from failing to passing.
  • Advanced from 5,091/5,907 to 5,092/5,907 exact fixtures.
  • Reduced failures from 816 to 815 (+1, zero regressions).
  • Lower shard remained identity-stable at 2,725/3,000 with the same 275 failures.
  • Upper shard advanced to 2,367/2,907; its only removed failure identity was constEnumPropertyAccess1.

Verification

  • ts_checker: 3,133/3,133
  • focused exact fixture: 1/1
  • smoke gate: 16/16
  • category gate: 86/86
  • baseline-aware gate: 586/586
  • full exact identity audit: 5,092/5,907

Source delta: 1 file, +14/-3. Commit author and committer are Chris <chrisbreuer93@gmail.com> with no coauthor trailers.

22 days ago

Parity checkpoint: type-only namespace value access

Pushed 2c49b6fa4 (fix(ts-checker): exclude type-only namespace values) to main.

This ports tsgo's type-space/value-space distinction for namespace imports:

  • named-export discovery remains broad enough to validate imports of type-only exports;
  • namespace value property access now rejects definitive type_only_decl and type_only_alias statuses;
  • runtime values and ambient const enums remain accessible;
  • missing/unknown resolutions retain the existing fallback scan.

This covers direct export type { A } and transitive export type * paths without changing ordinary namespace runtime exports.

Exact parity movement

  • Flipped namespaceMemberAccess from failing to passing.
  • Also flipped exportNamespace11 and exportNamespace12, which exercise the same rule through export type * re-exports.
  • Advanced from 5,092/5,907 to 5,095/5,907 exact fixtures.
  • Reduced failures from 815 to 812 (+3, zero regressions).
  • Lower shard remained identity-stable at 2,725/3,000 with the same 275 failures.
  • Upper shard advanced from 2,367/2,907 to 2,370/2,907; the only removed identities were the three fixtures above.

Verification

  • ts_checker: 3,134/3,134
  • focused exact fixture: 1/1
  • smoke gate: 16/16
  • category gate: 86/86
  • baseline-aware gate: 586/586
  • full exact identity audit: 5,095/5,907

Source delta: 1 file, +21/-1. Commit author and committer are Chris <chrisbreuer93@gmail.com> with no coauthor trailers.

22 days ago

Parity checkpoint: object-literal switch cases

Pushed 8511cf7b2 (fix(ts-checker): contextualize object switch cases) to main.

This ports the observable result of tsgo's reverse switch comparability elaboration:

  • fresh object-literal case expressions are checked for excess properties against the switch discriminant type;
  • ordinary case overlap checking and discriminant narrowing remain unchanged;
  • permissive targets such as empty object shapes still follow the existing excess-property exemptions.

The upstream switchStatements fixture now reports TS2353 on the name property of case { id: 12, name: '' } against class C, with the exact tsgo anchor and message.

Exact parity movement

  • Flipped switchStatements from failing to passing.
  • Advanced from 5,095/5,907 to 5,096/5,907 exact fixtures.
  • Reduced failures from 812 to 811 (+1, zero regressions).
  • Lower shard remained identity-stable at 2,725/3,000 with the same 275 failures.
  • Upper shard advanced from 2,370/2,907 to 2,371/2,907; its only removed identity was switchStatements.

Verification

  • ts_checker: 3,135/3,135
  • focused exact fixture: 1/1
  • smoke gate: 16/16
  • category gate: 86/86
  • baseline-aware gate: 586/586
  • full exact identity audit: 5,096/5,907

Source delta: 1 file, +19/-0. Commit author and committer are Chris <chrisbreuer93@gmail.com> with no coauthor trailers.

22 days ago

Parity checkpoint: noImplicitAny indexed-access variants

Pushed e60ef187d (fix(ts-parity): align indexed access option variants) to main.

This closes both sides of tsgo's noImplicitAny option matrix for literal element access:

  • the conformance loader now applies the noimplicitany=true|false value encoded by the selected variant baseline, matching its existing isolated/verbatim baseline overrides;
  • false mode permits a missing quoted element access and infers any;
  • true mode emits TS7053 with a nested TS2339 missing-property explanation instead of emitting TS2339 as the top-level diagnostic;
  • union-index TS7053 property chains retain their existing behavior.

Exact parity movement

  • Flipped propertyAccessStringIndexSignature from failing to passing.
  • Also flipped lateBoundAssignmentDeclarationSupport4 and lateBoundAssignmentDeclarationSupport5, which expected the same true-mode TS7053 shape.
  • Advanced from 5,096/5,907 to 5,099/5,907 exact fixtures.
  • Reduced failures from 811 to 808 (+3, zero regressions).
  • Lower shard advanced from 2,725/3,000 to 2,727/3,000 by removing the two late-bound fixtures.
  • Upper shard advanced from 2,371/2,907 to 2,372/2,907 by removing propertyAccessStringIndexSignature.

Verification

  • ts_checker: 3,136/3,136
  • focused exact fixture: 1/1
  • smoke gate: 16/16
  • category gate: 86/86
  • baseline-aware gate: 586/586
  • full exact identity audit: 5,099/5,907

Source delta: 2 files, +48/-4. Commit author and committer are Chris <chrisbreuer93@gmail.com> with no coauthor trailers.

22 days ago

Parity checkpoint: consumed generic yield contexts

Pushed ecbca98b8 (fix(ts-checker): diagnose generic yield contexts) to main.

This ports tsgo's TS7057 decision for consumed yield results in unannotated generators:

  • a yield passed to an any, unresolved, or unfixed type-parameter call argument has no concrete contextual next type and reports TS7057;
  • a concrete overload parameter supplies context and suppresses TS7057;
  • existing nested-yield and implicit-any identifier paths are deduplicated against the new call-context path;
  • the rule remains gated by noImplicitAny and by the existing unannotated-generator pass.

The generatorReturnTypeInference fixture now reports the exact TS7057 at f2(yield 1) while leaving f1(yield 1) clean.

Exact parity movement

  • Flipped generatorReturnTypeInference from failing to passing.
  • Advanced from 5,099/5,907 to 5,100/5,907 exact fixtures.
  • Reduced failures from 808 to 807 (+1, zero regressions).
  • Lower shard remained identity-stable at 2,727/3,000 with 273 failures.
  • Upper shard advanced from 2,372/2,907 to 2,373/2,907; its only removed identity was generatorReturnTypeInference.

Verification

  • ts_checker: 3,137/3,137
  • focused exact fixture: 1/1
  • smoke gate: 16/16
  • category gate: 86/86
  • baseline-aware gate: 586/586
  • full exact identity audit: 5,100/5,907

Source delta: 1 file, +49/-0. Commit author and committer are Chris <chrisbreuer93@gmail.com> with no coauthor trailers.

22 days ago

TypeScript parity progress: optional-chain false-branch narrowing

Pushed 805b32d83 (fix(ts-checker): preserve optional chain false branches).

What changed

  • Negative optional-chain discriminant comparisons now narrow receiver unions per variant instead of dropping every non-object constituent.
  • Nullish receivers survive when optional short-circuit undefined makes the comparison fail.
  • Optional members remain possible when absence can make the comparison fail.
  • Object variants are removed only when their discriminant definitely matches the compared literal.
  • Loose == null / != null comparisons account for JavaScript's undefined == null behavior.
  • Added paired || / && checker coverage so options?.a === false || options.b reports TS18048 while the true branch used by && still establishes a defined receiver.

Parity movement

  • Flipped controlFlowOptionalChain3 from failing to passing: +1 fixture
  • Exact tsgo parity: 5,100/5,907 -> 5,101/5,907
  • Remaining failures: 807 -> 806
  • Full failure-set comparison: only controlFlowOptionalChain3 was removed; zero additions/regressions
  • Source delta: 1 checker file, +92/-3

Verification

  • ts_checker: 3,138/3,138
  • Focused controlFlowOptionalChain2: 1/1
  • Focused controlFlowOptionalChain3: 1/1
  • Smoke gate: 16/16
  • Category gate: 86/86
  • Baseline-aware gate: 586/586
  • Full exact audit: 2,727/3,000 + 2,374/2,907 = 5,101/5,907

The worktree is otherwise clean apart from the pre-existing untracked .claude/ directory, which remains untouched.

22 days ago

TS parity update: inline JSDoc cast semantics

Pushed b5fafe9fdfadc3338849a963e7dda6828fbe38c7 (fix(ts-checker): apply inline JSDoc cast types) directly to main.

What changed

  • Recovered checked-JavaScript /** @type {...} */ (expr) casts from source wrappers after parser lowering removes comments and parentheses.
  • Applied nested casts inner-to-outer so the outer annotation wins, matching TypeScript.
  • Routed module-only JSDoc import types through existing export = target resolution before CommonJS object synthesis.
  • Kept type/value spaces distinct: import("./mod") resolves the exported instance type; typeof import("./mod") resolves the expression/static value type.
  • Added two focused checker regressions covering nested inline casts and export = import-type casts.

Scope

  • 1 file changed
  • 96 lines touched (+93/-3)
  • 2 checker regression tests added

Exact parity movement

  • Before: 5,101/5,907, 806 failures
  • After: 5,102/5,907, 805 failures
  • Flipped exact: assertionTypePredicates2
  • Full failure-identity diff: exactly that fixture removed, zero regressions added

Verification

  • ts_checker: pass
  • Focused exact assertionTypePredicates2: 1/1
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • Exact lower split: 2,727/3,000
  • Exact upper split: 2,375/2,907
  • Full inherited TypeScript exact audit: 5,102/5,907
22 days ago

TS parity update: positive guard definite-assignment flow

Pushed a98e02d8662ce29f52f3e60ef0b77da945993952 (fix(ts-checker): honor guard proofs in assignment flow) directly to main.

What changed

  • Branch-local TS2454 state now honors positive facts from declared type-predicate calls.
  • typeof equality/inequality guards clear pending state only in the branch where the tested runtime tag is proven.
  • Unary ! carries the branch polarity through to the guarded predicate.
  • Arbitrary boolean-returning calls are not treated as predicates; resolution uses existing fn_predicates metadata.
  • The guard-site TS2454, negative-branch TS2454, and outer post-if pending state are preserved.
  • Added a focused checker regression covering predicate and typeof positive branches.

Scope

  • 1 file changed
  • 67 lines added
  • 1 checker regression test added

Exact parity movement

  • Before: 5,102/5,907, 805 failures
  • After: 5,108/5,907, 799 failures
  • Flipped exact:
    • typeGuardEnums
    • typeGuardNarrowsPrimitiveIntersection
    • typeGuardNarrowsToLiteralType
    • typeGuardNarrowsToLiteralTypeUnion
    • typeGuardsInExternalModule
    • typeGuardsInGlobal
  • Full failure-identity diff: exactly those six fixtures removed, zero regressions added

Verification

  • ts_checker: pass
  • Focused exact guard cluster: 3/3
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • Exact lower split: 2,727/3,000
  • Exact upper split: 2,381/2,907
  • Full inherited TypeScript exact audit: 5,108/5,907
22 days ago

TS parity update: mismatched JSDoc heritage resolution

Pushed 93d649df06913f2b4101191223c85c5ee8c9a9c0 (fix(ts-checker): resolve mismatched JSDoc heritage types) directly to main.

What changed

  • Mismatched attached @extends / @augments targets now pass through the normal JSDoc type resolver.
  • An unresolved tag target emits TS2304 before the existing TS8023 heritage mismatch, matching upstream ordering and source position.
  • Declared alternate bases still emit only TS8023.
  • Matching JSDoc/class heritage names remain on the unchanged path.
  • Extended the existing checker mismatch regression to require the unresolved-name diagnostic.

Scope

  • 1 file changed
  • 4 lines added: 3 implementation lines and 1 regression assertion

Exact parity movement

  • Before: 5,108/5,907, 799 failures
  • After: 5,109/5,907, 798 failures
  • Flipped exact: extendsTagEmit
  • Full failure-identity diff: exactly that fixture removed, zero regressions added

Verification

  • ts_checker: pass
  • Focused exact extendsTagEmit: 1/1
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • Exact lower split: 2,727/3,000
  • Exact upper split: 2,382/2,907
  • Full inherited TypeScript exact audit: 5,109/5,907
22 days ago

TypeScript parity update: direct CommonJS aliases in JSDoc

Pushed b814f38ee335c4cbb29570e940bbee3b0f8a501c (fix(ts-checker): resolve direct require aliases in JSDoc) to main.

This ports tsgo behavior for whole-name JSDoc references backed by direct require() aliases. The checker now:

  • preserves a concrete local/late-bound JSDoc type before attempting module synthesis;
  • resolves module.exports = NamedClass and anonymous exported class expressions to their instance types;
  • falls back through export-assignment and existing CommonJS object synthesis only when needed;
  • keeps qualified require(...).member aliases on their established path;
  • verifies both member access and strict undefined rejection, so the parity win is not produced through any.

Exact parity movement

  • Before: 5,109/5,907 exact, 798 failures
  • After: 5,111/5,907 exact, 796 failures
  • Flipped: jsdocImportType, jsdocImportType2
  • Added failure identities: none
  • Lower split: 2,727/3,000, unchanged at 273 failures
  • Upper split: 2,384/2,907, improved to 523 failures

The first broad candidate exposed jsDeclarationsTypedefAndLatebound; the final ordering preserves its concrete local LazySet type, and it remains exact. varRequireFromTypescript also remains exact as the qualified-module guard.

Scope

  • 1 checker file
  • 84 additions, 4 removals
  • 1 focused regression test covering named and anonymous CommonJS class exports

Verification

  • ts_checker: 3,142/3,142
  • focused exact set: 8/8
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • exact corpus identity audit: 5,111/5,907, with exactly the two intended removals and zero additions

Next frontier: 796 exact-parity failures remain against the canonical typescript-go/tsgo corpus.

22 days ago

TypeScript parity update: invalid @augments diagnostic anchor

Pushed 25c57be889835e7cfd76aaf68c5d8e77bf0d5657 (fix(ts-checker): anchor invalid augments on declarations) to main.

This ports tsgo diagnostic placement for TS8022 when @augments/@extends is attached to a non-class function. Home previously pointed inside the JSDoc tag; it now anchors on the function declaration name while leaving the separate unattached-comment path unchanged.

Exact parity movement

  • Before: 5,111/5,907 exact, 796 failures
  • After: 5,112/5,907 exact, 795 failures
  • Flipped: jsdocAugments_notAClass
  • Added failure identities: none
  • Lower split: 2,727/3,000, unchanged at 273 failures
  • Upper split: 2,385/2,907, improved to 522 failures

Scope

  • 1 checker file
  • 2 additions, 2 removals
  • Existing unit assertion updated from the tag position to the declaration-name position

Verification

  • ts_checker: 3,142/3,142
  • focused exact: 1/1
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • exact corpus identity audit: 5,112/5,907, exactly one intended removal and zero additions

Next frontier: 795 exact-parity failures remain against the canonical typescript-go/tsgo corpus.

22 days ago

TypeScript parity update: weighted override spelling suggestions

Pushed 55d49646ffafd568ddde635f660ea9db782d1646 (fix(ts-checker): weight override spelling substitutions) to main.

This ports tsgo's spelling-suggestion distance for class override members. Home previously used ordinary case-insensitive Levenshtein distance; tsgo charges substitutions as two edits, insertions/deletions as one, and case-only changes as zero.

That distinction matters at short names: baz → bar has weighted distance 2 and exceeds the three-letter suggestion threshold, so JSDoc override diagnostics stay on TS4122. Longer doSomethang → doSomething remains within threshold and continues to use TS4123. The metric is scoped to override-member suggestions.

Exact parity movement

  • Before: 5,112/5,907 exact, 795 failures
  • After: 5,113/5,907 exact, 794 failures
  • Flipped: jsdocOverrideTag1
  • Preserved exact: override_js4
  • Added failure identities: none
  • Lower split: 2,727/3,000, unchanged at 273 failures
  • Upper split: 2,386/2,907, improved to 521 failures

Scope

  • 1 checker file
  • 40 additions, 1 removal
  • Documented weighted-distance helper plus one focused short-name regression test

Verification

  • ts_checker: 3,143/3,143
  • focused exact: 2/2
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • exact corpus identity audit: 5,113/5,907, exactly one intended removal and zero additions

Next frontier: 794 exact-parity failures remain against the canonical typescript-go/tsgo corpus.

22 days ago

TS parity: circular type-only alias cycles

Pushed 8bc137c47 (fix(ts-checker): resolve circular type-only aliases) directly to main.

What changed

  • Ported tsgo's alias-target cycle behavior for named type-only imports and re-exports.
  • Resolves relative virtual-module export aliases across files.
  • Collapses local export type { X as Y } aliases through the local import binding, matching tsgo's symbol indirection.
  • Emits TS2303 on the alias whose target closes the active resolution chain.
  • Deduplicates the diagnostic across all members of the same cycle.
  • Keeps runtime/value re-export cycles outside this rule.
  • Added focused tests for direct type-only re-export cycles and import-plus-local-re-export cycles.

Exact parity movement

  • Overall: 5,113 -> 5,115 / 5,907
  • Failures: 794 -> 792
  • Lower half: 2,727 / 3,000, unchanged at 273 failures
  • Upper half: 2,386 -> 2,388 / 2,907, failures 521 -> 519
  • Flipped fixtures: circular1, circular3
  • Failure identity audit: no additions or other removals
  • Scope: 187 insertions in one checker file, including two regression tests

Verification

  • Focused exact fixtures: 2/2
  • ts_checker: 3,145/3,145
  • ts_conformance: 1,374/1,374
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Full exact TypeScript corpus: 5,115/5,907

Next I am ranking the remaining 792 failures by the smallest coherent diagnostic families and will continue landing independently verified conventional commits.

22 days ago

TS parity: distinct ArrayBuffer and SharedArrayBuffer shapes

Pushed 13f217d2d (fix(ts-checker): model shared array buffer identity) directly to main.

What changed

  • Ported the structural ArrayBuffer and SharedArrayBuffer instance shapes from TypeScript's libs.
  • new ArrayBuffer(...) and new SharedArrayBuffer(...) now produce modeled instances instead of the same empty object type.
  • Added standard byteLength/slice members and distinct Symbol.toStringTag literal members, which makes the two buffer types structurally non-assignable like tsgo.
  • Added modern ArrayBuffer resizable/transferable members and SharedArrayBuffer growable members.
  • Preserved valid modern APIs caught by the broad default suite: transfer, detached, resizable, resize, maxByteLength, growable, and grow.
  • Added a focused regression test covering valid shared-buffer members, invalid .length, and SharedArrayBuffer-to-ArrayBuffer assignment.

The first default-suite run correctly exposed missing modern ArrayBuffer members in transferableArrayBuffer and resizableArrayBuffer; those shapes were completed before the final audit.

Exact parity movement

  • Overall: 5,115 -> 5,117 / 5,907
  • Failures: 792 -> 790
  • Lower half: 2,727 -> 2,729 / 3,000, failures 273 -> 271
  • Upper half: 2,388 / 2,907, unchanged at 519 failures
  • Flipped fixtures: assignSharedArrayBufferToArrayBuffer, useSharedArrayBuffer6
  • Failure identity audit: no additions or other removals
  • Remaining buffer diagnostic deltas in the exact dump: none
  • Scope: 67 insertions, 1 deletion in one checker file

Verification

  • Focused exact fixtures: 2/2
  • ArrayBuffer default regressions: 4/4
  • ts_checker: 3,146/3,146
  • ts_conformance: 1,374/1,374
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Full exact TypeScript corpus: 5,117/5,907

Next I am ranking the remaining 790 failures for another small tsgo behavior family.

22 days ago

TS parity: ES5 async arguments restriction

Pushed d9aad604a (fix(ts-checker): reject arguments in es5 async functions) directly to main.

What changed

  • Added TS2522 for arguments references in async functions/methods downleveled to ES5.
  • Restricts the diagnostic to a nearest enclosing function that is async and non-arrow; async arrows retain the existing TS2496 path.
  • Preserves the IArguments type after reporting so downstream call/assignment diagnostics still match.
  • Stops at ordinary nested function boundaries, where a fresh local arguments object is valid.
  • Added a focused regression test covering both the direct error and nested-function suppression.

Exact parity movement

  • Overall: 5,117 -> 5,118 / 5,907
  • Failures: 790 -> 789
  • Lower half: 2,729 -> 2,730 / 3,000, failures 271 -> 270
  • Upper half: 2,388 / 2,907, unchanged at 519 failures
  • Flipped fixture: asyncFunctionDeclarationCapturesArguments_es5
  • Failure identity audit: no additions or other removals
  • Scope: 43 insertions in one checker file

Residual diagnostic improvement

classStaticBlock6 still fails on unrelated parser/static-block diagnostics, but its previously missing TS2522 line now matches exactly. There are no remaining TS2522 plus/minus deltas in the full exact dump.

Verification

  • ts_checker: 3,147/3,147
  • ts_conformance: 1,374/1,374
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Focused exact: 1 fixture flipped + 1 residual diagnostic improved
  • Full exact TypeScript corpus: 5,118/5,907

Continuing from 789 remaining failures with the next smallest coherent tsgo rule.

22 days ago

TS parity: JSDoc constructor object-method this

Pushed 4d4bc8af3 (fix(ts-checker): close jsdoc constructor method this) directly to main.

What changed

  • A checked-JavaScript object-literal method carrying @constructor now keeps the object literal's closed this shape during contextual body checking.
  • Missing properties on both reads and writes report TS2339 once that shape is known.
  • The initial inference pass remains open, preserving ordinary checkJs expando behavior for untagged object methods.
  • Added a bound checked-JavaScript regression test using the upstream CRLF source shape, exact TS2339 prose, and the companion TS7009 assertion.

Exact parity movement

  • Overall: 5,118 -> 5,119 / 5,907
  • Failures: 789 -> 788
  • Lower half: 2,730 / 3,000, unchanged at 270 failures
  • Upper half: 2,388 -> 2,389 / 2,907, failures 519 -> 518
  • Flipped fixture: constructorTagOnObjectLiteralMethod
  • Failure identity audit: no additions or other removals
  • Scope: 49 insertions, 2 deletions in one checker file

Verification

  • ts_checker: 3,148/3,148
  • ts_conformance: 1,374/1,374
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Focused exact: 1/1
  • Full exact TypeScript corpus: 5,119/5,907

Continuing from 788 remaining failures with the next smallest coherent tsgo rule.

22 days ago

Parity checkpoint: checked-JS JSDoc arrow parameter context

Pushed 934b61837 (fix(ts-checker): contextualize jsdoc arrow parameters) to main.

What changed

  • Variable-level /** @type {...} */ function signatures now contextually type arrow/function initializer parameters inside their bodies, matching tsgo's checked-JS behavior.
  • Nested Closure-style function parameter types such as function((string)): string retain the whole callable type; their return colon is no longer mistaken for a named-parameter separator.
  • Added a bound regression covering both number-to-string body assignments and higher-order callback callability.

The first implementation flipped the target but exposed jsdocParseHigherOrderFunction as a new failure. The exact identity audit caught that trade before commit; the nested-function parser was corrected, and both focused upstream fixtures now pass.

Exact movement

  • Overall: 5,119 -> 5,120 / 5,907 exact passes
  • Failures: 788 -> 787
  • Lower shard: 2,730/3,000, 270 failures (unchanged)
  • Upper shard: 2,389 -> 2,390/2,907, failures 518 -> 517
  • Removed failure identity: jsdocTypeTagParameterType
  • Added failure identities: none
  • Guarded passing identity: jsdocParseHigherOrderFunction

Scope

  • 1 checker file
  • 46 insertions, 1 deletion

Verification

  • focused exact jsdocTypeTagParameterType: 1/1
  • focused exact jsdocParseHigherOrderFunction: 1/1
  • ts_checker: 3,149/3,149
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • exact corpus: 5,120/5,907, 787 failures
22 days ago

Parity checkpoint: JSDoc @satisfies property context

Pushed 06b65cb17 (fix(ts-checker): contextualize jsdoc satisfies properties) to main.

What changed

  • Checked-JS @satisfies object targets now feed matching member signatures into function-valued properties during signature construction.
  • Arrow/method parameters receive the real contextual type through the existing union/intersection/mapped/callable member lookup pipeline.
  • TS7006 is suppressed only when a concrete JSDoc contextual parameter type was actually selected.
  • The bound regression verifies the parameter HIR type is string, not merely that the diagnostic disappeared, and confirms excess-property checking still fires for an invalid sibling object.

The broader 15-fixture satisfies cluster now reports 9/15. checkJsdocSatisfiesTag3 also improved by losing its spurious standalone-method TS7006, but it remains in the failure set for a separate missing expected TS7006; it is not counted as closed.

Exact movement

  • Overall: 5,120 -> 5,121 / 5,907 exact passes
  • Failures: 787 -> 786
  • Lower shard: 2,730/3,000, 270 failures (unchanged)
  • Upper shard: 2,390 -> 2,391/2,907, failures 517 -> 516
  • Removed failure identity: checkJsdocSatisfiesTag13
  • Added failure identities: none

Scope

  • 1 checker file
  • 40 insertions, 1 deletion

Verification

  • focused exact checkJsdocSatisfiesTag13: 1/1
  • focused checkJsdocSatisfiesTag* cluster: 9/15
  • ts_checker: 3,150/3,150
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • exact corpus: 5,121/5,907, 786 failures
22 days ago

Parity checkpoint: inline @satisfies precedence

Pushed 2f68304e1 (fix(ts-checker): prioritize inline jsdoc satisfies context) to main.

What changed

  • A nested checked-JS @satisfies object target now takes precedence over an outer variable-level @type when deciding whether function-valued object members have contextual parameter types.
  • Members absent from the closer satisfies target no longer inherit context from the outer type.
  • The bound regression mirrors checkJsdocSatisfiesTag3 under noImplicitAny and proves the actual HIR split: f retains any and reports TS7006, while g receives the inline string context.

The focused checkJsdocSatisfiesTag* cluster advances from 9/15 to 10/15; the remaining five identities are unchanged.

Exact movement

  • Overall: 5,121 -> 5,122 / 5,907 exact passes
  • Failures: 786 -> 785
  • Lower shard: 2,730/3,000, 270 failures (unchanged)
  • Upper shard: 2,391 -> 2,392/2,907, failures 516 -> 515
  • Removed failure identity: checkJsdocSatisfiesTag3
  • Added failure identities: none

Scope

  • 1 checker file
  • 40 insertions

Verification

  • focused exact checkJsdocSatisfiesTag3: 1/1
  • focused checkJsdocSatisfiesTag* cluster: 10/15
  • ts_checker: 3,151/3,151
  • smoke: 16/16
  • category: 86/86
  • baseline-aware: 586/586
  • ts_conformance: 1,374/1,374
  • exact corpus: 5,122/5,907, 785 failures
22 days ago

Parity update: JSDoc @satisfies diagnostic anchors

Pushed bfc747d3f (fix(ts-checker): anchor jsdoc satisfies diagnostics) to main.

This change preserves the absolute source offset of leading JSDoc bodies while resolving both inline and declaration-level @satisfies tags. Fallback TS1360 diagnostics now anchor on the satisfies tag name, matching tsgo, while property-specific diagnostics keep their existing expression/property anchors.

Parity movement

  • Full exact corpus: 5,122/5,907 -> 5,123/5,907 passing
  • Total failures: 785 -> 784
  • Lower shard: 2,730/3,000, unchanged at 270 failures
  • Upper shard: 2,392/2,907 -> 2,393/2,907, failures 515 -> 514
  • Removed failure identity: checkJsdocSatisfiesTag4
  • Added failure identities: none
  • Focused checkJsdocSatisfiesTag* cluster: 10/15 -> 11/15

Scope

  • 1 checker file
  • 61 insertions, 15 deletions
  • Added a bound checker regression asserting TS1360 starts at the JSDoc tag name

Verification

  • Exact checkJsdocSatisfiesTag4: 1/1
  • Focused checkJsdocSatisfiesTag*: 11/15
  • ts_checker: 3,152/3,152
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Default ts_conformance: 1,374/1,374
  • Exact full corpus: 5,123/5,907

The remaining focused failures are checkJsdocSatisfiesTag1, checkJsdocSatisfiesTag11, checkJsdocSatisfiesTag12, and checkJsdocSatisfiesTag15; their TS1360 positions are now closer/correct where applicable, but they retain separate type/prose or extra-diagnostic gaps.

22 days ago

Parity update: effective JSDoc @satisfies duplicate ordering

Pushed d439f3da7 (fix(ts-checker): order jsdoc satisfies duplicates) to main.

This ports tsgo's effective-tag ordering for duplicate @satisfies checks: initializer-inline tags are collected before declaration-leading tags, and every effective tag after the first emits TS1223 at the tag name. This covers both repeated tags in one declaration comment and declarations that combine a leading tag with an inline initializer tag.

Parity movement

  • Full exact corpus: 5,123/5,907 -> 5,124/5,907 passing
  • Total failures: 784 -> 783
  • Lower shard: 2,730/3,000, unchanged at 270 failures
  • Upper shard: 2,393/2,907 -> 2,394/2,907, failures 514 -> 513
  • Removed failure identity: checkJsdocSatisfiesTag11
  • Added failure identities: none
  • Focused checkJsdocSatisfiesTag* cluster: 11/15 -> 12/15

Scope

  • 1 checker file
  • 75 insertions, 26 deletions
  • Replaced count-only duplicate detection with ordered tag reporting
  • Added a bound regression for both TS1223 ordering/position cases

Verification

  • Exact checkJsdocSatisfiesTag11: 1/1
  • Focused checkJsdocSatisfiesTag*: 12/15
  • ts_checker: 3,153/3,153
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Default ts_conformance: 1,374/1,374
  • Exact full corpus: 5,124/5,907

The remaining focused failures are now checkJsdocSatisfiesTag1, checkJsdocSatisfiesTag12, and checkJsdocSatisfiesTag15.

22 days ago

Parity update: contextual JSDoc @satisfies object literals

Pushed 7c3a923de (fix(ts-checker): preserve jsdoc satisfies literals) to main.

This change validates JSDoc @satisfies object literals property by property when their inferred object type has widened scalar members. Literal syntax is checked against each contextual target member, including union/intersection targets, before the widened structural fallback. It also matches tsgo's declaration-tag behavior: property and excess-property diagnostics remain active, but declaration-level missing keys do not synthesize the inline-expression TS1360 fallback.

That declaration-level rule intentionally follows tsgo's checked behavior rather than the divergent nested original-TypeScript TS2741 baseline.

Parity movement

  • Full exact corpus: 5,124/5,907 -> 5,126/5,907 passing
  • Total failures: 783 -> 781
  • Lower shard: 2,730/3,000, unchanged at 270 failures
  • Upper shard: 2,394/2,907 -> 2,396/2,907, failures 513 -> 511
  • Removed failure identities: checkJsdocSatisfiesTag1, checkJsdocSatisfiesTag12
  • Added failure identities: none
  • Focused checkJsdocSatisfiesTag* cluster: 12/15 -> 14/15

Scope

  • 1 checker file
  • 86 insertions
  • Added a bound regression for contextual literal members and declaration-level missing-key behavior

Verification

  • Both targeted exact fixtures pass
  • Focused checkJsdocSatisfiesTag*: 14/15
  • ts_checker: 3,154/3,154
  • Smoke: 16/16
  • Category: 86/86
  • Baseline-aware: 586/586
  • Default ts_conformance: 1,374/1,374
  • Exact full corpus: 5,126/5,907

checkJsdocSatisfiesTag15 is now the only remaining failure in the focused cluster.

22 days ago

TS parity checkpoint: direct JSDoc satisfies function types

Pushed 0e037b874 (fix(ts-checker): resolve jsdoc satisfies function types) to main.

What changed

  • Contextually type direct arrow initializers from leading JSDoc @satisfies signatures while preserving tsgo's uncontextualized function-declaration behavior.
  • Resolve sibling function signatures on demand for JSDoc typeof references used before the function's ordinary checker pass.
  • Reduce JSDoc Parameters<T> utility types to tuple types and enforce JSDoc-derived tuple targets positionally for array initializers.
  • Preserve source parameter names, including rest parameter names, when rendering JSDoc function signatures in diagnostics.
  • Accept unbraced @type T tags in the shared JSDoc parser, including the form used by typedefTagTypeResolution.

Exact parity movement

  • Full corpus: 5,126/5,907 -> 5,128/5,907
  • Failures: 781 -> 779
  • Lower shard: 2,730/3,000 unchanged; the complete 270-failure identity set is identical
  • Upper shard: 2,396/2,907 -> 2,398/2,907; failures 511 -> 509
  • JSDoc satisfies cluster: 14/15 -> 15/15

Removed failure identities:

  • checkJsdocSatisfiesTag15
  • typedefTagTypeResolution

Added failure identities: none.

Validation

  • checkJsdocSatisfiesTag15 exact focused run: 1/1
  • typedefTagTypeResolution exact focused run: 1/1
  • complete checkJsdocSatisfiesTag exact cluster: 15/15
  • ts_parser: 769/769
  • ts_checker: 3,155/3,155
  • default conformance smoke: 16/16
  • default conformance category: 86/86
  • default conformance baseline-aware: 586/586
  • default conformance package total: 1,374/1,374
  • exact corpus shards: 2,730/3,000 and 2,398/2,907

Scope: 122 insertions, 7 deletions across 2 files.

Current exact parity: 5,128/5,907 (86.81%), with 779 fixtures remaining.

22 days ago

TS parity checkpoint: JSDoc generic constraint anchors

Pushed 38a9e1c0e (fix(ts-checker): anchor jsdoc generic constraints) to main.

What changed

  • Carry raw JSDoc generic type-argument offsets alongside lowered argument types.
  • Anchor newly emitted TS2344 constraint diagnostics on the offending JSDoc argument for both nominal generic aliases and generic JSDoc typedefs.
  • Preserve existing relation logic and prose; only the source anchor changes from the following declaration to the exact type argument.
  • Strengthen the existing checker regression to pin both A<number> and B<number> argument offsets.

Exact parity movement

  • Full corpus: 5,128/5,907 -> 5,129/5,907
  • Failures: 779 -> 778
  • Lower shard: 2,730/3,000 unchanged; complete 270-failure identity set unchanged
  • Upper shard: 2,398/2,907 -> 2,399/2,907; failures 509 -> 508
  • checkJsdocTypeTag cluster: 4/10 -> 5/10

Removed failure identity:

  • checkJsdocTypeTag4

Added failure identities: none.

Validation

  • checkJsdocTypeTag4 exact focused run: 1/1
  • complete checkJsdocTypeTag exact cluster: 5/10, with one clean identity removal
  • ts_checker: 3,155/3,155
  • default conformance smoke: 16/16
  • default conformance category: 86/86
  • default conformance baseline-aware: 586/586
  • default conformance package total: 1,374/1,374
  • exact corpus shards: 2,730/3,000 and 2,399/2,907

Scope: 39 insertions, 6 deletions in 1 file.

Current exact parity: 5,129/5,907 (86.83%), with 778 fixtures remaining.

22 days ago

TS parity checkpoint: spaced Closure JSDoc function types

Pushed 0d0802015 (fix(ts-checker): parse spaced jsdoc function types) to main.

What changed

  • Recognize Closure-style JSDoc function signatures with whitespace after function, such as function (number): string.
  • Use the same parser rule for signature lowering, nested function parameters, function-declaration compatibility, return-type positions, and TS7014 missing-return detection.
  • Read function-type return annotations from an arrow/function expression's JSDoc-owning variable.
  • Suppress a redundant declaration-level TS2322 only for a direct JSDoc-typed variable function initializer whose body already reports the precise return mismatch.
  • Strengthen checker coverage for contextual parameter typing, TS7014, and invalid-call TS2345 on spaced Closure syntax.

During verification, an initially broader duplicate-suppression rule regressed contextuallyTypeCommaOperator02 and contextuallyTypeLogicalAnd02. The exact lower-shard audit caught both; the rule was narrowed to direct JSDoc-owned variable initializers, and the full contextuallyType family returned to 18/18 before commit.

Exact parity movement

  • Full corpus: 5,129/5,907 -> 5,130/5,907
  • Failures: 778 -> 777
  • Lower shard: 2,730/3,000 unchanged; complete 270-failure identity set unchanged
  • Upper shard: 2,399/2,907 -> 2,400/2,907; failures 508 -> 507
  • checkJsdocTypeTag cluster: 5/10 -> 6/10

Removed failure identity:

  • checkJsdocTypeTag2

Added failure identities: none.

Validation

  • checkJsdocTypeTag2 exact focused run: 1/1
  • complete checkJsdocTypeTag exact cluster: 6/10, one clean identity removal
  • contextual function regression family: 18/18
  • ts_checker: 3,155/3,155
  • default conformance smoke: 16/16
  • default conformance category: 86/86
  • default conformance baseline-aware: 586/586
  • default conformance package total: 1,374/1,374
  • exact corpus shards: 2,730/3,000 and 2,400/2,907

Scope: 53 insertions, 15 deletions in 1 file.

Current exact parity: 5,130/5,907 (86.84%), with 777 fixtures remaining.

22 days ago

TS parity progress: broad JSDoc callable semantics

Pushed 452441eae (fix(ts-checker): distinguish broad jsdoc callable types) to main.

What changed

  • Treats bare JSDoc ? as any, matching tsgo/TypeScript wildcard semantics.
  • Keeps @type {Function} and @type {function} broad rather than contextually typing arrow parameters, so their parameters correctly emit TS7006 under noImplicitAny.
  • Preserves concrete Closure/TypeScript callable annotations as contextual signatures, including multiline @type blocks.
  • Recognizes raw @type fallback markers only at real tag boundaries, avoiding false matches on longer text such as @typescript.
  • Preserves raw object versus Object spelling across repeated script var declarations so TS2403 renders and fires exactly.
  • Strengthens the existing checker regression around callable arity, wildcard assignment, broad-callable implicit-any, and wrapper/object redeclaration behavior.

Exact movement

  • Exact parity: 5,130/5,907 -> 5,131/5,907 (86.85% -> 86.86%)
  • Failures: 777 -> 776
  • checkJsdocTypeTag cluster: 6/10 -> 7/10
  • Removed failure identity: checkJsdocTypeTag1
  • Added failure identities: none
  • Lower split remains identity-exact at 2,730/3,000 with the same 270 failures.
  • Commit scope: 1 file, 82 insertions, 9 deletions.

Verification

  • Focused exact (checkJsdocTypeTag1, destructuringParameterDeclaration9): 2/2
  • JSDoc type-tag cluster: 7/10; remaining misses are Tag5, Tag6, and object-property Tag2
  • ts_checker: 3,155/3,155
  • Default conformance: 1,374/1,374
    • smoke: 16/16
    • category: 86/86
    • baseline-aware: 586/586
  • Full 5,907-case split identity audit completed against the tsgo-pinned TypeScript corpus.
22 days ago

TypeScript parity progress: JSDoc callable aliases

Pushed ecd66f833 (fix(ts-checker): resolve jsdoc callable aliases).

Logic ported

  • Bare @type references to single-signature callable typedefs now publish the typedef return contract onto the function signature.
  • Arrow-style JSDoc function types bind a following union as the return type, so (x: T) => A | B no longer becomes a union containing a partial function signature.
  • Callable-object aliases are counted structurally. Single signatures provide contextual typing; overloaded callable objects do not incorrectly contextually select the first overload.
  • Unsupported overloaded function annotations now emit exact TS8030 while preserving literal implementation return unions for downstream calls.
  • Explicit JSDoc return contracts are no longer overwritten by body inference.
  • Added a focused checker regression covering both a single callable alias and an overloaded callable object alias.

Parity movement

  • Pushed exact parity: 5,131 -> 5,132 / 5,907 (+1, 86.86% -> 86.88%)
  • Pushed failures: 776 -> 775
  • checkJsdocTypeTag cluster: 7/10 -> 8/10
  • Exact identity removed: checkJsdocTypeTag5
  • Exact identities added: none
  • Local combined tree measured 5,138 / 5,907; that includes six concurrent unstaged JSX closures and is not attributed to this commit.

Validation

  • checkJsdocTypeTag5: 1/1 exact
  • ts_checker: 3,156/3,156
  • default ts_conformance: 1,374/1,374
  • exact lower slice: 2,730/3,000, identity unchanged
  • exact upper slice: 2,408/2,907, only checkJsdocTypeTag5 removed
  • staged scope: 1 file, +125/-14

The remaining failures in this focused cluster are checkJsdocTypeTag6 and checkJsdocTypeTagOnObjectProperty2.

22 days ago

Pushed e93190680 (fix(ts-checker): enforce jsdoc function arity).

This ports the remaining declaration-side JSDoc @type arity behavior needed by tsgo:

  • reports TS8030 for non-callable annotations on named functions;
  • reports TS8030 when the annotated callable provides fewer parameters than the implementation;
  • keeps checked-JS parameters required when a JSDoc type tag is present;
  • emits type-aware TS2322 diagnostics for function-valued object properties, including exact source/target signatures.

Parity movement:

  • exact TypeScript parity: 5,132 -> 5,134 / 5,907 (+2, 86.88% -> 86.91%);
  • exact failures: 775 -> 773;
  • checkJsdocTypeTag cluster: 8/10 -> 9/10;
  • removed failure identities: checkJsdocTypeTag6, typeTagCircularReferenceOnConstructorFunction;
  • added failure identities: none;
  • local combined tree: 5,140 / 5,907, including six concurrent unstaged JSX closures that are not part of this commit.

Validation:

  • focused checkJsdocTypeTag6: 1/1 exact;
  • ts_checker: 3,157/3,157;
  • default ts_conformance: 1,374/1,374;
  • exact lower slice: 2,730/3,000, identity set unchanged;
  • exact upper slice: 2,410/2,907, with only the two named failures removed;
  • staged scope: 1 file, +79/-3.

The only remaining fixture in this ten-test cluster is checkJsdocTypeTagOnObjectProperty2.

22 days ago

Pushed 7eaedbabf (fix(ts-checker): contextualize jsdoc object properties).

This completes the ten-fixture checkJsdocTypeTag cluster by porting tsgo's object-property behavior:

  • carries a property's JSDoc callable type into method/arrow parameters and parameter defaults;
  • validates default initializers against the contextual parameter type;
  • keeps precise return-value diagnostics while suppressing redundant whole-function TS2322 reports;
  • models a pre-assignment shorthand read of an evolving uninitialized variable as undefined;
  • renders primitive assignment mismatches with the same concrete source and non-nullish target constituent as tsgo.

Parity movement:

  • exact TypeScript parity: 5,134 -> 5,135 / 5,907 (+1, 86.91% -> 86.93%);
  • exact failures: 773 -> 772;
  • checkJsdocTypeTag cluster: 9/10 -> 10/10;
  • removed failure identity: checkJsdocTypeTagOnObjectProperty2;
  • added failure identities: none;
  • local combined tree: 5,141 / 5,907 (87.03%), including six concurrent unstaged JSX closures outside this commit.

Validation:

  • focused checkJsdocTypeTagOnObjectProperty2: 1/1 exact;
  • ts_checker: 3,158/3,158;
  • default ts_conformance: 1,374/1,374;
  • exact lower slice: 2,730/3,000, identity set unchanged;
  • exact upper slice: 2,411/2,907, with only the named failure removed;
  • committed scope: 1 file, +130/-22.

The full exact identity audit added no regressions.

22 days ago

Reserved namespace declaration parity

Pushed 0307656ed (fix(ts-parser): recover reserved namespace names) to main.

This ports tsgo declaration-scan and recovery behavior for reserved identifiers in namespace paths:

  • contextual identifier names still commit namespace/module declarations;
  • a leading reserved name such as debugger recovers declare and namespace as value statements, then emits TS2819 and TS1005 at the upstream locations;
  • Home flattened dotted ambient namespaces now preserve the upstream allowance for an inner declare modifier;
  • malformed template module names remain on the dedicated TS1443 recovery path.

The first full identity audit caught a two-fixture regression in templateStringInModuleName and templateStringInModuleNameES6: the new invalid-name lookahead had shadowed the existing TS1443 branch. Reordering the recovery checks and adding a focused parser regression restored both before the final audit.

Attributable parity movement

  • exact passes: 5,135 -> 5,137 / 5,907 (+2, 86.96%)
  • exact failures: 772 -> 770
  • removed identities: ambientModuleDeclarationWithReservedIdentifierInDottedPath, ambientModuleDeclarationWithReservedIdentifierInDottedPath2
  • added identities: none
  • combined local tree: 5,143 / 5,907 (87.07%), including six concurrent JSX closures that were not part of this commit

Commit scope

  • packages/ts_parser/src/ts_parser.zig
  • 80 insertions, 4 deletions

Verification

  • parser: 763 / 763
  • checker: 3,158 / 3,158
  • default conformance: 1,374 / 1,374
  • smoke: 16 / 16
  • category: 86 / 86
  • baseline-aware: 586 / 586
  • focused ambient fixtures: 2 / 2
  • focused template recovery fixtures: 2 / 2
  • exact lower slice: 2,732 / 3,000
  • exact upper slice: 2,411 / 2,907

The exact failure-name comparison removed only the two targeted ambient namespace identities and introduced no new failures.

Sign in to comment on this issue.