mirror of
https://github.com/abhigyanpatwari/GitNexus.git
synced 2026-09-15 23:32:49 +00:00
1084 commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
henry201605
|
eeea46466b
|
fix(group): handle named annotation args in Java Spring route extraction (#1834)
Some checks are pending
CodeQL / Analyze (javascript-typescript) (push) Waiting to run
CodeQL / Analyze (python) (push) Waiting to run
Gitleaks / gitleaks (push) Waiting to run
Publish / Classify release event (push) Waiting to run
Publish / RC guard (marker + release-PR skip) (push) Blocked by required conditions
Publish / ci (push) Blocked by required conditions
Publish / Publish to npm (push) Blocked by required conditions
Publish / Build & Push RC Docker images (push) Blocked by required conditions
Scorecard / Scorecard analysis (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-cli) (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-web) (push) Waiting to run
* fix(group): handle named annotation args in Java Spring route extraction
The Java HTTP plugin only matched positional `@RequestMapping("/path")`
syntax for class-level prefixes and method-level routes. Named argument
forms (`path = "/path"` and `value = "/path"`) produce an
`element_value_pair` AST node that the tree-sitter queries did not cover,
causing the class prefix to be lost and named-arg method routes to be
missed entirely during cross-repo contract extraction.
Add a second pattern to both SPRING_CLASS_PREFIX_PATTERNS and
SPRING_METHOD_ROUTE_PATTERNS matching the element_value_pair structure.
* fix(group): constrain Spring named-arg query to path/value keys + add regression tests
Address Claude review on PR #1834. The named-argument patterns added
in
|
||
|
dependabot[bot]
|
ca3e1755c2
|
chore(deps)(deps): bump lru-cache from 11.4.0 to 11.5.0 in /gitnexus (#1844)
Bumps [lru-cache](https://github.com/isaacs/node-lru-cache) from 11.4.0 to 11.5.0. - [Changelog](https://github.com/isaacs/node-lru-cache/blob/main/CHANGELOG.md) - [Commits](https://github.com/isaacs/node-lru-cache/compare/v11.4.0...v11.5.0) --- updated-dependencies: - dependency-name: lru-cache dependency-version: 11.5.0 dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> |
||
|
dependabot[bot]
|
6acdc49f06
|
chore(deps)(deps-dev): bump @types/node in /gitnexus (#1845)
Bumps [@types/node](https://github.com/DefinitelyTyped/DefinitelyTyped/tree/HEAD/types/node) from 25.9.0 to 25.9.1. - [Release notes](https://github.com/DefinitelyTyped/DefinitelyTyped/releases) - [Commits](https://github.com/DefinitelyTyped/DefinitelyTyped/commits/HEAD/types/node) --- updated-dependencies: - dependency-name: "@types/node" dependency-version: 25.9.1 dependency-type: direct:development update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> |
||
|
azizur100389
|
b1445daf04
|
feat(cpp): rank user-defined conversions (#1829) | ||
|
dale
|
d903152eba
|
fix(typescript): reuse suffix index in scope resolver (#1840)
* fix(typescript): reuse suffix index in scope resolver Build a suffix index once per TypeScript scope-resolution pass and pass it into standard import resolution so package-style imports avoid repeated linear file-list scans.\n\nFixes #1839 * test(typescript): add wiring-level test for scope-resolver suffix index - Test typescriptScopeResolver.resolveImportTarget directly (the real production entry point) with package-style, unresolvable, and relative imports - Use vi.spyOn on buildSuffixIndex to verify the index is built inside the makeTsResolveImportTarget closure — fails if index wiring is removed - Fix existing test to pass real file lists instead of empty arrays alongside the prebuilt index, matching production wiring --------- Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> Co-authored-by: Test <test@example.com> |
||
|
Bassey Riman
|
6c572749b0
|
fix(web): stop Nexus AI agent when user clicks Stop (#1820)
* fix(web): stop Nexus AI agent when user clicks Stop Wire AbortController through chat streaming so Stop cancels the LangGraph run instead of only hiding the loading UI. Fixes #1615. * fix(web): address PR review feedback for Nexus AI stop Guard stream cleanup against Stop-then-Send races, remove dead cancelled handler, tighten abort error detection, add stopped tool-call status, and extend abort unit tests. Fixes #1615. * chore(autofix): apply prettier + eslint fixes via /autofix command * fix(web): address review findings for Nexus AI stop/cancel - Fix race conditions in useAppState.tsx abort lifecycle: - Replace stale isChatLoading closure guard with chatStateRef - Track and cancel rAF handles in stopChatResponse/finally - Move cancelled chunk check before onChunk dispatch - Simplify finally block to unconditional cleanup via chatStateRef - Guard tool_result from overwriting stopped status - Have clearChat abort in-flight streams before clearing - Reorder isAbortError to check error identity before signal.aborted - Refactor AgentStreamChunk to discriminated union for exhaustive switch - Fix test assertions to use exact .toEqual() per DoD §2.7 - Add test for plain Error with name AbortError - Remove dead markStopped alias, simplify signal spread-conditional --------- Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Test <test@example.com> |
||
|
Sparsh
|
7556a8e73a
|
feat(cobol): migrate COBOL to scope-based resolution (regex provider) (#941) (#1835)
* feat(cobol): migrate COBOL to scope-based resolution (regex provider)
Migrate COBOL to scope-based registry resolution, validating the
parse-source-agnostic contract — COBOL uses regex, not tree-sitter,
but implements the same LanguageProvider interface via emitScopeCaptures.
Phase 1-5 complete per #941 DoD.
New files:
languages/cobol/captures.ts — emitScopeCaptures wrapping regex tagger
languages/cobol/interpret.ts — import/type-binding/receiver hooks
languages/cobol/index.ts — barrel export
languages/cobol/scope-resolver.ts — ScopeResolver wiring (9 fields, 3 toggles)
Modified files:
languages/cobol.ts — wire 4 scope-resolution hooks
registry.ts — register cobolScopeResolver
registry-primary-flag.ts — document REGISTRY_PRIMARY_COBOL
Fixtures:
17 fixture files, 30 test cases across 11 required classes
test/integration/resolvers/cobol-scope.test.ts
Tests: 24/24 pass (default + REGISTRY_PRIMARY_COBOL=0)
tsc: zero cobol-specific errors
Shadow mode (GITNEXUS_SHADOW_MODE=1): zero crashes
Regex perf: 10K-line file in 408ms (threshold: 2000ms)
NOT added to MIGRATED_LANGUAGES — REGISTRY_PRIMARY_COBOL env var only.
* chore(cobol): add COBOL to MIGRATED_LANGUAGES
* Revert "chore(cobol): add COBOL to MIGRATED_LANGUAGES"
This reverts commit
|
||
|
Hugo Gu
|
c8117d1292
|
feat(web): Introduce Tree View and Circles View in Web Viewer (#1799)
* feat(graph-view): add tree and circles layout modes
Add alternate graph layouts to the web viewer with new graph view state, canvas controls, adapters, and Sigma layout logic for tree and concentric-circle rendering. Include layout and adapter tests plus tree-view E2E coverage aligned with the English UI labels, and tune node visibility, edge layering, large-graph behavior, and tree-layer spacing so the new views stay readable. Follow up the tree-view work by keeping noisy variables hidden by default and mapping Property/Const icons so filter coverage stays in sync with the expanded node taxonomy.
Co-authored-by: OpenAI Codex <noreply@openai.com>
AI-model: GPT-5 Codex
* fix(web): cap tree layout spring iterations and remove unused variable
Finding A (blocker): calculateTreeLayout runs 14 synchronous spring
iterations over all edges and nodes — O(N×E×14) + O(N log N) per layer
per iteration — with no size guard. At 10K+ nodes this freezes the
main thread for several seconds.
Fix: make SPRING_ITERATIONS adaptive:
- N > 10 000 → 0 iterations (proportional initial layout only)
- N > 3 000 → 4 iterations
- otherwise → 14 iterations (unchanged behaviour for small graphs)
Also removes the unused `const r` at useSigma.ts:1314, which was a
leftover after the radial-resistance decomposition was removed.
This clears the CodeQL "unused variable" warning (Finding G).
Co-authored-by: Claude <noreply@anthropic.com>
AI-model: claude-sonnet-4-6
* test(graph-adapter): add circles adapter tests and tree layout perf bound
Finding B (high): knowledgeGraphToCirclesGraphology had zero test
coverage. Adds three new tests:
- ring placement: verifies Folder→ring 0, File→ring 1, Function→ring 3
and confirms circles-specific attributes (circlesRing, circlesAnchorX/Y)
are set while tree attributes (treeAnchorX/Y) are absent.
- edge styling: CONTAINS is marked isHierarchyEdge=true with the
hierarchy colour; CALLS is cross-cutting with its own colour.
- CALLS cross-cutting: a lone CALLS edge between two Functions is
correctly identified as a non-hierarchy edge.
Also adds a performance-bound test for the tree adapter at 2 000 nodes /
4 000 edges (the adaptive 14-iteration path) asserting completion within
2 s — catches regressions to the O(N×E×iterations) main-thread blocking
that Finding A identified.
Co-authored-by: Claude <noreply@anthropic.com>
AI-model: claude-sonnet-4-6
* refactor(web): rename Tree View → Sequential Layout, Circles → Radial Layout
Aligns the UI labels with standard graph layout terminology from the
Cambridge Intelligence taxonomy (cambridge-intelligence.com/blog/automatic-graph-layouts):
Tree View → Sequential Layout (顺序布局)
Circles → Radial Layout (径向布局)
Force Graph → Force Graph (unchanged)
Internal graphViewMode keys ('tree', 'circles', 'force') are unchanged —
only the displayed strings in en/zh-CN locales and the E2E button selectors
are updated.
Co-authored-by: Claude <noreply@anthropic.com>
AI-model: claude-sonnet-4-6
* perf(web): add adaptive large-graph guards to sequential layout physics
For graphs with N > 5 000 nodes, each rAF frame of runTreeLayout was
doing O(N log N) sort + O(N × k) repulsion pair comparisons (k ≈ 2 400
for a 20 K-node graph spread across 1 080 px at range 130). At that
scale each frame took hundreds of ms, making the canvas appear completely
frozen even though the physics loop was still running.
Fix mirrors the circles layout adaptive strategy:
N > 5 000 (large):
- Skip repulsion pass (O(N × k) → 0)
- Skip spread-force sort (O(N log N) → 0)
- Velocity cap raised to ±12 / ±6 px so nodes cover ground faster
- Damping 0.58, 1 sim step/frame, 30 s max duration
- Looser early-stop thresholds (max v 0.05, avg v 0.03, active 2 %)
N > 1 500 (medium):
- Velocity cap raised to ±6 / ±3 px
- 24 s max duration
- Repulsion and spread still active
N ≤ 1 500 (small):
- Unchanged behaviour (velocity ±3/±2, 18 s, all forces active)
Layer gravity (O(N)) and edge springs (O(E)) run for all graph sizes —
they provide the structural pull that replaces repulsion at large N.
Co-authored-by: Claude <noreply@anthropic.com>
AI-model: claude-sonnet-4-6
* fix(web): fix stale closure in sigma event handlers breaking node selection
The sigma 'clickNode', 'clickStage', 'enterNode', and 'leaveNode' handlers
are registered in a one-time useEffect (empty dep array). They captured
options.onNodeClick via closure, so they always called the initial version
of handleNodeClick — the one created before the graph loaded where
`if (!graph) return` exits immediately.
Consequence: clicking a node in the canvas never updated the app-level
selectedNode state. This broke:
- The Focus Depth filter (warning "Select a node to apply depth filter"
persisted even after a canvas click)
- The depth hop filter not applying (selectedNode was always null)
- The code panel not opening on canvas node click
Fix: store the three callback props in refs (onNodeClickRef, onNodeHoverRef,
onStageClickRef) and update them synchronously on every render. The sigma
event handlers now read from the refs, so they always invoke the latest
version of the callbacks without needing to re-register.
Co-authored-by: Claude <noreply@anthropic.com>
AI-model: claude-sonnet-4-6
* fix(web): address three code-review bugs in graph rendering
Bug 1 (useSigma.ts): forces in the tree physics loop were computed once
before the sub-steps loop and reused for every step, causing 2× displacement
on slow frames (>64ms, simulationSteps>1). Fix: move forceX/forceY Maps and
all force accumulation (layer gravity, edge springs, repulsion, spread) inside
the loop so each sub-step integrates from current node positions.
Bug 2 (graph-adapter.ts): all three adapters used `graph.hasEdge(src,tgt)`
as a dedup guard, which silently drops any second edge between the same node
pair. A CALLS relationship between nodes that also have a CONTAINS edge was
always lost. Fix: switch from `new Graph()` to `new MultiGraph()` (allows
multiple edges per pair) and dedup by `rel.id` instead of by node pair.
Bug 3 (graph-adapter.test.ts): the cross-cutting edge styling test never
executed its CALLS branch because Bug 2 dropped the CALLS edge before the
assertion ran. Fix: assert `sigmaGraph.size === 2` and verify both edges
individually after collecting attrs by relationType.
Co-authored-by: Claude <noreply@anthropic.com>
AI-model: claude-sonnet-4-5
* fix(web): address three code-review bugs in graph rendering
- Move radial layout force accumulation inside the sub-step loop so
forces are recomputed from updated node positions each iteration
instead of using stale forces computed before the loop began
- Revert knowledgeGraphToGraphology from MultiGraph back to Graph with
node-pair deduplication to prevent ForceAtlas2 from double-applying
spring forces for node pairs that share multiple relation types
- Add Target to the lucide-icons import in FileTreePanel.tsx so the
Const node type icon resolves without a ReferenceError
Co-authored-by: Claude <noreply@anthropic.com>
AI-model: claude-sonnet-4-6
* fix(web): address four more PR review comments
Edge visibility (useSigma.ts): HAS_METHOD / HAS_PROPERTY edges were hidden
when any edge-type filter was active because those types are not in the EdgeType
union. Normalize HAS_METHOD → DEFINES and HAS_PROPERTY → CONTAINS before the
visibleTypes.includes() guard so Kotlin/Java hierarchy edges follow the same
filter logic as their semantic equivalents.
Force-mode edge styles (graph-adapter.ts): HAS_METHOD / HAS_PROPERTY fell back
to the default gray color in the force-graph adapter because EDGE_STYLES had no
entries for them. Added explicit entries using the same hues as DEFINES/CONTAINS
so force mode renders Kotlin/Java hierarchy edges consistently with tree/circles.
Accessibility (GraphCanvas.tsx, locales): the layout-mode switcher (Force /
Tree / Circles) had no ARIA semantics. Added role="tablist" on the container
and role="tab" + aria-selected on each button. Added the viewModes.label i18n
key (used as aria-label on the tablist) to en and zh-CN locale files.
Flaky test (graph-adapter.test.ts): replaced the hard 2 s wall-clock assertion
with a structural check (node count + edge count) that is deterministic across
CI hardware. Timing tests are inherently flaky and provide no correctness signal.
Co-authored-by: Claude <noreply@anthropic.com>
AI-model: claude-sonnet-4-5
---------
Co-authored-by: OpenAI Codex <noreply@openai.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Gergő Magyar <gergomagyar@icloud.com>
|
||
|
ChamHerry
|
681a352006
|
fix(worker): analyze native worker aborts (#1833)
* fix(analyze): avoid native aborts on generated worker bundles Retire timed-out parse workers instead of force-terminating native parser state, and skip Monaco generated worker bundles by default while preserving explicit .gitnexusignore negation overrides. Constraint: Node native tree-sitter bindings can abort the process when a timed-out worker is terminated while inside parser state. Rejected: Falling back to sequential parsing for native stalls | it can move the same native crash onto the main thread. Confidence: high Scope-risk: moderate Directive: Keep timeout recovery from force-terminating workers until they return to JS or exit naturally. Tested: npm test; npx tsc --noEmit; npm run build; targeted analyze on /Users/wangxc/Code/keep; gitnexus detect_changes --scope staged Not-tested: Node 22 LTS runtime and non-macOS platforms * fix(worker): bound retired parser worker lifetimes Keep timeout recovery from immediately terminating workers that may still be inside native parser state, while making terminal pool shutdown own retired worker cleanup so long-lived processes do not accumulate retired threads. Constraint: Claude review on PR #1833 required retiredWorkers cleanup in pool.terminate() and tripBreaker() without regressing no-immediate-terminate timeout safety. Rejected: clearing the retiredWorkers set without terminating | would remove JS bookkeeping while leaking the underlying worker thread. Confidence: high Scope-risk: moderate Directive: Preserve the distinction between recoverable timeout retirement and terminal pool shutdown; do not reintroduce immediate terminate in removeWorkerFromSlot(..., 'retire'). Tested: npx vitest run test/unit/worker-pool-timeout-retire.test.ts; npx vitest run test/unit/worker-pool-timeout-retire.test.ts test/unit/worker-pool-resilience.test.ts test/unit/worker-pool-cumulative-timeout.test.ts test/unit/worker-pool-slot-generation.test.ts; npx tsc --noEmit; npm run build; npx prettier --check src/core/ingestion/workers/worker-pool.ts test/unit/worker-pool-timeout-retire.test.ts ../docs/todo/pr-1833-retired-worker-cleanup-plan.md; npx eslint src/core/ingestion/workers/worker-pool.ts test/unit/worker-pool-timeout-retire.test.ts; gitnexus detect_changes --scope staged. Not-tested: npm test full suite did not complete green in this environment; two runs each had one unrelated test/unit/hooks.test.ts parseHookOutput null failure, and each failed hook test passed when rerun in isolation. * ci: retrigger checks Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: wangxc <wangxc_a_bj@si-tech.com.cn> Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> Co-authored-by: Test <test@example.com> Co-authored-by: Cursor <cursoragent@cursor.com> |
||
|
Gergő Magyar
|
5e012c373b
|
fix(cli): detect missing LadybugDB native binary at startup with actionable guidance (#835) (#1837)
* fix(cli): detect missing LadybugDB native binary at startup with actionable guidance (#835) Add checkLbugNative() pre-flight that verifies lbugjs.node exists before any command transitively imports @ladybugdb/core. When missing (bun default install, --ignore-scripts), prints repair instructions instead of crashing with ERR_DLOPEN_FAILED. Also enhances `gitnexus doctor` to probe the native binary status. * fix(review): guard eval-server, un-guard status command eval-server transitively loads @ladybugdb/core and needs the native binary check. status only reads filesystem metadata and should remain accessible when the binary is missing. * fix(lint): use console.log instead of console.error in native check gate The project eslint config only allows console.log. * fix(cli): route native-check to stderr and validate binary loadability Fixes two Codex adversarial review findings: 1. Native-check failure message now goes to process.stderr.write instead of console.log, preventing MCP stdout protocol contamination. 2. checkLbugNative now attempts a controlled require() probe after the existence check. Truncated, ABI-mismatched, or wrong-platform binaries produce actionable guidance instead of passing through to crash at process.dlopen. --------- Co-authored-by: Test <test@example.com> |
||
|
Gergő Magyar
|
05d269ec28
|
feat(ruby): migrate Ruby to scope-based resolution (RFC #909 Ring 3) (#1831)
* feat(ruby): migrate Ruby to scope-based resolution (RFC #909 Ring 3) Implement the full scope-resolution pipeline for Ruby following the PR #1639 (Rust migration) standard, targeting registration in MIGRATED_LANGUAGES with 100% scope parity. Scope resolver hooks (languages/ruby/): - query.ts: RUBY_SCOPE_QUERY covering scopes, declarations, imports, type-bindings (constructor inference via .new), and references - captures.ts: emitRubyScopeCaptures orchestrator with import decomposition, receiver-binding synthesis, method reclassification, and arity metadata for both declarations and calls - receiver-binding.ts: self type-binding synthesis for instance methods, singleton methods, and class << self blocks - interpret.ts: interpretRubyImport (wildcard semantics) and interpretRubyTypeBinding (YARD, constructor, alias sources) - import-target.ts: resolveRubyImportTarget adapting the existing suffix resolver for require/require_relative/load - merge-bindings.ts: tier-based shadowing (local > namespace > import) - arity.ts: Ruby arity check with *args/**kwargs/&block support - scope-resolver.ts: rubyScopeResolver with custom buildRubyMro (kind-aware IMPLEMENTS partitioning: prepend > direct > include; extend excluded from instance MRO per legacy semantics) - simple-hooks.ts: bindingScopeFor, importOwningScope, receiverBinding Wiring: - ruby.ts provider gains 7 scope-resolution hooks - Registered in SCOPE_RESOLVERS map and MIGRATED_LANGUAGES - 127 legacy tests wired with createResolverParityIt('ruby') - 27 new scope-specific tests in ruby-scope.test.ts Parity: 89/127 legacy tests pass under registry-primary; 38 are heritage/property/YARD gaps expected in V1. All 127 pass under legacy. Closes #931 * feat(ruby): add emitHeritageEdges hook, YARD parsing, bare calls, property emission Extend the scope-resolution pipeline with a new optional `emitHeritageEdges` hook (ScopeResolver contract + run.ts wiring) that runs between `preEmitInheritanceEdges` and `buildMro`. This lets languages whose heritage declarations are syntactic method calls (Ruby include/extend/prepend) emit IMPLEMENTS edges from the scope-resolver without touching the legacy pipeline. Ruby scope-resolution improvements: - Heritage: intercept include/extend/prepend in captures.ts, encode as special imports, emit IMPLEMENTS edges via emitHeritageEdges hook - Properties: intercept attr_accessor/attr_reader/attr_writer, emit Property nodes + HAS_PROPERTY edges via the same hook - Bare calls: add (body_statement (identifier)) capture to scope query, matching the legacy query pattern for zero-arity method calls - YARD parsing: second-pass comment scanner for @param/@return/@type annotations with findFollowingMethod that handles body_statement nesting - Query fixes: @declaration.trait for modules (was @declaration.module which normalizeNodeLabel didn't recognize), constant constructor bindings (SERVICE = UserService.new), call-return inference Parity: 114/127 legacy tests pass under registry-primary (up from 89). Remaining 13 are advanced type-inference chain resolution (compound receiver, cross-file return-type propagation, for-in element types). * feat(ruby): achieve 100% scope-resolution parity (127/127) Fix all 13 remaining type-inference failures: - Add expandsWildcardTo hook (expandRubyWildcardNames) so finalize can materialize individual bindings from require/require_relative wildcard imports, unblocking cross-file return-type propagation - Add member-call-return type binding synthesis in captures.ts for assignments like `x = obj.method()` — enables compound receiver chaining through member call return types - Add YARD @return support for attr_accessor/attr_reader/attr_writer calls, creating field-type bindings for chain resolution - Add @declaration.property captures alongside __property__ imports so properties register in localDefs → model.fields → write-access - Add constructor-return inference for methods ending with Foo.new() - Add for-loop variable type aliasing in scope query - Rebuild nodeLookup after emitHeritageEdges in run.ts so Property nodes created by the heritage hook are visible to downstream passes - Extend compound-receiver resolver to handle compound member-call rawNames with () and increase max depth from 4 to 8 - Extend receiver-bound-calls Case 3b for compound rawNames All 127 legacy Ruby tests pass under both REGISTRY_PRIMARY_RUBY=0 (legacy) and =1 (registry-primary). Ruby is now fully registered in MIGRATED_LANGUAGES with 100% scope parity. * test(ruby): add pipeline benchmark exercising heritage emission Synthetic Ruby codebases at 100/250/500 files with include + extend + prepend mixins, diamond mixin patterns (shared BaseMixin modules), attr_accessor properties, YARD annotations, and cross-file imports. Strict equality assertions verify exact IMPLEMENTS and HAS_PROPERTY edge counts: 4 IMPLEMENTS per class (include x2, extend, prepend) plus 1 per non-base mixin module, 3 HAS_PROPERTY per class. Dedup in emitRubyMixinEdges prevents double-counting when the worker path (repos >= 15 files) already created Property/IMPLEMENTS edges before scope-resolution runs. Scaling: 0.76x and 1.40x (both linear, well under 3x threshold). * ci: retrigger build * fix(ci): resolve format, registry-primary-flag, and sequential-mixin test failures - Run prettier on all changed files (captures.ts, run.ts, ruby-scope.test.ts, ruby.test.ts, ruby-pipeline-benchmark.test.ts) - Update registry-primary-flag.test.ts: use Swift (not in MIGRATED_LANGUAGES) instead of Ruby for the isolation and env-var mutation tests - Pin ruby-sequential-mixin.test.ts to REGISTRY_PRIMARY_RUBY=0 (legacy mode) since it tests inferImplicitReceiver + selectDispatch hooks that live in the legacy call-processor (gated off under registry-primary) --------- Co-authored-by: Test <test@example.com> |
||
|
Gergő Magyar
|
d5b2edddc4
|
fix(test): use retry cleanup in antigravity e2e to prevent ENOTEMPTY flake (#1838)
* fix(test): use retry cleanup in antigravity e2e to prevent ENOTEMPTY flake Replace bare `fsp.rm` / `fs.rmSync` in antigravity-hook-e2e.test.ts afterAll with `cleanupTempDir` / `cleanupTempDirSync` from test-db.ts which retry with backoff on transient filesystem errors. Also make `shouldSwallowCleanupError` swallow ENOTEMPTY on all platforms (was Windows-only). The CI failure on macOS was ENOTEMPTY on a deeply nested node-gyp cache directory inside the temp HOME — a cleanup-time race that retries usually resolve, but the final attempt must not crash the test suite if the race persists. * fix: restore fsp import needed for mkdtemp/mkdir --------- Co-authored-by: Test <test@example.com> |
||
|
Gergő Magyar
|
4870879b21
|
fix(wiki): add budget-aware grouping to prevent context overflow on large repos (#627) (#1832)
* fix(wiki): add budget-aware grouping to prevent context overflow on large repos (#627) When the grouping prompt exceeds 100k tokens (e.g. Apache TVM with ~2,378 files and ~306k estimated tokens), batch files by top-level directory and issue one LLM call per batch. Partial results are deterministically merged; any batch failure falls back to directory-based grouping. * fix(wiki): address review findings — exact assertions, progress fix, error logging - Replace bounds-only .toBeGreaterThan assertions with exact .toBe values - Add per-batch budget compliance assertion for sub-batch case - Add assertion that partial LLM results don't leak through nuclear fallback - Pass fixedPercent/percentRange to streamOpts in batched LLM calls - Log batch failure in onProgress before falling back to directory grouping - Strengthen mergeGroupings dedup test from .toContain to exact .toEqual * chore(autofix): apply prettier + eslint fixes via /autofix command * fix(wiki): prevent slug collisions and handle single-file oversize in batched grouping mergeGroupings now normalizes module keys by slug so case/punctuation variants ("API Routes" vs "API routes") merge into one module instead of producing colliding .md files. batchFilesForGrouping now truncates per-file symbol lists via binary search when a single file exceeds GROUPING_TOKEN_BUDGET, so every LLM request stays within the context window. * style(wiki): apply prettier formatting to generator.ts --------- Co-authored-by: Test <test@example.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> |
||
|
Gergő Magyar
|
c916c88361
|
feat(mcp): add limit/offset/summaryOnly pagination to impact tool (#1818)
* feat(mcp): add limit/offset/summaryOnly pagination to impact tool (#414) The impact tool returns unbounded byDepth arrays for hub symbols (base error classes, shared utilities), producing 140KB+ responses that get truncated by MCP clients. maxDepth alone does not help when most dependents are at depth 1. Add three new parameters: - summaryOnly: returns counts/risk/processes/modules without byDepth - limit: caps symbols per depth level (default 100) - offset: skips symbols for pagination Also adds byDepthCounts to all responses so agents can see total counts even when the symbol list is paginated or omitted. Closes #414 * fix(mcp): prevent pagination from silently truncating cross-repo impact Address review findings on #1818: - F1 (blocker): _runImpactBFS no longer defaults to limit 100 when limit is not set — only _impactImpl (MCP entry) applies the default. Internal callers (impactByUid, group impact) get complete results. GroupToolPort.impact interface gains optional limit param, and cross-impact.ts passes limit: 10000 for local UID collection. - F2 (blocker): tool description updated — byDepth is now documented as paginated, not 'all affected symbols'. - F3: impactByUid calls _runImpactBFS without limit, so Phase-2 neighbor results are no longer capped at 100. - F4: pagination metadata now appears when offset > 0 (head truncation), not just tail truncation. Pagination.limit is null when uncapped. - F5: limit/offset schema types changed from number to integer; Math.trunc applied in implementation as defense-in-depth. - F6: 7 new tests — multi-depth pagination, offset-only truncation, offset past end, float inputs, _runImpactBFS internal uncapped path, collectImpactSymbolUids with paginated vs complete data. * fix(mcp): NaN guard on pagination params, complete GroupToolPort interface - Add Number.isFinite guard to limit/offset in _runImpactBFS so NaN inputs fall through to uncapped/zero defaults instead of producing silent empty byDepth with no truncation signal. - Add offset and summaryOnly to GroupToolPort.impact interface to match the implementation and prevent silent param loss at the port boundary. - Replace bounds-only toBeLessThan assertion with exact byDepthCounts and pagination assertions per DoD §2.7. * fix(mcp): address remaining review findings for impact pagination - #3: Forward limit/offset/summaryOnly through callToolAtGroupRepo so group-mode MCP callers can use the new pagination params. - #4: Extract GROUP_LOCAL_PHASE_LIMIT constant from magic 10000 in cross-impact.ts with a comment explaining the intent. - #7: eval-server formatImpactResult uses byDepthCounts[depth] for the 'and N more' suffix instead of paginated slice length. - #8: Extract ImpactParams interface from duplicate inline type definitions in impact() and _impactImpl(). - #9: Add --limit, --offset, --summary-only CLI flags to the impact command with i18n help strings (en + zh-CN). - #10: Clarify in tool description that limit/offset apply per depth level, not per total result set. * chore(autofix): apply prettier + eslint fixes via /autofix command * @ fix(mcp): address Copilot review feedback on impact pagination - Sanitize limit/offset with Number.isFinite in _impactImpl to prevent NaN passthrough from bypassing the default limit of 100 - Omit pagination.limit field instead of emitting null when paginationLimit is Infinity, keeping the response schema consistent - Move GROUP_LOCAL_PHASE_LIMIT after all imports in cross-impact.ts - Stop forwarding limit/offset/summaryOnly to group-mode impact since runGroupImpact overrides limit with GROUP_LOCAL_PHASE_LIMIT for UID collection and does not re-paginate - Validate CLI parseInt results with Number.isFinite before passing to the backend, falling back to undefined so defaults apply - Use byDepthCounts to decide whether to render depth sections in formatImpactResult, handling empty pages from offset past end @ * @ fix(mcp): address code review findings on impact pagination - Fix formatImpactResult "N more" count: use Math.min(items.length, 12) instead of hardcoded 12, so paginated pages with <12 items show the correct remaining count - Detect summaryOnly responses (byDepth absent, byDepthCounts present) and show a summary-mode message instead of misleading "(0 items on this page — adjust offset)" per depth level - Document that limit/offset/summaryOnly are single-repo only and ignored in group mode (@groupName) in MCP tool schema descriptions - List byDepthCounts in summaryOnly description and note byDepth absence when summaryOnly is true - Remove unused limit/offset/summaryOnly from GroupToolPort.impact interface since they are never forwarded to group impact - Deduplicate parseInt calls in CLI tool.ts: extract to local variables with consistent optional-chain usage @ * chore(autofix): apply prettier + eslint fixes via /autofix command * @ fix(group): restore limit in GroupToolPort.impact interface cross-impact.ts passes limit: GROUP_LOCAL_PHASE_LIMIT through the GroupToolPort.impact interface for UID collection. Only offset and summaryOnly were truly unused — limit must stay. @ * @ docs: add limit/offset/summaryOnly to impact tool options in README @ --------- Co-authored-by: Test <test@example.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> |
||
|
Sparsh
|
966ddb981e
|
feat(cpp): thread base-specifier qualifier through dependent-base lookup (#1815) (#1819)
Some checks are pending
CodeQL / Analyze (javascript-typescript) (push) Waiting to run
CodeQL / Analyze (python) (push) Waiting to run
Gitleaks / gitleaks (push) Waiting to run
Publish / Classify release event (push) Waiting to run
Publish / RC guard (marker + release-PR skip) (push) Blocked by required conditions
Publish / ci (push) Blocked by required conditions
Publish / Publish to npm (push) Blocked by required conditions
Publish / Build & Push RC Docker images (push) Blocked by required conditions
Scorecard / Scorecard analysis (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-cli) (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-web) (push) Waiting to run
* feat(cpp): thread base-specifier qualifier through dependent-base lookup (#1815) captures.ts: add extractBaseLookupQualifier, fix isBaseDependent for qualified_identifier bases. two-phase-lookup.ts: qualifier storage, markCppDependentBase accepts qualifier, dedup index by nodeId (last-wins), V3 qualifier targeting (dormant). Infrastructure delivered: qualifier extraction, storage, dedup, isBaseDependent fix. V3 targeting dormant until qualifiedName computation fix reaches localDefs. Part of #1564. Infrastructure for #1815. * fix(cpp): three conservatism fixes for dependent-base lookup Fix 1 — Map collision in markCppDependentBase (line 83): Change innermost storage from Map<baseName, qualifier> to Map<baseName, Set<qualifier>> so multiple captures of the same dependent base name with different qualifiers don't collide. Fix 2 — Single-candidate bypass (lines 197-206): For qualified bases with only one candidate, verify namespace match before accepting. Unqualified bases still accept the unique candidate. Previously accepted regardless, creating false edges. Fix 3 — V3→V2 fallthrough (line 221): When a syntactic qualifier is present but no exact match is found, suppress rather than falling through to V2 prefix-heuristic. V2 only runs for truly unqualified bases, which is what it was designed for. All three are conservative bug fixes — turn false positives into suppression, not behavior changes. 250/250 tests pass both modes. --------- Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> |
||
|
dependabot[bot]
|
05151b1079
|
chore(deps)(deps): bump lru-cache from 11.3.6 to 11.4.0 in /gitnexus (#1826) | ||
|
dependabot[bot]
|
5458ddce77
|
chore(deps)(deps-dev): bump @types/node in /gitnexus (#1825) | ||
|
dependabot[bot]
|
3161938730
|
chore(deps)(deps-dev): bump vitest from 4.1.6 to 4.1.7 in /gitnexus (#1824) | ||
|
Gergő Magyar
|
d4449b4ec8
|
fix(lbug): resolve non-ASCII paths for KuzuDB on Windows (#1811) (#1817)
* fix(lbug): resolve non-ASCII paths to 8.3 short form on Windows (#1811) KuzuDB's native C++ layer uses ANSI file APIs (fopen) on Windows. When the repo path contains CJK or other non-ASCII characters, the UTF-8 bytes from Node.js are misinterpreted as the system's Active Code Page (e.g. GBK), producing a garbled path — "Error 3: The system cannot find the path specified." Add `toNativeSafePath()` which converts non-ASCII paths to their Windows 8.3 short-name form (all-ASCII) before passing them to the native layer. Applied to both the database open path and the COPY CSV paths. No-ops on non-Windows and on all-ASCII paths. Closes #1811 * test(lbug): add unit + integration tests for non-ASCII path handling (#1811) - Unit tests for toNativeSafePath: ASCII passthrough, non-Windows no-op, Windows short-path conversion, nonexistent-path fallback - Integration test: full initLbug + loadGraphToLbug round-trip with CJK characters in the storage path — runs on all platforms - Fix toNativeSafePath to reject cmd.exe output containing '?' chars (replacement for unrepresentable Unicode in the console code page) - Register integration test in vitest lbug-db project and cross-platform-tests.ts matrix * chore(autofix): apply prettier + eslint fixes via /autofix command * feat(lbug): junction fallback, tmpdir CSV staging, pool-adapter coverage (#1811) U1+U4: toNativeSafePath now tries 8.3 short path → NTFS junction fallback → diagnostic warning. Junctions target path.dirname(p) and reconstruct the leaf. Handles EEXIST races. Registers cleanup on exit/SIGTERM/SIGINT. Orphan scan on first call removes stale junctions from prior crashes. U2: loadGraphToLbug redirects csvDir to os.tmpdir() when storagePath contains non-ASCII on Windows, avoiding non-ASCII characters in COPY FROM paths entirely. U3: All 4 createLbugDatabase call sites in pool-adapter.ts now wrap dbPath with toNativeSafePath. * fix(test): fix CI failures from toNativeSafePath addition (#1811) - Fix lbug-non-ascii-path integration test: use CodeRelation (actual relationship table name) instead of CALLS - Add toNativeSafePath to lbug-config.js mocks in pool-wal-recovery and lbug-pool-win-fts-probe tests — pool-adapter now imports it * fix(lbug): sanitize path before cmd.exe shell expansion (CodeQL) Reject paths containing cmd.exe metacharacters (" % | & < > ^) before interpolating into the `for %I` short-path command. Prevents command injection via crafted path names. * fix(lbug): address code review findings in non-ASCII path implementation - U1: Use process.exit(0) on Windows instead of process.kill re-raise (SIGTERM forcefully kills on Windows, handlers never fire) - U2: Pass safePath to openWithLockRetry so sidecar sweep targets the path KuzuDB actually opened, not the original non-ASCII path - U3: Skip junction creation in worker threads (isMainThread guard) to prevent junction leaks from pool-adapter workers - U4: Replace existsSync with lstatSync in orphan scan to avoid 30s blocking on unreachable UNC network targets * chore(autofix): apply prettier + eslint fixes via /autofix command * fix(lbug): correct SIGTERM exit code and run Prettier (#1811) - Use exit code 143 (SIGTERM) / 130 (SIGINT) on Windows instead of 0 so termination is not masked as success - Run Prettier to fix formatting (CI Gate blocker) * fix(lbug): eliminate CodeQL command-injection taint in tryShortPath Pass the path via GITNEXUS_SP environment variable instead of interpolating it into the cmd.exe command string. The FOR loop reads %GITNEXUS_SP% from the environment, so the command text is entirely static — no user-controlled data in the shell command. Also removes CMD_UNSAFE_RE since the env var approach makes character-level sanitization unnecessary. --------- Co-authored-by: Test <test@example.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> |
||
|
Gergő Magyar
|
a229e8e77b
|
fix(build): skip build.js when running outside the monorepo (#1795) (#1816)
`scripts/build.js` assumes the monorepo sibling `gitnexus-shared`
exists. When a user runs `npm install` from within a global install
directory, the `prepare` lifecycle fires `build.js`, which calls
`execSync(tsc, { cwd: nonExistentPath })` — Node reports this as
the misleading `spawnSync /bin/sh ENOENT`.
Add an early guard: if `gitnexus-shared` is absent and `dist/`
already exists (published package context), exit cleanly. If neither
exists, print a helpful error pointing to the monorepo checkout.
Co-authored-by: Test <test@example.com>
|
||
|
Nilotpal Kashyap
|
50c6acb108
|
feat(setup): implement antigravity integration setup and hook adapter… (#1730)
* feat(setup): implement antigravity integration setup and hook adapter for gitnexus * docs(readme): list Antigravity in supported editors * test(setup-antigravity): pin platform per-test to fix Windows CI failure The MCP entry assertion expected `npx` directly, but on Windows `getMcpEntry()` wraps it as `cmd /c npx ...`, which broke the Windows runner. Pin platform to darwin in beforeEach so the existing assertion is deterministic, restore the descriptor in afterEach, and add a parity test for the win32 cmd-wrapper shape. * fix(antigravity): align hook adapter to Gemini CLI schema + fix Windows CI Rebase the Antigravity integration on the canonical Gemini CLI hooks contract (https://geminicli.com/docs/hooks/reference/), which is the documented schema Antigravity 2.0 inherits: - Hook adapter: replace PreToolUse/PostToolUse with the single AfterTool event. BeforeTool has no documented context-injection channel in the Gemini contract, so augmentation runs in AfterTool where hookSpecificOutput.additionalContext is the documented way to append text to the tool result the agent reads. Stale-index hints land in the same channel (so the agent sees them) and are mirrored to stderr for terminal users. Tool-name matcher updated to Gemini CLI snake_case (search_file_content|glob|run_shell_command). - Setup: write hooks to ~/.gemini/settings.json under canonical hooks.AfterTool[] (replaces the ad-hoc hooks.json top-level group). Polite-neighbor merge preserves existing user hooks. Also copy win-rm-list-json.ps1 alongside hook-db-lock-probe.cjs so the Windows MCP server ownership probe doesn't silently fail open. - Tests: 17 regression tests covering MCP write, win32 shape, hook schema, polite-neighbor merge, idempotency, adapter context emission, stale-index hint, and skill layout. - README: footnote documenting the AfterTool design choice and a link to the Gemini CLI hooks reference. Windows CI fix: installSkillsTo previously used glob('*.md') + glob('*/SKILL.md'), which returned zero matches under the Windows runner's temp paths (8.3 short-name like RUNNER~1). Replace with fs.readdir + dirent type checks — same behavior, no path quirks. This fixes the only failing Windows job on the PR. * fix(antigravity): address PR review — windowsHide, stale docs, dead code Addresses the production-readiness review findings on PR #1730: - F1 (blocker): add windowsHide:true to all four spawnSync sites in the Antigravity hook adapter (findCanonicalRepoRoot, runGitNexusCli's two branches, buildStaleIndexHint) so they don't flash console windows on Windows. Matches the fix #1794 already on main for the Claude hook. - F2 (blocker): update gitnexus/README.md editor table to say AfterTool and link the Gemini CLI hooks reference. The published README had drifted to the pre-c1872b4 PreToolUse + PostToolUse schema. - F3: rewrite the stale ~/.gemini block comment in setup.ts. It still described the old hooks.json + gitnexus group + grep_search design. - F4: remove grep_search dead code from extractPattern and its doc comment. The registered matcher is search_file_content|glob|run_shell_command, so grep_search would never be invoked. - F5: annotate timeout:10000 with a ms-unit comment noting Gemini CLI uses milliseconds (Claude Code uses seconds). - F6: add the GITNEXUS_DEBUG branch to extractAugmentContext for parity with the Claude adapter, so suppressed augment stderr is recoverable. - F7: stageAdapter test helper now copies win-rm-list-json.ps1 alongside the .cjs helpers, so the adapter's Windows lock-probe path isn't a silent fail-open in child-process smoke tests. * test(antigravity): add integration tests and register in cross-platform matrix Adds end-to-end coverage on top of the unit-level tests, per maintainer request: - test/integration/setup-antigravity.test.ts (10 tests): exercises the real setupCommand() against a temp HOME with ~/.gemini/antigravity/ present. Verifies mcp_config.json shape, ~/.gemini/settings.json AfterTool entry, adapter + helpers + win-rm-list-json.ps1 copy, baked-in cliPath rewrite (issue #108 regression class), skill layout, polite-neighbor merge against existing user hooks, idempotency, skip-when-absent, corrupt-file safety, and key preservation. - test/integration/antigravity-hook-e2e.test.ts (19 tests): runs the full install-then-execute flow — invokes setupCommand to lay down the adapter + helpers, then spawns the INSTALLED adapter as a real child process against a temp git repo + .gitnexus/. The source adapter cannot be spawned directly (it requires sibling .cjs helpers that only live in hooks/claude/); install-then-spawn mirrors the production codepath. Covers staleness detection across all five git mutation types, --embeddings propagation, polite skip on toolResponse.error / exit_code !== 0, augment crash-free behavior, cwd validation, corrupted/missing meta.json, unknown event names, empty stdin, and the no-.gitnexus deep-nested case. - scripts/cross-platform-tests.ts: registers all three antigravity test files (unit in PLATFORM_LOGIC, two integration files in SPAWN_CLI) so Windows and macOS CI exercise them on every run. * fix(antigravity): review fixes — dedup, silent-failure guard, type coercion, glob filter - Delete mergeGeminiSettingsHooks (verbatim copy of mergeHooksJsonc), replace call site with the original - Unify geminiHasGitnexusHook into hasGitnexusHook with commandFragment parameter; delete the duplicate - Guard against silent adapter-copy failure: verify the adapter file exists before registering the AfterTool hook entry in settings.json; surface helper copy errors instead of swallowing - Fix toolSucceeded type coercion: use Number() so string exit_code values from Gemini CLI are handled correctly - Align glob tool extractPattern with Claude adapter's restrictive regex filter (/[*\/]([a-zA-Z][a-zA-Z0-9_-]{2,})/) - Remove bounds-only toBeGreaterThan(0) assertion (DoD §2.7) - Add antigravity adapter to HOOK_FILES windowsHide regression list * chore(autofix): apply prettier + eslint fixes via /autofix command * chore: trigger CI --------- Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> Co-authored-by: Test <test@example.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> |
||
|
Copilot
|
baf57bec88
|
feat(rust): Migrate Rust to scope-based resolution (RFC #909 Ring 3) (#1639)
* Initial plan * feat: add Rust scope-resolution hooks (RFC #909 Ring 3) Implement the scope-based resolution pipeline for Rust, following the established pattern from Go and other migrated languages. New files in gitnexus/src/core/ingestion/languages/rust/: - query.ts: tree-sitter scope query covering scopes, declarations, imports, type bindings, and references - cache-stats.ts: parse cache hit/miss counters - import-decomposer.ts: decomposes use declarations into individual import captures (handles grouped, wildcard, renamed, re-exported) - receiver-binding.ts: synthesizes self type bindings for impl methods - interpret.ts: interprets captures into ParsedImport/ParsedTypeBinding - arity.ts: arity compatibility checker (no overloading in Rust) - merge-bindings.ts: local-shadows-import binding merge strategy - simple-hooks.ts: binding scope, import owning scope, receiver binding - import-target.ts: resolves Rust module paths (crate/super/self) - method-owners.ts: bridges impl block methods to struct defs - captures.ts: main emit function with import decomposition and self-binding synthesis - scope-resolver.ts: ScopeResolver implementation - index.ts: barrel re-exports Wiring changes: - rust.ts: add scope hook imports and properties to defineLanguage - registry.ts: register rustScopeResolver in SCOPE_RESOLVERS - registry-primary-flag.ts: add Rust to MIGRATED_LANGUAGES Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: magyargergo <11230420+magyargergo@users.noreply.github.com> * feat(LANG-rust): add scope-resolution hooks and register Rust ScopeResolver Implements RFC #909 Ring 3 deliverables: - query.ts: tree-sitter scope query for Rust - captures.ts: emitRustScopeCaptures with method reclassification - import-decomposer.ts: use statement decomposition (groups, renames, globs) - interpret.ts: interpretRustImport + interpretRustTypeBinding - import-target.ts: crate/module/super/self path resolution - receiver-binding.ts: self/&self/&mut self receiver synthesis - method-owners.ts: impl block → struct ownership bridging - arity.ts: no-overloading arity check - merge-bindings.ts: local > import > wildcard binding precedence - simple-hooks.ts: binding/import scope, receiver binding - scope-resolver.ts: ScopeResolver contract implementation - Wired into rustProvider (rust.ts) with scope hooks - Registered in SCOPE_RESOLVERS (pipeline/registry.ts) - NOT yet added to MIGRATED_LANGUAGES (29 advanced pattern tests pending) Agent-Logs-Url: https://github.com/abhigyanpatwari/GitNexus/sessions/8f14b730-79d4-4356-9505-325750d71f84 Co-authored-by: magyargergo <11230420+magyargergo@users.noreply.github.com> * Add Rust scope-resolution integration tests (RFC #909 Ring 3) Tests cover the core deliverables for the Rust scope-resolution pipeline: - impl blocks and trait implementations - Module resolution (crate::, super::, self::) - Struct fields and type bindings - Self/&self/&mut self receiver binding - Generic functions (V1 ignores generic args) - Grouped imports (use foo::{A, B}) - Renamed imports (use foo::Bar as Baz) - Arity checking (no overloading) - Struct literal constructor inference - Return type inference - Scoped/qualified calls (Foo::new()) - Enum declarations - Multiple impl blocks - Free function calls - Typed let bindings Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: magyargergo <11230420+magyargergo@users.noreply.github.com> * test(LANG-rust): add 28 scope-resolution integration tests Covers 15 test suites validating: impl blocks, trait impls, grouped imports, renamed imports, module resolution, receiver binding, arity filtering, struct literal inference, return type inference, qualified calls, struct fields, enums, multiple impl blocks, free calls, and typed let bindings. Agent-Logs-Url: https://github.com/abhigyanpatwari/GitNexus/sessions/8f14b730-79d4-4356-9505-325750d71f84 Co-authored-by: magyargergo <11230420+magyargergo@users.noreply.github.com> * feat(LANG-rust): add implicit crate path fallback, 35 scope tests - Import resolver now falls back to crate-relative for unqualified module paths (Rust 2015 edition compat) - Added 7 more test cases: re-exports, shadowing, closures, default trait methods (35 total, exceeding ≥30 requirement) Agent-Logs-Url: https://github.com/abhigyanpatwari/GitNexus/sessions/8f14b730-79d4-4356-9505-325750d71f84 Co-authored-by: magyargergo <11230420+magyargergo@users.noreply.github.com> * feat(rust): add Rust to MIGRATED_LANGUAGES with 100% scope-resolution parity 35/35 integration tests pass under both REGISTRY_PRIMARY_RUST=0 (legacy) and =1 (scope-resolution). Rust scope-resolution is now the default production call-resolution path. * test(rust): add pipeline benchmark matching PHP benchmark pattern Generates synthetic Rust codebases at 100/250/500 files with structs, impl blocks, traits, cross-module use declarations, and method calls. Measures wall-clock time, peak heap, and scaling ratios. Results: sub-linear scaling (0.79x ratio), 500 files in 3.8s with workers, 110MB peak heap. Gated by GITNEXUS_BENCH=1. * chore(autofix): apply prettier + eslint fixes via /autofix command * perf(rust): optimize captures, type normalization, and method-owner linking - Cache findEnclosingImpl result to avoid duplicate tree walk per function - Avoid double namedChild accessor call in struct field arity counting - Extract regex constants (REF_PREFIX_RE, PTR_PREFIX_RE) from hot normalization loops - Replace O(s) suffix-match scan with O(1) Map lookup in method-owner linking Benchmark: 500 files 3806ms → 2970ms (-22%), 250 files 2432ms → 1752ms (-28%) * fix(rust): resolve CodeQL alerts — file-system race and dead code - Remove existsSync+appendFileSync/writeFileSync TOCTOU in benchmark fixture generator; appendFileSync creates if missing - Remove always-true guard on computeRustCallArity return; narrow return type from number|undefined to number - Remove no-op .filter() in fixture generator * feat(rust): hoist impl return-type bindings to struct scope for chain resolution Synthesize a module-level duplicate of @type-binding.return captures for methods inside impl blocks. The scope-extractor's auto-hoist places these on the struct's Class scope, making them visible to the compound receiver chain resolver via classScopeByDefId. Without this, method return types are only on the impl block scope (which has no class-like def and is not indexed by classScopeByDefId), so chains like svc.get_user().save() cannot follow the intermediate return type. This is the structural prerequisite for chain resolution, pattern binding, and for-loop element-type parity (28 remaining tests). The cross-file return-type propagation step still needs wiring for full parity. * fix(rust): use implNode anchor for return-type hoisting — fixes chain resolution Use the enclosing impl_item node (not tree.rootNode) as the synthetic capture anchor. The scope-extractor's auto-hoist places bindings whose anchor matches the innermost scope on the parent scope. With implNode, the binding lands on the Module scope (parent of impl's Class scope), giving declaredAtScope the correct context for findClassBindingInScope to resolve the return type across the scope chain. Unlocks: chain calls (svc.get_user().save()), return-type inference, assignment chains, cross-file binding propagation, call-result binding, deep field chains — 28→27 failing tests. * feat(rust): add populateRangeBindings hook + .await query capture Implement populateRustRangeBindings (Phase 2 hook, same pattern as Go's populateGoRangeBindings) to populate type bindings that need runtime type lookup — for-loop element types, if-let/while-let captured patterns, match arm patterns, and struct destructuring field types. Also add tree-sitter query capture for let x = fn().await — unwraps await_expression to find the inner call_expression. 28→19 failing tests: fixes for-loop Tier 1c, .iter()/.into_iter(), async .await, if-let captured_pattern. * fix(rust): fix tuple_struct_pattern variable extraction + Result<T,E> raw type lookup - Skip wrapper type identifier when finding bound variable in tuple_struct_pattern (Some(user) was binding 'Some' not 'user') - Add lookupRawParameterType to read unstripped generic type from AST for Ok/Err pattern resolution (normalizeRustTypeName strips generics) 28→15 failing tests: fixes if-let Some, if-let Ok/Err, match arm patterns. * fix(rust): fix match_arm parent traversal + raw return type for for-loop calls - Walk up from match_arm through match_block to find match_expression for source variable extraction - Add lookupRawFunctionReturnType to find unstripped return type from AST for same-file for-loop call expression iterables 28→14 failing tests. * feat(rust): inject field type bindings on struct scopes for chain resolution Walk struct_item AST nodes and inject field types (e.g., address -> Address) as typeBindings on the struct's Class scope. The compound receiver chain resolver uses these to follow field chains like user.address.save(). Also fixes: match_arm parent traversal to match_expression, lookupFieldType to check typeBindings first. 28→11 failing tests: fixes field type chains, deep chains, struct destructuring. * feat(rust): cross-file return type lookup for for-loop call iterables Build allReturnTypes map across all parsedFiles in Phase 2 first pass, then use it to resolve for-loop iterables like `for x in get_fn()` when get_fn is defined in another file. 28→9 failing tests. * feat(rust): cross-file field type map for struct destructuring Build allFieldTypes map across parsedFiles in Phase 2 first pass. Used by processStructDestructuring to resolve `let Point { x, y } = p` when Point is defined in another file. 28→7 failing tests. * fix(rust): compound assignment write capture + pending assignment fixpoint - Add compound_assignment_expr query for +=, -=, etc. field writes - Add processPendingAssignments with 3-pass fixpoint for field access and method call result variable bindings (let addr = user.address, let city = addr.get_city()) 28→5 failing tests. * fix(rust): identity method return-type bindings for unwrap/expect chains Inject unwrap/expect/clone/as_ref/as_mut as return-type bindings on struct scopes that return the struct's own type. Since normalizeRustTypeName already unwraps Option<T> → T, calling .unwrap() on a value typed as T is semantically an identity — the return type equals the receiver type. 28→3 failing tests: fixes user.unwrap().save() and repo.unwrap().save() chains. * fix(rust): skip enum variant call-return bindings + cross-file pending assignments + identity alias - Skip Some/None/Ok/Err in @type-binding.call-return — these are enum variant constructors, not type names; let the annotation capture win - Add identifier alias handler in processPendingAssignments for `let alias = opt` chains - Cross-file field type and method return type lookup in pending assignment fixpoint via findFieldTypeAcrossFiles/findMethodReturnTypeAcrossFiles - Identity method bindings (unwrap/expect/clone) on struct scopes 155/156 tests pass (99.4%). Remaining: trait default method dispatch via MRO (repo.count() where count has default impl on Repository trait). * feat(rust): 100% scope-resolution parity — MRO with same-file IMPLEMENTS + trait default method reclassification - Add buildRustMro that includes same-file IMPLEMENTS edges in the MRO chain, so trait default methods (e.g., repo.count()) resolve through the struct → trait ancestry walk - Only add IMPLEMENTS to MRO when struct and trait are in the same file; cross-file trait calls require the trait to be imported (Rust semantics) - Reclassify function_item inside trait_item as @declaration.method so default trait methods register in the model's methods lookup 156/156 legacy parity tests pass. 35/35 scope tests pass. 0 regressions. * fix(rust): address code review findings — null guard, name collision, scope order - Fix Array.find() null guard: check === undefined not === null in processCapturedPattern (find() never returns null) - Fix allReturnTypes/allFieldTypes name collision: delete entry on second occurrence so colliding names (new, default, Config) produce no result rather than a wrong result - Fix lookupTypeInScopes: search function scope then module scope only, skip unrelated Class scopes that could shadow names from other functions --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: magyargergo <11230420+magyargergo@users.noreply.github.com> Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> Co-authored-by: Test <test@example.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> |
||
|
eddie.pan
|
5ce448a93a
|
feat(wiki): support local Claude and Codex providers (#1769)
* feat(wiki): support local Claude and Codex providers * chore(autofix): apply prettier + eslint fixes via /autofix command * fix(wiki): address local CLI provider review findings - Add subprocess timeout: LocalCLIConfig gains requestTimeoutMs, runLocalCLI sets a kill timer that rejects with an actionable error matching the HTTP timeout message format. --timeout is no longer silently ignored for claude/codex providers. - Add windowsHide: true to spawn() to prevent console window flash on Windows, matching cursor-client.ts behavior. - Skip GITNEXUS_MODEL env var for local providers so a user's OpenAI model name doesn't cross-contaminate claude/codex CLI invocations. Precedence for local providers: --model → savedLocalModel → ''. - Guard against empty stdout: reject with actionable error when CLI exits 0 but produces no output, preventing silent empty wiki pages. * fix(wiki): address deep-review findings in local CLI providers - Move empty-output guard from runLocalCLI to per-provider callers so Codex can read --output-last-message file even when stdout is empty - Merge existing config in interactive setup (local + Azure paths) to prevent saveCLIConfig from erasing previously saved API keys - Use StringDecoder for stdout/stderr to handle multi-byte UTF-8 chars split across pipe chunk boundaries - Distinguish ENOENT from non-zero exit in detectLocalCLI so users see auth guidance instead of misleading "CLI not found" when the binary exists but is not authenticated * test(wiki): add subprocess contract tests for local CLI providers Add 21 integration-level tests covering the Claude and Codex subprocess contracts that wiki-flags.test.ts mocks out: - Claude argv: -p, --output-format text, --no-session-persistence, --model conditional, stdin prompt content, CI=1, windowsHide:true - Codex argv: exec subcommand, --sandbox read-only, -c approval_policy, --output-last-message temp path, --cd, stdin marker, --model - Timeout: kill timer fires and rejects, no timer when unset - Codex file fallback: stdout used when file missing, error when both empty - detectLocalCLI: warn on non-ENOENT, silent on ENOENT - onChunk: cumulative byte count forwarded Also register the test in cross-platform-tests.ts SPAWN_CLI section and fix detectLocalCLI ENOENT detection logic (invert the check so non-ENOENT errors produce a warning). * fix(wiki): platform-aware process tree kill and Codex contract snapshot - Add killChildTree helper that uses taskkill /T /F /PID on Windows to terminate the entire process tree (including cmd.exe grandchildren), with fallback to child.kill() if taskkill fails or on non-Windows - Add Codex CLI flag contract snapshot test that locks the exact spawn args — any flag rename, reorder, or removal is caught immediately - Add Windows taskkill tests: success path asserts taskkill called with correct PID and /T /F flags, failure path verifies child.kill() fallback --------- Co-authored-by: eddie.pan2 <eddie.pan2@jtexpress.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> Co-authored-by: Test <test@example.com> |
||
|
Sparsh
|
1bbc876336
|
fix(cpp): dependent-base resolution across nested/inline namespaces (#1634) (#1814)
* fix(cpp): dependent-base resolution across nested/inline namespaces (#1634) Replace exact namespace-prefix match with prefix-contains filter capped at one level deeper, then accept only if exactly one candidate survives. Behavior change: - Derived<T> in ns::outer can now find Inner<T> in ns::outer::inner (nested namespace) or ns::v1 (inline namespace) via prefix walking - Global-scope deriving classes match any single-segment namespace - Sibling namespace collisions (e.g. detail::Inner vs public_api::Inner) correctly suppress when multiple candidates share the same simple name - Deep nesting (ns → ns.a.b) still suppresses (one-level cap) Fixtures added: pos: nested ns, this->f() -> 1 edge to inner::Inner::f neg: no Inner exists -> 0 edges inline: inline namespace variant -> 1 edge sibling-suppress: sibling collision -> 0 edges (ambiguity suppressed) Part of #1564. 64. * test: add deep-nesting suppression fixture, link #1815 in comment, unqualify inline fixture - Update code comment to reference follow-up issue #1815 instead of 'deferred to follow-up' - Inline fixture: drop explicit v1:: qualifier (exercise inline-expansion path more idiomatically as DoD intended) - Add deep-nesting suppression fixture (ns.a.b -> 0 edges) that pins the one-level cap as a documented invariant - Add legacy parity entry for deep-nesting fixture Part of #1564, #1634. --------- Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> |
||
|
Gergő Magyar
|
5e8690f992
|
feat(progress): add per-language progress reporting to scope-resolution phase (#1813)
* feat(progress): add per-language progress reporting to scope-resolution phase (#1741) The scope-resolution phase (which can run 74+ minutes on large Java/Kotlin repos) previously emitted zero progress updates, causing the CLI progress bar to freeze at ~49% with a stale "Parsing code" label — making users think the tool was stuck. - Add `scopeResolution` to PipelinePhase type and PHASE_LABELS - Add `onProgress` callback to `runScopeResolution` with per-file updates during the extract loop and sub-phase boundary markers (building scope model, resolving references, emitting edges) - Wire progress through `scopeResolutionPhase` with pre-counted file totals, per-language labels, and pipeline-wide percent mapping (90-95 internal) - Bump mro/communities/processes percent ranges to 95-100 to maintain monotonic progress after scope resolution - Add `scopeResolution` to mro's deps (latent ordering fix: mro reads EXTENDS edges that scope resolution writes via preEmitInheritanceEdges) * fix(progress): clamp overallRatio, fire final extract event, fix mro @deps JSDoc - Clamp overallRatio to [0,1] so percent never exceeds 95 when readFileContents drops files (langFileCount < totalScopeFiles) - Fire onProgress for the last file in the extract loop even when files.length is not divisible by progressInterval - Update mro @deps JSDoc to include scopeResolution * fix(progress): ensure bar redraws at every state transition - Fire initial 'extracting' event at file 0 so the sub-phase label appears immediately, not after progressInterval files - Emit a completion event at percent 95 when scope resolution finishes so the bar definitively reaches the phase ceiling before mro starts * feat(progress): improve UX with human-readable elapsed, language counter, cleaner labels - Format elapsed time as "5m 12s" / "1h 20m" instead of raw "(312s)" for all pipeline phases (CLI-wide improvement) - Add language counter "[1/3]" to scope-resolution detail so users know how many languages remain and which is active - Rename sub-phases for clarity: "building scope model" → "analyzing types", "emitting edges" → "linking symbols" - Remove nested parentheses from detail strings for cleaner display - Expand scope-resolution percent range from 5 to 8 points (90-98 internal → 54-59% display) for more visible bar motion - Re-allocate mro (98), communities (98-99), processes (99-100) * feat(progress): typed sub-phases, i18n locales, and test coverage - Extract ScopeResolutionSubPhase union type with exhaustive switch guard so adding a sub-phase without updating phase.ts is a compile error - Add scopeResolution key to en and zh-CN locale files so the web UI shows translated labels instead of raw message fallback - Extract formatElapsed to its own module with 7 boundary-value tests (0s, 59s, 60s, 3599s, 3600s, 3661s, 7323s) - Add runScopeResolution onProgress integration test proving sub-phase order (extracting → analyzing types → resolving references → linking symbols) and the 0-file early-return path --------- Co-authored-by: Test <test@example.com> |
||
|
Alaa Kaddour
|
efcab45560
|
feat(web): support GITNEXUS_BACKEND_URL env var for Docker deployments (#1286)
* feat(web): support GITNEXUS_BACKEND_URL env var for Docker deployments * fix(docker): escape inline script injection to prevent XSS and add server-level integration tests - Add jsonForScriptTag() that escapes <, >, & after JSON.stringify to prevent </script> breakout in inline config script - Sanitize rawBackendUrl in warning log to prevent log injection via newlines - Replace 5 duplicated-helper injection tests with 7 server-level HTTP integration tests that spawn the real docker-server.mjs with GITNEXUS_BACKEND_URL set - Add XSS-specific test: URL containing </script> must produce exactly 1 <script> tag - Add empty-string backendUrl frontend test - Improve Docker Compose Linux guidance with explicit <server-ip> example * fix(docker): harden log sanitization, fix error leak, fix killAndWait race - Broaden log sanitization regex from [\r\n] to [\x00-\x1f\x7f] to strip all C0 control characters including ANSI escape sequences - Replace error.message leak in 500 handler with generic string; log the real error server-side via console.error - Fix killAndWait TOCTOU race by registering exit listener before kill and adding post-kill exitCode guard * fix(docker): handle readFile race to resolve CodeQL file-system-race alert Wrap readFile in try/catch so the TOCTOU between stat() and readFile() is handled gracefully — if the file vanishes between the check and the read, return 404 instead of crashing. * @ fix(docker): eliminate TOCTOU race and format web components Replace the previous try/catch approach with fs.promises.open() to get a file handle, then use handle.stat()/readFile()/createReadStream() from the same fd — properly eliminates the CodeQL "file system race condition" alert by removing the window between stat() and read. Also runs prettier on the 5 web component files that were failing the format CI check. @ * chore(autofix): apply prettier + eslint fixes via /autofix command * chore: trigger CI * @ fix(docker): pass GITNEXUS_BACKEND_URL to the web container The env var was documented but commented out, so docker-server.mjs never received it and the config injection was dead. Uncomment the environment block with a passthrough default so users can set GITNEXUS_BACKEND_URL in .env or their shell for remote/custom deployments. @ * @ fix(docker): eliminate stat() to resolve CodeQL js/file-system-race CodeQL pairs any stat() (FileCheck) with a subsequent open() (FileUse) on an aliased path. The previous approach kept stat() for directory detection, which the analyzer flagged regardless of the fd-based reads. Replace stat() entirely with open() + handle.stat(). On Linux (Docker), open() succeeds for directories, so handle.stat().isDirectory() detects them without a standalone stat() call. This removes the FileCheck node from the data-flow graph, eliminating the alert at its source. @ * @ fix(docker): break CodeQL path alias chain between open() calls CodeQL js/file-system-race pairs two open() calls when their path arguments are data-flow aliased. The previous approach derived the fallback path from the request path (resolve(initialPath, index.html)), creating an alias chain the analyzer could trace. Restructure so the SPA fallback uses a module-level constant (spaFallback = resolve(root, index.html)) with zero data-flow from the request. The two open() calls now have provably independent path arguments, eliminating the FileCheck/FileUse pair. Also simplifies the logic: for an SPA, all non-file requests serve root/index.html — no directory/index.html detection needed since the client-side router handles subroutes. @ --------- Co-authored-by: Test <test@example.com> Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> |
||
|
Sparsh
|
73a6a5376e
|
fix(cpp): thread call-site types into qualified member lookup (#1632) (#1810)
Some checks are pending
CodeQL / Analyze (javascript-typescript) (push) Waiting to run
CodeQL / Analyze (python) (push) Waiting to run
Gitleaks / gitleaks (push) Waiting to run
Publish / Classify release event (push) Waiting to run
Publish / RC guard (marker + release-PR skip) (push) Blocked by required conditions
Publish / ci (push) Blocked by required conditions
Publish / Publish to npm (push) Blocked by required conditions
Publish / Build & Push RC Docker images (push) Blocked by required conditions
Scorecard / Scorecard analysis (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-cli) (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-web) (push) Waiting to run
* fix(cpp): thread call-site types into qualified member lookup (#1632) Widen Callsite (arity optional, add argumentTypes) and add optional callsite?: Callsite to ScopeResolver.resolveQualifiedReceiverMember. receiver-bound-calls.ts passes the ReferenceSite through structurally; resolveCppQualifiedNamespaceMember forwards it to narrowOverloadCandidates along with cppConversionRank, enabling exact-type and conversion-rank disambiguation across inline-namespace children. Behavior change: - outer::foo(42) where v1 declares foo(int) and v2 declares foo(double) now resolves to v1::foo (was: 0 edges, conservatively suppressed). - Same-name same-normalized-signature (e.g. foo(int) vs foo(long)) still suppresses at 0 edges via isOverloadAmbiguousAfterNormalization. - ADL using-import path (resolveAdlCandidates) unchanged — passes no callsite, narrowing degrades to existing pass-through behavior. Closes #1632. Part of #1564. * fix(cpp): update legacy parity expected-failure list for #1632 - Remove stale expected-failure entry for old diff-sigs test name (test now expects 1 edge; legacy DAG also emits 1 edge) - Add entry for normalized-signature ambiguity (int vs long) test - Rename describe block from 'conservative suppress' to 'distinct signatures resolved via call-site types' Verified both modes: REGISTRY_PRIMARY_CPP=1: 241/241 passed REGISTRY_PRIMARY_CPP=0: 194 passed, 47 skipped, 0 failed |
||
|
Lucas van Staden
|
1c4993251c
|
fix(php): synthesize module scope for namespace-less PHP files (.phtml) (#1801)
* fix(php): phtml scope synthesis with full-file range + O(1) Step 4 lookup (#1801, #1803) Address PR #1801 review findings and complete #1803 fix: scope-extractor.ts: - Synthetic Module scope uses full-file range (computed from existing drafts) so positionIndex containment works for top-level references in ERROR-root .phtml files - Orphan scope re-parenting done on drafts in extract() by replacing with new drafts — no mutation of readonly fields, no PHP-specific logic in shared buildScopeTree - Dead matchCount parameter removed from ensureModuleScope namespace-siblings.ts: - Step 4 parsedFiles.find() replaced with pre-built Map for O(1) lookup (was O(n²) with 16K files = ~256M comparisons) * test(php): add pipeline benchmark for scaling regression detection Synthetic PHP fixture generator (N files × M namespaces × K classes) with cross-namespace imports and calls. Measures wall-clock, peak heap, node/edge counts at 100/250/500 file scales with worker pool enabled. Results on current branch: - 100 files: 982ms, 65MB (9.8ms/file) - 250 files: 1310ms, 70MB (5.2ms/file) - 500 files: 2006ms, 92MB (4.0ms/file) - Scaling: sublinear (0.53x-0.77x ratio) Gated behind GITNEXUS_BENCH=1 so it does not run in normal CI. * chore: trigger CI * fix: prettier formatting + update scope-extractor test for synthesis behavior * fix: extend synthetic Module range to all captures + update integration test Address CI failure and review findings: - ensureModuleScope now computes range from ALL captures (scope, declaration, reference, type-binding) not just scope drafts. This ensures top-level references after the last inner scope are covered. - Update parse-worker-scope-integration test for synthesis behavior. - Update extract() docstring to document synthesis contract. --------- Co-authored-by: Test <test@example.com> Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> |
||
|
Gergő Magyar
|
06c3fb360d
|
fix(group): move manifest/workspace extraction before closeLbug (#1802) (#1807) | ||
|
Lucas van Staden
|
2006a3e5ac
|
fix(php): reduce memory during deferred-call accumulation and scope-resolution (#1800) | ||
|
Gergő Magyar
|
66f9ec8eff
|
feat(java): add Java to MIGRATED_LANGUAGES with 100% scope-resolution parity (#1805)
* feat(java): add Java to MIGRATED_LANGUAGES with 100% scope-resolution parity Route Java through the scope-resolution pipeline instead of the legacy single-threaded call processor, fixing the analyze hang on large Java codebases (issue #1741). Changes: - Add Java to MIGRATED_LANGUAGES (registry-primary-flag.ts) - Add tree-sitter queries for var type inference (call-result, alias, field-access, enhanced-for), instanceof/switch pattern bindings, and method references (User::getName, this::save, User::new) - Fix importedName to use simple class name instead of FQN so finalize binding materialization matches correctly - Implement buildJavaMro with IMPLEMENTS edge transitive closure for interface default method resolution - Implement populateJavaPackageSiblings for same-package implicit class visibility across files - Implement cross-file return-type mirroring from imported class files via populateRangeBindings hook - Add var type binding post-processing in captures.ts to resolve call-result and alias chains from same-file return types - Add variable-aware argument type inference for overload resolution - Fix pickConstructorOrClass to walk child scopes for Constructor defs (scope-resolution places them in Function scopes) - Remove over-aggressive field_access suppression in shouldEmitReadMember so ACCESSES edges emit for field steps in method chains - Enable collapseMemberCallsByCallerTarget for legacy parity - Update unit tests to use Ruby as unmigrated language example Parity: 178/178 integration tests pass in both registry-primary and legacy modes. * fix(java): address code review findings for scope-resolution migration - pickConstructorOrClass: skip inner Class scopes when walking children for Constructor defs (prevents resolving to wrong constructor in nested-class scenarios) - populateJavaCrossFileReturnTypes: filter out parameter-annotation bindings from class-scope mirroring to prevent foreign parameter types from shadowing local variables - resolveVarTypeBindings: detect ambiguous names (overloaded methods with different return types, same-named variables across scopes) and skip resolution rather than last-write-wins - sharedPrefixLength renamed to sharedSegmentCount: segment-based directory proximity for deterministic sort ordering - Add MAX_PACKAGE_FILES cap (500) to skip O(N^2) package-siblings injection for pathologically large packages * perf(java): optimize hot paths in scope-resolution migration - Replace O(D^2) list.some() dedup with O(1) Set lookup in populateJavaPackageSiblings binding injection - Replace queue.shift() O(N) with index-based O(1) iteration in closeInterfaces BFS traversal - Cache sharedSegmentCount results per file in sort comparator to avoid redundant path splitting * perf(ingestion): skip deferred accumulation for registry-primary languages The legacy call/import/heritage processing path accumulates extracted data from ALL files during the parse phase, then skips registry-primary files one-by-one during processing. For a 25K-file Java codebase this wastes ~150 MB holding calls that are never consumed. Gate the accumulation with a per-chunk file-path cache: calls, imports, heritage, constructor bindings, and assignments for registry-primary languages (Java, Python, TypeScript, Go, C#, C, C++, PHP, JavaScript, Kotlin) are no longer pushed into the deferred arrays. The scope- resolution pipeline handles these languages independently. Verified: 2258/2258 resolver integration tests pass across all languages. * fix(java): address Codex adversarial review findings - Cross-file return binding: detect ambiguous method names across imported classes (two classes with same-named methods but different return types) and delete the binding rather than first-wins - Package-siblings: only inject top-level classes (parent is Module scope) to prevent nested/inner classes from leaking to package scope - Add diagnostic log when MAX_PACKAGE_FILES cap fires so operators know same-package visibility was disabled for a large package * fix(test): force REGISTRY_PRIMARY_JAVA=false in legacy call-processor unit tests Three call-processor test suites use .java file paths to exercise legacy DAG features (MRO fast path, interface dispatch, class lookup fallback). Now that Java is in MIGRATED_LANGUAGES, the call-processor skips Java files. Force the flag off in beforeEach/afterEach so the legacy path runs, matching the existing Python pattern in the same file. --------- Co-authored-by: Test <test@example.com> |
||
|
Gergő Magyar
|
ac9a2ee12f
|
chore(ci): consolidate parity shards and narrow cross-platform matrix (#1798)
* chore(ci): reduce CI runner-minutes by consolidating parity and narrowing cross-platform
Scope-resolution parity previously spawned 9 separate GitHub Actions jobs
(one per migrated language), each doing full checkout + npm ci + build for
a single test file. Consolidate into one job running scripts/run-parity.ts
which loops through all migrated languages sequentially — same coverage,
~45 fewer runner-minutes of redundant setup per PR.
Cross-platform (Windows/macOS) previously ran the full 373-file test suite.
Narrow to 45 platform-sensitive files (native LadybugDB, process spawning,
path separators, worker threads, filesystem behavior). Full suite still runs
on Ubuntu with coverage.
Also adds 2 missing lbug integration tests (lbug-orphan-sidecar-recovery,
lbug-readonly-init) to the sequential lbug-db vitest project where they
belong, and rewrites TESTING.md to document all test lanes.
* fix: address code review findings on parity and cross-platform scripts
- Capture stderr in run-parity.ts (vitest writes diagnostics to stderr)
- Lower per-invocation timeout from 5min to 60s to stay within CI job limit
- Add --language flag validation (error on missing value)
- Add timeout diagnostic to run-cross-platform.ts catch block
- Add analyze-wal-checkpoint-failure.test.ts to lbug-db sequential project
- Expand cross-platform list: parser-loader, pipeline, pipeline-graph-golden,
setup-skills, cli/tool-no-index-stderr (51 files, was 45)
* fix: add shell:true for Windows npx resolution and simplify fs import
execFileSync('npx', ...) fails with ENOENT on Windows because npx is
npx.cmd — shell:true resolves this. Also replaces dynamic await
import('fs') with static import, and fixes timeout detection to use
err.killed instead of err.code.
* fix(ci): raise parity per-invocation timeout to 120s and job timeout to 30min
TypeScript and C++ resolver tests take 60-90s on CI runners, exceeding
the 60s per-invocation timeout. Raise to 120s. Also bump the job-level
timeout from 25 to 30 minutes for margin (realistic total is ~11 min).
* fix(ci): raise parity per-invocation timeout to 180s for C++ resolver
C++ resolver tests take 130-150s on CI runners due to template
metaprogramming, ADL, and SFINAE fixture volume. 120s was still too
tight. Realistic total across all 9 languages is ~12 min, well under
the 30-min job timeout.
* fix(ci): use stdio inherit for parity — no per-invocation timeout
Switch from piped stdio with per-invocation timeouts to stdio: 'inherit'.
Vitest output streams to CI console in real time, making failures
immediately visible. The CI job-level timeout (30 min) is the only
guard — no more artificial per-invocation timeouts that cut off slow
resolver tests like C++ (which genuinely takes 3+ minutes).
---------
Co-authored-by: Test <test@example.com>
|
||
|
ManniX-ITA
|
39e9b40136
|
fix(windows): pass windowsHide:true to every child_process spawn-family call (#1794)
* fix(hooks): pass windowsHide:true to every spawnSync to suppress flashing console windows on Windows
On Windows, every PostToolUse and Stop event from Claude Code (and
the Cursor integration variant) cold-spawns ``node`` / ``npx.cmd`` /
``git`` / ``lsof`` through ``child_process.spawnSync``. Without
``windowsHide: true`` in the options, Node's child_process module
asks ``CreateProcess`` to use ``STARTF_USESHOWWINDOW`` with
``SW_SHOWDEFAULT``, and a black console window flashes onto the
user's desktop for the duration of the call. Under active
editor / agent use this means a near-continuous stream of pop-up
windows — unusable in practice (reported live on a Windows 11
workstation running the gitnexus Claude plugin against an active
project; the flashes stack on the taskbar and steal focus from the
editor).
The Node fix is one option flag per spawnSync:
spawnSync(cmd, args, {
encoding: 'utf-8',
timeout,
cwd,
stdio: ['pipe', 'pipe', 'pipe'],
windowsHide: true, // <-- new
});
``windowsHide`` is a no-op on macOS/Linux (Node docs: "Hide the
subprocess console window that would normally be created on Windows
systems"), so the patch is platform-neutral and zero-risk on the
other two majors.
This commit touches every ``spawnSync`` call in the three sources
that ship the hook layer:
* gitnexus/hooks/claude/gitnexus-hook.cjs (4 sites)
* gitnexus/hooks/claude/hook-db-lock-probe.cjs (3 sites)
* gitnexus-claude-plugin/hooks/gitnexus-hook.js (6 sites)
* gitnexus-claude-plugin/hooks/hook-db-lock-probe.cjs (3 sites)
* gitnexus-cursor-integration/hooks/gitnexus-hook.cjs (3 sites)
Total: 19 spawn sites guarded. ``hook-lock.cjs`` / ``hook-lock.js``
don't spawn subprocesses; nothing else in the hooks/ dirs touches
``child_process``.
Verified on Windows 10 22H2 / Node 22.21 / gitnexus 1.6.5 by
installing the locally-built tarball and running an active Claude
Code session against a large mixed-language repo — no console
window appears for any hook fire (pre-fix: ~2-3 visible flashes per
edit). No behavioural change on Linux/macOS hosts.
* test(hooks): regression — every hook spawnSync paired with windowsHide:true
Source-level assertion that every ``spawnSync`` invocation in the
hook layer has a matching ``windowsHide: true`` in its options
object. Without the flag, Node's child_process module asks
CreateProcess to use STARTF_USESHOWWINDOW with SW_SHOWDEFAULT and
a black console window flashes onto the user's desktop for the
duration of each call — see the parent fix commit.
The check is source-level rather than behavioural because:
* the flag's effect is observable only on Windows;
* GitHub Actions runs vitest on Linux for the hook tests;
* regressing this is easy (every new spawnSync site has to remember
to add the flag), and a runtime check on a Windows-only CI leg
would still let a PR land on the main branch first.
Counts spawnSync occurrences and windowsHide:true occurrences per
file (in code, ignoring comments) and asserts equality. Five files
covered:
* gitnexus/hooks/claude/gitnexus-hook.cjs
* gitnexus/hooks/claude/hook-db-lock-probe.cjs
* gitnexus-claude-plugin/hooks/gitnexus-hook.js
* gitnexus-claude-plugin/hooks/hook-db-lock-probe.cjs
* gitnexus-cursor-integration/hooks/gitnexus-hook.cjs
Adding a new hook file requires updating the HOOK_FILES tuple. A
sanity assertion ``spawnCount > 0`` catches accidental deletion of
all spawn calls in a future refactor (would otherwise silently make
the count-equality assertion trivially true).
Sits next to the existing "no shell: true" and ".cmd extension"
regression tests in test/unit/hooks.test.ts — same shape, same
spirit.
* fix(src): extend windowsHide:true to every spawn-family call in cli/core/mcp/server
Companion to the hook-layer fix in this branch's first commit. The
same Windows console-window flash bug applies to every
``spawn`` / ``spawnSync`` / ``execFile`` / ``execFileSync`` /
``execFileAsync`` / ``execSync`` call in the source tree — not just
the hooks. The MCP local backend
(``src/mcp/local/local-backend.ts``) and the ``gitnexus serve`` git
helpers (``src/server/git-clone.ts``) are particularly bad because
they run from daemonized processes that have no parent console; the
spawned child auto-allocates one and it pops onto the user's
desktop. The CLI sites are less visible (the user is at a terminal
with an existing console; ``stdio: 'inherit'`` shares it) but the
flag is harmless there — windowsHide only suppresses NEW console
allocation, an inherited parent console is untouched. The visible
output of ``gitnexus analyze`` and friends is preserved verbatim.
The pre-existing fix at ``src/core/lbug/extension-loader.ts:96``
established the convention in this codebase. This commit applies it
uniformly.
Sites covered (21 new):
| File | Sites |
|---|---|
| src/cli/analyze.ts | 1 |
| src/cli/setup.ts | 2 |
| src/cli/wiki.ts | 3 |
| src/core/embeddings/embedder.ts | 1 |
| src/core/git-staleness.ts | 3 |
| src/core/run-analyze.ts | 1 |
| src/core/wiki/cursor-client.ts | 2 |
| src/core/wiki/generator.ts | 3 |
| src/mcp/local/local-backend.ts | 2 |
| src/server/git-clone.ts | 2 |
| src/core/lbug/extension-loader.ts | (already had it, untouched) |
Combined with the 19 hook sites from the first commit + the 1
pre-existing extension-loader site, the codebase now has uniform
``windowsHide: true`` on every spawn-family call.
Behavioural notes:
* ``windowsHide`` is documented by Node as a no-op on POSIX —
Linux/macOS hosts see byte-identical behaviour.
* ``stdio: 'inherit'`` callers (e.g. ``cli/wiki.ts:522`` opens the
editor in the user's terminal) keep their interactive UX. The
child inherits the parent's stdio handles; no new console is
allocated; the flag has nothing to hide.
* Piped callers (``stdio: ['pipe',…]``) continue to deliver every
byte of stdout/stderr back to the parent for the parent to log
/ process / re-print. No output is swallowed.
* ``execSync`` / ``execFileSync`` callers that previously had no
``stdio`` option (e.g. ``generator.ts:887`` ``execSync('git
rev-parse HEAD', { cwd })``) keep their default pipe semantics
(``.toString()`` still works) — windowsHide is added alongside
the existing ``cwd`` option.
Verified on Windows 10 22H2 / Node 22.21 by installing the locally
built tarball and exercising:
* MCP detect_changes via the local backend → no flash.
* gitnexus serve → no flash on git clone/clone-pull.
* gitnexus analyze interactively → output appears in terminal as
before, no extra window.
* test(windowsHide): extend regression to every spawn-family call in src/
Companion to the src/ patch. The hooks.test.ts regression now
covers 16 files (5 hooks + 11 source files), and asserts the
invariant for every spawn-family function — not just spawnSync.
Changes:
* Generalise countSpawnCalls() to also count spawn, execFile,
execFileSync, execFileAsync, execSync (the entire spawn-family
surface of child_process). Skip method calls (e.g. RegExp.exec)
via a negative-lookbehind on ``.``.
* Add SRC_FILES table with all 11 source-tree files that import
spawn-family functions from child_process.
* Loop over [...HOOK_FILES, ...SRC_FILES] so a regression in any
file fails the same test name.
* Tighten the assertion to ``hideCount >= spawnCount`` rather
than strict equality, because some sites (e.g. setup.ts:534
using execFileAsync via shell:true on Windows) may legitimately
add windowsHide to nested option objects in future refactors.
* Sanity gate ``spawnCount > 0`` per file catches a refactor
that deletes all spawn calls (would otherwise make the
assertion trivially true).
Manually exercised against the patched repo:
16 files, 28 total spawn-family calls, 28 windowsHide:true.
All pass.
The convention to keep this list in sync: every new file in
gitnexus/src/ that imports from 'child_process' must be added to
the SRC_FILES tuple. The cost is one line per file; the benefit
is the next contributor never has to think about windowsHide
again — the test will catch a miss before merge.
* style: prettier --write on storage/git.ts + hooks.test.ts
CI quality / format job flagged two formatting issues in the
merge-resolution commit: a long single-line options object in
storage/git.ts and similar in hooks.test.ts. prettier --write
fixes both with the project's standard wrap-and-trailing-comma
style. No semantic change.
* test(git): include windowsHide in toHaveBeenCalledWith assertion
The merge-resolution commit added windowsHide:true to the
'git rev-parse --is-inside-work-tree' execSync call in
src/storage/git.ts, but the matching strict-shape assertion in
git.test.ts:31-34 still expected the pre-patch two-key options
object {cwd, stdio}. vitest's toHaveBeenCalledWith does a deep
structural match, so the extra third key flipped the assertion
to fail.
Add windowsHide: true to the expected shape. Only this one
assertion is strict; the two siblings ('passes the correct cwd'
and the no-cwd-arg case) use expect.objectContaining and
expect.any(String) and remain green without modification.
* test(setup-codex): include windowsHide in execFile shape assertions
Same root cause as the git.test.ts fix on this branch: the windowsHide
patch added windowsHide:true to the execFile() options in
src/cli/setup.ts, but three strict-shape toHaveBeenCalledWith
assertions in setup-codex.test.ts still expected the pre-patch
{shell:true} / {shell:false} two-key options. vitest does a deep
structural match, so the extra key flipped the assertions to fail
on every CI matrix leg (ubuntu coverage + macos + windows).
Adding windowsHide:true alongside the existing 'shell' key in
all three sites.
* ci: retrigger checks
go-parity failed on a flaky onnxruntime-node postinstall network timeout
(AggregateError [ETIMEDOUT] in node ./script/install), which cascaded into
the CI Gate. No code change — empty commit to re-run the pipeline.
* fix(test): strengthen windowsHide regression assertions (PR #1794 review)
- Replace toBeGreaterThanOrEqual with exact toBe per DoD §2.7
- Remove unused `m` variable in countSpawnCalls (CodeQL finding)
- Add windowsHide: true to runGit test helper for consistency
---------
Co-authored-by: Gergő Magyar <gergomagyar@icloud.com>
Co-authored-by: ManniX-ITA <35522085+ManniX-ITA@users.noreply.github.com>
Co-authored-by: Test <test@example.com>
|
||
|
Léon Simmons
|
a8a8a3710d
|
fix(lbug): skip init lock and filesystem mutations for read-only opens (#1783) (#1784)
Some checks are pending
CodeQL / Analyze (javascript-typescript) (push) Waiting to run
CodeQL / Analyze (python) (push) Waiting to run
Gitleaks / gitleaks (push) Waiting to run
Publish / Classify release event (push) Waiting to run
Publish / RC guard (marker + release-PR skip) (push) Blocked by required conditions
Publish / ci (push) Blocked by required conditions
Publish / Publish to npm (push) Blocked by required conditions
Publish / Build & Push RC Docker images (push) Blocked by required conditions
Scorecard / Scorecard analysis (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-cli) (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-web) (push) Waiting to run
`doInitLbug` unconditionally called `acquireInitLock`, which creates
`${dbPath}.init.lock` inside the workspace. On a Docker `:ro` bind
mount this fails with EROFS.
The init lock prevents a TOCTOU race during DB creation — read-only
opens never create databases and don't need it. Split the init path:
- Read-only: skip path cleanup, init lock, orphan sidecar removal,
and mkdir. Go straight to preflightLbugSidecars (allowQuarantine:
false) then openLbugConnection with readOnly: true.
- Writable: unchanged behavior (lock, cleanup, open).
- Shadow-replay recovery: catch EROFS/EACCES/EPERM from the writable
fallback in ensureReadOnlyConnectionUsable and surface an actionable
error instead of a raw filesystem exception.
Includes integration test verifying read-only open never creates
lbug.init.lock on disk.
Fixes #1783
Co-authored-by: Gergő Magyar <gergomagyar@icloud.com>
|
||
|
azizur100389
|
eb69f667ab
|
feat(cpp): Add structured resolver suppression outcomes (#1785) | ||
|
Minidoracat
|
2b6e7ffbd9
|
fix(php): avoid Blade templates entering PHP analysis (#1790) | ||
|
Hugo Gu
|
2c066d46a1
|
test(cli): stabilize eval-server host checks (#1786) | ||
|
dependabot[bot]
|
fb94dba484
|
chore(deps)(deps-dev): bump tsx from 4.22.0 to 4.22.3 in /gitnexus (#1789) | ||
|
Hugo Gu
|
7fc797e2ce
|
feat: Support DeepSeek V4 API (#1594)
Some checks are pending
CodeQL / Analyze (javascript-typescript) (push) Waiting to run
CodeQL / Analyze (python) (push) Waiting to run
Gitleaks / gitleaks (push) Waiting to run
Publish / Classify release event (push) Waiting to run
Publish / RC guard (marker + release-PR skip) (push) Blocked by required conditions
Publish / ci (push) Blocked by required conditions
Publish / Publish to npm (push) Blocked by required conditions
Publish / Build & Push RC Docker images (push) Blocked by required conditions
Scorecard / Scorecard analysis (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-cli) (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-web) (push) Waiting to run
|
||
|
Gergő Magyar
|
51e667808a
|
feat(lang-kotlin): flip Kotlin to MIGRATED_LANGUAGES + close #1756 / #1757 (refs #1746) (#1782) | ||
|
dependabot[bot]
|
84ac88a741
|
chore(deps)(deps): bump qs from 6.14.2 to 6.15.2 in /gitnexus (#1791) | ||
|
ChamHerry
|
fc6007e70b
|
feat(i18n): make web and CLI language-aware (#1748) | ||
|
Copilot
|
87b91c821e
|
fix(lbug): add WAL checkpoint-threshold control (#1772)
* Initial plan * fix(analyze): add WAL auto-checkpoint CLI control and default-off behavior * test(analyze): share lbug auto-checkpoint parsing and align validation * fix(analyze): always enable lbug auto-checkpoint and expose threshold control * refactor(lbug): inline always-on auto-checkpoint constructor arg * fix(analyze): guide checkpoint-threshold on Ladybug WAL checkpoint IO failures * test(analyze): cover checkpoint IO guidance and add integration guard * fix(analyze): tighten checkpoint IO detection and remove test hook * fix(analyze): remove checkpoint test hook and tighten error matching * fix(analyze): rename to wal-checkpoint-threshold, raise default, add manual checkpoint driver with retry Address review feedback on PR #1772: - Rename CLI flag, env var, AnalyzeOptions field, recovery-hint tag, and parser/constants from lbug-* to engine-neutral wal-* (matches the existing WAL_RECOVERY_SUGGESTION / isWalCorruptionError convention). - Raise default threshold from -1 (Ladybug stock ~16 MiB) to 64 MiB so users on the default config no longer hit the original rename/remove race. - Align both READMEs to publish 67108864 (64 MiB) instead of 65536 (which would have made the crash more frequent). - Add wal-checkpoint-driver.ts: a periodic manual CHECKPOINT driver wrapped in a 3-attempt jittered retry (50/200/500 ms), driven from runFullAnalysis. Opt-out via GITNEXUS_WAL_MANUAL_CHECKPOINT=0. Moves the race window into a JS-controllable retry surface while keeping native auto-checkpoint on. - Move LBUG_CHECKPOINT_RENAME_RE / REMOVE_RE plus the predicate (renamed to isLbugCheckpointIoError) into lbug-config.ts alongside isWalCorruptionError. Predicate is now exported. Add a permissive fallback matcher and pin the matched Ladybug version in comments. - Warn instead of silently defaulting when GITNEXUS_WAL_CHECKPOINT_THRESHOLD is set to a non-empty unparseable value (closes the CLI-vs-env asymmetry). - Add a typed RecoveryHint string-literal union in cli-message.ts so future hint tags can't drift. - Add a real integration test under test/integration/ that triggers a Ladybug checkpoint IO failure via a pre-existing directory at the rename target (portable across platforms; no test-only injection hook). - Add small-disk / CI caveat (32 MiB secondary suggestion) to the recovery hint and README env-var rows. - Document CLI/env precedence in the analyze --help block. - Help placeholder: <value> -> <bytes>. - Rename analyze-lbug-auto-checkpoint.test.ts to use the new wal-* token. * chore(lbug): remove dead jitteredDelay helper and apply prettier - Drop unused `jitteredDelay` function flagged by CodeQL in PR #1772; the retry loop already inlines the same calculation with the injectable `randomImpl` so the helper was dead. Move the non-cryptographic-by-design comment next to the actual jitter site. - Apply `prettier --write` to wal-checkpoint-driver.ts and the new integration test to absorb the PR autofix bot's formatting findings. --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> Co-authored-by: Test <test@example.com> |
||
|
azizur100389
|
952ada70c5
|
feat(cpp): Resolve overloaded operator calls (#1754)
* feat(cpp): resolve overloaded operator calls * fix(cpp): tighten overloaded operator resolution --------- Co-authored-by: Gergő Magyar <gergomagyar@icloud.com> |
||
|
Gergő Magyar
|
060fe75715
|
docs(lang-kotlin): refresh scope-resolver JSDoc after #1758-#1763 landed (#1781)
The scope-resolver header comment claimed forced-mode passed 154/175 (88%) and listed smart casts, cross-file iterables, method chains, overload selection, virtual dispatch, and interface defaults as "remaining gaps". All six landed in PRs #1774-#1779. Forced mode now passes 175/175 (verified post-merge against `main`). Update the header to: - state the current forced-mode result accurately, - enumerate the closed sub-issues so future readers can trace each capability back to its PR, - and explicitly name the remaining flip blockers (#1755, #1756, #1757) so the next maintainer to look at this file knows exactly what's required before adding `Kotlin` to `MIGRATED_LANGUAGES`. Docs-only — no behavioral changes. Refs #1746. Co-authored-by: Test <test@example.com> |
||
|
Gergő Magyar
|
d15f8bef54
|
feat(ingestion): log deferred resolution progress when verbose (#1741) (#1773)
* feat(ingestion): log deferred resolution progress when verbose Add [deferred-profile] timing logs for post-chunk import, heritage, heritage-map, and legacy call resolution. Enabled on GITNEXUS_VERBOSE / analyze -v (and optionally GITNEXUS_PROFILE_DEFERRED) to diagnose analyze stalls on large repos (issue #1741). Co-authored-by: Cursor <cursoragent@cursor.com> * chore(autofix): apply prettier + eslint fixes via /autofix command * fix(ingestion): address PR #1773 production-readiness review Move deferred call progress logs after the registry-primary skip so sites= counts match files actually resolved. Only time buildHeritageMap when heritage records exist; otherwise log an explicit skip. Add wiring tests that assert [deferred-profile] emission from buildHeritageMap and processCallsFromExtracted. Snapshot GITNEXUS_PROFILE_DEFERRED env vars in analyze CLI isolation. Co-authored-by: Cursor <cursoragent@cursor.com> * fix(ingestion): address PR #1773 code-review findings P0 - Replace forbidden toBeGreaterThanOrEqual/toBeLessThan in profileElapsedMs test with exact-arithmetic vi.spyOn(hrtime.bigint) asserting .toBe(2.5) and .toBe(0). DoD §2.7 compliance. P2 - Use Number() (not parseInt) when parsing GITNEXUS_PROFILE_DEFERRED_SLOW_MS so scientific notation like '1e9' doesn't silently parse to 1 and turn the slow-file log into a per-file log storm. - Introduce startTimer(enabled): bigint | null and endTimer(start, format) helpers in deferred-resolution-profile.ts; refactor 6+ timing blocks in parse-impl.ts and call-processor.ts to use them. Removes the 0n sentinel that conflated 'disabled' with 'zero elapsed time' and let TS narrow correctly. - Split the call-processor file counter: filesProcessed (all iterated) vs resolvedFiles (post registry-primary skip). Key the every-N progress log and the start-of-phase log on resolvedFiles so mixed Python+JVM repos where the skipped language sorts first still emit 'calls 1/1 file=...' on the first non-skipped file. Adds a wiring test for the mixed-language ordering case. P3 - Restore the original isDev '🔗 E1: Seeded ...' logger.info line so log scrapers keyed on the emoji marker still match; emit the [deferred-profile] variant only when deferredProfile && !isDev. - Move tFile = startTimer(profileCalls) below the registry-primary skip so skipped files don't trigger an hrtime.bigint() call. - Document GITNEXUS_PROFILE_DEFERRED and GITNEXUS_PROFILE_DEFERRED_SLOW_MS in the README env-var table. * refactor(ingestion): extract parseTruthyEnv to shared utils (U5) Three narrow-form env-var truthy checkers (verbose.ts, registry-primary-flag.ts, deferred-resolution-profile.ts) each had their own `'1' | 'true' | 'yes'` parser with subtle divergences (trim or no trim, set vs disjunction). Consolidate on a single `parseTruthyEnv(raw)` helper in utils/env.ts — the module already serves as the centralization point for shared ingestion env constants. logger.ts's broader `isTruthyEnv` (negative-list, pino-debug convention) stays untouched — different intent, different semantics. New table-driven test at test/unit/env.test.ts covers case variants, whitespace, and rejection of falsy / unknown tokens. * refactor(ingestion): named constants for deferred-profile log gates (U6) Replace magic literals 10 / 100 / 3_000 / 5_000 in deferred-resolution-profile.ts with module-private named constants LOG_EVERY_N_VERBOSE, LOG_EVERY_N_PROFILE, DEFAULT_SLOW_MS_VERBOSE, DEFAULT_SLOW_MS. Not exported — internal tuning knobs. Pure refactor; existing tests assert the exact values and still pass unchanged. * fix(ingestion): pre-pass denominator for deferred call progress (U1, A1) The live per-file denominator in processCallsFromExtracted previously read `totalFiles - skippedRegistryPrimaryFiles` at log time. On mixed Python+JVM repos where the skipped language interleaves with the resolved one, the denominator drifts upward as the loop iterates — files iterated before later skips have been seen carry an inflated denominator. The live ratio only self-corrects after the final file has been classified. Fix: one-pass pre-count over byFile.keys() before the work loop computes resolvedTotal once. The denominator is then stable from the first emission onward. The pre-pass runs only on the enabled path (profileCalls=true) so the disabled path keeps zero extra work. Adds a wiring test exercising the alternating [ts, py, ts, py, ...] order that triggered the drift, asserting every emitted line uses `/4` and no other denominator slips through. * fix(ingestion): E1 enrichment log emits on both dev and profile flags (U2, A2) The post-chunk E1 enrichment log used `if (isDev) {...} else if (deferredProfile) {...}` which is mutually exclusive. On combined runs (NODE_ENV=development + GITNEXUS_PROFILE_DEFERRED=1) the [deferred- profile] line was silently swallowed — operators grepping that prefix saw a gap between wildcard-synth and heritage timings, while the inline comment promised dual emission. Fix: two independent `if` statements so both branches fire when both flags are set. The original emoji-prefixed `🔗 E1: Seeded` line keeps its phrasing for any dev-mode log scrapers that depend on the marker. Pinning test (parse-impl-e1-emission-shape.test.ts) reads the source and asserts (a) both branches exist as standalone `if` statements and (b) the closing `}` of the isDev branch is followed by `if`, not `else if`. Source-shape pins are the right test scope for a purely structural change — the regression we are guarding against is exactly how a future reader greps for it. * feat(ingestion): unresolved-side counters in heritage-map profile (U7) The existing maxNameCartesian / ambiguousHeritageRecords counters in buildHeritageMap only observed records where BOTH the child and parent name lookups resolved. On JVM monorepos the actual pathological case is one side empty (typically an unresolved external supertype with many same-named children, or vice versa) — those records were silently dropped from the metric. Add `unresolvedChildLookups` and `unresolvedParentLookups` in a separate `if (profileHeritage)` block placed immediately after the two `lookupClassByName` calls (so it observes the unresolved cases the length-guarded ambiguity block below cannot see). Both counters reuse the existing childDefs / parentDefs values — no additional lookups. Done-summary log extended to include the two new counters. Wiring test covers both directions (unresolved parent, unresolved child) plus the existing "both resolved" baseline now asserts the new counters report zero for that case. * fix(ingestion): endTimer formatter exception safety (U3) Wrap the format callback in endTimer in a try/catch so a throwing formatter (custom toString, JSON.stringify on a circular object, future heavier serializers) cannot abort the deferred resolution band. Observability code must never escalate to a load-bearing failure mode. On catch we emit a single `[deferred-profile] formatter error: …` line via logDeferredProfile and return; the caller's stage continues as if profiling had no-op'd for this timer. DoD §2.8 is satisfied — the failure is surfaced, not silently swallowed. Tests cover the four cases: happy path emits the formatted line, null start no-ops without invoking the formatter, throwing formatter is caught and surfaces one error line, non-Error throws are coerced via String() in the message. * fix(ingestion): defensive wrap + dropped-line counter for logDeferredProfile (U4) Wrap logger.info inside logDeferredProfile in a try/catch so a throwing underlying logger cannot abort the deferred resolution band. Pino with sync:false (the current SonicBoom destination) does not throw synchronously for `info(string)` calls, but first-use construction paths (pino-pretty resolve, level validation) and any future transport reconfiguration could. The wrap is belt-and-suspenders coverage; the counter makes silent failures visible. A module-private droppedLogLines counter accumulates dropped lines. Two helpers — getDeferredProfileDroppedCount() and resetDeferredProfileDroppedCount() — expose the counter. The handler deliberately does NOT call the failing logger; that would risk an infinite loop if the failure is steady-state. processCallsFromExtracted resets the counter at entry (so each analyze run gets a fresh count rather than accumulating across the process lifetime — relevant for the MCP server, eval harness, integration tests), and surfaces the count in the done-summary as `note: N profile log lines dropped (logger errors)` when greater than zero. DoD §2.8 (no silent diagnostic catches) is satisfied. Tests cover the helper API (zero at entry, idempotent reset) and the happy path; the catch arm is pinned via source-shape assertion since the logger Proxy can't be vi.spyOn'd directly (lazy `get` trap, no own-property to wrap). * docs(readme): clarify GITNEXUS_PROFILE_DEFERRED_SLOW_MS coercion (U8) The env-var row mentioned integer / scientific notation only, but the underlying parser (`Number(raw)` since the U2 fix in PR #1773) also accepts decimals like `.5` and hex like `0x10`. Document the actual acceptance set plus the non-finite / non-positive fallback so operators setting unusual values know what to expect. --------- Co-authored-by: Test <test@example.com> Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> |
||
|
Gergő Magyar
|
f72d9a99c6
|
fix(lang-kotlin): virtual dispatch via constructor type override (#1762) (#1778)
Closes #1762. `val animal: Animal = Dog(); animal.speak()` resolved to `Animal.speak` (or no edge to Dog at all) under `REGISTRY_PRIMARY_KOTLIN=1` because the Kotlin scope query emits BOTH an annotation type-binding (`animal -> Animal`) and a constructor- inferred type-binding (`animal -> Dog`). The generic scope-extractor ranks annotation sources higher than constructor-inferred sources (see `typeBindingStrength` in scope-extractor.ts), so the annotation always won and `animal.speak()` dispatched against the static type. Kotlin's virtual dispatch semantics expect the dynamic type — the overriding `Dog.speak` should win when the RHS is a constructor call, because that's what runs at runtime. Fix: in `emitKotlinScopeCaptures`, suppress the `@type-binding. annotation` capture when the underlying `property_declaration` has a `call_expression` value sibling. The constructor-inferred capture remains, becomes the sole binding for the variable, and receiver-bound resolution dispatches against the constructed class (and walks its MRO). This is intentionally Kotlin-specific — flipping precedence globally would change behavior for other languages whose static-type annotations are still the right binding when present. Kotlin is the language where the constructor RHS is the dispatch target by design. Verification (REGISTRY_PRIMARY_KOTLIN=1): - Forced-mode: 21 -> 20 failing of 175 (1 fewer; test 1715 in `test/integration/resolvers/kotlin.test.ts` now green). - Default-mode Kotlin: 175/175 unchanged. - Full resolver suite: 2216/2216 unchanged. - Remaining 20 failures are tracked by sibling sub-issues (#1758, #1759, #1760, #1761, #1763). Does NOT add Kotlin to MIGRATED_LANGUAGES per parent #1746 flip criteria. Closes #1762. Refs #1746. Co-authored-by: Test <test@example.com> |
||
|
Gergő Magyar
|
64efc202f6
|
fix(lang-kotlin): method-chain fixpoint receiver types (#1760) (#1776)
Closes #1760. Multi-step intra-file chains like val user = getUser() val addr = user.address val city = addr.getCity() city.save() produced no `CALLS` edge for `city.save()` because the Kotlin extractor only inferred property types for `simple_identifier` values (`val x = y`) and call expressions with simple-identifier callees (`val x = fn()`). Navigation expressions (`val addr = user.address`) and call expressions with navigation-expression callees (`val city = addr.getCity()`) returned null, leaving `addr` and `city` unbound — the chain broke two hops before `city.save()`. Implementation: - `collectKotlinClassMembers(rootNode)` indexes per-file class fields (primary-constructor `val`/`var` params + body property declarations) and method return types. Per-file scope matches the existing extractor design. - `inferKotlinPropertyType` gains two new cases: 1. `navigation_expression` value — receiver type via `localTypes`, field type via `classMembers.fields`. 2. `call_expression` with `navigation_expression` callee — receiver type via `localTypes`, method return type via `classMembers.methods`. Both return null when any link is unknown (safe / over-conservative). Verification (REGISTRY_PRIMARY_KOTLIN=1): - Forced-mode: 21 -> 20 failing of 175 (1 fewer; test 1491 in `test/integration/resolvers/kotlin.test.ts` now green). - Default-mode Kotlin: 175/175 unchanged. - Full resolver suite: 2216/2216 unchanged. - Remaining 20 failures are tracked by sibling sub-issues (#1758, #1759, #1761, #1762, #1763). Does NOT add Kotlin to MIGRATED_LANGUAGES per parent #1746 flip criteria. Closes #1760. Refs #1746. Co-authored-by: Test <test@example.com> |
||
|
Gergő Magyar
|
67cc4c6d94
|
fix(lang-kotlin): cross-file iterable return propagation (#1759) (#1775)
Two related bugs surfaced in REGISTRY_PRIMARY_KOTLIN=1 forced mode: 1. `import models.getRepo` silently resolved to `models/User.kt` (the first `.kt` file inside `models/` by iteration order) when no file was named after the symbol. `findKotlinFile` returned a single directory child as a fallback, so the importer's module-scope mirror only ever picked up the first arbitrary candidate — `getUser → User` landed but `getRepo → Repo` never did, and downstream `repo.save()` resolution fell through to no edge. 2. `for (x in importedCallable())` produced no for-loop type binding when the callee's return type lived in another file, because `inferKotlinIterableElementType`'s call-expression arm consulted only the local file's `returnTypes` map. Fix: - Split `findKotlinFile` into `findKotlinExactOrSuffix` (exact / suffix match only) and `findKotlinDirectoryChild` (legacy single-child fallback). Add `findKotlinPackageFiles` returning every `.kt`/`.kts` file inside a package directory. The resolver now fans out the stripped path through `findKotlinExactOrSuffix → findKotlinPackageFiles`, returning a `readonly string[]` candidate set. The finalize pass walks each candidate and picks the one whose `localDefs` actually export the imported name — exactly the multi-target contract `FinalizeHooks.resolveImportTarget` already supports. - `inferKotlinIterableElementType` for `call_expression` now falls back to the callee's identifier text when the local return-type map has no entry. `propagateImportedReturnTypes` chain-follows `loopvar → callee → ElementType` once the imported `callee → Element` mirror lands at module scope (which now works thanks to fix #1). Verification (REGISTRY_PRIMARY_KOTLIN=1): - Forced-mode: 21 -> 18 failing of 175 (3 fewer; tests 487, 1242, 1251 in test/integration/resolvers/kotlin.test.ts now green). - Default-mode Kotlin: 175/175 unchanged. - Full resolver suite: 2216/2216 unchanged (incl. `kotlin-calls` `util.OneArg.writeAudit` regression check at line 176). - Remaining 18 failures are tracked by sibling sub-issues (#1758, #1760, #1761, #1762, #1763). Does NOT add Kotlin to MIGRATED_LANGUAGES per parent #1746 flip criteria. Closes #1759. Refs #1746. Co-authored-by: Test <test@example.com> |
||
|
Gergő Magyar
|
a3e7dfa8a6
|
fix(lang-kotlin): interface default method dispatch via implements-split MRO (#1763) (#1779)
Closes #1763. `user.validate()` on `class User : Validator` resolved to no edge under REGISTRY_PRIMARY_KOTLIN=1 when validate() was a default method declared on the Validator interface: class User(val name: String) : Validator interface Validator { fun validate(): Boolean = true } fun run() { val user = User("alice"); user.validate() } The generic `buildMro` walks EXTENDS edges only. Kotlin classes implement interfaces via IMPLEMENTS edges (per the parsing-processor), so the implementor's MRO never picked up the interface's default methods — `findOwnedMember(User, validate)` returned undefined and no fallback walked to Validator. Fix: replace `defaultLinearize` with a Kotlin-specific MRO builder modeled after PHP's `buildPhpMro` (trait composition): 1. Run the generic `buildMro` (EXTENDS-only). 2. Collect direct IMPLEMENTS edges as class -> interface[] map. 3. For each class, walk its EXTENDS-MRO ancestors AND its own IMPLEMENTS edges to seed interface candidates, then BFS-close to pick up transitive interface inheritance (interface A : B). 4. Append the interface closure to the class's MRO (after the EXTENDS chain — Kotlin requires explicit override on conflict, so this ordering is a safe approximation for method lookup). 5. Classes with no EXTENDS but with IMPLEMENTS edges (the #1763 fixture shape) get their MRO seeded directly from their interfaces. Verification (REGISTRY_PRIMARY_KOTLIN=1): - Forced-mode: 21 -> 20 failing of 175 (1 fewer; test 2062 in `test/integration/resolvers/kotlin.test.ts` now green). - Default-mode Kotlin: 175/175 unchanged. - Full resolver suite: 2216/2216 unchanged. - Remaining 20 failures are tracked by sibling sub-issues (#1758, #1759, #1760, #1761, #1762). Does NOT add Kotlin to MIGRATED_LANGUAGES per parent #1746 flip criteria. Closes #1763. Refs #1746. Co-authored-by: Test <test@example.com> |