GitNexus/gitnexus/scripts/cross-platform-tests.ts
Gergő Magyar 01af833875
fix(scope): preserve lexical bindings in HTTP wrapper detection (#3632)
* docs(scope): clarify lexical binding ownership and wrapper provenance

Checkpoint for #3630. Publishing the implemented source, regression tests, and benchmark harness as a draft PR; full-suite and final scaling validation remain in progress.

* fix(scope): publish #3630 docs/plans

* fix(scope): publish #3630 gitnexus/bench/lexical-binding-provenance

* fix(scope): publish #3630 gitnexus/bench/lexical-binding-provenance/fixtures

* fix(scope): publish #3630 gitnexus/bench/lexical-binding-provenance/results

* fix(scope): publish #3630 gitnexus/src/core/ingestion

* fix(scope): publish #3630 gitnexus/src/core/ingestion/languages/typescript

* fix(scope): publish #3630 gitnexus/src/core/ingestion/languages

* fix(scope): publish #3630 gitnexus/src/core/ingestion/pipeline-phases

* fix(scope): publish #3630 gitnexus/src/core/ingestion/route-extractors

* fix(scope): publish #3630 gitnexus/src/core/ingestion/workers

* fix(scope): publish #3630 gitnexus/test/fixtures/lang-resolution/fetch-wrapper-target-identity/lib

* fix(scope): publish #3630 gitnexus/test/fixtures/lang-resolution/fetch-wrapper-target-identity

* fix(scope): publish #3630 gitnexus/test/fixtures/lang-resolution/fetch-wrapper-target-identity/packages/http

* fix(scope): publish #3630 gitnexus/test/fixtures/lang-resolution/fetch-wrapper-target-identity/packages/http/src

* fix(scope): publish #3630 gitnexus/test/fixtures/lang-resolution/fetch-wrapper-target-identity/server

* fix(scope): publish #3630 gitnexus/test/fixtures/lang-resolution/fetch-wrapper-target-identity/web

* fix(scope): publish #3630 gitnexus/test/integration/resolvers

* fix(scope): publish #3630 gitnexus/test/unit

* fix(scope): publish #3630 gitnexus/test/unit/group

* fix(scope): publish #3630 gitnexus/test/unit/scope-resolution

* fix(scope): publish #3630 gitnexus/test/unit/scope-resolution/typescript

* fix(scope): publish #3630 gitnexus-shared/src/scope-resolution

* fix(scope): treat wrapped assignment targets as writes

`(app) = new Map()`, `(app as T) = x` and `app! = x` assign the same binding as
`app = x`, but route-fact invalidation and the shared pattern walker stopped at
the wrapper, so Express provenance survived the write and produced a false
route. Recurse through parenthesized, as/satisfies and non-null targets.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(routes): resolve wrappers through overloads, declarator lists and barrels

Three wrapper shapes lost their identity:

- Bodyless overload signatures were indexed as separate definitions sharing
  the implementation's binding, so the binding looked ambiguous and both local
  and imported calls were rejected. Signatures now declare no callable.
- Facts for `const base = 0, request = (url) => fetch(url)` followed the first
  declarator. The captured name now selects its own declarator.
- An import resolved to a workspace barrel stopped at index.ts. Workers now
  record named, star and imported-then-exported re-exports, and resolution
  follows them to the declaring module: local and named exports shadow
  `export *`, star exports never carry `default`, and conflicting star exports
  block the name. No repository-wide name fallback is restored.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* test(scope): assert closure ownership against the declaration

The closure-write assertion compared one lookup with itself, which also passes
when both are undefined. Compare it with the outer declaration and the shadowing
parameter, and keep the memoization check explicit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* refactor(routes): share lexical helpers and release per-file HTTP contexts

- Share binding keys, parameter patterns and transparent expression wrappers
  from lexical-bindings instead of four local copies; drop the unused
  moduleStem export; name the inferred-source predicates in lookup.
- Key the per-file HTTP context by its tree rather than its ParsedFile.
  Workers retain ParsedFiles for the whole chunk, which kept every file's AST
  and traversal maps alive until the chunk flushed.
- Index re-exports by name, memoize barrel walks per starting export, and keep
  facts without import targets as-is instead of copying them.
- Move TS-only wrapper cases into the TS describe.each instead of branching
  inside the per-language loop.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(routes): keep wrapper identity for HOCs, CommonJS, object keys and fetch packages

Four definition shapes that main linked by name lost their wrapper identity:

- `const load = useCallback((url) => fetch(url))` and curried HOCs produced no
  facts because definition lookup stopped at the call; it now follows the
  callable argument, as the structural definition capture does.
- `module.exports = { request }`, `exports.x = …` and `module.exports.x = …`
  were not exports, so destructured `require` callers missed the wrapper.
  CommonJS export assignments are recorded when `module`/`exports` are the
  unshadowed globals, including functions defined inline in the export object.
- An object-literal key or method name was resolved as a lexical binding and
  collided with a same-named module wrapper. Only binding identifiers now name
  a definition's binding.
- `import fetch from 'node-fetch'` no longer counted as fetch. Imports of known
  fetch implementations (node-fetch, cross-fetch, isomorphic-fetch, the fetch
  member of undici and whatwg-fetch) are fetch evidence again; parameters and
  other local names still are not.

Reference identifiers are also resolved only when their name is bound to a
callable or class region; other identifiers cannot add a reachability edge.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* test(routes): drive the forwarding fallback case from data and cover re-export merges

Replace the boolean-switched forwarding fallback case with table rows instead
of branching inside the test, and assert that barrel re-export facts survive
worker sub-batch merges.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): reject stale or dirty call-graph exports

The committed comparison named a candidate revision that is not on this branch,
and nothing stopped it. Comparison now requires both exports to come from clean
checkouts and the candidate export to come from the checkout running the
comparison.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): regenerate the call-graph comparison on the reviewed head

Clean baseline 565e50276 and clean candidate 8a5f7b738 on the frozen 432-file
corpus (fingerprint 88be1faf...): exactly one CALLS target changes, the
annotated closure probe `declaredOwner.writeAndRead` now calls `Base.m`
instead of `Foo.m`. The other 431 CALLS and all 11 METHOD_OVERRIDES edges are
unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): emit export provenance and review notes from the comparison

The comparison report carried export provenance and the expected-change
explanation only as hand-added fields, which a regeneration dropped. Emit
both exports' revisions, dirty flags and fingerprints, the unchanged CALLS
and METHOD_OVERRIDES counts, and an optional `--note` explanation directly.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): record calibrated scaling reports and the regenerated comparison

Duration-calibrated-v2 runs (5 warmups, 15 measured batches, 500 ms pilot
target) pinned to the same CPU, baseline then candidate back to back:

- Baseline 565e50276 exceeds the 1.8 normalized ceiling on six arms at 4x
  input: position-index 7.92, position-deep-completed 4.65, nested-helpers
  4.82, nested-helper-chain 14.05, reverse-chain 4.31, name-collisions 2.61
  (medians).
- Candidate 94f1a055b passes --check: every arm stays at or below 1.32 at 4x
  for both minimum and median, with correct semantic guards on every arm.

The comparison is regenerated from clean exports with its provenance and
review note emitted by the tool: one expected CALLS target change, 431 CALLS
and 11 METHOD_OVERRIDES unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* refactor(scope): replace recursion with explicit stacks in new traversals

The binding work added five recursive traversals whose depth follows input
shape: region summaries (callable nesting), barrel resolution (re-export
chain length), pattern binding names (destructuring depth), route-fact write
invalidation (pattern depth), and the position index's max-end tree. Each
now uses an explicit stack with the same visiting order and results.

Regression cases cover the inputs the recursive versions could not handle:
a 20,000-deep destructuring pattern and a 50,000-module re-export chain both
threw RangeError before this change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* perf(routes): avoid a full-tree script walk when building lexical bindings

Route-fact extraction ran 1.78x slower than main on a 512-function file
(87 ms vs 49 ms) because createLexicalBindings walked every AST node in script
to detect collapsed parameter/body environments, which only functions with
parameter defaults or computed keys can have.

- Find functions with the native type search, skip plain parameter lists
  without any subtree search, and walk only the formals and bodies of the
  rest. Detection is unchanged, except that functions inside syntax-error
  regions are now also checked, which can only add fail-closed ambiguity.
- Read each node's id and start position once per lookup instead of
  repeatedly crossing into native getters.

The same extraction now takes about 56 ms against 51 ms on main.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* perf(routes): walk wrapper evidence with a tree cursor

The wrapper context visited every node through namedChildren, which
materializes each child as a SyntaxNode and records a parent edge by node
id. A tree cursor reads node types and offsets natively, so only the node
types the handlers inspect become SyntaxNodes. Declared identifiers are
keyed by start offset, and the walk records the ancestor facts later
queries need (stored binding, default export, declarator of a name)
instead of a parent map.

On the 512-wrapper scaling input, context construction drops from about
58 ms to 48 ms.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): regenerate the call-graph comparison on the cursor walk

The candidate export at 6061ca71b has the same fingerprint as the previous
candidate: the traversal rewrite changes no edge. The one expected target
change and the 431 unchanged CALLS and 11 METHOD_OVERRIDES remain.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): record final scaling reports on the cursor walk

Baseline 565e50276 and candidate 61de364ff ran back to back on CPU 23
with the duration-calibrated-v2 protocol. Every candidate check stays at
or below a 1.32 normalized factor against the 1.8 ceiling; the baseline
fails six of nine arms. Workload fingerprints agree except the documented
name-collision false positive in the baseline.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs(plan): checkpoint the final review, scaling and suite results for #3630

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(routes): treat export aliases of one wrapper as one star-export target

Star-export candidates were deduplicated by the resolved export name, so
`export { request, request as send }` reached through two barrels counted
as two targets and blocked the import, dropping the consumer's FETCHES
edge. Candidates are now keyed by the declaration each export name names,
matching ECMAScript ResolveExport's binding comparison; distinct
declarations exported under the same name stay ambiguous.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): regenerate scaling and call-graph reports on the alias fix

Baseline 565e50276 and candidate 4d62bd045 ran back to back on CPU 23.
Every candidate check stays at or below 1.22 against the 1.8 ceiling;
the baseline still fails six of nine arms. The CALLS/METHOD_OVERRIDES
export is unchanged (fingerprint 4db24f57), so the comparison only
records the new candidate revision.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs(plan): record the export-alias fix and the regenerated reports for #3630

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(scope): keep the type a write gives an undeclared name

A constructor write to a name no scope declares (`client = new ApiClient()`
in sloppy JavaScript) declares nothing, so the lookup stays absent. The
carried constructor fact was dropped on that exit, which lost the receiver
type main had inferred for later reads in the writing scope. The absent
result now carries it, and findReceiverTypeBinding uses it before the
namespace fallbacks. Only complete-claim providers carry such a fact.

Tests pin the carried type per inferred source, the undeclared write, and
preservePosition across a module redirect.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(routes): close the wrapper identity gaps found in review

- CommonJS wrappers captured as their assignment (`exports.x = function`,
  `module.exports.x = () =>`) produce facts again: findFunction follows the
  assignment's right side.
- Name-only fallback considers only definitions that can be the target: an
  unresolvable import names a module-level export, an undeclared global a
  top-level function. Nested and unexported helpers no longer erase a
  wrapper's edges.
- Only possible wrappers and module-level functions are emitted, and the
  routes phase builds the index only when routes and candidate calls exist.
- A rebound binding (later assignment, `var` initializer over a function
  declaration, reassigned export) proves no wrapper across files.
- `(<T>x) = ...` is a write; transparent wrappers skip type arguments.
- Calls inside recovered syntax are still collected; instance-field
  initializers resolve after the module body, so a later declaration is
  visible.
- Local exports are always recorded, so a star export conflicting with a
  non-callable export is ambiguous, and fully resolved barrel frames are
  shared across walks.
- HTTP hooks run per file under a guard; a failing file or one without a
  ParsedFile still records its function names so fallback declines.
- namesFetch is iterative, large patterns no longer spread into push, the
  facts cache is keyed by the name node, and the provider context cache
  checks its ParsedFile.

Tests cover each case; the behavior rows fail on the previous source.
The TS-only call cases no longer branch inside the test file, call targets
are asserted exactly, the function-type list is drift-checked, and the
persistence test joins the cross-platform worker list.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs(routes): describe wrapper fallback, barrel and recall boundaries

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): count the new CommonJS fixture's dropped write site

`lib/cjs-member.js` in the fetch-wrapper target-identity fixture writes
`module.exports.arrowRequest`, a member write whose receiver is not typed.
That adds one `receiver-unresolved` write (190 -> 191 all-kind drops,
19 -> 20 writes). Call drops and shape-arm states are unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): regenerate scaling and call-graph reports on the review fixes

Baseline 565e50276 and candidate 528ea1478 ran back to back on CPU 23.
Every candidate check stays at or below 1.27 against the 1.8 ceiling; the
baseline still fails six of nine arms. The CALLS/METHOD_OVERRIDES export
is unchanged (fingerprint 4db24f57), so the comparison only records the
new candidate revision.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs(plan): checkpoint the review fixes and regenerated reports for #3630

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* refactor(routes): share CommonJS checks and trim per-node work

- `isCommonJsGlobal`/`isModuleExports` in lexical-bindings replace the two
  copies in fetch-wrapper and framework-route-facts (both resolve the same
  scope from the same start position).
- LEXICAL_FUNCTION_TYPES is derived from the TypeScript FUNCTION_NODE_TYPES
  instead of a third hand-kept copy; the drift test still pins it.
- `unwrap` reads UNWRAPPED_TYPES, the set the cursor walk mirrors, and the
  class-body region is no longer shadowed under the name `deferred`.
- framework-route-facts reads each node's type once per loop iteration.
- The position index returns the latest-starting scope when it already
  contains the point, before the max-end tree search; that entry is the
  search's rightmost match.
- carried() returns a shared empty set for argument forms that cannot carry
  a value, the resolution file set is built without intermediate arrays,
  and the routes phase reads the call list once.

No behavior change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): regenerate scaling and call-graph reports on the refactor

Baseline 565e50276 and candidate 3db8657ba ran back to back on CPU 23.
Every candidate check stays at or below 1.32 against the 1.8 ceiling; the
baseline still fails six of nine arms. The CALLS/METHOD_OVERRIDES export
is unchanged (fingerprint 4db24f57). The later main merge (e2a92dab8)
touches no measured file.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs(plan): checkpoint the simplification pass and its reports for #3630

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Address PR review feedback (#3632)

- Destructuring defaults (`{ fetch = globalThis.fetch } = {}`, renamed,
  array and local forms) are alias evidence for their leaf binding and
  no longer count a second initialization, which also invalidated
  bindings like `[send = other] = values`.
- `export default (request)` records the default export: the value is
  unwrapped before the identifier check.
- The HTTP fallback records and the collapsed-environment test row now
  say what they stand for.

Regression rows cover each form in JS, TS and TSX and fail on the
previous source; two fixture consumers cover imported wrappers.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore(bench): refresh lexical binding reports after review fixes

Regenerate clean baseline and candidate measurements at c263c4f72.
All candidate scaling checks pass, the semantic delta remains one CALLS
target, and receiver-resolution counts match the existing baseline.

* docs(plan): checkpoint refreshed review-fix measurements

Record c263c4f72 review fixes, the regenerated scaling and semantic
results, and the passing receiver and CI checks.

---------

Co-authored-by: Gergo Magyar <gergomagyar0@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-20 11:36:06 +01:00

441 lines
25 KiB
TypeScript

/**
* Cross-platform test subset runner.
*
* Runs only the tests that exercise platform-sensitive behavior on
* Windows and macOS. The full suite runs on Ubuntu; this narrows the
* cross-platform matrix to tests that actually vary across OSes.
*
* Categories included:
* - Platform-specific logic (path.sep, process.platform guards)
* - Native addon loading (LadybugDB, tree-sitter)
* - Process spawning and shell behavior
* - Filesystem locking and temp-dir behavior
* - Worker threads (real, not mocked)
* - CLI end-to-end tests
*
* When adding a new test that uses platform-varying APIs (native addons,
* child_process with real spawning, filesystem locking, path.sep), add
* it to the appropriate section below.
*
* Usage:
* npx vitest run $(npx tsx scripts/cross-platform-tests.ts)
* # or via the package script:
* npm run test:cross-platform
*/
// Platform-specific logic tests — contain explicit process.platform guards
// or test behavior that differs across operating systems
const PLATFORM_LOGIC = [
'test/unit/setup.test.ts',
'test/unit/setup-jsonc.test.ts',
'test/unit/setup-codex.test.ts',
'test/unit/setup-antigravity.test.ts',
'test/integration/setup-uninstall-roundtrip.test.ts',
'test/unit/resolve-invocation.test.ts',
// CLI-spawn entry-point resolution; its path-separator assertion (cli[/\\]index)
// must exercise the Windows backslash branch, so run it on the OS matrix (#2394).
'test/unit/cli-entry.test.ts',
'test/unit/platform-capabilities.test.ts',
// The tsconfig loader rebases `paths` targets through `path.resolve`, so the
// wildcard suffix it must recognise is `/*` on POSIX and `\*` on Windows. It
// only looked for `/*`, and every alias target came back as `src*` on
// Windows while the Ubuntu run stayed green — so this file has to run where
// the separator differs.
'test/unit/tsconfig-index.test.ts',
// The unit half of the same rebasing rule. Fixture-free and pathApi-injectable
// (every separator assertion passes an explicit `path.win32` / `path.posix`),
// so unlike the fixture suite above it fails on EVERY runner when the
// normalisation is removed rather than only on windows-latest. Registered
// beside its fixture sibling so the two halves stay discoverable as one group.
'test/unit/tsconfig-rebase-target.test.ts',
// The gitnexus-plan safe writer resolves every name through a per-platform
// backend: Linux anchors through /proc/self/fd, macOS resolves lexically and
// verifies each step against descriptors it holds open. Publication is link(2)
// on both. #2905 shipped the Darwin backend after the suite had silently
// skipped on every non-Linux runner, so this file must run on the OS matrix or
// the macOS half is unverified by construction — and the flag, trailing-
// separator and hard-link fixtures assert kernel behaviour that only a real
// Darwin kernel can confirm. Windows is refused by the capability gate; the
// suite asserts that refusal rather than skipping it.
'test/unit/evidence-provenance-helper.test.ts',
// Windows drive-letter case variance in the analyzer runner-identity path
// fields (#2668): normalizeAnalyzerRootPath is a POSIX no-op, so the
// "identity path fields are normalizer-stable" fixpoint guard only bites on
// the windows-latest matrix — it must run there, not just in the Ubuntu
// full-suite where it's trivially green. Deliberately the split-out
// normalization file, NOT analyzer-identity.test.ts: the latter's fixture
// tests compare identity fields against raw temp-dir paths and fail on macOS,
// where /var/... realpaths to /private/var/....
'test/unit/analyzer-identity-path-normalization.test.ts',
// `isInside` containment guard vs Windows cross-drive paths: path.relative
// returns the absolute target across drives, so the guard needs isAbsolute.
// Fixture-free and pathApi-injectable, so it is portable to every runner.
'test/unit/analyzer-identity-is-inside.test.ts',
// `\\?\` extended-length prefix normalization (#2667): fixture-free and
// platform-injectable (every assertion passes an explicit 'win32'), so like the
// is-inside guard above it is portable to every runner and its assertions run
// identically here and on Ubuntu. Registered alongside its two siblings so the
// Windows path-handling guards stay discoverable as one group. Same
// mixed-prefix relativize hazard as is-inside, reached through a
// caller-supplied path.
'test/unit/windows-long-path-prefix.test.ts',
// getconf page-size probe: explicit process.platform gate (win32 short-circuit)
// plus a live-probe test whose only real non-4K coverage is macos-arm64's
// 16 KiB pages — the exact hardware class #1231 targets (#2424 review).
'test/unit/lbug-config-pagesize.test.ts',
'test/unit/worker-pool-windows-quarantine.test.ts',
'test/unit/lbug-pool-fts-load.test.ts',
// U7 arm B: Windows FTS names a vendor-neutral OpenSSL/VC++ prerequisite
// and must never recommend borrowing Git for Windows DLLs. The file's
// assertions are unconditional so a skip-only suite cannot stay green.
'test/integration/fts-windows-dependency.test.ts',
// Remedy-text suites: discoverability only. They pass explicit platform
// strings into pure classifiers and drive mocked rejections with hardcoded
// literals, so they assert identically on every runner. Registering them
// here does not claim Windows-specific behavioral coverage.
'test/unit/extension-load-error.test.ts',
'test/unit/fts-degraded-warning.test.ts',
// Vendored-root symlink containment uses realpathSync + path.relative; a
// prefix-only leak follows a Windows junction / POSIX symlink out of
// vendor/. Ubuntu-only would leave that guard unverified on the OS matrix.
'test/unit/lbug-extension-loader.test.ts',
// Global registry writes use the platform-specific index-lock backend
// (Windows named pipe, Linux socket, or macOS file lock). This includes the
// overlapping-registration regression from #2716 on every OS matrix.
'test/unit/repo-manager.test.ts',
'test/unit/repo-manager-finalize-invariant.test.ts',
'test/unit/git-utils.test.ts',
'test/unit/hooks.test.ts',
'test/unit/hook-db-lock-probe.test.ts',
'test/unit/cursor-hook.test.ts',
'test/unit/factory-plugin.test.ts',
'test/unit/sidecar-recovery.test.ts',
'test/unit/pool-wal-recovery.test.ts',
'test/unit/lbug-adapter-wal-schema.test.ts',
'test/unit/detect-changes-worktree.test.ts',
'test/unit/eval-server-bind-restriction.test.ts',
'test/unit/ignore-service.test.ts',
'test/unit/group/bridge-db.test.ts',
'test/unit/group/bridge-db-edge.test.ts',
'test/unit/group/fs-utils.test.ts',
'test/unit/onnxruntime-node-resolver.test.ts',
// Windows cmd.exe arg-quoting + compose-and-spawn for the npm install (#2372):
// the quoting rules and win32 single-string spawn shape are OS-sensitive, so
// exercise them on real windows-latest. The spawn-shape/path tests force their
// platform branch and derive expected paths via the real fns, so they pass on
// any host (see the platform stubs + resolve() in the test file).
'test/unit/embedding-runtime-install.test.ts',
// Real-spawn arg-delivery round-trip: proves the install spawn delivers args
// to the child intact on each platform — win32 via the cmd.exe -> .cmd %* ->
// node chain (real cmd.exe, not just our model), macos/linux via the no-shell
// array form. Runs on every platform (the ubuntu suite covers Linux; this
// registration adds windows + macos).
'test/unit/embedding-install-arg-delivery.test.ts',
// Structural FTS-extension classifier against REAL binaries (#2374): on this
// matrix `process.execPath` / `lbugjs.node` are a real PE (windows) and Mach-O
// (macos), so the header parsing is proven on genuine binaries, not synthetic
// buffers (the ubuntu suite covers the ELF path).
'test/integration/extension-binary-real.test.ts',
// Server repo resolver branches on path shape (path.isAbsolute, backslash
// detection) and canonicalizePath/realpathSync, all of which differ between
// POSIX and Windows — the fail-closed path-claim semantics must hold on the
// real windows-latest path implementation (#2419/#2420).
'test/unit/server-api-repo-resolution.test.ts',
// #3073: cwd-based repository selection canonicalizes real paths, compares
// platform separators/case, and rejects nested Git-boundary fallthrough.
'test/unit/calltool-dispatch.test.ts',
// Preview capabilities use real file symlinks, canonical temp paths, atomic
// replacements and physical-target drift checks on each host filesystem.
'test/unit/rename-preview-plans.test.ts',
'test/unit/rename-plan.test.ts',
// The index write-lock (#2658) selects its backend by process.platform — the
// OS socket lock (Windows named pipe / Linux abstract socket) vs the file
// fallback — and its socket-backend describe block is gated to linux/win32.
// The Ubuntu suite only proves the Linux abstract-socket path, so run it here
// to exercise the Windows named-pipe backend and the macOS file fallback on
// their real platforms (#2658 review H3).
'test/unit/index-lock.test.ts',
];
// Native LadybugDB integration tests — exercise the @ladybugdb/core
// N-API addon which has known platform-specific behavior (Windows
// file-lock lag after close, macOS N-API destructor segfaults)
const LBUG_NATIVE = [
// Inline-handler findings must survive native persistence and cold explain
// readers on each OS; passing the Ubuntu-only taint suite is insufficient.
'test/integration/taint-explain.test.ts',
// Real indexing/MCP previews include tracked/untracked edits, file-symlink
// retargets and staging failures; the helper runs in a separate native process.
'test/integration/rename-preview-plan.test.ts',
'test/integration/rename-provider-matrix.test.ts',
'test/integration/rename-provider-c.test.ts',
'test/integration/rename-provider-cobol.test.ts',
'test/integration/rename-provider-cpp.test.ts',
'test/integration/rename-provider-csharp.test.ts',
'test/integration/rename-provider-dart.test.ts',
'test/integration/rename-provider-go.test.ts',
'test/integration/rename-provider-java.test.ts',
'test/integration/rename-provider-javascript.test.ts',
'test/integration/rename-provider-kotlin.test.ts',
'test/integration/rename-provider-objective-c.test.ts',
'test/integration/rename-provider-php.test.ts',
'test/integration/rename-provider-python.test.ts',
'test/integration/rename-provider-ruby.test.ts',
'test/integration/rename-provider-rust.test.ts',
'test/integration/rename-provider-swift.test.ts',
'test/integration/rename-provider-typescript.test.ts',
'test/integration/rename-provider-vue.test.ts',
'test/integration/rename-provider-zig.test.ts',
'test/integration/pdg-explicit-owner.test.ts',
'test/integration/skip-fts.test.ts',
'test/integration/lbug-core-adapter.test.ts',
// Native COPY, cached vectors and cold shared-store readers vary by platform.
'test/integration/embedding-identity-integrity.test.ts',
'test/integration/shared-store-embedding-reuse.test.ts',
// Real endpoint retries include native index writes and cold database reads.
'test/integration/embedding-input-isolation.test.ts',
'test/integration/lbug-vector-extension.test.ts',
'test/integration/lbug-pool.test.ts',
'test/integration/lbug-pool-stability.test.ts',
'test/integration/lbug-lock-retry.test.ts',
'test/integration/lbug-open-retry.test.ts',
'test/integration/lbug-close-handle-release.test.ts',
'test/integration/lbug-orphan-sidecar-recovery.test.ts',
'test/integration/lbug-interrupted-checkpoint-recovery.test.ts',
'test/integration/lbug-checkpoint-accounting.test.ts',
'test/integration/lbug-incremental-search.test.ts',
// Retained identities and reader metadata must agree on every native platform.
'test/integration/incremental-method-integrity.test.ts',
'test/integration/go-incremental-method-owners.test.ts',
'test/integration/query-symbol-integrity.test.ts',
'test/integration/mcp-index-generation.test.ts',
'test/integration/fts-repair-warm-session.test.ts',
'test/unit/lbug-pool-pinning.test.ts',
'test/unit/hybrid-search.test.ts',
'test/integration/lbug-readonly-init.test.ts',
'test/integration/lbug-non-ascii-path.test.ts',
// Cross-repo trace e2e: builds two real lbug indexes + a real bridge and
// opens them through the pool adapter (native addon + bridge file locking).
// Windows is skipped in-file (describeReopen) due to the bridge reopen lock.
'test/integration/group/cross-trace-e2e.test.ts',
'test/integration/group/graphql-resolve-symbol.test.ts',
'test/integration/local-backend.test.ts',
'test/integration/bm25-symbol-hydration.test.ts',
'test/integration/local-backend-calltool.test.ts',
'test/integration/search-core.test.ts',
'test/integration/search-pool.test.ts',
'test/integration/fts-description-search.test.ts',
'test/integration/staleness-and-stability.test.ts',
'test/integration/analyze-wal-checkpoint-failure.test.ts',
// Publication, cold readers and sidecar handling depend on native filesystem behavior.
'test/integration/analyze-atomic-swap.test.ts',
'test/unit/incremental-write-integrity.test.ts',
'test/unit/lbug-publication.test.ts',
'test/integration/fts-stemmer-sweep.test.ts',
'test/integration/lbug-multiwriter-deadlock.test.ts',
// #2409 batched incremental writeback: chunked IN-list DETACH DELETEs +
// backslash quote escaping against the REAL native engine — the failing
// environment for #2409 was Windows, so the write pattern must be proven
// on the windows-latest native addon, not just Ubuntu.
'test/integration/lbug-delete-nodes-for-files.test.ts',
// #2409 defect 2: dirty-flag recovery parks lbug.wal/.shadow (rename next
// to a live native DB, rm-then-rename over an existing parked copy) before
// any open — rename semantics are exactly what differs on Windows.
'test/unit/incremental-dirty-recovery.test.ts',
// #2623: the incremental writeback must load VECTOR before the CodeEmbedding
// join-delete, and the blocked path must escalate instead of crashing. The
// win32 VECTOR gate was removed in the same PR, so this ordering must be
// proven on the windows-latest native addon, not just Ubuntu. Budget: ~25s
// on Linux → expect ~2min on the slowest Windows shard.
'test/unit/incremental-vector-extension-ordering.test.ts',
// #2841: the FTS half of that same gate, plus the both-extensions-blocked
// case — and it needs this matrix for two reasons the VECTOR sibling above
// does not cover. The reported failure environment is a machine where the
// extension stopped LOADING, which is the #2374 class and Windows-reported
// (the same reason fts-extension-e2e.test.ts is registered below), so the
// FTS-unavailable branch has to run on a real Windows/macOS runner rather
// than only on Ubuntu where FTS always loads. And its both-blocked case is
// gated on GITNEXUS_REQUIRE_VECTOR=1, which ci-tests.yml requires in both
// coverage and platform jobs so an unavailable VECTOR fails loudly. Budget: four real analyze runs, so expect it to sit alongside the
// VECTOR sibling's ~87s Windows measurement.
'test/unit/incremental-index-extension-dml-gate.test.ts',
];
// Process spawning and CLI tests — exercise child_process with real
// process spawning, which behaves differently across platforms (shell
// quoting, path resolution, signal handling)
const SPAWN_CLI = [
'test/integration/cli-e2e.test.ts',
'test/integration/cli-limit-e2e.test.ts',
'test/integration/hooks-e2e.test.ts',
'test/integration/skills-e2e.test.ts',
// Spawns the real CLI across hermetic HOME/USERPROFILE homes to exercise the
// FTS extension lifecycle — the #2374 bug was Windows-reported, so this must
// run on the Windows/macOS matrix, not just the Ubuntu full suite.
'test/integration/fts-extension-e2e.test.ts',
'test/integration/fts-vendored-root-seam.test.ts',
'test/integration/server-http-startup.test.ts',
'test/integration/mcp/server-startup.test.ts',
'test/integration/analyze-heap-oom-e2e.test.ts',
'test/integration/group/group-cli.test.ts',
'test/integration/cli/tool-no-index-stderr.test.ts',
// Real CLI spawn + directory symlinks for the update-notice parent/child path.
'test/integration/cli/update-notice.test.ts',
'test/integration/setup-skills.test.ts',
'test/integration/setup-antigravity.test.ts',
'test/integration/antigravity-hook-e2e.test.ts',
'test/unit/local-cli-subprocess.test.ts',
'test/unit/runner-exec-tail.test.ts',
// Real cross-process single-writer lock coordination (#2658): child processes
// contend for the lock and race to reclaim a dead holder. Process spawning,
// kernel socket auto-release (Win named pipe / Linux abstract socket), and the
// FILE-backend rename-steal reclaim (macOS/BSD default) all vary across OSes —
// the exact behaviors the Windows/macOS matrix must prove. macOS timing first
// exposed a file-backend double-admit race here (#2658 review); the reclaim is
// now judgment-verified so a live holder is never displaced.
'test/integration/analyze-index-lock-concurrency.test.ts',
// The per-group sync lock (R9), same class of guarantee one level up: real
// child processes contend for one group's lock while this process runs a real
// `syncGroup`, and the CLI case spawns the real command. Everything that
// varies here is platform-owned — which backend `selectBackend()` picks
// (Windows named pipe / Linux abstract socket / macOS file lock), kernel
// auto-release on SIGKILL vs. the file backend's pid-liveness reclaim, and
// `mkdir` over an occupied path. The fail-closed cases pin
// GITNEXUS_INDEX_LOCK_BACKEND=file so the filesystem branch is exercised on
// every OS rather than only where it is the default; no case is skipped on
// any platform, because a skipped case turns "a sync that cannot be protected
// does not run" into a claim that holds on Ubuntu only.
'test/integration/group/group-sync-lock-concurrency.test.ts',
// The three `dist/` module-load closure guards, all built on the shared
// child-process probe in `test/helpers/module-load-probe.ts`. That probe IS
// the platform-varying part: it spawns `process.execPath` in array form,
// clears NODE_OPTIONS, addresses its target via `pathToFileURL` (Windows needs
// the `file:///C:/...` form — a bare absolute path is not a valid ESM
// specifier there), and renders every result through a `path.sep`→POSIX
// normalisation the anchors and offender regexes depend on. None of that is
// proven anywhere else.
//
// Cheap: measured on the Windows runner at 448 ms, 53 ms and sub-second. An
// earlier attempt to register them still turned the matrix red — not from
// their own cost, but because vitest sharded by file COUNT, so inserting any
// file re-partitioned the list and happened to cluster `cli-e2e` (621 s) with
// `cli-limit-e2e` (75 s) on one shard. The split is weight-aware now
// (`scripts/cross-platform-shard.ts`), so a cheap file can no longer move a
// heavy one.
//
// #2802: MCP startup must not eagerly load the analyze-only language
// provider registry or the group contract extractors.
'test/integration/mcp/startup-language-closure.test.ts',
// PR #1383: `cli/mcp.js`'s static-import closure must stay leaf-only so no
// native binding initialises before the stdout sentinel installs.
'test/integration/mcp/import-closure.test.ts',
// #2091/#2093/#2116: the scope-resolution registry must not load the optional
// tree-sitter grammars at import time. The offender regexes match grammar
// paths with either separator, which only the Windows runner proves.
'test/integration/optional-grammars/registry-import-closure.test.ts',
];
// Worker threads tests — exercise real worker_threads which have
// platform-specific behavior (thread spawning, IPC, exit handling)
const WORKER_THREADS = [
// Native callback parsing and exact caller/summary ownership (#3489).
'test/integration/cfg/interproc-taint.test.ts',
'test/integration/function-local-identity.test.ts',
'test/unit/community-processor.test.ts',
'test/integration/worker-pool.test.ts',
'test/integration/parse-impl-quarantine-cache-skip.test.ts',
// Real parse workers and both on-disk parse stores must retain lexical
// local-import ownership on each platform's native grammar/runtime pair.
'test/integration/local-import-worker-parity.test.ts',
'test/integration/resolvers/local-import-persistence.test.ts',
// The same worker and store path for fetch-wrapper facts and their resolution.
'test/integration/resolvers/fetch-wrapper-persistence.test.ts',
];
// Tree-sitter native addon smoke tests — verify that native grammars
// load correctly on each platform (binary compatibility, .node loading)
const NATIVE_ADDON_SMOKE = [
'test/integration/tree-sitter-languages.test.ts',
'test/integration/parsing.test.ts',
'test/integration/pipeline.test.ts',
'test/integration/pipeline-graph-golden.test.ts',
'test/unit/parser-loader.test.ts',
'test/unit/parser-loader-abi.test.ts',
];
// Filesystem behavior tests — exercise operations that vary across
// platforms (CRLF, symlinks, permissions, temp dirs)
const FILESYSTEM = [
// The deletion-guard cases in this file require real Windows path semantics.
'test/unit/canonicalize-path-long-path-prefix.test.ts',
'test/unit/storage-resolver.test.ts',
// Dart package metadata is captured during the shared scan on every platform.
'test/unit/dart-package-imports.test.ts',
// Verify the same capture is reused by native parsing and scope resolution.
'test/integration/resolvers/dart.test.ts',
// Cargo membership uses path normalization, descriptor validation, symlinks,
// and Rust native parsing (including long Windows source strings).
'test/unit/scope-resolution/rust-cargo-targets.test.ts',
// The durable ParsedFile store's prune tolerates a chunk directory it cannot
// delete (#3204). The failures that motivate it — held handles, read-only
// mounts — are Windows- and macOS-flavored, and the permission-based case
// skips itself where chmod cannot block a delete, so run it everywhere.
'test/unit/parsedfile-store.test.ts',
'test/integration/filesystem-walker.test.ts',
'test/integration/analyze-directory-symlinks.test.ts',
'test/integration/watch-filesystem.test.ts',
'test/integration/markdown-processor-crlf.test.ts',
'test/integration/ignore-and-skip-e2e.test.ts',
// Pins that the bridge pairing verdict is measured before the database is
// opened. The property it protects is about mtime behavior across OS and
// filesystem, and the alternative — really opening the bridge — cannot run on
// Windows at all (in-process write→read reopen of the same bridge.lbug is a
// documented limitation). Running it on every platform is the whole point:
// Windows is where an unverified assumption about mtime would hurt most.
'test/unit/group/bridge-pairing-precedes-open.test.ts',
// The raw-control-byte guard reads every tracked text file `git ls-files`
// reports — 4893 of them — and decides membership from the git path, which is
// always `/`-separated no matter what the host separator is. Both halves of
// that are platform-varying: the collector basename-matches with
// `path.posix.basename` against `git ls-files -z` output while the reads go
// through `path.join`, so on Windows the same string is consumed under two
// separator conventions in one pass, and only a real windows-latest run
// proves they agree. It is also the file-count-heaviest read loop in the
// suite, so it is where a per-file filesystem cost (NTFS + Defender, or
// macOS's slower stat path) would show up first. No case is skipped on any
// platform: a guard that only holds on Ubuntu is not a guard on the file
// whose NUL it exists to catch. Budget: the heaviest single case is one
// 4893-file pass — 2.3 s on a slow virtualised filesystem, 0.34 s on a local
// disk — against a 30 s testTimeout.
'test/unit/source-control-bytes.test.ts',
// Auto-sync reads a cloned `.gitnexusrc` through the symlink/hard-link guard,
// and clone recovery uses real `git` plus temp dirs. Ubuntu coverage alone
// would never create the Windows file symlink (`type: 'file'`) or run the
// git-config failure path on windows-latest / macos-latest.
'test/unit/gitnexus-rc-embeddings.test.ts',
'test/unit/git-clone.test.ts',
];
const ALL_CROSS_PLATFORM = [
...PLATFORM_LOGIC,
...LBUG_NATIVE,
...SPAWN_CLI,
...WORKER_THREADS,
...NATIVE_ADDON_SMOKE,
...FILESYSTEM,
];
// When invoked directly, print the file list for vitest consumption
if (process.argv[1]?.endsWith('cross-platform-tests.ts')) {
console.log(ALL_CROSS_PLATFORM.join('\n'));
}
export {
ALL_CROSS_PLATFORM,
PLATFORM_LOGIC,
LBUG_NATIVE,
SPAWN_CLI,
WORKER_THREADS,
NATIVE_ADDON_SMOKE,
FILESYSTEM,
};