test(parity): add node:process granular suite by proggeramlug · Pull Request #1331 · PerryTS/perry

proggeramlug · 2026-05-22T10:47:22Z

Comprehensive node:process node-suite (there was none before). 41 granular cases covering the deterministic surface, validated against node --experimental-strip-types. Tests use the global process (the primary API). This branch also carries #1371's runtime fixes (merged in), so it lands tests and impl together.

Passing (24/41)

object shape · platform/arch · pid · ppid · env reads + Object.keys · cwd() · version + versions.node/v8 · argv shape · nextTick ordering/returns-undefined · hrtime() tuple + bigint · memoryUsage() · uptime() · stdout/stderr write + fds · global identity · method-value typeof · env writes · argv0/execPath/title (bold = fixed by #1371).

Gaps (17/41) — each tracked by its own granular issue

method gaps: cpuUsage node:process — implement process.cpuUsage() #1347, release node:process — implement process.release #1348, execArgv node:process — implement process.execArgv #1349, exitCode node:process — implement process.exitCode (default undefined + set/get) #1350, nextTick(cb,…args) node:process — process.nextTick(cb, ...args) drops trailing args #1351
EventEmitter API node:process — implement the EventEmitter API (on/emit/listeners/removeListener/...) #1372 · umask node:process — implement process.umask() #1373 · abort node:process — implement process.abort() #1374 · emitWarning node:process — implement process.emitWarning() #1375 · resourceUsage/getActiveResourcesInfo node:process — implement resourceUsage() + getActiveResourcesInfo() #1376 · availableMemory/constrainedMemory node:process — implement availableMemory() + constrainedMemory() #1377 · features node:process — implement process.features #1378 · config node:process — implement process.config #1379 · allowedNodeEnvironmentFlags node:process — implement process.allowedNodeEnvironmentFlags (Set) #1380 · versions sub-fields node:process — populate process.versions sub-fields (uv, modules, ...) #1381

Every failure is recorded in known_failures.json pointing at its specific issue (no umbrellas). No version bump / CHANGELOG.

proggeramlug · 2026-05-22T14:09:11Z

Final-pass update: suite is now 52 cases, 28 pass. Added kill/chdir/debugPort/argv[1] (pass) and report/connected/memoryUsage.rss/getBuiltinModule/loadEnvFile/sourceMapsEnabled/title-set (gaps → #1395–#1401). Every failure maps to a granular issue in known_failures.json; cross-checked no stale/untracked.

…sweep (#1414) Rolls up 26 PRs that merged to main post-v0.5.1023 without version bumps: - node:crypto gap-fixes (#1386 #1393 #1394 #1402 #1405): randomInt, timingSafeEqual, getHashes/getCiphers, sha224/sha384, base64 digest, Buffer hash input, no-arg digest() → Buffer, pbkdf2Sync digest arg, scryptSync. - node:perf_hooks (#1321 + #1328 #1342 coverage): performance + User Timing + PerformanceObserver native impl, granular node-suite + edge-case coverage. - #1090 GC checkpoint runtime work (#1324). - #1311 geisterhand on iOS (#1316 #1383 #1384 #1385). - #1312 process.env.X (unset) is nullish undefined (#1314). - #1319 thread-safety hardening for cross-thread runtime statics. - #1322 exact-head GC evidence packet. - #1323 wasm timers dispatch through mem_call bridge (#1329). - #1317 node:timers/promises shadow-segfault fix (#1326). - #1330 node:process suite (#1331). - #1292 bcrypt.hash() returns String (#1307). - #1293 fastify .json()/.body external-fastify dispatch (#1308). - #1296 app pattern performance gaps. - #1297 diagnostics_channel parity. - #1301 iOS App Groups capability (#1313). - #1318 #1325 os/methods/modern-methods static dispatch. - #1315 expanded Node parity test coverage. - #1382 ui-ios stdlib pump for async fetch. - #1392 ui-wasm reactive state + setText (#1404).

…test global-timer pollution CI flake on #1331 (`event_pump::tests::wait_returns_when_timer_due` returning in 426µs instead of ~50ms). `js_wait_for_event`'s wait budget is derived from GLOBAL timer state and `SERIAL` only serializes the event_pump tests against each other — a concurrently-running non-serialized test (a timer/promise test in another module) can inject a sooner deadline or notify and shorten any single measurement. Pollution never lengthens a wait past its 50ms budget, so a working wait blocks ~50ms on at least one un-polluted attempt. Replace the single-shot measurement with the file's established retry-until-clean shape (same as `sustained_budget_zero_spin_is_throttled`): drain leaked timer/notify state at the start of each attempt, schedule our 50ms timer, measure, and accept the test as soon as one attempt lands in [40ms, 500ms). Stays deterministic against a real budget bug (which would return early on *every* attempt) while tolerating the cross-test race. Verified: full multi-threaded `cargo test -p perry-runtime` passes (wait_returns_when_timer_due OK); 5× targeted stress all green.

…e_parity.sh First exhaustive node:process suite on main: 60 granular cases (39 pass / 22 fail) covering shape, platform, env, cwd, version + versions sub-fields, argv, nextTick, hrtime + bigint, props, memory + rss + available, cpu + threadCpuUsage, uptime, stdout/stderr fds + write, global identity, release, exec-argv, exit + exit-code, events (on/emit/listeners), methods (umask/abort/emit-warning/kill/chdir/get-builtin-module/load-env-file/ source-maps-enabled/dlopen/ref-unref/exit), resource (usage + active), features, config, ipc (connected/send-disconnect), report, allowed flags, title, sub-fields, unix uid/gid, module-load-list, finalization, exception uncaught-capture, shapes ppid/debug-port/method-types/process-object/ module-load-list, exit-code, hrtime tuple form. 22 failing cases tracked in known_failures.json with granular per-gap issue references. Method-value reflection (#1343), env-write (#1344), hrtime() tuple (#1345), and argv0/execPath/title (#1346) were originally fixed in #1371 but that PR merged into the #1331 branch only — #1331 was rebased onto current main and those fixes were dropped; the issues are reopened. Adds run_module_parity.sh: thin wrapper that runs ./run_parity_tests.sh --suite node-suite --module <m> for a list of modules (default: process + perf_hooks) and prints a combined per-module + total summary.

…e_parity.sh First exhaustive node:process suite on main: 60 granular cases (39 pass / 22 fail) covering shape, platform, env, cwd, version + versions sub-fields, argv, nextTick, hrtime + bigint, props, memory + rss + available, cpu + threadCpuUsage, uptime, stdout/stderr fds + write, global identity, release, exec-argv, exit + exit-code, events (on/emit/listeners), methods (umask/abort/emit-warning/kill/chdir/get-builtin-module/load-env-file/ source-maps-enabled/dlopen/ref-unref/exit), resource (usage + active), features, config, ipc (connected/send-disconnect), report, allowed flags, title, sub-fields, unix uid/gid, module-load-list, finalization, exception uncaught-capture, shapes ppid/debug-port/method-types/process-object/ module-load-list, exit-code, hrtime tuple form. 22 failing cases tracked in known_failures.json with granular per-gap issue references. Method-value reflection (#1343), env-write (#1344), hrtime() tuple (#1345), and argv0/execPath/title (#1346) were originally fixed in onto current main and those fixes were dropped; the issues are reopened. Adds run_module_parity.sh: thin wrapper that runs ./run_parity_tests.sh --suite node-suite --module <m> for a list of modules (default: process + perf_hooks) and prints a combined per-module + total summary.

Node's `process.dlopen(module, filename, flags?)` is the native-addon (.node) loader. Perry was reading it as a bare number, so `typeof process.dlopen === "function"` returned false and call sites crashed with "value is not a function". Perry statically links every dependency at compile time — there's no dynamic loader to wire up. The closest no-op is to return undefined: optional-dep wrappers that feature-detect on `typeof process.dlopen === "function"` first see a "loaded" non-error and fall back to their pure-JS path; real addon-loading attempts surface as undefined exports downstream. - Call lowering (`native_module.rs`): folds to `Expr::Undefined`, sitting next to the `ref` / `unref` / `setSourceMapsEnabled` / `getBuiltinModule` arms. - Typeof fold (`lower_expr.rs`): adds `dlopen` to the same process-method-as-function list. Test: test-parity/node-suite/process/methods/dlopen.ts (already in the granular suite from #1331) now passes — byte-identical to `node --experimental-strip-types`.

Node exposes `process.hasUncaughtExceptionCaptureCallback()` (boolean getter) + `process.setUncaughtExceptionCaptureCallback(fn|null)` (installs/clears a single capture callback that intercepts uncaught exceptions before the `uncaughtException` event fires). Perry was reading both as bare numbers — `typeof` lied with `"number"` and any call crashed with "value is not a function". Perry doesn't expose the underlying capture hook, so both behave as no-ops: - `hasUncaughtExceptionCaptureCallback()` → `Expr::Bool(false)`. - `setUncaughtExceptionCaptureCallback(fn|null)` → `Expr::Undefined`. Both names are added to the typeof-fold list in `lower_expr.rs` so feature-detection (`typeof process.X === "function"`) returns `"function"`, sitting next to the `ref` / `unref` / `setSourceMapsEnabled` / `getBuiltinModule` / `dlopen` arms. Test: test-parity/node-suite/process/exception/uncaught-capture.ts (in the granular suite from #1331) now passes — byte-identical to `node --experimental-strip-types`.

…1344) (#1504) Pre-fix `process.env.X = v` hit the generic `js_object_set_field_by_name` path and stored on a cached ProcessEnv backing object, but the read side (`process.env.X` → `EnvGet`) consults `std::env::var` directly. The value never round-tripped: child processes inherited the unmodified parent env, and even `console.log(process.env.X)` immediately after the assignment came back `undefined`. Route `PropertySet { object: ProcessEnv, property, value }` through a new `js_setenv(name, value)` runtime helper that: - coerces the value to a string via `js_jsvalue_to_string` (matches Node's `process.env.PORT = 8080` → `"8080"` coercion), - writes via `std::env::set_var(name, str_value)` — exactly the same store the existing `js_getenv` read side consults. Add a matching `js_removeenv(name)` for the future `delete process.env.X` hookup (HIR side is already in flight in #1331's branch). Closes #1344.

…1505) Node 20.12+ exposes `process.loadEnvFile(path?)` to read a `.env` file from disk and add its KEY=value entries to process.env. Perry was reading it as a bare number, so `typeof process.loadEnvFile === "function"` returned false and call sites crashed with "value is not a function". Perry today doesn't persist `process.env.X = v` writes (tracked in #1344), so eagerly loading a `.env` would be moot. Returning undefined is the closest no-op for the probe-and-call pattern. Real `.env` loading is tracked separately — this PR closes the typeof / call-doesn't-crash gap. - Call lowering (`native_module.rs`): folds to `Expr::Undefined`, sitting next to the `ref` / `unref` / `setSourceMapsEnabled` / `getBuiltinModule` / `dlopen` / `hasUncaughtExceptionCaptureCallback` / `setUncaughtExceptionCaptureCallback` arms. - Typeof fold (`lower_expr.rs`): adds `loadEnvFile` to the same process-method-as-function list. Test: test-parity/node-suite/process/methods/load-env-file.ts (in the granular suite from #1331) now passes — byte-identical to `node --experimental-strip-types`.

…1478) (#1507) Node's `performance.markResourceTiming(info, ...)` records a PerformanceResourceTiming entry built from a PerformanceTimingInfo- shaped object (paired with `clearResourceTimings` / `setResourceTimingBufferSize`). Perry was reading the property as undefined — `typeof performance.markResourceTiming === "function"` returned false and any call crashed. Perry doesn't track the resource-timing buffer yet, so the call is a no-op returning undefined. The accompanying read of `performance.getEntriesByType("resource")` still returns an empty array, matching the no-recorded-entries shape. - HIR dispatch (`module_static.rs`): adds `markResourceTiming` to the perf_hooks NativeMethodCall list. - Codegen (`perf_hooks.rs`): folds the call to undefined, sitting next to the `clearResourceTimings` / `setResourceTimingBufferSize` arms. - Property-read typeof (`object/native_module.rs`): adds `("perf_hooks", "markResourceTiming")` to the callable-export list so `typeof performance.markResourceTiming === "function"` matches Node and `const f = performance.markResourceTiming` is a callable bound-method closure. Test: test-parity/node-suite/perf_hooks/resource-timing/mark-resource-timing.ts (in the granular suite from #1331) now passes — byte-identical to `node --experimental-strip-types`.

…1479) (#1516) Node returns `"[object Performance]"` from `Object.prototype.toString.call(performance)` — that's the spec'd host-tag for the `Performance` interface. Perry was returning the generic `"[object Object]"` because the perf_hooks namespace is a `NATIVE_MODULE_CLASS_ID`-tagged ObjectHeader: it never reaches the class-level `Symbol.toStringTag` hook (those are keyed by user-class class_id). - `read_native_module_name(obj_ptr)` helper in `object/native_module.rs` decodes the module-name string stored in field 0 of the namespace ObjectHeader. - `js_object_to_string` in `object/mod.rs`: after the class-level hook misses, if `class_id == NATIVE_MODULE_CLASS_ID` look up the module name and map it via a new `native_module_to_string_tag` table. Today the only entry is `perf_hooks → "Performance"`; future modules with a defined Node tag (`url → "URL"`, `events → "EventEmitter"`, etc.) slot in here. Test: test-parity/node-suite/perf_hooks/shapes/to-string-tag.ts (in the granular suite from #1331) now passes — byte-identical to `node --experimental-strip-types`. `performance[Symbol.toStringTag]` reading "Performance" directly remains a separate gap (Symbol-keyed property dispatch on namespace objects); the spec-compliant toString path is now covered.

proggeramlug mentioned this pull request May 22, 2026

node:process — gaps surfaced by the granular node-suite (method typeof, env write, hrtime(), argv0/execPath/title) #1330

Closed

proggeramlug force-pushed the feat/process-node-suite branch 2 times, most recently from 11ff9b1 to bab130d Compare May 22, 2026 11:42

proggeramlug mentioned this pull request May 22, 2026

feat(process): #1330 wire method values, env writes, hrtime(), argv0/execPath/title #1371

Merged

proggeramlug force-pushed the feat/process-node-suite branch from a084b37 to a83ac81 Compare May 23, 2026 09:36

proggeramlug force-pushed the feat/process-node-suite branch from a83ac81 to 63014c4 Compare May 23, 2026 09:39

proggeramlug merged commit 8510555 into main May 23, 2026
17 checks passed

proggeramlug deleted the feat/process-node-suite branch May 23, 2026 09:45

proggeramlug mentioned this pull request May 23, 2026

feat(process): implement process.dlopen as no-op stub (#1409) #1501

Merged

2 tasks

proggeramlug mentioned this pull request May 23, 2026

feat(process): implement uncaught-exception capture API (#1406) #1503

Merged

2 tasks

proggeramlug mentioned this pull request May 23, 2026

feat(process): implement process.loadEnvFile as no-op stub (#1399) #1505

Merged

2 tasks

proggeramlug mentioned this pull request May 23, 2026

feat(perf_hooks): implement performance.markResourceTiming as no-op (#1478) #1507

Merged

2 tasks

proggeramlug mentioned this pull request May 23, 2026

feat(perf_hooks): tag performance namespace as [object Performance] (#1479) #1516

Merged

2 tasks

Navigation Menu

Uh oh!

Uh oh!

test(parity): add node:process granular suite - #1331

test(parity): add node:process granular suite#1331
proggeramlug merged 1 commit into
mainfrom
feat/process-node-suite

proggeramlug commented May 22, 2026 •

edited

Loading

Uh oh!

proggeramlug commented May 22, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Uh oh!

Conversation

proggeramlug commented May 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Passing (24/41)

Gaps (17/41) — each tracked by its own granular issue

Uh oh!

proggeramlug commented May 22, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

proggeramlug commented May 22, 2026 •

edited

Loading