module: remove --experimental-transform-types - #61803

marco-ippolito · 2026-02-13T16:37:18Z

This PR removes the --experimental-transform-types flag and all related code, tests, and documentation.

nodejs-github-bot · 2026-02-13T16:37:23Z

Review requested:

GeoffreyBooth · 2026-02-13T17:03:11Z

process.features.typescript now only returns a boolean

I think we should add process.features.stripTypes that returns a boolean. This should be a separate PR so that it can be backported.

Then the question becomes what to do with process.features.typescript. I think having a feature called typescript only makes sense if the overall feature has various modes, like currently it's either off (false), strip or transform. But we might someday add other TypeScript-related features like type checking, or tsconfig.json support, or support for various features within tsconfig.json; and these might be individual features to be toggled on or off (like maybe a user enables tsconfig.json paths support but not type checking) and so a single string value for an overall mode doesn't make sense. We could make process.features.typescript a Record<string, boolean> so that there's something like process.features.typescript.strip === true and other features add onto there, or treat each new TypeScript-related feature as a top-level key on process.features.

For the purposes of this PR, perhaps the simplest thing to do would be to just keep process.features.typescript as it is but remove transform as a possible return type. Then a redesign of process.features.typescript can happen in a dedicated PR.

marco-ippolito · 2026-02-13T17:04:51Z

User can run --no-strip-types so typescript might not be enabled

aduh95 · 2026-02-13T17:45:37Z

process.features.typescript now only returns a boolean.

I wouldn't do that:

it's perfectly workable as is, as a string
if we end up re-introducing --transform-type or another type of typescript handling, it will put us in an awkard position.

Why risk it?

styfle · 2026-02-13T18:08:38Z

if we end up re-introducing --transform-type or another type of typescript handling, it will put us in an awkard position.

If we think thats a possibility in the future, then we should keep process.features.typescript as a string and also keep the options object in stripTypeScriptTypes() so we can allow strip but throw and error for other modes.

TLDR - keep this semver-major change as breaking only for those using transform but anyone using strip should not have a breaking change. Does that make sense?

codecov · 2026-02-13T18:34:53Z

Codecov Report

❌ Patch coverage is 66.66667% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.66%. Comparing base (da5efc4) to head (1708c57).
⚠️ Report is 35 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/internal/modules/typescript.js	75.00%	1 Missing ⚠️
src/compile_cache.cc	0.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #61803      +/-   ##
==========================================
+ Coverage   88.84%   89.66%   +0.82%     
==========================================
  Files         674      676       +2     
  Lines      204957   205951     +994     
  Branches    39309    39434     +125     
==========================================
+ Hits       182087   184670    +2583     
+ Misses      15088    13426    -1662     
- Partials     7782     7855      +73

Files with missing lines	Coverage Δ
lib/internal/bootstrap/node.js	`98.96% <100.00%> (-0.01%)`	⬇️
src/compile_cache.h	`100.00% <ø> (ø)`
src/node_options.cc	`76.32% <ø> (-0.05%)`	⬇️
lib/internal/modules/typescript.js	`92.30% <75.00%> (-5.18%)`	⬇️
src/compile_cache.cc	`79.09% <0.00%> (-1.64%)`	⬇️

... and 210 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

GeoffreyBooth · 2026-02-13T19:58:22Z

This PR's goal is to remove transform-types. So in this PR, I think just remove the transform response from process.features.typescript. Any other rethinking or redesign of that property can come in a follow-up PR.

marco-ippolito · 2026-02-23T16:09:35Z

Applied suggestions, only remove transform related stuff

ljharb

LGTM overall

marco-ippolito · 2026-02-23T23:29:43Z

Cna someone help me understand the CI failure?

Renegade334 · 2026-02-25T21:42:11Z

Cna someone help me understand the CI failure?

The --[experimental-]strip-types CLI flag needs to be reintroduced here:

https://github.com/nodejs/node/blob/50a77dbaa26fd58460d1bee1de317047d1462a5f/test/parallel/test-runner-cli.js#L67-L68

aduh95 · 2026-02-26T08:06:35Z

Cna someone help me understand the CI failure?

The --experimental-strip-types CLI flag needs to be reintroduced here:

That doesn't really make sense, type stripping has been unflagged for a while, that flag is now a no-op. The issue seems to be that the following branch is no longer needed:

https://github.com/nodejs/node/blob/50a77dbaa26fd58460d1bee1de317047d1462a5f/test/parallel/test-runner-cli.js#L71-L74

`processTypeScriptCode` does not produce source maps at all for type stripping. Given that the transform mode was removed, there is no need to keep the source map support around it. Signed-off-by: Chengzhong Wu <cwu631@bloomberg.net> PR-URL: #63738 Refs: #61803 Reviewed-By: Marco Ippolito <marcoippolito54@gmail.com> Reviewed-By: Joyee Cheung <joyeec9h3@gmail.com> Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: Trivikram Kamat <trivikr.dev@gmail.com>

Node 26 removed the "transform" value (nodejs/node#61803); the getter returns "strip" by default or false under --no-strip-types. Code feature-detecting TypeScript support against Node 26 checks for "strip", so returning "transform" breaks strict comparisons.

… MCP negotiation test; provider-view cross-process coverage (integration) (#554) * fix(scripts): choose the TypeScript transform flag Node actually supports (Node 26 drops --experimental-transform-types) runScript spawned every plain .ts script under node --experimental-transform-types. Node 26 removed the flag (nodejs/node#61803) and rejects it as a bad option (exit code 9), so the Verify (Node 26) leg failed on every main push. typeScriptTransformFlags (core/runtime.ts) decides from process.allowedNodeEnvironmentFlags: the transform flag where the binary accepts it (Node 22, 24), nothing on Node 26, which strips types unflagged. Unit-tested against the flag sets of each release line. * chore: name #554 in the changeset * fix(scripts): name --strip-types on Node 26 so an inherited NODE_OPTIONS=--no-strip-types cannot switch TypeScript loading off Codex review on #554: with no command-line flag the child inherited the environment's --no-strip-types and failed on every typed .ts source. The helper now picks the first flag the binary accepts, strongest first: --experimental-transform-types (22, 24), then --strip-types (26). Covered by a script-dispatch test that sets the version-appropriate switch in NODE_OPTIONS and expects the source run to succeed regardless. * test(routes): prove the provider request view across built surfaces; fix framework-mode.md ordering (from #556) * fix(install): apply the operator .env layer before plugin modules evaluate and below manifest env defaults (#469) Two findings from the #538 self-review. Precedence: a host merges the stdio server's manifest `env` block into the child environment, so the shell could not tell a manifest default from a host export and reserved both — manifest env beat the file, contrary to the documented `manifest < .env < .env.local < process.env`. The emitted stdio entry now embeds the server's normalized `env` block as build-time literals and `applyOperatorEnv` takes it as `manifestEnv`: a present variable is reserved only when its value differs from the embedded default, so a passed-through default yields to the file while a host or operator export is kept. An operator export equal to the default is indistinguishable from the pass-through and yields too; a default carrying a path token never equals its expanded value and is always kept. Host manifests are unchanged. Import timing: the layer was a statement after the consumer imports, and ESM evaluates static imports first, so module-level `process.env` reads in hook handlers and CLI route/provider modules never saw the file. A dynamic `import()` after the statement does not help either — Rspack inlines a single-chunk bundle into one scope and places the dynamic target ahead of the static imports. The layer is now a generated virtual module (`agent-bundle/launch-env-layer`) that every stdio entry, hook wrapper, and artifact CLI bin imports first, with the server module, handler, routes, providers, and state definition as static imports after it; the build marks generated modules side-effectful so a consumer `"sideEffects": false` cannot drop the bare import. The MCP shell's `loadEntry` becomes a static import for the same reason, so the console guard now covers the factory call and the running server rather than the module's top-level evaluation. Tests build each shell through the real pipeline and run it under node with a `process.env` read at module top level: manifest-only key takes the file, host-exported key keeps the host value, absent key takes the file, `AGENT_BUNDLE_ENV_FILE=none` restores the previous behaviour. * chore: drop the tracked .superpowers scratch notes and ignore the folder * fix(mcp): install the stdout guard in the stdio entry's first import so module-scope writes never reach the protocol stream (#469) The env-precedence follow-up made the generated stdio entry import the server module statically so the operator .env layer lands by import order — but that put the module's top level ahead of the console guard that `runGeneratedStdioMcpEntry` installs in the shell body. A `console.log` or `process.stdout.write` at module scope in a consumer's server or tool module reached stdout, which carries JSON-RPC framing, contradicting the documented guarantee that redirection precedes the consumer module's evaluation. The stdio shell now imports a generated prelude (`agent-bundle/stdio-prelude`) as its first import: it calls `redirectConsoleToStderr` from `agent-bundle/mcp-entry`, then applies the operator .env layer with the server's manifest env defaults. Hook wrappers and the artifact CLI bin keep the env-only layer (`agent-bundle/launch-env-layer`) — they legitimately write stdout. The guard has one implementation: `redirectConsoleToStderr` returns the guard already installed (recognised by `process.stdout.write` still being its redirect) instead of stacking a second, which would capture the redirect as the original and restore stdout to stderr; the lifecycle adopts the prelude's guard and restores raw stdout from it before serving. Tests: a built stdio entry whose server module writes `console.log('hello')` and `process.stdout.write('raw\n')` at module scope, driven by a real stdio client through initialize, tools/list, and tools/call, asserts both land on stderr (fails on the previous code: stderr held only the factory-time line); the entry-shell unit tests pin the prelude as the stdio entry's first import and the env-only layer for hook wrappers and the CLI bin; the mcp-entry unit test pins guard adoption and re-install after restore. * fix(mcp): adopt the installed stdout guard regardless of write identity so a consumer wrapper cannot stack a second guard (#469) Adoption by identity (`process.stdout.write === redirectedWrite`) broke the moment a consumer module wrapped `process.stdout.write` at module scope: the lifecycle's `redirectConsoleToStderr()` saw a foreign function, installed a second guard with the wrapper recorded as the original, and restoring for the protocol stream handed stdout to the wrapper — which still forwarded to the first redirect, so every JSON-RPC frame left on stderr and the client hung in initialize. The rule is now: while a guard is installed, `redirectConsoleToStderr()` returns it whatever `process.stdout.write` has become; `restoreProtocolStdout()` restores the real original the guard owns, writes one stderr line if a module replaced the write in the meantime (the replacement is discarded — stdout is the protocol channel and wrapping it is unsupported), and clears the installed guard so a later call installs anew. Tests: the mcp-entry unit test wraps the redirect, adopts the same guard, restores to the real stdout, and installs fresh afterwards (fails on a677371 at the adoption step); the packed stdio test's server module now also wraps `process.stdout.write` at module scope and the real client still completes initialize, tools/list, and tools/call with the wrapper's output and the warning on stderr (hangs to timeout on a677371). * fix(mcp): make restoreProtocolStdout once-only so a stale or repeated restore cannot clobber a fresh guard (#469) Two holders of the same guard could restore twice: after the first restore and a fresh install, the stale restore overwrote the fresh redirect with the old original while `installedGuard` still named the fresh guard, so adoption returned a guard that was no longer installed. A plain double restore also emitted the foreign-wrapper warning twice. The guard now records that it has restored and returns immediately on later calls.

…#1364) Node 26 removed `--experimental-transform-types` (nodejs/node#61803), so on Node 26 the dev scripts in `packages/spotlight` exit before starting the sidecar: ``` $ pnpm dev:server node.exe: bad option: --experimental-transform-types ``` `dev` and `dev:server` now use `--experimental-strip-types`. The sidecar sources have no syntax that needs a transform (enums, namespaces, parameter properties), so stripping types is enough. The transform flag already required Node 22.7+, and `--experimental-strip-types` exists since 22.6, so this doesn't narrow which Node versions can run the dev scripts. Tested locally from `packages/spotlight` (`node --experimental-strip-types src/run.ts`, after `pnpm install`): - Node 26.12.0: before, exits immediately with `bad option`. After, `Spotlight listening on 8969`. - Node 22.12.0: `Spotlight listening on 8969`. Only these two dev scripts change. Builds and the published CLI are unaffected. The repo's lint-staged hook also reformatted the `"files"` array in the same `package.json` onto one line (`["dist"]`).

Node.js 26 removed the flag, so passing it now fails with "bad option" (exit code 9). Stop forwarding it to Node.js, drop it from the command-line switch docs, document the removal as a 46.0 breaking change, and remove the spec and fixture that exercised it. The spec that checks TypeScript-only syntax is not transformed stays. Ref: nodejs/node#61803 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU

* chore: bump node in DEPS to v26.12.1 * chore: bump node in DEPS to v26.13.0 * fix(patch): build: add simdutf dir to include path in GN build Keep upstream's simdutf include dir in the node_js2c include_dirs and add "tools" next to it. The patch's FixedArray::Get(context, i) change in node_builtins.cc is no longer needed: upstream made the same change in nodejs/node#61898. Ref: nodejs/node#65382 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): vfs: integrate with CJS and ESM module loaders Upstream moved the native legacyMainResolve call from esm/resolve.js into the loaderMethods table in internal/modules/helpers.js, so pass the asar-aware fileExists callback from there instead. Ref: nodejs/node#63653 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): src: let embedders supply a builtin code cache without a snapshot BuiltinLoader::RefreshCodeCache() now merges with insert_or_assign upstream, so drop that hunk from the patch. Ref: nodejs/node#65352 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): src: move import.meta initializer to native land PathHelpersLazyGetter moved from node_modules.cc to module_wrap.cc and gained a sibling ImportMetaResolveLazyGetter; switch both to Holder(). Also follow the DatabaseSyncLimits -> DatabaseLimits rename in node_sqlite.cc (nodejs/node#65988) and the Null(isolate) cleanup in node_constants.cc (nodejs/node#59805). Ref: nodejs/node#60603 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): src: fix usage of deprecated V8 API Upstream switched to Get/SetContinuationPreservedEmbedderDataV2(), which Chromium's V8 marks deprecated in favor of the unsuffixed names. Use the unsuffixed getter and setter. Ref: nodejs/node#60174 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): src: rename legacy trace event headers The TRACING_CATEGORY_NODE* macros moved from node_internals.h to tracing/trace_event.h, so define NODE_TRACE_CATEGORY_ENABLED there. Agent::CreateDefault() picks the legacy agent again when V8_USE_PERFETTO is unset (nodejs/node#64053), so only the Perfetto branch is gated on NODE_USE_PERFETTO_SDK. node.gyp now lists the Perfetto sources in node_tracing_perfetto_sources. Add them, with V8's tracing target and config, to the GN libnode target. The removed perfetto backport patch used to carry that wiring. Ref: nodejs/node#64565 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): src: let embedders exempt linked bindings from the addon permission Upstream added EnvironmentFlags::kNoAddonPermissionForLinkedBindings around the same check. Keep the electron_* exemption alongside it and note in the patch description that the patch can be dropped once Electron sets the new flag. Ref: nodejs/node#66067 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * chore: remove upstreamed patches * support_v8_sandboxed_pointers.patch Ref: nodejs/node#62237 * refactor_attach_cppgc_heap_on_v8_isolate_creation.patch Ref: nodejs/node#58070 * test_accomodate_v8_thenable_stack_trace_change_in_snapshot.patch Ref: nodejs/node#59805 * api_remove_deprecated_getisolate.patch Ref: nodejs/node#59805 Ref: nodejs/node#60223 * api_delete_deprecated_fields_on_v8_isolate.patch Ref: nodejs/node#60488 * api_promote_deprecation_of_v8_context_and_v8_object_api_methods.patch Ref: nodejs/node#60602 * src_refactor_module_wrap_cc_to_update_fixedarray_get_params.patch Ref: nodejs/node#61898 * src_refactor_wasmstreaming_finish_to_accept_a_callback.patch Ref: nodejs/node#61898 * src_stop_using_v8_propertycallbackinfo_t_this.patch Ref: nodejs/node#61898 * fix_add_externalpointertypetag_to_v8_external_api_calls.patch Ref: nodejs/node#61898 * src_stop_using_removed_v8_promise_reject_events.patch Ref: nodejs/node#62336 * src_backport_the_perfetto_trace_agent.patch Ref: nodejs/node#64565 Ref: nodejs/node#64721 * test_account_for_libc_rm_error_on_electron_linux.patch Ref: nodejs/node#66065 * fix_do_not_pass_removed_jspi_v8_flag.patch Ref: nodejs/node#60014 * test_drop_removed_wasm_rab_integration_flag.patch The v26 version of nodejs/node#65845 never added the flag. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * chore: update patches (trivial only) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * chore: bump @types/node to v26 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * chore: update filenames.auto.gni Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * chore: remove deleted specs Upstream deleted parallel/test-promise-swallowed-event and parallel/test-warn-multipleResolves when the multipleResolves event reached end-of-life. Ref: nodejs/node#58707 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: clear crash dumps with fs.rmSync instead of recursive rmdir Node.js v25 moved the fs.rmdir recursive option to end-of-life, so fs.rmdirSync(dir, { recursive: true }) now throws and @types/node v26 no longer accepts the option. Ref: nodejs/node#58616 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: listen on IncomingMessage unions through EventEmitter With @types/node v26, http.IncomingMessage declares typed `on` overloads that do not unify with Electron.IncomingMessage's, so collectStreamBodyBuffer() no longer type-checks when calling `on` on the union. Ref: Unable to locate reference Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: pass an ArrayBuffer-backed view to subtle.digest in asar wrapper @types/node v26 types webcrypto's BufferSource as views over an ArrayBuffer only, so a Uint8Array<ArrayBufferLike> no longer type-checks. Archive reads never use a SharedArrayBuffer, so narrow the type at the call site. Ref: Unable to locate reference Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * node#64565: src: rename legacy trace event headers Node.js renamed src/tracing/trace_event_common.h to trace_event_legacy.h and moved the legacy trace macros out of trace_event.h into trace_event_legacy_inl.h, so feed both to electron_generate_node_defines instead of the removed header. The generated push/undef set matches the previous one apart from header guards, NODE_TRACE_CATEGORY_ENABLED and INTERNAL_DECLARE_SET_TRACE_VALUE_INT, which Node.js no longer defines. Ref: nodejs/node#64565 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): src: add perfetto trace agent Upstream's trace_event_perfetto.h includes the amalgamated "perfetto.h" SDK header, which Chromium's Perfetto does not provide, so libnode fails to compile. Include the Perfetto public headers Chromium ships instead (as the removed backport patch did), add the tracing/trace_config includes agent_perfetto.h relied on, and skip Node's TrackEvent category definitions when Chromium's built-in categories are already in scope. Ref: nodejs/node#64565 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): src: handle indexed properties in `process.env` Upstream added indexed interceptors for process.env. Migrate the new EnvSetterIndexed and EnvDefinerIndexed callbacks to PropertyCallbackInfo<Boolean> like their named counterparts, since Chromium's V8 only accepts the Boolean-returning indexed setter and definer callbacks. Ref: nodejs/node#60826 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: read the whole heap snapshot stream Since Node.js 26, readable.read() returns one buffered chunk at a time instead of concatenating everything buffered, so takeHeapSnapshot() parsed only the first 64 KiB of the snapshot. Read readableLength bytes, as upstream test/common/heap.js now does. Ref: nodejs/node#60441 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: use --trace-warnings in the node-options-parsing crash case Node.js 26 throws ERR_MISSING_OPTION('--permission') when a boolean permission flag such as --allow-addons is set without --permission (previously only the fs flags were checked), so the utility process in this fixture exited before reaching options.js. Repeat --trace-warnings instead; the case still checks that disallowed NODE_OPTIONS are stripped from a list with repeated options. Ref: nodejs/node#62072 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): restore user timing trace events The usertiming.js hunk that emits node.perf.usertiming trace events for performance.mark() and performance.measure() lived in the perfetto backport patch, which was removed as upstreamed. Upstream Node.js never had it (Electron added it in #50591), so carry it in fix_use_embedder_perfetto_client_when_available.patch. Ref: #50591 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: trigger the V8 sandbox fatal error from V8 directly Since Node.js 26, node::Buffer::New() copies memory allocated outside the V8 sandbox into it instead of handing it to v8::ArrayBuffer::NewBackingStore(), so the external-ab addon no longer crashed the utility process and "emits 'error' when fatal error is triggered from V8" timed out. Call NewBackingStore() from the addon so the same v8_ArrayBuffer_NewBackingStore fatal error is still raised. Ref: nodejs/node#62237 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: delete directories with rm instead of recursive rmdir Node.js 26 removed the recursive option of fs.rmdir(), fs.rmdirSync() and fsPromises.rmdir() (DEP0147), so they now throw ERR_INVALID_ARG_VALUE. Use rmSync() and promises.rm(), and put an archive in the directory so the tests delete what their titles say. Ref: nodejs/node#58616 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * node#64894: worker: add support for Web Workers Node.js 26 adds a Web Worker API and its pre-execution deletes globalThis.Worker unless --experimental-web-worker is set, without checking noBrowserGlobals, so renderers and web workers with Node integration lost Blink's Worker. Pass the flag in those processes, as is already done for --experimental-eventsource; Node's own Worker is not installed under kNoBrowserGlobals. Ref: nodejs/node#64894 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * node#61803: module: remove --experimental-transform-types Node.js 26 removed the flag, so passing it now fails with "bad option" (exit code 9). Stop forwarding it to Node.js, drop it from the command-line switch docs, document the removal as a 46.0 breaking change, and remove the spec and fixture that exercised it. The spec that checks TypeScript-only syntax is not transformed stays. Ref: nodejs/node#61803 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * node#65589: src: disable V8 external memory reasonable size check Node.js now passes --external-memory-max-reasonable-size=0 to V8 in every process that initializes it, which changes FlagList::Hash() and so the code-cache key. Add the flag to the build-time js2c code cache flavors that initialize Node.js and to electron_xcache's default flags, so their caches keep being accepted (xcache.spec.ts failed on the flag hash). Ref: nodejs/node#65589 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: route Node.js 26's one-shot fs jobs through the asar wrapper Node.js 26 writes files (fs.writeFile, fs.appendFile and their fsPromises versions) with a native WriteFileJob, and copies directory trees for fs.cp() / fsPromises.cp() with a native CpDirJob, instead of going through binding.open() and the fs.promises walk the asar wrapper already handles. Writes into archives therefore failed with ENOTDIR instead of EACCES (and missed unpacked entries), and fs.cp() of an archive directory failed with ENOTDIR. Give binding.WriteFileJob the same archive checks as open(), wrap fs.writeFile() for archive paths since lib/fs.js captured the native job at startup, and copy archive trees for CpDirJob with the walk cpSync() already uses, which now also stamps directories with preserveTimestamps like the native walk. Ref: nodejs/node#65489 Ref: nodejs/node#65488 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): do not leak NodePlatform's default TracingController The NodePlatform hunk that owns the v8::TracingController it creates when the embedder passes none lived in the perfetto backport patch, which was removed as upstreamed, but upstream still allocates it with a bare new. With the perfetto agent, ELECTRON_RUN_AS_NODE passes no controller, so LeakSanitizer failed every run-as-node spec on ASan builds. Carry the hunk in fix_use_embedder_perfetto_client_when_available.patch. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: keep fsPromises.stat() and lstat() native outside archives The asar wrapper replaced fsPromises.stat() and fsPromises.lstat() with promisified callback versions for every path, so they ran as callback requests (FSREQCALLBACK) instead of fsPromises' own. Node.js 26's fsPromises.cp() checks its paths with them before handing the copy to a native job, and parallel/test-fs-cp-async-destination-appears-late tells the two apart. Use the wrapper only for paths inside archives. Ref: nodejs/node#65488 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: match Node.js 26 recursive readdir in the asar wrapper Node.js 26 accepts Buffer paths for recursive readdir, keeps parentPath a Buffer for them and for the 'buffer' encoding, and walks the tree breadth-first, following links to directories without stat()ing every entry. The asar wrapper's own walks (which also descend into archives) broke on Buffer paths and differed in order and parentPath, failing the new parallel/test-fs-readdir-recursive-* tests. Replace the three walks with one breadth-first walk shaped like Node's, which still lists and descends into archives as directories, and lets errors from any directory reach the caller. Ref: nodejs/node#65487 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * node#66138: src: add --process-timeout=N Node.js starts the --process-timeout watchdog in NodeMainInstance::Run(), which ELECTRON_RUN_AS_NODE does not use, so the option was accepted but never applied: parallel/test-process-timeout-* failed, and the fixtures that block the main thread hung the Node.js test job until it was cancelled. Start and stop the watchdog around the event loop in NodeMain() the same way. Ref: nodejs/node#66138 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: close the FileHandle in the stale asar descriptor spec Node.js 26 made closing a FileHandle on garbage collection an error (DEP0137 end-of-life). The 'handle' variant of "does not capture a descriptor number that was closed behind its back" dropped its FileHandle, so the renderer running the asar specs later died with an uncaught ERR_INVALID_STATE and every remaining asar spec failed with "Render frame was disposed". Close the handle like the other variants close their descriptors. Ref: nodejs/node#58536 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: allow ArrayBuffers above 2 GiB in Node.js-enabled isolates Node.js 26 allocates Buffer.allocUnsafe(), Buffer.allocUnsafeSlow() and the copies made by AdoptIntoBackingStore() through the isolate's ArrayBuffer::Allocator (nodejs/node#61480, nodejs/node#62237). Electron's sandboxed-pointers patch for Node.js 24 sent those through ArrayBuffer::Allocator::NewDefaultAllocator() instead, which reaches Electron's in-sandbox allocator, and that serves requests beyond PartitionAlloc's single-allocation cap (2 GiB - 2 MiB) from sandbox pages. The isolates of the browser process, the utility-process node service and ELECTRON_RUN_AS_NODE use gin's ArrayBufferAllocator, which has no such fallback, so with Node.js 26 a Buffer.allocUnsafe() of more than 2 GiB - 2 MiB in those processes failed with ERR_MEMORY_ALLOCATION_FAILED, where Electron 45 and 46 allowed up to 32 GiB - 1 (parallel/test-fs-write-buffer-large). Give those isolates an allocator that delegates to gin's partition up to the cap and serves larger requests from sandbox pages, sharing the page bookkeeping with the in-sandbox allocator. Buffer.alloc() and new ArrayBuffer() above the cap now work there as well. The page map is only locked while a page allocation is live. Renderers are unaffected: Blink's isolate allocator already allows giga allocations. Ref: nodejs/node#61480 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: include node_watchdog.h only where --process-timeout is wired node_watchdog.h chooses between its POSIX and Windows members with Node's __POSIX__ define, which Node's own targets set but Electron's do not, so including it from node_includes.h broke every electron_lib source on Linux ("unknown type name 'BOOL'"). Include it in node_main.cc alone, with __POSIX__ set around it on POSIX so the class matches Node's own build. Ref: nodejs/node#66138 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: let Node.js 26's virtual file system answer recursive readdir Node.js 26 adds an experimental virtual file system that fs functions consult (vfsState.handlers) before the real one. The asar wrapper walks recursive readdir itself and never asked it, so with a VFS mounted fs.readdir(..., { recursive: true }) and its sync and promise forms listed the real path behind the mount (parallel/test-vfs-reserved-root, parallel/test-vfs-zip-provider-readdir-recursive failed with ENOTDIR). Ask the VFS first, as Node does. Ref: nodejs/node#63537 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: do not log terminated Electron init scripts as errors CompileAndCall() logged an error whenever running one of Electron's bundled init scripts threw, including when execution was terminated because the environment was being torn down. A Worker that is stopped, or whose process exits, while it is still bootstrapping is not an error. On Node.js 26 the unref()'d Worker spawned by module-hooks/test-async-loader-hooks-with-worker-permission-allowed reaches that point while node_init runs, and the log line broke the test's empty-stderr check. Only log real exceptions. Ref: Unable to locate reference Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix(patch): expect the Wasm Global setter WPT to pass with Chromium V8 Chromium's V8 passes "Calling setter without argument" in the Wasm jsapi WPT global/value-get-set.any.js, which Node's own V8 still fails and lists as an expected failure, so the run failed on an unexpected pass. Ref: Unable to locate reference Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: honor --no-node-snapshot under ELECTRON_RUN_AS_NODE ELECTRON_RUN_AS_NODE accepted --no-node-snapshot but always created its isolate from the embedded Node startup snapshot, and Node worker threads inherit that snapshot from their parent. Since Node.js 26 a worker deserializes the whole bootstrapped context from it, so parallel/test-worker-nearheaplimit-deadlock, which passes --no-node-snapshot so that its 0 MB-heap worker does not run out of memory during that deserialization, crashed with SIGSEGV. Skip the snapshot when Node's options turn it off, as node::Start() does; the process then boots from the V8 context snapshot like a build without a Node snapshot. The browser process and the node service parse Node's options only after creating their isolate, so they are unaffected. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * build: fetch Perfetto trace_processor_shell for the Node.js tests nodejs/node#66068 (test: enable trace_events tests with perfetto) moved the trace_events tests from test/parallel into a new test/trace_events suite. When Node.js is built with Perfetto (process.config.variables .v8_use_perfetto, which Electron's build sets), those tests write binary .pftrace files and convert them back to JSON with Perfetto's trace_processor_shell, through test/common/trace_events.js. Without the tool, all 17 trace_events/* tests in nn-test fail with "trace_processor_shell is missing". The same PR added tools/perfetto/get_trace_processor, which upstream's Makefile and CI use to fetch it. Run that script in the Node.js test job. It downloads the Perfetto release matching the vendored deps/perfetto/VERSION (v58.2) from Perfetto's GitHub releases to tools/perfetto/trace_processor_shell, which is the default path the test helper resolves, so TRACE_PROCESSOR_SHELL_PATH is not needed. Cache the binary keyed on that VERSION file. Upstream pins only the version, not a checksum. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: disable Node.js source phase import and process-timeout report tests Electron always passes --no-js-source-phase-imports. In renderers Blink owns that V8 flag and freezes it before Node.js starts, and the other process types match it, so the flag is consistent across processes and the embedded code cache, which is keyed on V8's flag hash, stays valid. nodejs/node#60364 made Node.js honour that opt-out instead of always enabling the flag. The es-module/test-esm-wasm-*source-phase* tests are already disabled for the same reason. Node.js 26 brings more tests that need source phase imports: - es-module/test-esm-wasm-jitless (nodejs/node#66533) calls import.source() and fails with "SyntaxError: Unexpected identifier 'source'". - The wasm/jsapi WPT suite, test/wpt/test-wasm-jsapi.mjs (nodejs/node#59034), is new to Electron with this upgrade. Its esm-integration/source-phase*.tentative cases use import.source() and fail with the same SyntaxError. Its module/moduleSource.tentative cases expect the %AbstractModuleSource% intrinsic, which V8 only installs when the flag is on. Each case runs as an .any.html and an .any.worker.html variant, so that is 8 entries. nodejs/node#66138 added --process-timeout and report/test-report-process-timeout. nodejs/node#66428 split out report/test-report-process-timeout-worker. Both validate the generated report with test/common/report.js, which expects header.componentVersions to equal process.versions. In Electron, process.versions also has chrome and electron, so the comparison fails. The other report tests that use this helper are already disabled for the same reason. The process timeout itself works, and both tests get as far as the report check. The WPT cases are listed as wpt/<test file>/<case id>, which is how tools/test.py names them and matches --skip-tests. Teach --validateDisabled to check the test file for these entries. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: resolve V8's NewBackingStore by Electron's export name on Windows The external-ab addon now calls v8::ArrayBuffer::NewBackingStore() itself, but on Windows it is built with MSVC's standard library while Electron is built with Chromium's libc++, and MSVC mangles return types into symbol names. electron.exe exports the function as returning std::__Cr::unique_ptr, so the addon's import of the std::unique_ptr variant never resolved, the utility process died without a V8 fatal error and "emits 'error' when fatal error is triggered from V8" timed out on windows-x64. Look the function up by Electron's exported name there instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: only reference std-free V8 exports from the external-ab addon On macOS and Linux the addon is built against a different libc++ ABI namespace than Electron's std::__Cr, so v8::ArrayBuffer::New(isolate, std::shared_ptr<BackingStore>) is an import that cannot bind. It only went unnoticed because NewBackingStore() raises the fatal error first. Stop after NewBackingStore(), whose mangled name has no standard library types, and throw if it ever returns, as on Windows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * Revert "build: fetch Perfetto trace_processor_shell for the Node.js tests" This reverts commit 8920a7c. Node.js uses trace_processor_shell only to read back the trace file that a process started with --trace-event-categories writes. Electron never writes that file: Chromium owns tracing, and Electron does not start Node.js's file-based tracing agent. Every trace_events/* test fails before it reaches trace_processor_shell, because no trace file exists or, in test-trace-events-api, because getEnabledCategories() is undefined in the child. Those tests are disabled again in the next commit, so no Node.js test needs trace_processor_shell, and downloading it would only add a network fetch to the job. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: disable the Node.js trace_events tests in their new directory nodejs/node#66068 (test: enable trace_events tests with perfetto) moved 17 test-trace-events-* tests out of test/parallel into a new test/trace_events suite. That took them out from under Electron's "parallel/test-trace-events" entry, which --skip-tests matches as a substring. Electron has never run these tests. They were disabled one by one when Electron started running the Node.js tests (1d06f67) and folded into the parallel/ prefix in #49467. They start Node.js with --trace-event-categories and read back the trace file it writes. Electron does not start Node.js's file-based tracing agent, because Chromium owns tracing, so no trace file is ever written. In nn-test, all 17 fail on a missing trace file, and test-trace-events-api also finds getEnabledCategories() undefined in its child. Add the matching "trace_events/test-trace-events" prefix. It covers all 17 files in test/trace_events. Keep the parallel/ entry, which still matches the 13 test-trace-events-* tests that stayed in test/parallel. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: copy directory trees out of archives without blocking fs.cp() nodejs/node#65488 (fs: copy directory trees for fs.cp() on the thread pool) made fs.cp() and fsPromises.cp() copy a new destination directory with a native CpDirJob. The job walks the tree on the thread pool and calls ondone once. Electron's override for archive sources ran the synchronous fs.cpSync() walk in a single process.nextTick, so an asynchronous cp of a directory out of an archive blocked the event loop until the whole tree had been copied. Give the job an asynchronous version of the walk. It goes through the fs.promises overrides, so packed files use the existing asynchronous copyFile path, and it keeps the same ordering, force/errorOnExist, symlink and preserveTimestamps handling and errors as the sync walk. The job still reports through ondone exactly once. fs.cpSync() keeps the synchronous walk. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: trigger the utility process V8 fatal error from the testing binding The utilityProcess 'error' event from #43774 reports any V8 fatal error. Its spec raised one with the external-ab native addon, which handed V8 memory from outside the sandbox. Node.js 26 makes every Buffer site sandbox compatible, so that no longer happens through Node.js APIs. The earlier commits on this branch made the addon call V8 directly, and that needed ABI workarounds on Windows and for libc++. As Deepak suggested in review, drop the addon and its workspace entries. Add triggerV8FatalError() to the electron_common_testing binding, which is only built when DCHECKs are on. It fails a V8 API check by setting an internal field on a plain object, so it does not depend on the V8 sandbox. The spec keeps #43774's checks on the event type, location and report trigger, also checks the report's event message, and now runs wherever the testing binding exists, including Windows on Arm. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * test: wait for the utility process exit only after its 'error' event events.once(emitter, 'exit') also listens for 'error' and rejects with its first argument. The previous commit started waiting for 'exit' before the 'error' event arrived, so the utility process's 'FatalError' rejected that promise and the spec failed with "Unknown Error: FatalError". Wait for 'exit' after 'error' has been received, as the spec from #43774 did. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * refactor: use one d8-style PartitionAlloc allocator inside the V8 sandbox Following Deepak's review, replace the sandbox-page fallback added for nodejs/node#61480 and the older InSandboxAllocator with a single PAInSandboxAllocator, modelled on d8's. nodejs/node#61480 moved Buffer.allocUnsafe() onto the isolate's ArrayBuffer allocator. gin's partition caps a single allocation at partition_alloc::MaxAllocationSize() (2 GiB - 2 MiB), so Buffers above that size failed. The previous fix served those requests from whole sandbox pages and tracked them in a separate map. PAInSandboxAllocator owns a PartitionAlloc partition in the configurable pool, which gin::V8Initializer::Initialize() places inside the V8 sandbox. The partition is created lazily, after the pool exists. It has allow_giga_allocations set and allocates with kAllowGigaAllocations, along with kNoOverrideHooks and kNoMemoryToolOverride, as gin does. A single ArrayBuffer can therefore reach MaxGigaAllocationSize() (16 GiB - 2 MiB on 64-bit) without the page fallback. It is both the ArrayBuffer allocator for JavascriptEnvironment isolates and V8's in-sandbox allocator, so Node.js worker_threads isolates use it through NewDefaultAllocator(). Builds without the V8 sandbox keep gin's allocator. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU * fix: reject instead of throwing in asar fs.promises.stat/lstat wrappers The new fs.promises.stat/lstat wrappers call splitPath() before deciding whether to use fsPromises' own implementation. For a URL, splitPath() runs getValidatedPath(), which throws for a non-file: URL, so the error escaped synchronously instead of rejecting the returned promise as it did with util.promisify(fs.stat) on main and as Node's fsPromises does. Make the wrappers async so every input error rejects, and add a spec for it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU --------- Co-authored-by: electron-roller[bot] <84116207+electron-roller[bot]@users.noreply.github.com> Co-authored-by: Claude <noreply@anthropic.com>

* chore: bump node in DEPS to v26.12.1 Co-authored-by: electron-roller[bot] <84116207+electron-roller[bot]@users.noreply.github.com> * chore: bump node in DEPS to v26.13.0 Co-authored-by: electron-roller[bot] <84116207+electron-roller[bot]@users.noreply.github.com> * fix(patch): build: add simdutf dir to include path in GN build Keep upstream's simdutf include dir in the node_js2c include_dirs and add "tools" next to it. The patch's FixedArray::Get(context, i) change in node_builtins.cc is no longer needed: upstream made the same change in nodejs/node#61898. Ref: nodejs/node#65382 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): vfs: integrate with CJS and ESM module loaders Upstream moved the native legacyMainResolve call from esm/resolve.js into the loaderMethods table in internal/modules/helpers.js, so pass the asar-aware fileExists callback from there instead. Ref: nodejs/node#63653 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): src: let embedders supply a builtin code cache without a snapshot BuiltinLoader::RefreshCodeCache() now merges with insert_or_assign upstream, so drop that hunk from the patch. Ref: nodejs/node#65352 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): src: move import.meta initializer to native land PathHelpersLazyGetter moved from node_modules.cc to module_wrap.cc and gained a sibling ImportMetaResolveLazyGetter; switch both to Holder(). Also follow the DatabaseSyncLimits -> DatabaseLimits rename in node_sqlite.cc (nodejs/node#65988) and the Null(isolate) cleanup in node_constants.cc (nodejs/node#59805). Ref: nodejs/node#60603 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): src: fix usage of deprecated V8 API Upstream switched to Get/SetContinuationPreservedEmbedderDataV2(), which Chromium's V8 marks deprecated in favor of the unsuffixed names. Use the unsuffixed getter and setter. Ref: nodejs/node#60174 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): src: rename legacy trace event headers The TRACING_CATEGORY_NODE* macros moved from node_internals.h to tracing/trace_event.h, so define NODE_TRACE_CATEGORY_ENABLED there. Agent::CreateDefault() picks the legacy agent again when V8_USE_PERFETTO is unset (nodejs/node#64053), so only the Perfetto branch is gated on NODE_USE_PERFETTO_SDK. node.gyp now lists the Perfetto sources in node_tracing_perfetto_sources. Add them, with V8's tracing target and config, to the GN libnode target. The removed perfetto backport patch used to carry that wiring. Ref: nodejs/node#64565 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): src: let embedders exempt linked bindings from the addon permission Upstream added EnvironmentFlags::kNoAddonPermissionForLinkedBindings around the same check. Keep the electron_* exemption alongside it and note in the patch description that the patch can be dropped once Electron sets the new flag. Ref: nodejs/node#66067 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * chore: remove upstreamed patches * support_v8_sandboxed_pointers.patch Ref: nodejs/node#62237 * refactor_attach_cppgc_heap_on_v8_isolate_creation.patch Ref: nodejs/node#58070 * test_accomodate_v8_thenable_stack_trace_change_in_snapshot.patch Ref: nodejs/node#59805 * api_remove_deprecated_getisolate.patch Ref: nodejs/node#59805 Ref: nodejs/node#60223 * api_delete_deprecated_fields_on_v8_isolate.patch Ref: nodejs/node#60488 * api_promote_deprecation_of_v8_context_and_v8_object_api_methods.patch Ref: nodejs/node#60602 * src_refactor_module_wrap_cc_to_update_fixedarray_get_params.patch Ref: nodejs/node#61898 * src_refactor_wasmstreaming_finish_to_accept_a_callback.patch Ref: nodejs/node#61898 * src_stop_using_v8_propertycallbackinfo_t_this.patch Ref: nodejs/node#61898 * fix_add_externalpointertypetag_to_v8_external_api_calls.patch Ref: nodejs/node#61898 * src_stop_using_removed_v8_promise_reject_events.patch Ref: nodejs/node#62336 * src_backport_the_perfetto_trace_agent.patch Ref: nodejs/node#64565 Ref: nodejs/node#64721 * test_account_for_libc_rm_error_on_electron_linux.patch Ref: nodejs/node#66065 * fix_do_not_pass_removed_jspi_v8_flag.patch Ref: nodejs/node#60014 * test_drop_removed_wasm_rab_integration_flag.patch The v26 version of nodejs/node#65845 never added the flag. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * chore: update patches (trivial only) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * chore: bump @types/node to v26 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * chore: update filenames.auto.gni Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * chore: remove deleted specs Upstream deleted parallel/test-promise-swallowed-event and parallel/test-warn-multipleResolves when the multipleResolves event reached end-of-life. Ref: nodejs/node#58707 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: clear crash dumps with fs.rmSync instead of recursive rmdir Node.js v25 moved the fs.rmdir recursive option to end-of-life, so fs.rmdirSync(dir, { recursive: true }) now throws and @types/node v26 no longer accepts the option. Ref: nodejs/node#58616 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: listen on IncomingMessage unions through EventEmitter With @types/node v26, http.IncomingMessage declares typed `on` overloads that do not unify with Electron.IncomingMessage's, so collectStreamBodyBuffer() no longer type-checks when calling `on` on the union. Ref: Unable to locate reference Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: pass an ArrayBuffer-backed view to subtle.digest in asar wrapper @types/node v26 types webcrypto's BufferSource as views over an ArrayBuffer only, so a Uint8Array<ArrayBufferLike> no longer type-checks. Archive reads never use a SharedArrayBuffer, so narrow the type at the call site. Ref: Unable to locate reference Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * node#64565: src: rename legacy trace event headers Node.js renamed src/tracing/trace_event_common.h to trace_event_legacy.h and moved the legacy trace macros out of trace_event.h into trace_event_legacy_inl.h, so feed both to electron_generate_node_defines instead of the removed header. The generated push/undef set matches the previous one apart from header guards, NODE_TRACE_CATEGORY_ENABLED and INTERNAL_DECLARE_SET_TRACE_VALUE_INT, which Node.js no longer defines. Ref: nodejs/node#64565 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): src: add perfetto trace agent Upstream's trace_event_perfetto.h includes the amalgamated "perfetto.h" SDK header, which Chromium's Perfetto does not provide, so libnode fails to compile. Include the Perfetto public headers Chromium ships instead (as the removed backport patch did), add the tracing/trace_config includes agent_perfetto.h relied on, and skip Node's TrackEvent category definitions when Chromium's built-in categories are already in scope. Ref: nodejs/node#64565 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): src: handle indexed properties in `process.env` Upstream added indexed interceptors for process.env. Migrate the new EnvSetterIndexed and EnvDefinerIndexed callbacks to PropertyCallbackInfo<Boolean> like their named counterparts, since Chromium's V8 only accepts the Boolean-returning indexed setter and definer callbacks. Ref: nodejs/node#60826 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: read the whole heap snapshot stream Since Node.js 26, readable.read() returns one buffered chunk at a time instead of concatenating everything buffered, so takeHeapSnapshot() parsed only the first 64 KiB of the snapshot. Read readableLength bytes, as upstream test/common/heap.js now does. Ref: nodejs/node#60441 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: use --trace-warnings in the node-options-parsing crash case Node.js 26 throws ERR_MISSING_OPTION('--permission') when a boolean permission flag such as --allow-addons is set without --permission (previously only the fs flags were checked), so the utility process in this fixture exited before reaching options.js. Repeat --trace-warnings instead; the case still checks that disallowed NODE_OPTIONS are stripped from a list with repeated options. Ref: nodejs/node#62072 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): restore user timing trace events The usertiming.js hunk that emits node.perf.usertiming trace events for performance.mark() and performance.measure() lived in the perfetto backport patch, which was removed as upstreamed. Upstream Node.js never had it (Electron added it in #50591), so carry it in fix_use_embedder_perfetto_client_when_available.patch. Ref: #50591 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: trigger the V8 sandbox fatal error from V8 directly Since Node.js 26, node::Buffer::New() copies memory allocated outside the V8 sandbox into it instead of handing it to v8::ArrayBuffer::NewBackingStore(), so the external-ab addon no longer crashed the utility process and "emits 'error' when fatal error is triggered from V8" timed out. Call NewBackingStore() from the addon so the same v8_ArrayBuffer_NewBackingStore fatal error is still raised. Ref: nodejs/node#62237 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: delete directories with rm instead of recursive rmdir Node.js 26 removed the recursive option of fs.rmdir(), fs.rmdirSync() and fsPromises.rmdir() (DEP0147), so they now throw ERR_INVALID_ARG_VALUE. Use rmSync() and promises.rm(), and put an archive in the directory so the tests delete what their titles say. Ref: nodejs/node#58616 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * node#64894: worker: add support for Web Workers Node.js 26 adds a Web Worker API and its pre-execution deletes globalThis.Worker unless --experimental-web-worker is set, without checking noBrowserGlobals, so renderers and web workers with Node integration lost Blink's Worker. Pass the flag in those processes, as is already done for --experimental-eventsource; Node's own Worker is not installed under kNoBrowserGlobals. Ref: nodejs/node#64894 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * node#61803: module: remove --experimental-transform-types Node.js 26 removed the flag, so passing it now fails with "bad option" (exit code 9). Stop forwarding it to Node.js, drop it from the command-line switch docs, document the removal as a 46.0 breaking change, and remove the spec and fixture that exercised it. The spec that checks TypeScript-only syntax is not transformed stays. Ref: nodejs/node#61803 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * node#65589: src: disable V8 external memory reasonable size check Node.js now passes --external-memory-max-reasonable-size=0 to V8 in every process that initializes it, which changes FlagList::Hash() and so the code-cache key. Add the flag to the build-time js2c code cache flavors that initialize Node.js and to electron_xcache's default flags, so their caches keep being accepted (xcache.spec.ts failed on the flag hash). Ref: nodejs/node#65589 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: route Node.js 26's one-shot fs jobs through the asar wrapper Node.js 26 writes files (fs.writeFile, fs.appendFile and their fsPromises versions) with a native WriteFileJob, and copies directory trees for fs.cp() / fsPromises.cp() with a native CpDirJob, instead of going through binding.open() and the fs.promises walk the asar wrapper already handles. Writes into archives therefore failed with ENOTDIR instead of EACCES (and missed unpacked entries), and fs.cp() of an archive directory failed with ENOTDIR. Give binding.WriteFileJob the same archive checks as open(), wrap fs.writeFile() for archive paths since lib/fs.js captured the native job at startup, and copy archive trees for CpDirJob with the walk cpSync() already uses, which now also stamps directories with preserveTimestamps like the native walk. Ref: nodejs/node#65489 Ref: nodejs/node#65488 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): do not leak NodePlatform's default TracingController The NodePlatform hunk that owns the v8::TracingController it creates when the embedder passes none lived in the perfetto backport patch, which was removed as upstreamed, but upstream still allocates it with a bare new. With the perfetto agent, ELECTRON_RUN_AS_NODE passes no controller, so LeakSanitizer failed every run-as-node spec on ASan builds. Carry the hunk in fix_use_embedder_perfetto_client_when_available.patch. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: keep fsPromises.stat() and lstat() native outside archives The asar wrapper replaced fsPromises.stat() and fsPromises.lstat() with promisified callback versions for every path, so they ran as callback requests (FSREQCALLBACK) instead of fsPromises' own. Node.js 26's fsPromises.cp() checks its paths with them before handing the copy to a native job, and parallel/test-fs-cp-async-destination-appears-late tells the two apart. Use the wrapper only for paths inside archives. Ref: nodejs/node#65488 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: match Node.js 26 recursive readdir in the asar wrapper Node.js 26 accepts Buffer paths for recursive readdir, keeps parentPath a Buffer for them and for the 'buffer' encoding, and walks the tree breadth-first, following links to directories without stat()ing every entry. The asar wrapper's own walks (which also descend into archives) broke on Buffer paths and differed in order and parentPath, failing the new parallel/test-fs-readdir-recursive-* tests. Replace the three walks with one breadth-first walk shaped like Node's, which still lists and descends into archives as directories, and lets errors from any directory reach the caller. Ref: nodejs/node#65487 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * node#66138: src: add --process-timeout=N Node.js starts the --process-timeout watchdog in NodeMainInstance::Run(), which ELECTRON_RUN_AS_NODE does not use, so the option was accepted but never applied: parallel/test-process-timeout-* failed, and the fixtures that block the main thread hung the Node.js test job until it was cancelled. Start and stop the watchdog around the event loop in NodeMain() the same way. Ref: nodejs/node#66138 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: close the FileHandle in the stale asar descriptor spec Node.js 26 made closing a FileHandle on garbage collection an error (DEP0137 end-of-life). The 'handle' variant of "does not capture a descriptor number that was closed behind its back" dropped its FileHandle, so the renderer running the asar specs later died with an uncaught ERR_INVALID_STATE and every remaining asar spec failed with "Render frame was disposed". Close the handle like the other variants close their descriptors. Ref: nodejs/node#58536 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: allow ArrayBuffers above 2 GiB in Node.js-enabled isolates Node.js 26 allocates Buffer.allocUnsafe(), Buffer.allocUnsafeSlow() and the copies made by AdoptIntoBackingStore() through the isolate's ArrayBuffer::Allocator (nodejs/node#61480, nodejs/node#62237). Electron's sandboxed-pointers patch for Node.js 24 sent those through ArrayBuffer::Allocator::NewDefaultAllocator() instead, which reaches Electron's in-sandbox allocator, and that serves requests beyond PartitionAlloc's single-allocation cap (2 GiB - 2 MiB) from sandbox pages. The isolates of the browser process, the utility-process node service and ELECTRON_RUN_AS_NODE use gin's ArrayBufferAllocator, which has no such fallback, so with Node.js 26 a Buffer.allocUnsafe() of more than 2 GiB - 2 MiB in those processes failed with ERR_MEMORY_ALLOCATION_FAILED, where Electron 45 and 46 allowed up to 32 GiB - 1 (parallel/test-fs-write-buffer-large). Give those isolates an allocator that delegates to gin's partition up to the cap and serves larger requests from sandbox pages, sharing the page bookkeeping with the in-sandbox allocator. Buffer.alloc() and new ArrayBuffer() above the cap now work there as well. The page map is only locked while a page allocation is live. Renderers are unaffected: Blink's isolate allocator already allows giga allocations. Ref: nodejs/node#61480 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: include node_watchdog.h only where --process-timeout is wired node_watchdog.h chooses between its POSIX and Windows members with Node's __POSIX__ define, which Node's own targets set but Electron's do not, so including it from node_includes.h broke every electron_lib source on Linux ("unknown type name 'BOOL'"). Include it in node_main.cc alone, with __POSIX__ set around it on POSIX so the class matches Node's own build. Ref: nodejs/node#66138 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: let Node.js 26's virtual file system answer recursive readdir Node.js 26 adds an experimental virtual file system that fs functions consult (vfsState.handlers) before the real one. The asar wrapper walks recursive readdir itself and never asked it, so with a VFS mounted fs.readdir(..., { recursive: true }) and its sync and promise forms listed the real path behind the mount (parallel/test-vfs-reserved-root, parallel/test-vfs-zip-provider-readdir-recursive failed with ENOTDIR). Ask the VFS first, as Node does. Ref: nodejs/node#63537 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: do not log terminated Electron init scripts as errors CompileAndCall() logged an error whenever running one of Electron's bundled init scripts threw, including when execution was terminated because the environment was being torn down. A Worker that is stopped, or whose process exits, while it is still bootstrapping is not an error. On Node.js 26 the unref()'d Worker spawned by module-hooks/test-async-loader-hooks-with-worker-permission-allowed reaches that point while node_init runs, and the log line broke the test's empty-stderr check. Only log real exceptions. Ref: Unable to locate reference Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix(patch): expect the Wasm Global setter WPT to pass with Chromium V8 Chromium's V8 passes "Calling setter without argument" in the Wasm jsapi WPT global/value-get-set.any.js, which Node's own V8 still fails and lists as an expected failure, so the run failed on an unexpected pass. Ref: Unable to locate reference Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: honor --no-node-snapshot under ELECTRON_RUN_AS_NODE ELECTRON_RUN_AS_NODE accepted --no-node-snapshot but always created its isolate from the embedded Node startup snapshot, and Node worker threads inherit that snapshot from their parent. Since Node.js 26 a worker deserializes the whole bootstrapped context from it, so parallel/test-worker-nearheaplimit-deadlock, which passes --no-node-snapshot so that its 0 MB-heap worker does not run out of memory during that deserialization, crashed with SIGSEGV. Skip the snapshot when Node's options turn it off, as node::Start() does; the process then boots from the V8 context snapshot like a build without a Node snapshot. The browser process and the node service parse Node's options only after creating their isolate, so they are unaffected. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * build: fetch Perfetto trace_processor_shell for the Node.js tests nodejs/node#66068 (test: enable trace_events tests with perfetto) moved the trace_events tests from test/parallel into a new test/trace_events suite. When Node.js is built with Perfetto (process.config.variables .v8_use_perfetto, which Electron's build sets), those tests write binary .pftrace files and convert them back to JSON with Perfetto's trace_processor_shell, through test/common/trace_events.js. Without the tool, all 17 trace_events/* tests in nn-test fail with "trace_processor_shell is missing". The same PR added tools/perfetto/get_trace_processor, which upstream's Makefile and CI use to fetch it. Run that script in the Node.js test job. It downloads the Perfetto release matching the vendored deps/perfetto/VERSION (v58.2) from Perfetto's GitHub releases to tools/perfetto/trace_processor_shell, which is the default path the test helper resolves, so TRACE_PROCESSOR_SHELL_PATH is not needed. Cache the binary keyed on that VERSION file. Upstream pins only the version, not a checksum. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: disable Node.js source phase import and process-timeout report tests Electron always passes --no-js-source-phase-imports. In renderers Blink owns that V8 flag and freezes it before Node.js starts, and the other process types match it, so the flag is consistent across processes and the embedded code cache, which is keyed on V8's flag hash, stays valid. nodejs/node#60364 made Node.js honour that opt-out instead of always enabling the flag. The es-module/test-esm-wasm-*source-phase* tests are already disabled for the same reason. Node.js 26 brings more tests that need source phase imports: - es-module/test-esm-wasm-jitless (nodejs/node#66533) calls import.source() and fails with "SyntaxError: Unexpected identifier 'source'". - The wasm/jsapi WPT suite, test/wpt/test-wasm-jsapi.mjs (nodejs/node#59034), is new to Electron with this upgrade. Its esm-integration/source-phase*.tentative cases use import.source() and fail with the same SyntaxError. Its module/moduleSource.tentative cases expect the %AbstractModuleSource% intrinsic, which V8 only installs when the flag is on. Each case runs as an .any.html and an .any.worker.html variant, so that is 8 entries. nodejs/node#66138 added --process-timeout and report/test-report-process-timeout. nodejs/node#66428 split out report/test-report-process-timeout-worker. Both validate the generated report with test/common/report.js, which expects header.componentVersions to equal process.versions. In Electron, process.versions also has chrome and electron, so the comparison fails. The other report tests that use this helper are already disabled for the same reason. The process timeout itself works, and both tests get as far as the report check. The WPT cases are listed as wpt/<test file>/<case id>, which is how tools/test.py names them and matches --skip-tests. Teach --validateDisabled to check the test file for these entries. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: resolve V8's NewBackingStore by Electron's export name on Windows The external-ab addon now calls v8::ArrayBuffer::NewBackingStore() itself, but on Windows it is built with MSVC's standard library while Electron is built with Chromium's libc++, and MSVC mangles return types into symbol names. electron.exe exports the function as returning std::__Cr::unique_ptr, so the addon's import of the std::unique_ptr variant never resolved, the utility process died without a V8 fatal error and "emits 'error' when fatal error is triggered from V8" timed out on windows-x64. Look the function up by Electron's exported name there instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: only reference std-free V8 exports from the external-ab addon On macOS and Linux the addon is built against a different libc++ ABI namespace than Electron's std::__Cr, so v8::ArrayBuffer::New(isolate, std::shared_ptr<BackingStore>) is an import that cannot bind. It only went unnoticed because NewBackingStore() raises the fatal error first. Stop after NewBackingStore(), whose mangled name has no standard library types, and throw if it ever returns, as on Windows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * Revert "build: fetch Perfetto trace_processor_shell for the Node.js tests" This reverts commit 8920a7c. Node.js uses trace_processor_shell only to read back the trace file that a process started with --trace-event-categories writes. Electron never writes that file: Chromium owns tracing, and Electron does not start Node.js's file-based tracing agent. Every trace_events/* test fails before it reaches trace_processor_shell, because no trace file exists or, in test-trace-events-api, because getEnabledCategories() is undefined in the child. Those tests are disabled again in the next commit, so no Node.js test needs trace_processor_shell, and downloading it would only add a network fetch to the job. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: disable the Node.js trace_events tests in their new directory nodejs/node#66068 (test: enable trace_events tests with perfetto) moved 17 test-trace-events-* tests out of test/parallel into a new test/trace_events suite. That took them out from under Electron's "parallel/test-trace-events" entry, which --skip-tests matches as a substring. Electron has never run these tests. They were disabled one by one when Electron started running the Node.js tests (1d06f67) and folded into the parallel/ prefix in #49467. They start Node.js with --trace-event-categories and read back the trace file it writes. Electron does not start Node.js's file-based tracing agent, because Chromium owns tracing, so no trace file is ever written. In nn-test, all 17 fail on a missing trace file, and test-trace-events-api also finds getEnabledCategories() undefined in its child. Add the matching "trace_events/test-trace-events" prefix. It covers all 17 files in test/trace_events. Keep the parallel/ entry, which still matches the 13 test-trace-events-* tests that stayed in test/parallel. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: copy directory trees out of archives without blocking fs.cp() nodejs/node#65488 (fs: copy directory trees for fs.cp() on the thread pool) made fs.cp() and fsPromises.cp() copy a new destination directory with a native CpDirJob. The job walks the tree on the thread pool and calls ondone once. Electron's override for archive sources ran the synchronous fs.cpSync() walk in a single process.nextTick, so an asynchronous cp of a directory out of an archive blocked the event loop until the whole tree had been copied. Give the job an asynchronous version of the walk. It goes through the fs.promises overrides, so packed files use the existing asynchronous copyFile path, and it keeps the same ordering, force/errorOnExist, symlink and preserveTimestamps handling and errors as the sync walk. The job still reports through ondone exactly once. fs.cpSync() keeps the synchronous walk. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: trigger the utility process V8 fatal error from the testing binding The utilityProcess 'error' event from #43774 reports any V8 fatal error. Its spec raised one with the external-ab native addon, which handed V8 memory from outside the sandbox. Node.js 26 makes every Buffer site sandbox compatible, so that no longer happens through Node.js APIs. The earlier commits on this branch made the addon call V8 directly, and that needed ABI workarounds on Windows and for libc++. As Deepak suggested in review, drop the addon and its workspace entries. Add triggerV8FatalError() to the electron_common_testing binding, which is only built when DCHECKs are on. It fails a V8 API check by setting an internal field on a plain object, so it does not depend on the V8 sandbox. The spec keeps #43774's checks on the event type, location and report trigger, also checks the report's event message, and now runs wherever the testing binding exists, including Windows on Arm. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * test: wait for the utility process exit only after its 'error' event events.once(emitter, 'exit') also listens for 'error' and rejects with its first argument. The previous commit started waiting for 'exit' before the 'error' event arrived, so the utility process's 'FatalError' rejected that promise and the spec failed with "Unknown Error: FatalError". Wait for 'exit' after 'error' has been received, as the spec from #43774 did. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * refactor: use one d8-style PartitionAlloc allocator inside the V8 sandbox Following Deepak's review, replace the sandbox-page fallback added for nodejs/node#61480 and the older InSandboxAllocator with a single PAInSandboxAllocator, modelled on d8's. nodejs/node#61480 moved Buffer.allocUnsafe() onto the isolate's ArrayBuffer allocator. gin's partition caps a single allocation at partition_alloc::MaxAllocationSize() (2 GiB - 2 MiB), so Buffers above that size failed. The previous fix served those requests from whole sandbox pages and tracked them in a separate map. PAInSandboxAllocator owns a PartitionAlloc partition in the configurable pool, which gin::V8Initializer::Initialize() places inside the V8 sandbox. The partition is created lazily, after the pool exists. It has allow_giga_allocations set and allocates with kAllowGigaAllocations, along with kNoOverrideHooks and kNoMemoryToolOverride, as gin does. A single ArrayBuffer can therefore reach MaxGigaAllocationSize() (16 GiB - 2 MiB on 64-bit) without the page fallback. It is both the ArrayBuffer allocator for JavascriptEnvironment isolates and V8's in-sandbox allocator, so Node.js worker_threads isolates use it through NewDefaultAllocator(). Builds without the V8 sandbox keep gin's allocator. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> * fix: reject instead of throwing in asar fs.promises.stat/lstat wrappers The new fs.promises.stat/lstat wrappers call splitPath() before deciding whether to use fsPromises' own implementation. For a URL, splitPath() runs getValidatedPath(), which throws for a non-file: URL, so the error escaped synchronously instead of rejecting the returned promise as it did with util.promisify(fs.stat) on main and as Node's fsPromises does. Make the wrappers async so every input error rejects, and add a spec for it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vf2yMuJ7XxvNktpKuiMPzU Co-authored-by: Claude <noreply@anthropic.com> --------- Co-authored-by: trop[bot] <37223003+trop[bot]@users.noreply.github.com> Co-authored-by: electron-roller[bot] <84116207+electron-roller[bot]@users.noreply.github.com> Co-authored-by: Claude <noreply@anthropic.com>

marco-ippolito added the semver-major PRs that contain breaking changes and should be released in the next major version. label Feb 13, 2026

nodejs-github-bot added c++ Issues and PRs that require attention from people who are familiar with C++. lib / src Issues and PRs involving general changes in the lib/ or src/ directories. needs-ci PRs that need a full CI run. labels Feb 13, 2026

marco-ippolito force-pushed the rip-transform-type branch from 2ea1afe to c62af40 Compare February 13, 2026 16:59

ShogunPanda approved these changes Feb 17, 2026

View reviewed changes

absidue mentioned this pull request Feb 19, 2026

feat: add support for --experimental-transform-types electron/electron#49711

Merged

6 tasks done

marco-ippolito force-pushed the rip-transform-type branch from c62af40 to 70303e1 Compare February 23, 2026 16:07

marco-ippolito force-pushed the rip-transform-type branch from 70303e1 to 50a77db Compare February 23, 2026 16:42

benjamingr approved these changes Feb 23, 2026

View reviewed changes

styfle approved these changes Feb 23, 2026

View reviewed changes

ljharb approved these changes Feb 23, 2026

View reviewed changes

himself65 approved these changes Feb 23, 2026

View reviewed changes

GeoffreyBooth approved these changes Feb 23, 2026

View reviewed changes

mcollina mentioned this pull request Jul 15, 2026

Expose the full Amaro loader API via 'node:module' #64518

Open

coderabbitai Bot mentioned this pull request Jul 17, 2026

Overnight integration #1: trusted-state (real kill-9) + effect-protocol + git-hardening + hermetic-rehearsal + durable effect-event store aumara-xyz/aukora-alpha#127

Closed

cirospaciari mentioned this pull request Jul 25, 2026

node:module: stripTypeScriptTypes strip mode + StrippedTypeScript compile cache (+3 tests) oven-sh/bun#35517

Draft

robobun mentioned this pull request Aug 6, 2026

process: report features.typescript as "strip" to match Node 26 oven-sh/bun#37014

Open

This was referenced Aug 6, 2026

Upgrade to Next.js 16.3, Instant Navigations, and landing Lighthouse wins TommyBez/skillsboard#93

Merged

Add an async context data test for @fedify/express fedify-dev/fedify#985

Merged

This was referenced Aug 10, 2026

config loading breaks on node 26 with typescript 7 (stripTypeScriptTypes transform mode removed) expo/expo#48724

Closed

expo config fails to load app.config.ts on node 26 (stripTypeScriptTypes transform mode removed) expo/expo#48725

Closed

robobun mentioned this pull request Aug 13, 2026

node:module: implement stripTypeScriptTypes oven-sh/bun#32206

Closed

expo-bot mentioned this pull request Aug 18, 2026

fix(require-utils): Fix Node 26 stripTypeScriptTypes fallback call expo/expo#48826

Merged

1 task done

coderabbitai Bot mentioned this pull request Aug 23, 2026

Add response-handling tests for @fedify/h3 fedify-dev/fedify#1007

Merged

CrazyCoder mentioned this pull request Aug 27, 2026

Test scripts cannot run on Node 26: --experimental-transform-types was removed in v26.0.0 4fuu/pi-pwsh#8

Closed

ScriptedAlchemy mentioned this pull request Sep 4, 2026

fix: Node 26 transform flag; operator .env precedence + import order; MCP negotiation test; provider-view cross-process coverage (integration) ScriptedAlchemy/agent-bundle#554

Merged

vincent-peng mentioned this pull request Sep 25, 2026

fix(tests): keep native catalog import working on Node 26 steipete/summarize#492

Merged

3 tasks done

kdy1 mentioned this pull request Oct 30, 2026

feat!: use SWC Next for stripping and remove transform mode nodejs/amaro#507

Draft

oristeiner1313-source mentioned this pull request Oct 30, 2026

fix(scripts): explain the pinned Node requirement when transform type stripping is unavailable morluto/rea#8159

Draft

4 of 22 tasks

mikinovation mentioned this pull request Nov 1, 2026

chore(deps): update dependency node to v26 mikinovation/mikinovation#445

Merged

1 task

Navigation Menu

Repository navigation

module: remove --experimental-transform-types - #61803

module: remove --experimental-transform-types#61803
nodejs-github-bot merged 1 commit into
nodejs:mainfrom
marco-ippolito:rip-transform-type

marco-ippolito commented Feb 13, 2026 •

edited

Loading

Uh oh!

nodejs-github-bot commented Feb 13, 2026

Uh oh!

GeoffreyBooth commented Feb 13, 2026

Uh oh!

marco-ippolito commented Feb 13, 2026

Uh oh!

aduh95 commented Feb 13, 2026

Uh oh!

styfle commented Feb 13, 2026

Uh oh!

codecov Bot commented Feb 13, 2026 •

edited

Loading

Uh oh!

GeoffreyBooth commented Feb 13, 2026

Uh oh!

marco-ippolito commented Feb 23, 2026

Uh oh!

ljharb left a comment

Uh oh!

marco-ippolito commented Feb 23, 2026

Uh oh!

Renegade334 commented Feb 25, 2026 •

edited

Loading

Uh oh!

aduh95 commented Feb 26, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

12 participants

Conversation

marco-ippolito commented Feb 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nodejs-github-bot commented Feb 13, 2026

Uh oh!

GeoffreyBooth commented Feb 13, 2026

Uh oh!

marco-ippolito commented Feb 13, 2026

Uh oh!

aduh95 commented Feb 13, 2026

Uh oh!

styfle commented Feb 13, 2026

Uh oh!

codecov Bot commented Feb 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

GeoffreyBooth commented Feb 13, 2026

Uh oh!

marco-ippolito commented Feb 23, 2026

Uh oh!

ljharb left a comment

Choose a reason for hiding this comment

Uh oh!

marco-ippolito commented Feb 23, 2026

Uh oh!

Renegade334 commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

aduh95 commented Feb 26, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

12 participants

marco-ippolito commented Feb 13, 2026 •

edited

Loading

codecov Bot commented Feb 13, 2026 •

edited

Loading

Renegade334 commented Feb 25, 2026 •

edited

Loading