node:child_process largely stubbed — 52 radar tests fail with 'value is not a function' #1780

Activity

1. proggeramlug

   added

   bugConfirmed defect or regressionConfirmed defect or regression

   parityCompatibility gap with Node.js, ECMAScript, or the supported ecosystemCompatibility gap with Node.js, ECMAScript, or the supported ecosystem

   on May 25, 2026
2. proggeramlug commented on May 26, 2026

   

   proggeramlug

   on May 26, 2026

   ContributorAuthor

   More actions

   Partial fix in #1827 (PR), plus scope mapping:

   Fixed: child_process.exec(cmd[, options], callback) — js_child_process_exec was a null stub and the codegen lowered the callback then discarded it, so exec(cmd, cb) never fired its callback. exec now runs the command (shell, like execSync) and invokes the callback with (err, stdout, stderr) (immediate-callback model, same as the async fs wrappers). The callback is located whether it sits in the options slot (exec(cmd, cb)) or the third slot (exec(cmd, options, cb)); non-zero exit yields an Error; the no-callback form still returns the stdout string. execSync / spawnSync already worked.

   Still open — the larger part of this ticket: the dominant node-core child_process pattern is the streaming API:

   const child = spawn(cmd, args);
   child.stdout.on('data', d => …);   // child.stdout is currently a string → "(string).on is not a function"
   child.on('exit'|'close'|'error', …);

   js_child_process_spawn / exec (the streaming forms) are explicit null-returning stubs ("needs event loop integration"). A real ChildProcess object — stdout/stderr as Readable streams, stdin Writable, EventEmitter (spawn/exit/close/error/message), pid/kill(), and IPC for fork — is a sizeable feature requiring subprocess + event-loop integration. Most of the 52 radar fails bottom out there. Suggest splitting that into its own ticket (spawn-with-streams) and keeping this for the audit.
3. proggeramlug

   added 4 commits that reference this issue on May 26, 2026

   fix(hir,runtime): builtin-prototype .call/.apply (#1777), fs.readdir …

   ...

   1e277eb

   fix(hir,runtime): builtin-prototype .call/.apply (#1777), fs.readdir …

   ...

   3740225

   feat(child_process): streaming spawn() ChildProcess + execFile/execFi…

   ...

   e9f6993

   feat(child_process,util): ChildProcess/Stream exports (#1856) + promi…

   ...

   43a4bfd
4. proggeramlug commented on May 27, 2026

   

   proggeramlug

   on May 27, 2026

   ContributorAuthor

   More actions

   Knocked out the remaining easy stubs in PR #1926 (runtime-only):

   • Options were ignored — spawn/exec/execFile/execFileSync/spawnSync/execSync now honor { cwd, env, shell } (the options pointer was already threaded to js_child_process_spawn_streams, just never read). env replaces the environment wholesale per Node semantics; shell:true/shell:"<path>" joins command+args into a <shell> -c line.
   • Listener removal was a no-op — removeListener/off now rebuild the listener array without the matching closure; removeAllListeners([event]) clears one event or all.

   New test_issue_1780_options.ts is byte-identical to node --experimental-strip-types. The follow-ups #1856 (ChildProcess/Stream exports) and #1857 (promisify exec/execFile) closed via #1861.

   Still open (the large remaining piece): fork + IPC (send/disconnect/'message', channel/connected semantics). Perry has no async subprocess reactor, so a real fork needs subprocess + event-loop integration — worth its own ticket rather than folding into this one. Leaving #1780 open to track that.
5. proggeramlug

   added a commit that references this issue on May 27, 2026

   feat(child_process): honor cwd/env/shell options + real listener remo…

   ...

   bc2d44d
6. proggeramlug commented on May 27, 2026

   

   proggeramlug

   on May 27, 2026

   ContributorAuthor

   More actions

   The original failure cluster here — 52 radar tests failing with value is not a function because spawn()'s .stdout was a string — is fixed: streaming spawn() (#1858), exec callback (#1828), execFile/execFileSync (#1858), ChildProcess/Stream exports + promisify(exec/execFile) (#1861, closing #1856/#1857), and cwd/env/shell options + listener removal (#1926). test_parity_child_process is byte-identical to Node and out of known_failures.json.

   The remaining surface is split into granular tickets (per the "no umbrella issues" preference) and this umbrella is closed:

   • node:child_process — fork() + IPC (send/disconnect/'message'/channel) unimplemented #1933 — fork() + IPC (send/disconnect/'message'/channel)
   • node:child_process — async subprocess reactor: live streaming stdout/stderr, stdin.write(), kill() on a running child #1934 — async subprocess reactor (live streaming stdout/stderr, stdin.write(), kill() on a running child)
   • node:child_process — exec/execFile failure Error missing .code/.signal/.killed/.cmd #1935 — exec/execFile failure Error missing .code/.signal/.killed/.cmd
   • node:child_process — spawnSync result missing pid/signal/output/error fields #1936 — spawnSync result missing pid/signal/output/error
   • node:child_process — exec/execFile/sync ignore the encoding option (no Buffer stdout/stderr) #1937 — exec/execFile ignore the encoding option (no Buffer stdout/stderr)
   • node:child_process — execSync/execFileSync do not throw on non-zero exit #1938 — execSync/execFileSync don't throw on non-zero exit

   Closing in favor of those.
7. proggeramlug

   closed this as completedon May 27, 2026
8. proggeramlug

   added 2 commits that reference this issue on May 27, 2026

   feat(node:child_process): async spawn reactor, fork()+IPC, exec error…

   ...

   34492ce

   feat(node:child_process): async spawn reactor, fork()+IPC, exec error…

   ...

   a309836