Repository navigation
os doctor: a pnpm -v that fails for any reason (Corepack unable to fetch its pinned or latest pnpm) is reported "✗ pnpm Not found" with "Install pnpm: npm install -g pnpm@10.28.1", though pnpm is installed #24464
Description
Filing gate ①, class (a): a public door's output says something false, measured. reach: os doctor for any developer whose pnpm runs through Corepack and cannot reach the registry in a project with no packageManager pin (offline, behind a proxy, or rate-limited). The door is in packages/cli, domain:cli by the lane table.
Filed by domain:cli seat 1 (seat post #6024) · zhuangjianguo · session session_01SYEJqDL3NXxrJRBVqaMu9T, from #24419's dev report (PR #24455, out_of_scope_findings[0]). ⛔ Not a claim. Grade, lane and area:* are triage's.
What happens (measured on PR #24455's head bd0651500)
- The setup: a
tmpdir()project with nopackageManager, Corepack shims onPATH, and the registry answering429. tsx bin/run-dev.js doctorprints✗ pnpm Not foundand→ Install pnpm: npm install -g pnpm@10.28.1, then exits1. Corepack's stack trace for the failed fetch is on stderr, above the report.- pnpm is installed. What failed was Corepack's fetch of the version it wanted.
Where
packages/cli/src/commands/doctor.ts, the pnpm probe (about :2098–:2112 on origin/main): execSync('pnpm -v') inside a try, whose catch maps EVERY failure to "Not found" with the install hint. A non-zero exit from a pnpm that exists, such as a Corepack fetch error or a version mismatch refusal, is not "Not found", and the prescribed npm install -g pnpm@10.28.1 does not fix it.
What done looks like
- The probe tells "no
pnpmon PATH" (ENOENT) apart from "pnpmran and failed". - For the second, it reports the failure with the child's own first line and a fix that matches it: for a Corepack fetch failure, the network or the
packageManagerpin, not a reinstall. - A test drives both: no binary, and a stub
pnpmonPATHthat exits non-zero with a Corepack-shaped message.
Dedupe
- MCP
search_issues"os doctor pnpm Not found corepack registry offline misleading install hint": 0 hits. - "doctor pnpm probe catch every failure reported as not installed": 1 hit, [finding] check:pm-governed-merges reports an uninstalled worktree as a shrunken self-test battery — exit 1 with a confident wrong diagnosis where its siblings exit 3 NOT MEASURED #15272 (closed, a different gate's misdiagnosis). Neither is this defect.
- The control query reaches the misdiagnosis family ([finding] check:pm-governed-merges reports an uninstalled worktree as a shrunken self-test battery — exit 1 with a confident wrong diagnosis where its siblings exit 3 NOT MEASURED #15272).
- Dedupe words: os doctor pnpm Not found corepack; doctor install pnpm hint misleading; doctor pnpm -v catch-all.
Generated by Claude Code
Activity
objectstack-fleet commentedon Oct 19, 2026
Triage: p3 ·
bug·domain:cli·area:devpath·pm:queue· triage seat ·session_01AavokzJ5DndAwitDXvKy4U· 2026-10-19T05:56ZTriage seat (objectstack-wide, seat post #6015). ⛔ Not a claim, ⛔ not a dispatch.
origin/main:packages/cli/src/commands/doctor.ts:2100runsexecSync('pnpm -v'), and itscatchreports'Not found'(:2110) for any failure. The premise holds.packageManagerpin. The hint is wrong, but nothing is broken.Generated by Claude Code
objectstack-fleet commentedon Oct 19, 2026
Claim: PM loop round 5 ·
domain:cliseat 1 · p3 · 2026-10-19T06:46ZSession:
session_01SYEJqDL3NXxrJRBVqaMu9TAccount:
zhuangjianguo(the seat's linked user asGET /useranswers it; always the card's assignee)Branch:
claude/issue-24464-doctor-pnpm-probeWorktree:
objectstack-issue-24464Domain:
domain:cliSeat:
domain:cli#1(seat post #6024)File surface (read on
origin/maine11ae4b6b; line numbers move; stop on a breach and explain it in the report):packages/cli/src/commands/doctor.ts: the environment probes'execSyncsites,pnpm -v(about:2100, the card's defect),tsc -v(about:2117) andgit --version(about:2191). Each one'scatchmaps every failure to "Not found". The pnpm probe must tell "nopnpmon PATH" apart from "pnpmran and failed". The other two are the same family: they move only if the same classifier covers them, and otherwise each one's reason goes in the PR body.packages/cli/test/(or beside the source, by the package's convention). It drives a missing binary and a stubpnpmonPATHthat exits non-zero with a Corepack-shaped message.patchchangeset on@objectstack/cli.packages/spec/**; any otherdoctor.tscheck; any exit code other than the one this probe's verdict already drives.Container & model:
S,mode:subagent,model: default tier(dispatch-gates.mjs --tier: "no path-derived mandate")Clause-②: no
Basis for the Clause-② line: a diagnostic's message and fix text. No accept set, export or published shape moves.
Responsibility: this lane's own code produces the risk (
packages/cli/src/commands/doctor.ts). No platform path covers it. It reaches any developer whosepnpmruns through Corepack and cannot fetch its version, in a project with nopackageManagerpin. It was measured on PR test(cli): tmpdir fixtures that run pnpm reach no registry (dev-no-watch pin, doctor readers, init leg) #24455's head.Thread-read: 6260709044
Serial constraints cleared:
packages/cli(migrate and serve files), and none touchesdoctor.ts.pm:dispatchedcard's thread namesdoctor.ts.ai.models()TSDoc and the/ai/modelsdocs row still say "Models this environment offers", which the spec describe now says as "this system" #24456, touches nopackages/clifile.Generated by Claude Code
objectstack-fleet commentedon Oct 19, 2026
os-dev-report
{
"issue": 24464,
"status": "done",
"branch": "claude/issue-24464-doctor-pnpm-probe",
"pr": "#24493",
"session": "session_01SYEJqDL3NXxrJRBVqaMu9T (subagent mode: the parent session id, as the relay derived it)",
"premise_still_valid": true,
"summary": "Premise holds on origin/main e11ae4b: doctor.ts's pnpm probe ran execSync('pnpm -v') and its catch answered 'Not found' + 'npm install -g pnpm@10.28.1' for every failure; tsc -v and git --version had the same catch-all. Measured first (Node 22.22, dash, the probe's own execSync call): a missing pnpm is the shell's exit 127, never Node's ENOENT. But exit 127 also comes from a pnpm on PATH whose node is missing ('/usr/bin/env: node: No such file or directory') and from a non-executable pnpm ('Permission denied'). Real Corepack 0.34.0 against a local 429 server exits 1, and its first stderr line is Node's FILE:LINE header, not the error. So the fix is a PATH lookup that runs only after a probe fails: the classifier asks whether a file named pnpm is on the PATH the probe ran with, where /bin/sh would look; on win32 it tries cwd then PATH with each PATHEXT extension, never an extensionless file, mirroring cmd.exe. Not found keeps today's row byte for byte. Ran-and-failed reports the child's first meaningful line (the first '...Error: ' line, else the first non-blank stderr line, else stdout's). The fix is the Corepack one (network/proxy or a packageManager pin, 'reinstalling pnpm will not fix that') when the stderr carries Corepack's own fetch-failure text, and the neutral 'run pnpm -v and read its error' otherwise. tsc and git moved under the same classifier with no change beyond message and fix; statuses are unchanged (pnpm error, TypeScript/Git warning), so no exit code moves. All edits are inside the claim's file surface: doctor.ts (exported pnpmCheck/typescriptCheck/gitCheck/probeTool/classifyToolProbeFailure/isOnShellPath/firstMeaningfulLine/isCorepackFetchFailure/toolFailedFix/COREPACK_FETCH_FIX, none re-exported from the package entry, so no published export moves), one new test file, and a patch changeset with 'Clause-②: no'. Draft PR #24493 at 023dd79, body written once at creation, read back byte-identical (12232 bytes) with assignee zhuangjianguo. Not measured on a Windows host; the win32 branch is pinned by four unit cases driven with platform 'win32' on Linux, and the PR body names that branch and why. Deviation: origin/main advanced 4 commits after the branch point (to 4bd7215: packages/lint, packages/spec, plugin-security, docs, an ADR; none under packages/cli). The branch was not merged with them before the PR opened; the AGENTS.md section 10 recheck scoped to the overlap is empty, and CI's merge ref covers the joint state.",
"tests": "All on 023dd79. (1) New packages/cli/test/doctor-tool-probe-failure.test.ts (unit tier, no child_process import; the test tsconfig program counts it, 1 hit with tsc --listFilesOnly): pnpm --filter @objectstack/cli exec vitest run --maxWorkers=2 test/doctor-tool-probe-failure.test.ts -> exit 0, "Tests 17 passed (17)". Every row case runs the real probe (execSync through /bin/sh) with PATH set to a test-owned directory: empty, or holding one #!/bin/sh stub that uses builtins only. Nothing else is on that PATH, so no real pnpm, Corepack or registry is reachable. (2) pnpm --filter @objectstack/cli exec vitest run --project unit --maxWorkers=2 -> VERDICT command-exit 0, "Test Files 312 passed (312)", "Tests 4627 passed (4627)". The integration tier is declared to CI: the diff touches no spawn entry and no integration-tier file. (3) pnpm --filter @objectstack/cli typecheck -> exit 0 (tsc --noEmit plus check:test-typecheck OK). pnpm --filter @objectstack/cli build -> exit 0. eslint --no-inline-config over the two changed .ts files -> exit 0, 0 errors, 0 warnings, neither file ignored. (4) CLI before/after, measured with tsx bin/run-dev.js doctor in an empty tmpdir under env -i. Before = doctor.ts restored from e11ae4b into the tree; restore proven by blob 7d2cb9411fc8 == HEAD. Results: no pnpm on PATH -> stdout and stderr byte-identical before and after, "Not found" plus the install fix. Real Corepack shim with a local 429 registry -> before "Not found" plus the install fix; after "Error: Server answered with HTTP 429 when performing the request to http://127.0.0.1:44765/pnpm/latest; ..." plus the Corepack fix. Stub ERR_PNPM_UNSUPPORTED_ENGINE -> after: its line plus the neutral fix. Working toolchain control -> byte-identical. Exit 1 in all 8 runs. (5) Ablation, after committing the fix: scripts/ablation-replace.mjs swapped the anchor "if (!isOnShellPath(binary, options)) return { outcome: 'not-found' };" for "if (binary) return ...not-found..." (today's catch-all). Anchor 1->0, replacement 0->1, blob 7d2cb9411fc8->a59ad6a54cbc. Result: "Tests 8 failed | 9 passed (17)"; the 8 are exactly the ran-and-failed cases. Restored: blob == HEAD, git diff HEAD empty. The test imports ../src/commands/doctor relatively, so there was no dist leg.",
"gates": "Local, all on 023dd79, each exit code captured before any pipe. (a) node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack: 68 derived families. 66 exited 0 on the first run. pnpm check:dual-build-cjs-loads and pnpm check:i18n-coverage first exited 3 (PREREQUISITE NOT MET: packages outside the cli dependency closure were unbuilt). Re-run once those had dist: both exit 0 ('111 published require entry point(s) across 69 package(s) load'; 'check-i18n-coverage: OK (14 config(s), 621 baselined untranslated string(s), none new)'). (b) dispatch-gates.mjs --ran over the 68 recorded '... :: exit N' lines: exit 0, 'Run reconciliation - 68 derived, 68 run, 0 NOT-MEASURED, 0 UNRUN' (a DERIVED zero). (c) The 11 declared-wide families (init-service-contract, live-db-isolation, meta-type-normalized, optional-error-sink, resume-authority-declared, route-envelope, runner-env-posture, settings-bind-window, startup-registry-verdict, verify-stand-in, wildcard-fallthrough): all exit 0. (d) pnpm lint (eslint . --no-inline-config, the full union), run after the final commit: VERDICT command-exit 0, no findings. Conflict noted: os-dev.md says a repo-wide scan is CI's and never owed, while the dispatch asked for it as this lane's union. I ran it because running it is not forbidden; that does not make it owed. (e) @objectstack/cli typecheck, unit tests and changed-file eslint: see tests. A raw-control-byte self-scan of the three changed files found none, and check:nul-bytes (in the derived set) exited 0. (f) Not run locally: check:platform-checklist --changed-from (no exported symbol was renamed or deleted, only added). CI on PR #24493 at report time (head 023dd79): every check-run in_progress except "No other open PR may claim the same issue" success and "Packed-tarball smoke (opt-in)" skipped. Check Changeset, Lint ·, Type Check · and Governed Surface Queue Guard were in_progress, not yet concluded. Not waited for, per contract.",
"mcp_calls": "0 (no MCP GitHub tool used; reads went through gh api single-card REST reads)",
"api_writes": "2 relay strokes (fleet-write repository_dispatch, POST /repos/objectstack-ai/objectstack/dispatches, each HTTP 204). Stroke 1: pr_create, which the relay ran as objectstack-fleet[bot]: POST /repos/objectstack-ai/objectstack/pulls (draft forced) and POST /repos//issues/24493/assignees [zhuangjianguo]; run 39217452199 success. Stroke 2: this os-dev-report comment, POST /repos//issues/24464/comments. Plus git push, which is not a REST write (empty-branch probe, then 2 commits). No labels written: the dispatch named none, and a changeset exists, so no skip-changeset.",
"open_questions": [],
"out_of_scope_findings": [
"carrier: none (承接者:无) · noted in PR #24493 Acceptance notes, not filed: the three probes run execSync with no timeout, so a pnpm -v that hangs would hang os doctor. Not observed: Corepack failed fast on a refused or rate-limited registry in every measured run. No reach: measured.",
"carrier: none (承接者:无) · noted in PR #24493 Acceptance notes, not filed: the not-found install hint names pnpm@10.28.1, while this repository's own packageManager is pnpm@10.31.0. Left alone because the not-found row was required to stay byte-identical. Drift, not a reproduced defect."
]
}
Generated by Claude Code
objectstack-fleet commentedon Oct 19, 2026
ACCEPT → PR #24493 comment
6262267322·domain:cliseat 1 · sessionsession_01SYEJqDL3NXxrJRBVqaMu9T· 2026-10-19T07:44Z · on head023dd79d0. The probes tell "not on PATH" (by a PATH lookup, since shell exit127does not discriminate) apart from "ran and failed". A miss keeps today's row byte for byte; a hit reports the child's own error line and a fix that matches it. Statuses and exit code are unchanged, and the ablation reds exactly the 8 ran-and-failed cases. Two Acceptance notes, no filing. Landing waits on every check of this head.Generated by Claude Code
objectstack-fleet commentedon Oct 19, 2026
Landed: PR #24493 →
959ee317b·domain:cliseat 1 ·zhuangjianguo· sessionsession_01SYEJqDL3NXxrJRBVqaMu9T· 2026-10-19T08:24Zmerged: true, merge commit959ee317b, merged 2026-10-19T08:22Z through the merge queue (enqueued 2026-10-19T08:02Z)origin/main959ee317bis an ancestor oforigin/main, with a single parent (squash). The 3 files and +458 / −50 match what was reviewed at023dd79d0gh-readonly-queue/main/*pr-24493*head remains6261405871→ report6262235637→ ACCEPT6262267322on PR fix(cli): os doctor tells a tool missing from PATH apart from one that ran and failed #24493 → pointer6262277768. No cross-lane declaration was owed.os doctor'spnpm,tscandgitprobes tell "not on PATH" apart from "ran and failed", by a PATH lookup after a failure. The lookup mirrors/bin/sh, and on Windowscmd.exewithPATHEXT; shell exit127does not discriminate.packageManagerpin) when Corepack could not fetch, and "run it and read its error" otherwise.pnpm@10.28.1has drifted.completedon the merge.pm:dispatchedis removed in the stroke after this comment.Generated by Claude Code