Repository navigation
Description
[agent] Found by the scheduled npm bug-hunt routine (ledger #302).
Summary
When .socket/vendor/state.json is lost (deleted, or dropped by a bad merge) while package-lock.json still wires the vendored tarballs, socket-patch vendor --revert keeps the artifacts, which is the correct fail-safe. It then warns vendor_orphan_still_wired with this remedy:
a project lockfile still points at .socket/vendor/npm/, which no ledger entry owns; the artifacts were kept (run
socket-patch repairto re-adopt them into the ledger, then revert again)
repair can't do that. It exits 1 with:
Cannot repair vendored artifact … the vendor ledger (.socket/vendor/state.json) has no entry for it, and repair does not rebuild the ledger from lockfiles; restore state.json from version control …
After repair, vendor --revert prints the same warning again ("Reverted 0 vendored packages.", exit 0). Following the advice goes round in a loop, and the project stays vendored.
Impact
Low severity, but it's a dead end for a user who doesn't have state.json in version control. The fail-safe parts all hold: in every step below, a cold-cache npm ci still installs patched bytes, and no artifact is deleted. Only the remedy is wrong. It contradicts CLI_CONTRACT.md (repair: "redownload missing/corrupt vendored artifacts (never re-synthesizing a lost ledger)"). It's in the same family as #900 and #977 (a suggested remedy that is a no-op).
Repro (Linux, npm 10.9.4, local mock of the patch API serving free patches for left-pad@1.3.0 and ms@2.1.3)
mkdir p && cd p
echo '{"name":"p","version":"1.0.0","dependencies":{"left-pad":"1.3.0","ms":"2.1.3"}}' > package.json
npm install
socket-patch scan --mode vendored --yes $API # exit 0, both vendored
rm .socket/vendor/state.json # ledger lost
socket-patch vendor --revert --yes $API
# Warning: … which no ledger entry owns; the artifacts were kept (run `socket-patch repair` to re-adopt them into the ledger, then revert again)
# Reverted 0 vendored packages. exit 0
socket-patch repair --yes $API
# Error: Cannot repair vendored artifact … repair does not rebuild the ledger from lockfiles … exit 1
socket-patch vendor --revert --yes $API # same warning again, exit 0Expected vs actual
- Expected: a remedy that works here. Either the one
repair,rollbackandscan --prunealready give ("restore .socket/vendor/state.json from version control, or restore the lockfile withgit checkout -- <lockfile>"), or arepairthat actually re-adopts the entry, as the warning promises. - Actual: the warning points at
repair, which by contract and by its own message refuses this exact state.
Every other command reports this state consistently (main 05ecc6e, run twice):
| command | exit | message |
|---|---|---|
vendor --revert |
0 | vendor_orphan_still_wired → "run socket-patch repair to re-adopt" (wrong) |
repair |
1 | "repair does not rebuild the ledger … restore state.json" |
rollback |
1 | "restore .socket/vendor/state.json from version control …" |
scan --mode vendored --prune |
1 | "restore .socket/vendor/state.json from version control" |
vendor --check |
1 | vendor_ledger_missing … "restore state.json" |
OS × version
| OS | npm | result |
|---|---|---|
| Linux | 10.9.4 (Node 22.22) | fail (x2) |
macOS and Windows weren't probed. The message is built in shared CLI code, so it isn't OS-dependent, and probably not npm-specific either (sweep_orphan_vendor_dirs is ecosystem-generic). Not bisected.
Suspect code
crates/socket-patch-cli/src/commands/vendor.rs:4071(thevendor_orphan_still_wiredtext) vscrates/socket-patch-cli/src/commands/vendored_backend/repair.rs:184(repair's refusal).
Backlog review — 2026-10-08
Priority: P1 → P3. Missing-ledger recovery advice loops, but artifacts are retained and fresh installs remain patched. The report explicitly says only the remedy is wrong.
Activity
mikolalysenko commentedon Oct 7, 2026
[agent] Triage: priority:p1 (npm). Not a duplicate: same family as #977 (a remedy that can't work), but a different code path (the
vendor_orphan_still_wiredremedy text vsrepair's no-ledger refusal). No open PR references it.Generated by Claude Code
mikolalysenko commentedon Oct 9, 2026
v5 triage: P3, not a release blocker. Correct the stale remedy when convenient, but reconstruction after a lost vendor ledger is outside the v5 happy-path gate. Keep P3; do not reintroduce ledger reconstruction.
This follows the maintainer's release scope: one normally completing CLI instance, prioritizing valid-lockfile patch/install behavior, compatibility, and actionable CLI UX.
mikolalysenko commentedon Oct 28, 2026
[agent] Re-triage from the scheduled npm bug-hunt routine (ledger #302): fixed on main
b787454by 675fec1 (#1739), which replaced therun socket-patch repair to re-adopt themremedy.Repro on npm 8.19.4 (v2 lock, root
ms@2.1.3and nestedms@2.0.0, both vendored and committed), thenrm .socket/vendor/state.json:vendor --revert --json: exit 0, both entriesskipped. The reason now reads "…which no ledger entry owns; the artifacts were kept (restore the missing entries in .socket/vendor/state.json from version control (if a git merge kept one side of state.json, merge both sides'entries), then revert again)".repair --json:vendor_ledger_missing, "restore state.json from version control and re-runsocket-patch repair, or restore the lockfile (git checkout -- <lockfile>) and re-vendor". It no longer points back at the revert.git checkout -- .socket/vendor/state.json, thenvendor --revert) removes both entries, exit 0.The advice no longer loops, so I'm closing this as completed.
Generated by Claude Code