Replace the external `@pnpm/registry-mock` (Verdaccio) test dependency with an in-repo, in-process registry that serves package fixtures to **both** the pacquet Rust tests and the pnpm CLI (Jest) tests. No separately managed registry process is needed.
### How it works
- **Fixtures** live at `registry/.fixtures/packages/<name>/<version>/…`, moved verbatim from [`pnpm/registry-mock`](https://github.com/pnpm/registry-mock) (keyed by each `package.json`'s `name`+`version`).
- **`pnpm-registry-fixtures`** builds verdaccio-shaped storage from those fixtures; the in-tree **`pnpm-registry`** crate serves it.
- Files whose names differ only by case (`@pnpm.e2e/with-same-file-in-different-cases`) and `bundleDependencies` trees are composed **in memory** by the builder, since neither can be committed to the working tree.
- **pacquet**: `pacquet-testing-utils`' `TestRegistry` starts the server lazily (once per process) in proxy mode, serving `@pnpm.e2e` fixtures locally and falling through to the npm uplink for real packages (`is-positive`, `is-negative`, …) — matching how registry-mock behaved.
- **pnpm CLI**: the `with-registry` Jest `globalSetup` builds storage from the fixtures via the new `pnpm-registry-prepare` binary (built from source in the Test CI job) and serves it with `pnpm-registry`. `REGISTRY_MOCK_PORT` / `REGISTRY_MOCK_CREDENTIALS` / `getIntegrity` now come from `@pnpm/testing.registry-mock`.
### Result
`@pnpm/registry-mock` is removed from every manifest, the catalog, and `packageExtensions`; `cargo test` / `cargo nextest run` / `just test` and the pnpm CLI Jest suites all run registry-backed tests without launching Verdaccio.
## 1. Support `nodeLinker: hoisted` on the fresh-lockfile install path (pacquet)
Closes#11871. Until now pacquet's `Install::run` hard-refused `nodeLinker: hoisted` without a checked-in lockfile (`ERR_PNPM_…UNSUPPORTED_FRESH_INSTALL_NODE_LINKER`).
- Extracted a shared `run_hoisted_linker` helper from the frozen path's hoisted branch (walker → `link_hoisted_modules` → `SymlinkDirectDependencies { link_only: true }` → `pkg_root_by_key` → walker-skip folding), so both install paths run identical logic.
- Fresh path now threads `node_linker` + `supported_architectures`, hands `CreateVirtualStore` the real linker (populating `cas_paths_by_pkg_id`), branches on `is_hoisted`, and returns `hoisted_locations` so `.modules.yaml` round-trips.
- Removed the guard and the dead `UnsupportedFreshInstallNodeLinker` error variant.
Ported upstream's `hoistedNodeLinker/install.ts` into `crates/cli/tests/hoisted_node_linker.rs` (real tests for the core layout, no-lockfile, `externalDependencies`, `autoInstallPeers`, and `hoistingLimits`; the rest stubbed as `known_failures` against `pnpm add`/update (#433) and build-phase (#11870) gaps), and ticked the boxes in `plans/TEST_PORTING.md`.
## 2. Add the `hoistingLimits` setting (pnpm CLI **and** pacquet)
Revives the stale #6468 (closes#6457) and brings both stacks to parity. `hoistingLimits` mirrors yarn's `nmHoistingLimits`: `none` (default — hoist as far as possible), `workspaces` (hoist only as far as each workspace package), `dependencies` (hoist only up to each workspace package's direct deps). It was previously a programmatic-only option in pnpm (no config surface) and a pacquet-only raw-map yaml field.
**pnpm CLI:** `config/reader` (`types.ts` enum + `Config.ts` + `configFileKey.ts`), `installing/linking/real-hoist`'s new `getHoistingLimits` (mode → the `@yarnpkg/nm` hoister's per-locator border map), and the install/add/recursive command option lists. Tests: `hoistedNodeLinker/install.ts` (`dependencies` mode) + `real-hoist` `getHoistingLimits` unit tests. Changeset included (minor).
**pacquet:** replaced the raw-map config with the same enum; added `get_hoisting_limits` (port of `getHoistingLimits`); and **fixed `real-hoist`'s border semantics** — a name in the limits marks a *border* whose descendants stay nested beneath it, not a leaf to block. (The earlier leaf-blocking behavior was the divergence flagged while porting; its unit tests were rewritten to the corrected semantics.)
Replaces the unmaintained `enquirer` package with `@inquirer/prompts` for all interactive CLI prompts. Fixes the `update -i` scrolling overflow bug where long choice lists were clipped in the terminal.
Fixes#6643
## User-facing changes
- **`pnpm update -i` / `pnpm update -i --latest`**: Scrolling now works correctly when many packages are available; the new library uses visual-line-aware pagination via `usePagination`
- **`pnpm audit --fix -i`**: Same scrolling fix for vulnerability selection
- **`pnpm approve-builds`**: Interactive build approval prompts updated
- **`pnpm patch`**: Version selection and "apply to all" prompts updated
- **`pnpm patch-remove`**: Patch removal selection updated
- **`pnpm publish`**: Branch confirmation prompt updated
- **`pnpm login`**: Credential prompts updated
- **`pnpm run` / `pnpm exec`** (with `verifyDepsBeforeRun=prompt`): Confirmation prompt updated
## Internal changes
- `OtpEnquirer` DI interface changed from `{ prompt }` to `{ input }`
- `LoginEnquirer` DI interface changed from `{ prompt }` to `{ input, password }`
- `enquirer` removed from catalog and all 8 package.json files
- `@inquirer/prompts` v8.4.3 added to catalog and all 8 package.json files
- Removed `OtpPromptOptions` and `OtpPromptResponse` exports from `@pnpm/network.web-auth` (no longer needed)
---------
Co-authored-by: Zoltan Kochan <z@kochan.io>
`pnpm install` (non-frozen) used to react to `ERR_PNPM_TARBALL_INTEGRITY` by logging the error, silently re-resolving from the registry, and overwriting the locked integrity. The lockfile's integrity was effectively advisory by default — a compromised registry, proxy, or republished version could substitute attacker-controlled content on a clean machine even though the project shipped a committed `pnpm-lock.yaml`.
Integrity mismatches against the lockfile now fail by default.
The **only** opt-in is **`pnpm install --update-checksums`** — a new flag, narrowly scoped to refreshing the locked integrity values. Mirrors yarn's flag of the same name. A warning still prints when the bypass takes effect so the rewrite stays auditable.
`--force` and `pnpm update` deliberately do **not** bypass the integrity check. They are routine refresh operations; silently overwriting a locked integrity in those flows would erase the protection a committed lockfile is supposed to provide. `--frozen-lockfile` behavior is unchanged. `--fix-lockfile` keeps its documented purpose (filling in missing lockfile entries) and is also not a bypass. Combining `--frozen-lockfile` with `--update-checksums` errors out — frozen mode refuses to rewrite the lockfile, which is exactly what `--update-checksums` is for.
`--update-checksums` also bypasses the resolver's on-disk metadata cache fast path (`pickPackage.ts:271`, `pick_package.rs:531`). Without that, a stale on-disk packument that already contained the pinned version would short-circuit the registry entirely and the flag would silently no-op on dev machines. With the gate, every first-encounter goes through a conditional GET; the in-memory cache is left alone so second-and-onward references within the same install still hit cached fresh data (one network round-trip per *unique* package, not per reference).
## Reported by
Reported privately via the security channel. The reproduction:
1. Publish `example-package@1.0.0` with content `v1` and install with pnpm; lockfile records the `v1` integrity.
2. Replace the registry's tarball+metadata for the same `1.0.0` with content `v2`.
3. On a clean store/cache, run `pnpm install`. Before this fix, pnpm logged `ERR_PNPM_TARBALL_INTEGRITY` but exited 0 with `v2` installed and the lockfile rewritten to the new integrity. After this fix, the same install exits non-zero.
## Prior art
- **npm** ([sebhastian](https://sebhastian.com/npm-err-code-eintegrity/)): hard-fails with `EINTEGRITY`. No dedicated override flag — recovery is `npm cache clean --force`, manually editing the lockfile, or deleting it.
- **yarn** ([Sean C Davis](https://www.seancdavis.com/posts/fix-yarn-integrity-check-failed/)): hard-fails with "Integrity check failed". Has a dedicated **`yarn install --update-checksums`** flag — pnpm now adopts the same name.
## Pacquet parity
Pacquet was already fail-hard on integrity mismatch by default (no auto-repair path to remove). This PR brings the rest of the surface into line so `pnpm install --update-checksums` keeps working when pacquet is the materialization target, and `pacquet install --update-checksums` behaves identically standalone:
- New `--update-checksums` flag on `pacquet install` (`crates/cli/src/cli_args/install.rs`), plumbed through `Install` and `InstallWithFreshLockfile` into the resolver.
- When the flag is set, pacquet skips the frozen-lockfile fast path and routes through the fresh-resolve path so locked integrity values get rewritten from the registry.
- `--frozen-lockfile + --update-checksums` errors with `pacquet_package_manager::frozen_lockfile_with_outdated_lockfile`, mirroring pnpm's `ERR_PNPM_FROZEN_LOCKFILE_WITH_OUTDATED_LOCKFILE`.
- `pacquet_tarball::verify_checksum_error` now carries a help hint pointing at `--update-checksums` and calling out the supply-chain implication, matching the updated pnpm `TarballIntegrityError`.
- The disk fast-path gate is mirrored in `crates/resolving-npm-resolver/src/pick_package.rs:531`, with the flag threaded from `ResolveOptions` → `PickPackageOptions`.
* feat(registry-access): extract setDistTag and dogfood from tests
Add `@pnpm/registry-access.commands#setDistTag` — the low-level PUT to
`/-/package/:pkg/dist-tags/:tag`. The CLI `dist-tag add` handler now
calls it instead of issuing the fetch inline.
Tests in this monorepo now use a thin new package
`@pnpm/testing.registry-mock` (REGISTRY_MOCK_PORT + REGISTRY_MOCK_CREDENTIALS
baked in) that delegates to `setDistTag`, replacing `addDistTag` from
`@pnpm/registry-mock`. That dropped helper relied on
`anonymous-npm-registry-client` and a verdaccio-era
fetch-then-DELETE-then-PUT dance that is no longer needed against
pnpm-registry.
39 test files swapped from `@pnpm/registry-mock` to
`@pnpm/testing.registry-mock`.
* fix: move setDistTag to its own package to break tsconfig project-reference cycle
testing/registry-mock → registry-access.commands → releasing/commands
→ installing/commands → installing/deps-installer → testing/registry-mock.
Extract setDistTag into @pnpm/registry-access.set-dist-tag (only depends
on @pnpm/error, @pnpm/network.fetch, @pnpm/npm-package-arg). Both
@pnpm/registry-access.commands and @pnpm/testing.registry-mock import
from it. Cycle gone.
* feat(registry-access): extract addUser helper, dogfood from login + tests
Add @pnpm/registry-access.add-user — a small helper that PUTs to
/-/user/org.couchdb.user:<name> and returns { token }. The CLI's
classicLogin (pnpm login fallback path) now calls it, and tests
use it via @pnpm/testing.registry-mock instead of the legacy
addUser from @pnpm/registry-mock.
Swapped 3 call sites: globalSetup.js, installing/deps-installer's
auth.ts, and pnpm/test/dlx.ts. AddUserHttpError exposes status +
text + parsed-json-if-applicable + headers so the CLI can still
do its OTP detection. One webauth-OTP login test mock had to be
adjusted to provide its body via `text` (JSON-stringified) rather
than `json` only, since the helper consumes the body via `text()`.
* refactor: consolidate set-dist-tag + add-user helpers into one @pnpm/registry-access.client package
One shared package is better than splitting per endpoint. Future endpoints
(publish, deprecate, etc.) can land here without another wrapper.
No behavioral change — same setDistTag and addUser exports as before,
just under one roof. Callers updated: registry-access.commands,
auth.commands, testing.registry-mock.
* fix(registry-access): sort imports
* fix(installing.commands): clarify "loose mode" wording in minimumReleaseAge log
The log line printed when pnpm auto-adds entries to
`minimumReleaseAgeExclude` referred to internal "loose mode" terminology,
which doesn't appear in the docs and isn't discoverable. Point users at
the actual setting name they need to flip.
Closes#11747
* Update installing/commands/src/policyHandlers.ts
Co-authored-by: Zoltan Kochan <z@kochan.io>
* fix(installing.commands): name the value in minimumReleaseAgeStrict log hint
Change "set minimumReleaseAgeStrict to gate these updates with a prompt"
to "set minimumReleaseAgeStrict to true to ..." so the value is explicit.
---------
Co-authored-by: shiminshen <16914659+shiminshen@users.noreply.github.com>
Co-authored-by: Zoltan Kochan <z@kochan.io>
* fix: cap lockfile verification memory and add trustLockfile opt-out
Verifying a multi-thousand-entry lockfile against `minimumReleaseAge`
or `trustPolicy: no-downgrade` retained every fetched packument in a
per-install cache for the entire install. On large workspaces this
OOM'd CI runners with a 2GB heap cap. Project both caches down to just
the fields each check reads (per-version trust evidence + the `time`
map for trust; package-level `modified` + version-name set for the
abbreviated shortcut) so the bulk packument is GC'd as soon as the
fetch returns.
Also adds a `trustLockfile` setting (default `false`) that skips the
verification pass entirely for environments where the lockfile is
already part of the trusted base. Mirrored in pacquet. Closes#11860.
* perf: share resolver packument cache with the lockfile verifier
The verifier kept its own per-install dedup Maps and re-fetched every
packument the resolver had already pulled during the same install.
Plumb the resolver's per-install `PackageMetaCache` through to the
verifier (via `createNpmResolutionVerifier` / `build_resolution_verifiers`)
so a name already in the resolver's LRU short-circuits the verifier's
disk/network round-trip — fast path only, the cached document is
projected for the trust check so the verifier's memory footprint stays
bounded.
In pnpm, `installing/client` now constructs one LRU and hands it to
both `createResolver` and `createResolutionVerifiers`. In pacquet, the
`InMemoryPackageMetaCache` is lifted to `Install::dispatch` and passed
to both `build_resolution_verifiers` and `InstallWithFreshLockfile`.
* fix(installing.commands): forward `pnpm install` flags to pacquet
When the install engine is delegated to pacquet via configDependencies,
pnpm hard-coded the args to `install --frozen-lockfile --reporter=ndjson`
and silently dropped the user's other CLI flags. `pnpm install --no-runtime`
therefore still installed the workspace's runtime devDependency, clobbering
the Node version the surrounding tooling had set up — visible as the
`Verify Node version` failure on PR #11765 where setup-pnpm provisions
Node 24.0.0 but pacquet then materializes node 24.6.0.
Pacquet's `install` subcommand already mirrors pnpm's surface for the
common flags (`--no-runtime`, `--prod`, `--dev`, `--no-optional`,
`--node-linker`, `--offline`, `--prefer-offline`, `--cpu`/`--os`/`--libc`).
Forward the user's argv verbatim when the command is `install`/`i`;
`add`/`update`/`dedupe` still don't forward — their flag surfaces don't
line up with pacquet's `install`.
* fix(installing.commands): pass --ignore-manifest-check to pacquet
`pnpm up` / `add` / `remove` were aborting with
`pacquet_package_manager::outdated_lockfile` whenever pacquet was
declared in `configDependencies`. After resolving and writing the
updated lockfile, pnpm hands materialization off to pacquet but
hasn't yet written the post-mutation `package.json` — that write
happens after `mutateModules` returns. Pacquet's frozen-lockfile
freshness gate then saw the new lockfile paired with the
pre-mutation manifest and refused to install.
Pass pacquet's new `--ignore-manifest-check` flag (pacquet PR #11811)
on every delegation. The flag is narrow: it only skips
`satisfies_package_manifest`. Settings drift like `overrides` is
still enforced, and pnpm already re-validated the lockfile before
delegating, so re-checking the manifest here was redundant work that
only ever fired false positives on the mutate-then-materialize path.
Requires a pacquet release that ships the flag; bump
`PACQUET_VERSION` in `pnpm/test/install/pacquet.ts` once it does, or
the existing e2e tests will fail against pacquet 0.2.2-9 (which
doesn't recognize the flag and clap would reject).
Closes#11797.
---
Written by an agent (Claude Code, claude-opus-4-7).
* fix: update pacquet in tests
* fix(installing.commands): strip positionals + always-injected flags when forwarding to pacquet
`collectForwardedFlags` checked `argv[0] === 'install'` to find the
command token to strip. Any global flag the user typed before `install`
(e.g. `--config.registry=...` in the e2e test) shifted the token out
of position, so the function returned the full argv and pacquet saw
`install` twice — `error: unexpected argument 'install' found`.
Use the parsed argv that `@pnpm/cli.parse-cli-args` already produced:
`remain` lists positionals (the `install`/`i` token and nothing else
on this code path, since `isInstallCommand` is only true when no
package params are present), and `original` preserves the user's
exact tokens. Drop positionals + the flags we always inject
(`--reporter=ndjson`, `--frozen-lockfile`, `--ignore-manifest-check`)
so clap doesn't reject duplicates either.
`original` over `cooked` deliberately: nopt's `cooked` splits
`--key=value` into two tokens, which would break pacquet's
`--config.<key>=<value>` parser (it requires the `=` form).
* fix(installing.commands): make argv.cooked/remain optional on InstallCommandOptions
Widening these to required broke test fixtures elsewhere (publish/pack/
deprecate/dist-tag/deploy) that construct minimal `argv: { original }`
options for code paths that never reach pacquet. Only the pacquet
delegation actually reads `remain`, so make the two new fields optional
on the shared options type and supply a default at the runPacquet call
site. The runtime path through main.ts already populates all three.
* fix(installing.commands): strip any user-supplied --reporter when forwarding to pacquet
Pacquet's `--reporter` is a clap value option with last-value-wins
semantics, so `pnpm install --reporter=silent` (or
`--reporter silent` two-token form) reached pacquet and overrode
the `--reporter=ndjson` pnpm injects, breaking the NDJSON-to-
streamParser plumbing the default reporter depends on. The previous
filter only matched the exact `--reporter=ndjson` token.
Walk argv with a lookahead so both `--reporter=<value>` and
`--reporter <value>` are dropped without consuming an adjacent flag.
* fix(installing.commands): drop negated/value forms of always-injected flags
`collectForwardedFlags` only matched the exact positive tokens
`--frozen-lockfile` and `--ignore-manifest-check`, so a user typing
`pnpm install --no-frozen-lockfile` (or `--frozen-lockfile=false`)
forwarded the negation to pacquet, which then saw both our injected
`--frozen-lockfile` and the user's `--no-frozen-lockfile` and crashed
clap with "unexpected argument".
Match every shape the user can write the same flag in: positive,
`--no-` negated, and any `=value` form. Can't blindly strip `--no-`
either way — pacquet has flags whose literal name starts with `no-`
(`--no-runtime`, `--no-optional`); those must still forward.
The user's `--no-frozen-lockfile` intent is honored upstream — pnpm
did a fresh resolve before delegating; pacquet's role here is just
lockfile-driven materialization, which is always frozen.
* fix(installing.commands): match positionals by index, hide reporter from dropped-flags warning
`collectForwardedFlags` matched positionals via `new Set(argv.remain)`,
which strips by value: a flag value that happened to equal a
positional token (e.g. `pnpm install --node-linker install`) was
wrongly dropped from the forwarded list, costing pacquet the value
of `--node-linker`. Walk `argv.original` with a subsequence pointer
into `argv.remain` so only the actual positional indexes get skipped.
`collectDroppedFlags` still surfaced `--reporter foo` / `--reporter=foo`
in the "may not be honored" warning on `add`/`update`/`dedupe`, but
pnpm honors reporter selection itself before delegation — so the
warning was misleading. Route both helpers through the same
`isAlwaysInjected` check and consume `--reporter` and its value the
same way `collectForwardedFlags` already does.
* fix: handle release-age policy in global installs
* refactor: dedupe global policy-callback wiring
Collapse setupPolicyHandlers + createResolutionPolicyManifestUpdater into
one createGlobalPolicyCallbacks helper used by both global add and global
update entry points.
---
Written by an agent (Claude Code, claude-opus-4-7).
---------
Co-authored-by: Zoltan Kochan <z@kochan.io>
When `configDependencies` declares pacquet (under either the unscoped `pacquet` or the scoped `@pnpm/pacquet` alias), pnpm delegates the fetch / import / link / build phases of an install to the pacquet Rust binary. Pnpm keeps owning dependency resolution — pacquet's resolver isn't ready yet — and hands pacquet a freshly-written lockfile to materialize.
Covered install shapes:
- frozen install (`tryFrozenInstall` → pacquet, no resolve needed)
- default isolated `nodeLinker` (`installInContext`: lockfileOnly resolve via JS, then pacquet)
- hoisted `nodeLinker` (same resolve-then-materialize shape)
- workspace partial install (subset of workspace projects mutated)
- agent-server install (`@pnpm/agent.client` resolves, pacquet materializes)
```yaml
# pnpm-workspace.yaml
configDependencies:
"@pnpm/pacquet": "^0.2.0" # or unscoped `pacquet`
```
## How it works
- `installing/commands/src/runPacquet.ts` resolves the platform binary via `createRequire(realpath(.pnpm-config/<name>/package.json))` — same algorithm the upstream wrapper uses, but skipping the JS shim's extra Node startup.
- Pacquet's NDJSON stderr is forwarded through `@pnpm/logger`'s global `streamParser` so `@pnpm/cli.default-reporter` renders its events the same way it renders pnpm's own. Non-JSON stderr lines pass through verbatim.
- A few pnpm-side log emits (`importing_done` placeholder, `pnpm:summary`) are suppressed when pacquet will take over so the reporter doesn't close streams or lock in empty diffs before pacquet's real events arrive. Pacquet's duplicate `pnpm:progress status:resolved` events are filtered on the resolve-then-materialize paths so the reporter doesn't double-count.
- `installing/deps-installer/src/install/index.ts` gates the delegation on a `runPacquet?: () => Promise<void>` callback in `StrictInstallOptions`. The CLI layer in `installing/commands/src/installDeps.ts` constructs the callback, threaded through both the single-project and workspace-recursive paths.
- The `pacquet` and `@pnpm/pacquet` npm packages ship the same JS shim from `pacquet/npm/pacquet/scripts/generate-packages.mjs`; per-platform binaries stay under the existing `@pacquet/<plat>-<arch>` scope and aren't duplicated.
Three coordinated changes that close the silent-bypass gap in loose `minimumReleaseAge` mode AND the discover-by-loop UX problem in strict mode (#10488), plus a parallel hardening of the lockfile verifier:
1. **Auto-collect into `minimumReleaseAgeExclude` (loose mode)** — fresh resolutions that fall back to a version newer than the cutoff are auto-recorded into the workspace manifest's `minimumReleaseAgeExclude`. A single info message lists what was persisted. The workspace manifest writer dedupes against existing entries.
2. **Lockfile verifier runs in loose mode too** — `createNpmResolutionVerifier` no longer gates on `minimumReleaseAgeStrict`. With auto-collect keeping the exclude list explicit, every accepted-immature pin must be on the list — same contract strict mode enforces. Lockfiles produced under a weaker (or absent) policy that still hold immature entries are rejected the same way strict mode would.
3. **Strict mode prompts on the aggregate set instead of throwing on the first** — the resolver always collects every immature direct and transitive in one pass; the install command's `handleResolutionPolicyViolations` checkpoint decides what to do with the set. Interactive (TTY) prompts the user once with the full list (default = No) and asks whether to add them all to `minimumReleaseAgeExclude` and proceed. Approve → install continues, persisted at the end. Decline → resolution aborts before the lockfile, package.json, or modules dir is touched. Non-interactive (CI) keeps `ERR_PNPM_NO_MATURE_MATCHING_VERSION` as the exit code but lists every offending entry instead of just the first one the resolver happened to hit.
4. **The lockfile verifier now also covers `trustPolicy: 'no-downgrade'`.** The same post-resolution gate that re-checks `minimumReleaseAge` on lockfile entries now re-runs `failIfTrustDowngraded` for every npm-registry entry whose name isn't on `trustPolicyExclude`. The two checks share a single full-metadata fetch per package, so the extra coverage doesn't cost an extra round trip when both policies are active. Resolver-time trust checks still run as before — this just closes the gap when an entry bypasses resolution (peek path, `--frozen-lockfile`, restored CI cache).
The steady-state flows:
- **Loose mode, `pnpm add foo@immature`**: lockfile clean, verifier no-op, resolver picks via lowest-version fallback, `foo@immature` lands in `minimumReleaseAgeExclude`, install succeeds. Subsequent `pnpm install --frozen-lockfile` in CI verifies against the populated list and succeeds.
- **Strict mode (interactive), security bump to `next@15.5.9`**: resolver collects `next@15.5.9` AND every immature `@next/swc-*@15.5.9` shim. pnpm prompts once with the full list. User approves → install completes, all entries persisted in `pnpm-workspace.yaml`. CI then runs the populated config cleanly.
- **Strict mode (non-interactive / CI)**: aborts with `ERR_PNPM_NO_MATURE_MATCHING_VERSION` listing every immature entry's `name@version` and publish time — no more discover-by-loop dance.
- **Teammate commits a poisoned lockfile**: single-policy batches reject with `ERR_PNPM_MINIMUM_RELEASE_AGE_VIOLATION` (or `ERR_PNPM_TRUST_DOWNGRADE`); a batch that trips both policies escalates to the generic `ERR_PNPM_LOCKFILE_RESOLUTION_VERIFICATION` and lists each entry's per-policy code in the breakdown.
### Implementation
- The npm resolver always falls back to the lowest matching version when no mature version satisfies the range, and flags the result with `ResolveResult.policyViolation` instead of throwing `NO_MATURE_MATCHING_VERSION`. `deferImmatureDecision` and `strictPublishedByCheck` are gone — every caller (install, dlx, outdated, self-update) inspects the violation and decides what to do.
- `policyViolation` flows from `ResolveResult` → `PackageResponse.body.policyViolation` → a shared accumulator in `ResolutionContext` → the `resolutionPolicyViolations` field on `resolveDependencyTree`'s return → out through `mutateModules` / `addDependenciesToPackage` to the install command.
- The violation type lives in `@pnpm/resolving.resolver-base` as `ResolutionPolicyViolation`; the npm resolver exports the two built-in codes (`MINIMUM_RELEASE_AGE_VIOLATION_CODE`, `TRUST_DOWNGRADE_VIOLATION_CODE`) as constants so consumers reference one source of truth.
- `handleResolutionPolicyViolations` runs between `resolveDependencyTree` and `resolvePeers` — the resolver-agnostic checkpoint where the install command's plan prompts (TTY) or aborts (no-TTY) with the full violation list.
- `setupPolicyHandlers` (in `installing/commands/src/policyHandlers.ts`) composes per-policy handlers behind a uniform plan interface: each handler has its own `handleResolutionPolicyViolations` (filter by code, decide what to do) and `pickManifestUpdates` (return a typed `WorkspaceManifestPolicyUpdates` patch the install command spreads into `updateWorkspaceManifest`). Today the only registered handler is `createMinimumReleaseAgeHandler` — strict + TTY prompts via `enquirer`, strict no-TTY throws `ERR_PNPM_NO_MATURE_MATCHING_VERSION` with every entry listed, loose mode auto-persists at the tail. Strict + `--no-save` is rejected up-front via `ERR_PNPM_STRICT_MIN_RELEASE_AGE_REQUIRES_SAVE`. Future policies plug in via a sibling factory + push into the handlers list, with no changes to `installDeps.ts` / `recursive.ts`.
- `installDeps` / `recursive` drain `pickManifestUpdates` after install and spread the patch into `updateWorkspaceManifest`. Plain `pnpm install` (no `--update`, no params) now still updates the workspace manifest when any handler contributes a patch. The `install` command's CLI schema gained `save: Boolean` so `--no-save` actually flows through to `opts.save = false` instead of being silently dropped by nopt.
- `makeResolutionStrict` (in `installing/client`) wraps a `ResolveFunction` and rethrows any `policyViolation` as a `PnpmError`. Used by `dlx` and `self-update` under strict `minimumReleaseAge` OR `trustPolicy: 'no-downgrade'`, since one-shot callers have nowhere to defer a violation to. Violation-code → error-code mapping lives in one place so future violation kinds get consistent UX.
- `createNpmResolutionVerifier` extends its check to `trustPolicy: 'no-downgrade'` — same per-entry fan-out, same cache key, sharing the full-metadata fetch with the maturity check. Trust-fetch errors now propagate up so the violation reason carries the underlying message (network code, 404 detail) instead of a generic "metadata is unavailable".
- `verifyLockfileResolutions`'s aggregate throw uses the per-policy code when every violation in the batch shares it, and escalates to a generic `LOCKFILE_RESOLUTION_VERIFICATION` (with per-entry codes in the breakdown) for mixed batches.
- The pnpm agent path refuses installs under `trustPolicy: 'no-downgrade'` (`ERR_PNPM_TRUST_POLICY_INCOMPATIBLE_WITH_AGENT`) — the agent has no server-side counterpart to that check yet, so silently allowing it would land a lockfile the local verifier would later reject. `minimumReleaseAge` is forwarded to the agent and enforced server-side, so that combination is fine.
### Pacquet parity
Pacquet only carries a stub reference to `minimumReleaseAgeExclude` (see `pacquet/crates/package-manager/src/version_policy.rs`); the broader `minimumReleaseAge` and `trustPolicy` policies aren't ported yet, so this feature is outside pacquet's current surface area. It'll come along when pacquet ports the policies.
### Closes
- Closes#10488 (resolves the discover-by-loop dance for security bumps without needing `withTransitives`).
Closes#11687.
## What
Cache the result of the post-resolution lockfile verification gate (#11583) so repeat installs against an unchanged lockfile skip the per-package registry round trips entirely. Persisted as JSON Lines at `<cacheDir>/lockfile-verified.jsonl`.
The cache layer is policy-neutral. Today there's one verifier (`minimumReleaseAge`); future resolver-side verifiers (jsr trust, attestation, …) plug in by declaring their own `policy` slot and `canTrustPastCheck` comparator — no install-side changes.
## Why
#11583 re-hits the registry on every install for every locked (name, version) pair. On warm/repeat installs where the lockfile hasn't moved, that's a stack of per-package round trips with nothing to show for them. This change makes the steady-state case effectively free without weakening the protection — the gate still runs in full whenever the lockfile changes, any verifier's policy tightens, or no record exists.
## How
### Cache lookup, in order
The cache is **indexed by content hash** so git worktrees with identical lockfile bytes share a cache entry. A secondary path-keyed index drives the same-machine stat shortcut.
1. **`stat()` shortcut** — when a previous record for this exact `lockfilePath` matches today's `size + mtime + inode`, trust the cached hash without reading anything. Zero I/O beyond the stat. Microseconds.
2. **Content lookup** — hash the in-memory lockfile (not the file bytes — we already have the parsed object) and look up by content hash. Catches worktrees (same content, different path) and CI checkouts (same content, reset stat). On hit, append a refreshed path/stat entry so the next install at this path takes the stat shortcut.
3. **Any active verifier rejects the cached `policy`** — run the full gate.
4. **No record** — run the full gate.
The in-memory object is hashed with `hashObject` from `@pnpm/crypto.object-hasher` (streaming, key-order-stable).
### Record shape
```json
{
"lockfile": {
"hash": "<sha256 base64>",
"path": "/abs/path/to/pnpm-lock.yaml",
"size": 154,
"mtimeNs": "1736245123000000000",
"inode": "12345"
},
"verifiedAt": "2026-05-17T...",
"policy": { "minimumReleaseAge": 1440 }
}
```
`policy` is the union of every active verifier's `policy` contribution. Verifiers checking the same logical policy (e.g. `minimumReleaseAge` honored by multiple registries) name it the same and share the slot — no resolver namespacing.
### File semantics
- **Sync fs throughout** — the cache is consulted once before verification fan-out and recorded once after. No concurrent install work to overlap with; keeping the call sites straight-line.
- **JSONL appends are atomic** on POSIX/NTFS, so parallel pnpm processes (monorepo installs, CI matrices sharing a cache) write without coordination. Latest record per `(path, hash)` tuple wins on read.
- **Bounded file** — capped at ~1000 entries; compaction is triggered by a single `stat()` of the cache file (1.5 MiB byte budget) so we never parse the file on the steady-state path. When triggered, the tail is rewritten via tempfile + rename.
- **No record on rejection** — a failing verification deliberately doesn't write a record; the next install must rerun the gate.
- **Single hash per install** — the in-memory hash is computed lazily and reused: `tryLockfileVerificationCache` returns the precomputed stat+hash to `recordVerification` on a miss, and the stat-shortcut hit forwards the cached record's hash unchanged.
## Plumbing
The verifier contract changed alongside the cache to make this composable without install-side knowledge of each policy:
- **`@pnpm/resolving.resolver-base`** — `ResolutionVerifier` is now `{ verify, policy, canTrustPastCheck }` (was a bare function in #11583). Each resolver-side verifier owns its policy snapshot and the comparator that decides whether a cached policy is still trustworthy.
- **`@pnpm/resolving.npm-resolver`** — `createNpmResolutionVerifier` returns the new shape: `policy: { minimumReleaseAge }`, `canTrustPastCheck` reads `minimumReleaseAge` from the merged cached bag.
- **`@pnpm/resolving.default-resolver`** — `createResolutionVerifier` (singular, returning a combined function) → `createResolutionVerifiers` (plural, returning a `ResolutionVerifier[]`). No combinator; each verifier handles its own protocol short-circuit inside `verify`, so dispatch happens naturally at the install side.
- **`@pnpm/installing.client`** — `Client.verifyResolution?` → `Client.resolutionVerifiers: ResolutionVerifier[]`. Same rename propagates through `@pnpm/store.connection-manager`, `@pnpm/testing.temp-store`, and `StrictInstallOptions`.
- **`@pnpm/installing.deps-installer`** — new `verifyLockfileResolutionsCache.ts` (`tryLockfileVerificationCache` + `recordVerification`). `verifyLockfileResolutions` takes the verifier list plus `cacheDir` + `lockfilePath` as flat options; the cache fires when both are present, otherwise the gate runs without memoization. The dedup key for in-flight candidates includes a serialization of `resolution` so two entries sharing a (name, version) but pinned via different protocols don't collapse.
Breaking but safe — `@pnpm/resolving.npm-resolver` hasn't been released since #11583 introduced the verifier abstraction, so no downstream consumer is on the old shape.
## Tests
- **17 unit tests** in `verifyLockfileResolutionsCache.ts`: cache miss/hit, stat shortcut, size mismatch falling through to hash lookup, hash-fallback on reset stat, content change with matching size, stricter/weaker policy, missing-field policy rejection, multi-verifier policy merge (shared field stored once), worktree case (same content, different path), JSONL append semantics, malformed-line tolerance.
- **12 integration tests** in `verifyLockfileResolutions.ts`: dedup of peer/patch-suffix variants, distinct-resolution dedup at the same (name, version), stable violation ordering, the 20-entry cap, multi-verifier fan-out (first failure wins), cache short-circuit on a passing run, no cache write on a rejecting run, empty-verifier-list passthrough.
- **1 e2e test** in `pnpm/test/install/minimumReleaseAge.ts`: bundled CLI plumbing — install once to seed the lockfile, enable `minimumReleaseAge` + `cacheDir`, install again, assert the cache file lands at `<cacheDir>/lockfile-verified.jsonl` with the documented record shape.
- Existing `minimumReleaseAge` (13) and `frozenLockfile` (12) suites still pass.
Closes#10438.
## What
Re-verify every entry in `pnpm-lock.yaml` against the policies the resolver chain was configured with — today: `minimumReleaseAge` in strict mode — right after the lockfile is loaded from disk and before any tarball is fetched. A locked version that fails the policy aborts the install with `ERR_PNPM_MINIMUM_RELEASE_AGE_VIOLATION`; `minimumReleaseAgeExclude` is honored.
## Why
The policy only fires while pnpm is *choosing* a version. Once a version is pinned in the lockfile — e.g. a developer disabled the policy locally and committed a fresh dependency, or a CI cache restored a stale lockfile — every later `pnpm install` (including `--frozen-lockfile` and `pnpm fetch`) installs it without re-checking, which defeats the supply-chain protection the setting is supposed to provide.
The threat model is **a lockfile someone else resolved**, not local resolution: local resolution is already covered by the resolver's own per-version filter. bun fixed the same shape of bug in [oven-sh/bun#30526](https://github.com/oven-sh/bun/pull/30526); this PR is the pnpm side.
## How
The fix introduces a generic `ResolutionVerifier` abstraction in the resolver chain — each resolver factory can ship a sibling verifier factory, exactly the way each resolver ships a `resolve` function. Today there's one verifier (npm); the shape leaves room for future ones (jsr, attestation-based, etc.) without changing the install-side interface.
- **`@pnpm/resolving.resolver-base`** exports the `ResolutionVerifier` / `ResolutionVerification` types — the shared contract.
- **`@pnpm/resolving.npm-resolver`** exports `createNpmResolutionVerifier`. Returns `undefined` when no policy is active, so callers can cheaply decide whether to iterate at all. When active, it inspects each lockfile entry, handles `minimumReleaseAgeExclude`, routes through named-registry prefixes (built-ins like `gh:` merged in), and uses `fetchFullMetadataCached` to fetch full registry metadata — decoupled from the resolver pipeline so neither `peekManifestFromStore` nor abbreviated metadata can hide the publish timestamp.
- **`@pnpm/resolving.default-resolver`** exports `createResolutionVerifier`, a combinator that asks each underlying verifier (today: npm) if it has work and returns `undefined` when none does. Designed so that adding more verifiers later doesn't change the install side.
- **`@pnpm/installing.client`** exposes `verifyResolution` on `Client`, built from the same `fetchFromRegistry` / `getAuthHeader` the resolver chain already uses — **no second fetcher is constructed**.
- **`@pnpm/store.connection-manager`** and **`@pnpm/testing.temp-store`** surface `verifyResolution` alongside the store controller they hand back, so it reaches `mutateModules` through the existing plumbing.
- **`@pnpm/installing.deps-installer`** gains one option on `StrictInstallOptions`: `verifyResolution?: ResolutionVerifier`. `mutateModules` invokes `verifyLockfileResolutions(ctx.wantedLockfile, opts.verifyResolution)` **once**, right after `getContext` returns the on-disk lockfile and before any path branches. When the verifier is `undefined`, the call is a no-op. The iteration is policy-neutral: dedupes by `(name, version)`, applies `pLimit(16)`, sorts violations stably, caps the printed list at 20 with an `…and N more` summary, throws a `PnpmError` carrying the verifier-supplied error code.
The error includes a recovery hint that points at `pnpm clean --lockfile` followed by `pnpm install` — the safe way to throw away a poisoned lockfile and rebuild from fresh resolution.
## Tests
- **9 unit tests** for `verifyLockfileResolutions` against a mock `ResolutionVerifier` — dedup, aggregation, stable ordering, the 20-entry cap, no-op behavior, the verifier-supplied error code surfacing in `PnpmError`.
- **13 integration tests** in `installing/deps-installer/test/install/minimumReleaseAge.ts` via the real `install()` entry — `testDefaults()` wires `verifyResolution` from `createTempStore` → `createClient`, so the npm verifier runs end-to-end at the install boundary. Covers the rejection scenario, `minimumReleaseAgeExclude`, the strict-mode toggle, the existing `minimumReleaseAge` resolver-side suite, and a `pnpm add` scenario where a pre-existing entry would otherwise survive resolution.
- **3 e2e tests** in `pnpm/test/install/minimumReleaseAge.ts` against the bundled CLI: rejection path with the right `ERR_PNPM_*` code and `pnpm clean --lockfile` hint in output, `minimumReleaseAgeExclude` honored, and the strict-off path (which now requires an explicit `minimumReleaseAgeStrict: false` since the config reader auto-enables strict mode when `minimumReleaseAge` is set).
- Existing `frozenLockfile` suite (12 tests) and npm-resolver suite (179 tests) still pass.
---------
Co-authored-by: Zoltan Kochan <z@kochan.io>
Adds a `--no-runtime` flag (config: `runtime: boolean`, default `true`) that suppresses install of runtime entries declared via `devEngines.runtime` (the `runtime:` protocol) **without modifying the lockfile**.
The lockfile keeps the runtime entry, so frozen-lockfile validation still passes; only the runtime fetch and `.bin` linking are skipped. Useful in CI matrices where the runtime is provisioned externally (e.g. via `pnpm runtime -g set node <version>`) before `pnpm install` runs.
The existing `--runtime-on-fail=ignore` is unsuitable for this case: it mutates the manifest and regenerates the lockfile to drop the runtime entry, which trips frozen-lockfile validation. The two flags are orthogonal and serve different purposes.
### Implementation
The hook lives in the lockfile filter stage:
- `lockfile/filtering/src/filterImporter.ts` — strips `runtime:` refs from the importer's deps maps when `skipRuntimes` is set.
- `lockfile/filtering/src/filterLockfileByImportersAndEngine.ts` — new `skipRuntimes?: boolean` option; runtime-protocol direct deps are dropped before they enter `pickedPackages`, so they never reach the dep graph or bin-linker. Applies to all runtimes (`node`, `deno`, `bun`) since they share the `runtime:` protocol prefix.
The option is plumbed through `installing/deps-restorer`, `installing/deps-installer`, and `installing/commands` to the user-facing `pnpm install --no-runtime` flag.
### Example
```json
// package.json
{
"devEngines": {
"runtime": {
"name": "node",
"version": "22.13.0",
"onFail": "download"
}
}
}
```
Local dev: `pnpm install` — installs node 22.13.0 as before.
CI matrix entry:
```yaml
- run: pn runtime -g set node ${{ matrix.node }}
- run: pn install --no-runtime
```
The lockfile is unchanged; the matrix's externally-provisioned node is used.
- Upgrade `@pnpm/semver-diff` and `@pnpm/colorize-semver-diff` to v2, which expose the helpers as named exports.
- Update the call sites in `@pnpm/deps.inspection.commands` and `@pnpm/installing.commands` from `semverDiff.default(...)` / `colorizeSemverDiff.default(...)` to plain `semverDiff(...)` / `colorizeSemverDiff(...)`.
- Refactor `buildPkgChoice` in `getUpdateChoices.ts` to build the row as a `string[]`. Previously the row was an object whose values relied on `nextVersion` being inferred as `any` (a side effect of the broken `.default` access poisoning the type) — that masked `outdatedPkg.current` and `outdatedPkg.workspace` being `string | undefined`. With the v2 named imports the types tighten up, and `Object.values(lineParts)` would no longer assign cleanly to `string[]`.
The previous v1 packages exported their helpers as `module.exports.default = fn`, so `.default(...)` only worked through the legacy CJS interop — and it broke under Node.js ESM (which is what the Jest runner uses with `--experimental-vm-modules`). Most of the `deps/inspection/commands` outdated tests had been silently failing on `main` with `TypeError: semverDiff.default is not a function`; this change brings them back.
Closes#11488.
`pnpm fetch` writes forced-empty `hoistPattern: []` and `publicHoistPattern: []` into `.modules.yaml` (because its `virtualStoreOnly` install path skips hoisting). In v10 the follow-up `pnpm install` ignored these unless the user had explicitly set a hoist-pattern in their config. v11's [#11199](https://github.com/pnpm/pnpm/pull/11199) removed that explicit-config gate, so `validateModules` now always sees the empty patterns as a hoist-pattern change and purges `node_modules` — slow on every CI run, and per the bug report sometimes leaves the modules dir in an `ERR_MODULE_NOT_FOUND` state on subsequent runs.
The fix marks `.modules.yaml` with a new `virtualStoreOnly: true` field after a fetch. `validateModules` recognizes this flag as "incomplete install state" and skips the `PUBLIC_HOIST_PATTERN_DIFF` / `HOIST_PATTERN_DIFF` comparisons. The next install then completes the missing post-import linking in place rather than purging. The flag is dropped from `.modules.yaml` once a normal install runs.
A genuine hoist-pattern change (without a fetch in between) still triggers the purge as before — verified manually with `publicHoistPattern` in `pnpm-workspace.yaml`.
`pnpm dlx` (and `pnpx`/`pnx`/`pnpm create`) now mirrors the `pnpm add -g` flow when the launched package's transitive deps have install scripts:
- dlx overrides `strictDepBuilds: false` for its install so the v11 default no longer turns ignored builds into an `ERR_PNPM_IGNORED_BUILDS` error. Without this, `pnpx @google/gemini-cli` (and similar — `node-pty`, `@github/keytar`) failed outright and forced users to retry with `--allow-build=<pkg>` for every offending dependency.
- After install, dlx detects skipped builds via `getAutomaticallyIgnoredBuilds` and runs the same interactive `approve-builds` prompt as `pnpm add -g`. In non-interactive mode the install is committed with builds skipped, matching `pnpm add -g` in CI; users who need those scripts can re-invoke with `--allow-build=<pkg>` to force a fresh cache key.
- If the install errors for unrelated reasons (network, etc.) the partially-populated prepare directory is removed so the next dlx run starts clean.
Closes#11444.
### Plumbing
- Exports `getAutomaticallyIgnoredBuilds` from `@pnpm/building.commands` so dlx can detect skipped builds without re-implementing modules-yaml reading.
- Adds `strictDepBuilds` (optional) to `InstallCommandOptions` — already accepted at runtime via the spread, this just makes it explicit at the type level so callers can override it.
Drops `getNodeExecPath`, `getNodeExecPathInBinDir`, and
`getNodeExecPathInNodeDir` along with their now-unused `which` dependency.
None of these helpers were referenced anywhere in the codebase.
* chore: upgrade @typescript/native-preview to 7.0.0-dev.20260421.2
- Add explicit `types: ["node"]` to the shared tsconfig because tsgo
20260421 no longer auto-acquires `@types/*` from `node_modules`.
- Refactor test files to explicitly import jest globals (`describe`,
`it`, `test`, `expect`, `beforeEach`, etc.) from `@jest/globals`
instead of relying on `@types/jest` ambient declarations. Under the
new tsgo build, `import { jest } from '@jest/globals'` shadows the
ambient `jest` namespace, breaking `@types/jest`'s `declare var
describe: jest.Describe;` globals.
- Add `@jest/globals` to each package's devDependencies where tests
now import from it, and add `@types/node` to packages that need it
but were relying on hoisted resolution.
- Replace `fail()` calls with `throw new Error(...)` since `fail` is
no longer globally available.
* chore: fix remaining tsgo type-strictness errors
- Strip `as <PnpmType>` casts on objects passed to toMatchObject /
toStrictEqual / toEqual; @jest/globals rejects the typed objects
(which include AsymmetricMatchers) vs. the repo-specific type.
- Type `jest.fn<...>()` explicitly where the mock's signature matters
for toHaveBeenCalledWith.
- Replace `beforeEach(() => X)` with `beforeEach(() => { X })` so the
return value is void, as the stricter jest typing requires.
- Use `expect.objectContaining({...})` in one place where the full
expected object triggered stricter type resolution.
- Cast `prompt.mock.calls` arg through `as unknown as Record<...>[]`
for patch.test.ts's nested-array matchers.
- Fix off-by-one `<reference path>` in pnpm/test/getConfig.test.ts
that only surfaced now.
- Move `@jest/globals` from devDependencies to dependencies in the
two `__utils__` packages that import it from `src/`.
- Clean up unused imports from the @jest/globals migration.
* chore: address Copilot review on #11332
- Move misplaced `@jest/globals` imports to the top import block in
checkEngine, run.ts, and workspace/root-finder tests where the
script dropped them below executable code.
- Replace `try { await x(); throw new Error('should have thrown') } catch`
in bins/linker, lockfile/fs, and resolving/local-resolver tests with
`await expect(x()).rejects.toMatchObject({...})`. The old pattern
swallowed an unrelated `throw` if the under-test call silently
succeeded, which would fail on the catch-block assertion with a
misleading message.
Library packages had `prepublishOnly: pn compile`, which expands to
`tsgo --build && pn lint --fix`. During `pn release` that runs eslint
against ~150 packages for no benefit — the code has already been linted
in CI and the release flow's upfront compile has already built dist/.
Switch lib prepublishOnly to a bare `tsgo --build` so the safety-net
compile stays but the per-package eslint cost is gone.
## Summary
Adds an opt-in **pnpm agent** server that resolves dependencies server-side and streams only the files missing from the client's content-addressable store.
- **`@pnpm/agent.server`** — multi-process HTTP server (Node.js `cluster`) with SQLite-backed metadata and file caches
- **`@pnpm/agent.client`** — streams an NDJSON response, dispatches worker threads to fetch files while the server is still resolving
- **New config**: `agent` in `pnpm-workspace.yaml` (opt-in)
## How it works
1. Client reads integrity hashes from its local store index
2. Sends `POST /v1/install` with dependencies + store integrities
3. Server resolves the dependency tree using pnpm's `install({ lockfileOnly: true })`, with a SQLite-backed `PackageMetaCache` for fast repeat resolution
4. As each package resolves, a wrapped `storeController.requestPackage` looks up its files and immediately streams digests the client is missing (NDJSON `D` lines)
5. Client reads the stream line by line; digest batches fill up and dispatch worker threads to `POST /v1/files` — file downloads overlap with server-side resolution
6. After resolution, server sends index entries (`I` lines) and lockfile (`L` line)
7. Client writes index entries to store, then runs headless install with a wrapped `fetchPackage` that calls `readPkgFromCafs` with `verifyStoreIntegrity: false` (files are trusted from the agent)
8. `/v1/files` response is gzip-streamed (274MB → ~80MB) — server pipes through `createGzip`, worker pipes through `createGunzip`, parsing and writing files to CAFS as data arrives
## Performance
1351-package project, cold local store, warm server (localhost):
| Scenario | Time |
|----------|------|
| Vanilla pnpm install (cold OS cache) | ~48s |
| Vanilla pnpm install (warm OS cache) | ~34s |
| With pnpm agent (consistent) | **~33s** |
### Key optimizations
1. **SQLite metadata cache** — server-side resolution drops from ~3.4s to ~0.9s
2. **SQLite file store** — consistent read performance regardless of OS file cache state
3. **Streaming `/v1/install`** — file digests stream during resolution, downloads start before resolution finishes
4. **Gzip-streamed `/v1/files`** — whole-stream gzip (274MB → ~80MB), significant savings on remote servers
5. **Worker-thread streaming HTTP** — workers pipe gzip → parse → write to CAFS as data arrives, no buffering
6. **No rehashing** — server-provided digests used directly, skipping 33K SHA-512 computations
7. **No re-verification** — wrapped `fetchPackage` calls `readPkgFromCafs` with `verifyStoreIntegrity: false`
8. **Direct `writeFileSync` with `wx`** — no stat + temp + rename
9. **Pre-packed msgpack** — server sends raw store index buffers, client writes directly to SQLite
10. **WAL checkpoint** — ensures store index entries written by agent are visible to headless install's worker threads
## Usage
Start the server:
```bash
node agent/server/lib/bin.js
```
Configure in `pnpm-workspace.yaml`:
```yaml
agent: http://localhost:4873
```
* feat: add runtimeOnFail setting
Adds a `runtimeOnFail` config setting ('ignore' | 'warn' | 'error' |
'download') that overrides the `onFail` field on `devEngines.runtime`
and `engines.runtime` in the root project's package.json. This makes
it possible to opt into (or out of) runtime auto-download without
changing the project manifest.
* fix: skip runtime download when version is missing
Without a version, convertEnginesRuntimeToDependencies would write
`runtime:undefined` into the manifest. Warn and skip instead.
* feat: apply runtimeOnFail override during install
The config reader override only mutates the context's rootProjectManifest,
but installDeps reads the manifest fresh via tryReadProjectManifest and
findWorkspaceProjects. Apply the override there too so `runtimeOnFail`
actually affects what gets installed. Adds an e2e test covering both
download and ignore overrides through the real CLI bundle.
