* feat(manager/mise): set depType for tools and task tools
Allow packageRules to match mise deps via matchDepTypes:
- tools for top-level [tools]
- task-tools for tasks.*.tools
* test(manager/mise): override RUST_197 depType in task extract cases
* Make the task tool dynamic
* Restore formatting
* Restore another formatting
* feat(config): add commitTrailers option
Renovate's only way to add git trailers today is `commitBody`, but it's
broken on Gerrit: Renovate appends its own `Renovate-Branch`/`Change-Id`
block after the body, so user trailers end up in a non-final message
block and are not real trailers per git-interpret-trailers rules.
This adds a `commitTrailers: string[]` option (each entry a full
`Key: value` line), standardizes Gerrit's own trailers through the same
mechanism so user trailers and platform trailers merge into one final
block, and verifies the result at runtime with `git interpret-trailers`
(commitlint's trailer-exists rule uses the same approach). The
`:gitSignOff` preset is migrated from `commitBody` to `commitTrailers`
so sign-off works correctly on Gerrit too.
* Simplify a bit
* Simplify docs even more
* Simplify another bit
* Add comment about rawExec
* Drop runtime commit trailer verification
Bypass simple-git only to emit a proactive warn log is not worth the
extra surface area. Config validation already enforces Key: value form.
* test(git): cover string commit message with trailers
Codecov patch failed on the partial ternary branch that normalizes a
string message before appending trailers.
* test(gerrit): fix trailers test after scm rework
Align the commitTrailers merge test with the post-rework Gerrit SCM:
use fakeSha helpers and drop the removed force flag.
* Improve wording on gitSignOff docs
* Validate compiled commitTrailers as well
* Improve trailers example
* Use isString instead
* Update lib/util/git/index.ts
Co-authored-by: Sebastian Poxhofer <secustor@users.noreply.github.com>
* Update lib/workers/repository/onboarding/branch/create.ts
Co-authored-by: Sebastian Poxhofer <secustor@users.noreply.github.com>
* Format code
* Fix type-check
* Fix tests
---------
Co-authored-by: Sebastian Poxhofer <secustor@users.noreply.github.com>
* feat(manager/mise): run mise lock in safe mode without allowedUnsafeExecutions
When `constraints.mise` pins a version that supports mise safe mode
(`MISE_SAFE=1`), run `mise lock` with safe mode enabled instead of
requiring `mise` in `allowedUnsafeExecutions`. Safe mode is a hard
boundary against project config executing code (templates exec/read_file,
_.source, hooks, tasks, asdf scripts, plugin installs are all refused),
so it — not the allowlist — protects the host.
The existing allowlisted path is unchanged: allowlisted repos run
`mise lock` exactly as before, without forcing safe mode, so configs
that legitimately execute code during locking keep working.
Safe mode also removes the need to pre-install node/npm/golang/ruby as
resolver tools, since npm version listing uses mise's built-in registry
HTTP client and go runs with GOTOOLCHAIN=local.
Refs: jdx/mise#11146 (safe mode), jdx/mise#11145 (lock --bump)
* fix: address oxlint and pin safe-mode min version to 2026.7.12
- logger.warn requires a static message with interpolated values in the
metadata object (oxlint rule); move the version into metadata
- set MISE_SAFE_MODE_MIN_VERSION to 2026.7.12, the expected first release
containing safe mode (merged after v2026.7.11)
* feat(manager/mise): use mise lock --bump for lock file maintenance
lockFileMaintenance means "update to the latest versions", but plain
`mise lock` only refreshes metadata for already-locked versions. Use
`mise lock --bump` so fuzzy selectors (e.g. node = "22") advance to the
latest matching version. `--bump` requires the same minimum mise version
as safe mode, so it is gated on the mise constraint and falls back to a
plain refresh otherwise.
Generalizes the version-gate helper (miseConstraintGuarantees) shared by
safe mode and --bump.
* refactor(manager/mise): detect mise version at runtime instead of via constraints
Determine safe-mode / --bump support by running `mise version` (which
does not load or execute project config) rather than requiring a pinned
`constraints.mise`. This works off whatever mise is actually installed,
so no extra configuration is needed.
The probe is skipped on the allowlisted, non-maintenance path, where the
version cannot change the outcome — keeping that hot path unchanged.
* feat(manager/mise): skip mise trust in safe mode
Safe mode now makes an untrusted config fully inert and no longer
requires trust to load it (jdx/mise#11151), so on the safe-mode path
`mise trust` is unnecessary and is skipped — `mise lock` runs with
`MISE_SAFE=1` alone. The allowlisted path is unchanged and still runs
`mise trust`.
* feat(manager/mise): confirm safe mode min version against v2026.7.12 release
MISE_SAFE=1 safe mode and `mise lock --bump` first shipped in mise
v2026.7.12 (released 2026-07-23), confirming the assumed
MISE_SAFE_MODE_MIN_VERSION. Resolve the pending-release TODO and align
the test version fixture with the real release date.
Validated end-to-end against the real v2026.7.12 binary: `mise version`
output matches the parser, `MISE_SAFE=1 mise lock` refreshes an untrusted
config without trust, `mise lock --bump` advances fuzzy selectors, and
project [env] is ignored under safe mode.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* test(manager/mise): cover mise version probe failure path
Add a test for the detectMiseVersion catch branch (exec throwing), so the
non-allowlisted path falls back to conservative behavior. Restores 100%
patch coverage for the mise manager.
* Update lib/modules/manager/mise/readme.md
Co-authored-by: Sebastian Poxhofer <secustor@users.noreply.github.com>
* Update lib/modules/manager/mise/readme.md
Co-authored-by: Sebastian Poxhofer <secustor@users.noreply.github.com>
* docs(mise): fix safe mode section link
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Sebastian Poxhofer <secustor@users.noreply.github.com>
fix(platform/gitlab): retry MR creation on transient source_branch race
GitLab occasionally rejects a merge request POST with
`400 {"message":{"source_branch":["does not exist"]}}` immediately after
a branch push succeeds, because the freshly pushed ref is not yet visible
to the API due to Gitaly eventual consistency. The request was not
retried, leaving an orphaned branch until the next run.
Extend the existing `calculateRetryDelay` hook (which already retries
transient `409 Resource lock` POST errors) to also retry this
`400 source_branch does not exist` case, reusing got's bounded
exponential backoff.
* feat(lint): add renovate/prefer-fs-util and renovate/prefer-luxon rules
Two shared-util substitution rules, both at warn level:
- renovate/prefer-fs-util flags value imports of fs, node:fs,
fs/promises, node:fs/promises, fs-extra and fs-extra/esm in lib/
source files outside lib/util/fs/, pointing to the scoped helpers
from lib/util/fs. Spec files and type-only imports are exempt.
A config-only change via no-restricted-imports was considered but
rejected: that rule is at error severity globally and per-path warn
severity is not expressible, so the 13 existing call sites would
break the baseline.
- renovate/prefer-luxon flags new Date(...) and Date.now() in lib/
source files, pointing to Luxon per best-practices. Date/time
utility modules (lib/util/date, lib/util/pretty-time,
lib/util/timestamp) and spec files (vi.setSystemTime etc.) are
exempt.
Baseline: 13 prefer-fs-util and 45 prefer-luxon warnings, zero errors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* refactor: use Luxon instead of new Date() at remaining call sites
Fixes all renovate/prefer-luxon violations for `new Date(...)`:
- `new Date().toISOString()` -> `DateTime.utc().toISO()` (byte-identical
output) in the http cache provider, schedule logging and PR cache.
- `new Date()` -> `DateTime.now().toJSDate()` for the git commit-date
fallback and the rubygems `syncedAt` markers (same `Date` values).
- rust-release-channel now computes the Rust 1.0 release date and nightly
version dates via `DateTime.utc(...)`; previously `new Date(y, m, d)`
used the local timezone while the epoch used UTC, so the comparison was
timezone-dependent at the boundary date. Tests run with TZ=UTC, where
behavior is unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* refactor: use lib/util/fs helpers in repository worker
Fixes renovate/prefer-fs-util violations outside bootstrap code:
- `fs.ensureDir()` -> `ensureDir()` from lib/util/fs (same fs-extra call
underneath; it no-ops on an empty dir name, so the repository worker
spec now simulates the init failure via a rejected `initRepo` instead
of relying on `fs.ensureDir('')` throwing).
- `fs.remove(privateCacheDir())` -> `cachePathExists()` + `rmCache()`,
keeping the remove-if-present semantics while asserting cacheDir
containment.
Two call sites keep raw imports with reasoned disable comments:
- lib/util/modules.ts synchronously enumerates Renovate's own installed
module directories at import time.
- lib/util/cache/package/impl/sqlite.ts receives cacheDir as an explicit
argument during global initialization, before GlobalConfig (which the
scoped cache helpers resolve against) contains cacheDir.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(lint): promote prefer-fs-util and prefer-luxon to error
Both rules are narrowed to what is actually accepted practice before
being promoted:
prefer-luxon no longer flags `Date.now()`: it returns UTC-based epoch
milliseconds used for elapsed-time measurement and TTL arithmetic, which
is timezone-independent by construction. The mined reviewer feedback
behind this rule ("use Luxon for date arithmetic outside util") targets
timezone-dependent date construction/formatting, which cannot occur with
a raw millis number; the rule keeps flagging `new Date(...)`.
prefer-fs-util exempts code that legitimately needs raw fs access
because the scoped lib/util/fs helpers resolve against GlobalConfig
localDir/cacheDir, which does not exist yet or does not apply there:
- lib/logger/: log-file streams at user-configured system paths during
bootstrap; lib/util/fs also depends on the logger (import cycle).
- lib/workers/global/: creates baseDir/cacheDir themselves and
reads/writes user-supplied paths before GlobalConfig is seeded.
- lib/config-validator.ts: standalone CLI without GlobalConfig.
- lib/util/git/: owns the localDir lifecycle (clone, emptyDir, symlinks)
and private-key files with sync cleanup on process exit.
Also fixes a pre-existing tsc error in prefer-fs-util.js: `importKind`
is a TypeScript-ESTree extension missing from eslint's base AST types.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* refactor(fs): make rmCache ignore missing paths
All rmCache callers use it for cleanup, where a missing path is a
no-op, so bake force: true in instead of guarding call sites with
an existence check.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* refactor(linters): exempt lib/instrumentation from prefer-fs-util
`lib/instrumentation/file-exporter.ts` landed on main while this branch
was open and imports `node:fs/promises` directly.
The exporter writes to a user-configured system path
(`RENOVATE_TRACING_FILE_EXPORTER_PATH`) outside `localDir`/`cacheDir`,
so the scoped helpers do not apply. `lib/instrumentation` is also
deliberately kept out of the module graph of instrumented libraries
(see the `⚠️` note in `lib/renovate.ts`), so it must not pull in
`lib/util/fs` and its config/logger dependencies. Exempt the directory,
as is already done for `lib/logger`.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* docs: shorten doc
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
As part of follow-up changes, we'll need to call the checks in multiple
places, so should prefactor these into helper methods.
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
As part of follow-up changes, we'll use these functions to centrally
determine as to whether to execute `minimumReleaseAge` or
`minimumConfidence` checks against a given updateType.
For now, we can wire this into our documentation tests, to make sure
that this stays in sync with `isMinimumReleaseAgeApplicable`.
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
As noted in #30714, and to make the Minimum Release Age page a
"one-stop-shop", we can generate the documentation accordingly.
Using an orange "maybe" for support helps nudge folks to checking which
registry they're using, and noting that it still doesn't necessarily
work, even if the Datasource supports it.
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
When GitHub returns 'GH013: Repository rule violations found' during
a push, the failure is currently silently swallowed by handleCommitError
whenever the changeset touches a file under .github/workflows/. The
existing swallow was intended for the narrow case of an App token
missing the 'workflows' scope, but it also masks unrelated ruleset
violations such as:
- GITHUB PUSH PROTECTION 'File path is restricted'
- Secret scanning push protection
- Signed commit / branch-naming / tag-deletion rulesets
Self-hosted operators cannot see these failures in aggregated log
reports because Renovate logs at info level, marks the branch as
aborted, and continues. Repositories that are effectively 100%
blocked by a ruleset appear healthy.
Detect the 'GH013' token before the workflow-files silent-swallow and
throw a CONFIG_VALIDATION error - mirroring the treatment already
applied to GH003 (force-push denied) and 'protected branch hook
declined'. This flags the repository as failing, populates
validationError / validationMessage so the failure reason is visible
in the repository summary, and lets any downstream reporting pick it
up as a real failure instead of an info-level abort.
Adds tests to lib/util/git/errors.spec.ts covering:
1. A GH013 Push Protection error whose changeset is entirely under
.github/workflows/ is thrown as CONFIG_VALIDATION with the
expected validationSource / validationError / validationMessage.
2. A generic non-GH013 workflow-file remote rejection still returns
null (existing swallow behaviour preserved).
Felipe Santos