* feat(manager): pass extracted constraints to updateArtifacts
`extractedConstraints` were collected during extraction but only ever read
by `fetch.ts` for the lookup. At artifact time every manager had to derive
the same tool constraints a second time, and when the updated package file
no longer carried the constraint there was nothing to fall back to.
`generateBranchConfig()` now merges the `extractedConstraints` of all
upgrades of a branch onto the branch config, which reaches `updateArtifacts()`
through `patchConfigForArtifactsUpdate()`. `UpdateArtifactsConfig` declares
the field and documents what it contains.
The new `resolveToolConstraint()` helper implements the ladder once:
user `constraints` > an optional artifact-time derivation > the extracted
value. The derivation is ranked above the extracted value on purpose: it
reads the package files as the branch changed them, so poetry (python is an
updatable dependency there), composer, pipenv (`pipenv` can be a dev
package) and pep621 (`requires-python` is an updatable dependency) would
otherwise regress to a stale constraint. pipenv, pep621 uv/pdm, poetry and
composer are migrated to the helper and keep their derivation.
Behaviour change: these managers now use the extract-time constraint as a
last resort instead of running unconstrained, for example on lock file
maintenance where pipenv deletes `Pipfile.lock` before deriving.
`execute-post-upgrade-commands.ts` is unaffected: it reads
`upgrade.constraints` per upgrade, and the extracted values are kept in a
separate field rather than folded into `constraints`, so post-upgrade task
tool constraints still come from user config only.
Not migrated, because their constraints are never extracted, so the
fallback would be dead code: bundler (`ruby`/`bundler`), pixi (`pixi`),
pep621 uv (`required-version`) and npm's node ladder, whose last step
already reads `package.json` engines.
* refactor(manager): resolve tool constraints through resolveToolConstraint()
Every manager now obtains the constraint of a tool it runs through
`resolveToolConstraint()` instead of reading `config.constraints` itself, so
the precedence "user config, then a value derived from the updated files, then
the constraints collected during extraction" applies everywhere. The helper
treats an empty string as not set at every step and accepts the
`PostUpdateConfig` shape, whose `constraints` may be `null`.
Behaviour changes that go beyond the new fallback to extracted constraints:
- `mix` applies its hardcoded erlang default only after the extracted
constraint, so an erlang version collected at extraction can win over it.
- `kustomize` decides `HELM_EXPERIMENTAL_OCI` from the fully resolved helm
constraint; `generateHelmEnvs()` takes that value instead of the config.
- `npm` (yarn): an upgrade of yarn without a new value now consults the user
constraint before deriving one from package.json or the lock file.
- The "Using <tool> constraint from config" debug lines in bundler, composer,
copier and pip-compile are gone, as the helper is the only reader of the
config now.
Helpers that read the config synchronously became async: bundler
`getBundlerConstraint()`, copier `getPythonVersionConstraint()` and
`getCopierVersionConstraint()`, pip-compile `get*VersionConstraint()` and
`getToolVersionConstraint()`, gomod `deriveGoToolchainConstraints()` and
`getUpdateImportPathCmds()`, mise `getMiseLockToolConstraints()`. composer
`getPhpConstraint()` is removed.
Each converted manager has a test showing the extracted constraint reaching
the executed tool, and, where a derive callback exists, that the derived value
still wins over it.
* refactor(linters): add prefer-resolve-tool-constraint rule
Reports member reads and destructuring of `constraints` and
`extractedConstraints` from a manager config in `lib/modules/manager/**`, so
that tool constraints are resolved through `resolveToolConstraint()` in
`lib/modules/manager/util.ts` only. Writes and the same properties on package
files, which extraction has to populate, are not reported. The helper itself
and spec files are exempt in `.oxlintrc.json`.
* refactor(manager): inline thin resolveToolConstraint wrappers
Drop copier/utils.ts and the pip-compile getPipToolsVersionConstraint /
getUvVersionConstraint helpers, which only forwarded to
resolveToolConstraint(). Callers now call it directly.
* docs(development): note lock file maintenance in tool constraints section
* refactor(manager): address review on tool constraint resolution
composer only runs extractConstraints() when a user constraint is missing,
via Lazy. UpdateArtifactsConfig no longer re-declares constraints and
extractedConstraints, which ToolConstraintsConfig already provides.
The warning under the port example dates from #22007 (May 2023), when
a matchHost with a port but no scheme was parsed into an empty host and
the rule applied to every host. Since #29487 (37.421.0) massageHostUrl
prepends https:// to any matchHost that has no scheme and contains a
port or a path, so that warning no longer describes the behavior and
contradicts the note under hostRules.matchHost, which only mentions
paths.
Replace the warning with a note that states the current behavior for
ports and paths, and mention ports in the hostRules.matchHost note.
Renovate deployments that are deployed across multiple organisations may
take advantage of Inherited Config as a means to decentralise some
decisions and/or preferences that their organisations would like to use
for some self-hosted configuration.
Similarly, the decision of which hosts to allowlist may be something
that a self-hosted administrator /may/ want to allow their organisation
admins to decide.
We can make this an opt-in capability to indicate that the self-hosted
administrator trusts the Inherited Config moreso, and allows someone who
can write to it the ability to allowlist access to internal hosts that
may not be intended.
If a self-hosted administrator sets `allowInternal: false`, this takes
precedence, allowing a "veto", and a self-hosted admin's `headers` are
still preferred, too.
It's not possible for Inherited Config to set grants for `extends`, as
config resolution has already occurred, so this still needs to be
performed at the self-hosted config level.
This trust is transitive: `hostRules` defined in presets the
inherited config `extends` are registered at the same inherited
tier once resolved, so this option must only be enabled when those
preset sources are trusted too.
Co-authored-by: Claude Fable 5 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
To harden Renovate from being able to access internal hosts, we can add
a level of enforcement on all outgoing requests.
We can split this into two types of traffic:
- where the response becomes configuration (i.e. `extends`)
- where the response does not become configuration (i.e.
`customDatasources`)
To harden this, we enforce on every outbound HTTP request:
- check the resolved URL before anything, including cache lookups, can
proceed
- catching IP-literal hosts in every spelling the WHATWG URL parser normalizes
- blocking known Cloud metadata IPs, too
- a `dnsLookup` wrapper validates every address a hostname resolves to,
on every new connection, which also defeats DNS rebinding
- a `beforeRedirect` hook re-runs the URL check on each redirect
target, with the new host getting its own grant lookup rather than
inheriting the original host's verdict
We default this enforcement to `warn`, to not break existing deployments
and instead give them visibility into required configuration before a
future major version (major+1) enables `block` by default.
We'll also remove allow in a future major version (major+2), to give
time for admins to migrate over.
When a request is blocked, this degrades like an unreachable registry,
so a Renovate run isn't aborted by a blocked request.
As part of this, we can also provide a worked example of the different
cases that do/do not need allowing via `hostRules`.
Requests through an admin-configured proxy resolve DNS at the proxy, so
`dnsLookup` never runs for them, so we can resolve the hostname locally
and validate every address before the request instead.
This doesn't catch everything, especially as the proxy may resolve the
name to a different result, and if we can't resolve the name - possible
in locked-down corporate environments - we fail open.
Child processes aren't currently able to route traffic through
Renovate's HTTP layer, so are unrestricted.
Co-authored-by: Claude Fable 5 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
As part of follow-up changes, we're looking to tighten the way that
Renovate handles outgoing HTTP traffic.
Before we do this, we can introduce a new config option,
`hostRules[].allowInternal`, that allows a self-hosted administrator to
mark an internal host as "trusted" for access.
This wires in the new global-only config option, `allowInternal`, which
implies a new internal field `internalHostGrant`.
We'll add full documentation in a future commit, when the full
enforcement is in place, as there are some wider changes to enforce it
fully.
Co-authored-by: Claude Fable 5 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
As part of follow-up changes, we're looking to tighten the way that
Renovate handles outgoing HTTP traffic, and a self-hosted administrator
disabling a host needs to be something a repository cannot undo.
In a similar change to
4732da7d2b, we can resolve `enabled` per
trust tier, so that a repository or preset rule can no longer re-enable
- or disable - a host the administrator's own rules set, which was
previously possible with a longer `matchHost`.
Within a tier nothing changes: the most specific rule still wins, so a
repository disabling a host of its own keeps working, as does an
administrator masking their own broad rule with a narrower one.
This is a behaviour change for repository config that previously relied
on overriding an admin rule's `enabled`.
Co-authored-by: Claude Fable 5 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
* feat(config): add experimental exitCodeForErrors option
* feat(config): give unknown-error its own exit code
* docs: restore exit codes 1 and 2 to the table
* feat(automerge): attempt branch automerge before falling back to a PR on merge queue rejection
* refactor(automerge): ask the platform instead of parsing merge queue rejection messages
* feat(gitlab): merge MRs through merge trains via the platform merge queue API
* feat(gitlab): report MRs waiting on a merge train through isPrInMergeQueue
* test(gitlab): drop targetBranch from mergePr calls
* Apply suggestion from @secustor
* refactor(gitlab): validate the merge train car response with a Zod schema
* feat(github): add experimental support for using merge queues without auto merge
* test(github): use vi.stubEnv for merge queue env flag
* fix(github): do not re-enqueue PRs already waiting in the merge queue
* refactor(github): reuse merge queue detection from #45252 for isBranchMergeQueueEnabled
* refactor(github): read the base branch from the PR instead of extending MergePRConfig
* feat(github): enable merge queue support without the RENOVATE_X_GITHUB_MERGE_QUEUE flag
* Apply suggestion from @secustor
* feat(manager): share OCI Helm chart resolution across managers
Add getOciChartDep() to helmv3/oci.ts and use it in argocd, fleet, flux,
helmfile, helmsman, helmv3, kustomize, sveltos, terraform and vendir.
- set pinDigests: false for OCI charts in all of them
- apply registryAliases to OCI charts in argocd, helmfile and helmsman
- keep the registry port in terraform helm_release repositories
- skip fleet OCI charts without a version instead of using 'undefined'
* docs(manager/helmv3): correct the OCI chart pinDigests rationale
* refactor(manager/helmfile): drop redundant OCI repository url guard
The repository schema already requires url to be a string, so the falsy
branch could not be reached by parsed input and left the patch partially
covered.
Mend's GitLab.com support has been around for a little bit, but we've
not explicitly documented it.
While we're doing this, we should make sure the outdated references to
GitLab security is updated in line with the current state of affairs.
With some additional research from Claude Opus 5, but with Claude Sonnet
5 driving the changes.
We can also make sure to clarify that folks may want to use Mend-hosted
against GitLab.com, and that there are security implications when
running Renovate on GitLab (more generally).
Researched-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
* fix(manager): handle registryAliases with shell variable defaults
Allows `registryAliases` keys to contain shell-style variables with
empty (or other) defaults, e.g. `${DEPENDENCY_PROXY:-}` or
`${CI_REGISTRY:-}`. The same image string is then resolved both
online (variable expanded) and offline (variable empty).
Changes in `dockerfile/extract.ts` (used by `gitlabci` too):
* Match the alias key as either `${name}/...` (existing) or `${name}...`
when `name` already ends in a non-identifier boundary character such
as `}`. Bare-identifier keys like `$CI_REGISTRY` continue to require
the slash boundary so they no longer overmatch keys that begin with
the same prefix (e.g. `$CI_REGISTRY` vs `$CI_REGISTRY_IMAGE`).
* Use `ensureTrailingSlash` to handle alias values that already include
a trailing slash.
* Preserve the original `name` and path structure in `dep.depName`
instead of substringing only after the alias `value`. This keeps the
variable in the depName so it survives a `splitImageParts` round-trip
and the resulting branch name comes out clean.
Adds `sanitizeDepName` regex passes in
`lib/workers/repository/updates/flatten.ts` so any leftover
`${VAR}`/`${VAR:-...}` placeholder is stripped from the depName before
it becomes a branch name (otherwise we'd get
`renovate/${dependency_proxy}python-3.x` instead of
`renovate/python-3.x`).
Tests cover:
* `${VAR:-}/image` (with slash) and `${VAR:-}image` (no slash, requires
the boundary form).
* Overlapping prefixes: `$CI_REGISTRY` vs `$CI_REGISTRY_IMAGE`.
* Registry alias with port (`localhost:5000/repo`).
* `sanitizeDepName` strips `${VAR}` / `${VAR:-}` / bare `$VAR`.
Reproducer: https://github.com/Wayneoween/renovate-36293-repro
* revert(manager/gitlabci): keep regex group name as 'prefix'
Reverting the unrelated `prefix` -> `rawPrefix` rename in `depProxyRe`.
Per review on #36293, this rename is out of scope for the registry-alias
fix and just adds review surface.
* refactor(manager/dockerfile): narrow alias boundary check to '}'
Per review on #36293. The previous regex `/[^A-Za-z0-9_]$/` matched any
non-word ending, which is broader than the documented case (alias keys
ending in `}` from `${VAR}` / `${VAR:-...}`). Replace with a plain
`name.endsWith('}')` check, which also avoids re-building a regex on
every iteration of the alias loop.
Existing tests (`'variables with default and slash'`,
`'variables with default no slash'`, `'alias with digest only'`)
continue to exercise the `}`-boundary path; the `'overlapping prefixes'`
case continues to exercise the slash-only path.
* chore(manager/dockerfile): trim comment and TODO library/ strip edge
Per review on #36293:
* Drop the verbose explanation block above the depName-restore branch;
the spec rows already document the expected shape.
* Add a TODO above the `dep.depName?.startsWith(valueWithSlash)` guard
noting a pre-existing edge case: when the inner `getDep` strips a
`library/` prefix (or similar) the depName no longer starts with
`valueWithSlash`, so the alias-rooted depName is not restored. Out of
scope for this PR.
* fix(manager/dockerfile): handle empty registry alias value
- strip trailing slash after ${VAR} placeholders in sanitizeDepName
- treat an empty registry alias value as Docker Hub instead of '/'
- clarify registry alias docs wording
* refactor(manager/dockerfile): drop redundant alias boundary guard
- `matchedAtBoundary` is only read behind `!matchedWithSlash`
- add docker-compose test for `${VAR:-}` registry aliases
* test(manager/dockerfile): pin special-prefix registry alias behavior
`getDep` strips `library/`, `amd64/` and `arm64/` as special prefixes, so
an alias whose value is one of those loses its alias-rooted `depName`.
Both the slash form and the variable form behave this way on `main`.
Private GitLab projects need a `hostType: "go"` host rule with a username and password so Renovate can authenticate the go-get request it uses to resolve a module's source repository. Without it GitLab returns a truncated go-import path for the unauthenticated request and the version lookup 404s.
Came up in discussion #37624 after the resolution change in #36963.
As part of changes we've recently made to the merging of
`hostRules[].headers`, it highlighted that a repository - or a preset
it extends - could lead to a header an admin has configured for one host
to be sent to another.
This happened due to the way we additively merged i.e.
`Object.assign({}, ...matchedHeaders)`.
This doesn't require the repository to know a credential is being
set (for a specific host), or to act maliciously - any header of their
own for that host is enough to bring the administrator's masked one
back, with nothing logged or failing to show it happened.
To fix this, we can:
- apply admin headers last, as they are "trusted"
- stop repositories from being able to unset an admin-set header
- retain the existing logic for the "most specific rule" being used
- make sure that `trusted` isn't a field that could be injected by a
repository, and make sure we explicitly only set it where we know
we're using self-hosted administrator config
- make `filterAllowedHeaders` drop `headers` if every header name was
denied, so `{}` isn't returned and used to suppress headers when
merged
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
Previously, if there was a `hostRule[].headers` value for a given host
set in both global self-hosted configuration and repo config, the repo
config would overwrite /all/ headers set by the admin.
To resolve this, we can now correctly merge the keys, with the repo
config taking priority.
This still only takes into account the `allowedHeaders`, so a repository
cannot bypass admin settings.
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
For some time, admin configuration for `hostRules[].headers` was
enforced through `allowedHeaders`, but no validation feedback was given
to administrators, as it was silently ignored when `applyHostRule` runs.
To provide visibility to administrators who were setting headers that
were silently being ignored, we can instead provide validation at the
time we set our global host rules.
In a future change, we'll align the behaviour of `allowedHeaders` and
`allowedEnv` to make it so admin-controlled `hostRules` aren't bound by
the `allowedHeaders` flag.
We can also make it clear in the documentation that this is the
behaviour.
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
As a further level of defence-in-depth for fixes on top of `allowedEnv`
and `allowedHeaders` enforcement, we can also make sure that we're
performing full config validation for the resolved configuration.
This allows us to report any configuration errors/warnings when we
include the full resolved configuration, without necessarily blocking
the full Renovate run, unless using `configValidationError`.
Co-authored-by: Claude Opus 4.8 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
Lock file maintenance deletes package-lock.json and regenerates it from
scratch, and a single npm install pass can produce a lock file that is out
of sync with package.json, so npm ci rejects the branch. Always run the
install command a second time for lock file maintenance, mirroring the
existing double run for remediations, instead of relying on users opting
in to the npmInstallTwice postUpdateOption. Because branch creation and
rebase share the same lock file generation path, this covers both.
We dropped the "bot" from our terminology long ago, but it looks like
we've got some outdated docs + code that reference it.
We should remove this to keep things consistent.
A `fix`, as we're changing the commit author's name, and log lines.
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
As noted in #36073, having support for the different protocols for
exporting OpenTelemetry data would be useful, especially if this is
leading to users seeing issues when they have auto-injected
configuration being set.
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
As noted in #43761, we had recently set `CI=true` for the Renovate and
Containerbase orgs on Mend-hosted apps due to a pnpm v11 requirement:
ERR_PNPM_ABORTED_REMOVE_MODULES_DIR_NO_TTY Aborted removal of
modules directory due to no TTY
If you are running pnpm in CI, set the CI environment variable to
"true", or set "confirmModulesPurge" to "false".
Similarly in #43606, it was noted that we should do this for `mise lock`
to mark the workspace as "trusted".
Although we went with a different situation, there will be cases where
this is useful for tools being executed, for instance when we invoke `yarn`.
We also make sure to keep our documentation up-to-date as we add to the
specific overrides.
With help from Claude Sonnet and Opus to:
- perform some of the mechanical updates as part of this change
- ensure that the Docker sidecar always has `hardcodedProcessEnv` passed
into it
- catch a needed change to `exposeAllEnv` handling
Co-authored-by: Claude Opus 5 <jamie.tanna+claude-code@mend.io>
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
* feat(config): add `overrideDescription` option
Presets which extend many other presets used an `ignoreDeps: []` sentinel
to keep their own one-line summary in the onboarding PR instead of one line
per extended preset. Replace that hack with an explicit `overrideDescription`
option, which is consumed and removed while presets are resolved.
The resolved descriptions are unchanged. Repositories which set
`ignoreDeps: []` in their own config no longer lose all preset descriptions
in their onboarding PR.
* fixup! feat(config): add `overrideDescription` option
* fixup! feat(config): add `overrideDescription` option
* fixup! feat(config): add `overrideDescription` option
Adds a custom oxlint rule renovate/v8-ignore-no-count that flags the
count form '/* v8 ignore next <N> */'. The count is not honored by
vitest's ast-v8-to-istanbul hint parser, which captures only the
keyword (if/else/next/file); 'next' always exempts exactly the next
AST node. Trailing counts therefore suggest a line range that is never
applied and mislead readers.
The rule is autofixable (drops the count, preserving actual behavior)
and has been applied to all existing occurrences.
Jamie Tanna