* fix(util/exec): pick tool releases compatible with resolved node version
* refactor(util/exec): pass resolved runtimes as an object
* refactor(util/exec): match tools against their containerbase parent
* test(util/exec): run real constraints filtering for parent versions
`install-tool` installs a version and also makes it the active one. The
memo skipped `install-tool` for any tool version already installed in the
run, so after A -> B -> A the last command ran on B. Track the active
version per tool instead and skip only when the requested version is
already active.
Co-authored-by: Claude Opus 5.5 <michael.kriese+claude-code@mend.io>
When running Renovate with `binarySource=install` on a repository with
many branches, the amount of time running `install-tool` - on already
installed tools - can take a significant percentage of the overall run.
For instance, on one repository we have instrumented with OpenTelemetry,
we see:
- initial `install-tool golang`: 8.7s
- follow-up `install-tool golang`: median 635ms, max 2.4s
However, because we run `install-tool golang` 640 times, this adds up to
24.8/29.2m (85%) of the repo's run time, which is an overwhelming
percentage when we're not even re-installing the tool.
To handle this more effectively, we can add an in-memory cache of all
the tools (and resolved versions) that we've installed in a given run.
We can't do this with `binarySource=docker`, as it has a fresh container
each time.
Co-authored-by: Claude Sonnet 5 <jamie.tanna+claude-code@mend.io>
* test: replace manual process.env mutation with vi.stubEnv
Enable Vitest's `unstubEnvs` so env stubs are reverted automatically
before every test, and convert the manual `process.env.X = ...` /
`delete process.env.X` pairs in specs to `vi.stubEnv()`.
This removes the hand-written `afterEach` cleanup blocks and the
`const OLD_ENV = process.env` / `process.env = { ...OLD_ENV }`
save-restore dance, which silently leaked state between tests when a
test set a variable without a matching delete.
Left as-is (incompatible with `unstubEnvs`, which unstubs before each
test, not after):
- module top-level env writes (cocoapods, gomod, post-update, pip-compile
and bundler artifacts specs)
- `beforeAll` env writes -- the two in `util/http` moved to `beforeEach`
* test: stop asserting on ambient environment variables
Three specs asserted on values that the surrounding environment can
override, so they only passed on a machine that happened not to set them:
- `util/exec/env`: `getChildProcessEnv()` forwards every entry of
`basicEnvVars`, but the spec only seeded a subset. Any other forwarded
variable that was set ambiently (e.g. a lowercase `http_proxy`) leaked
into the `toEqual` assertion. Clear all of them first.
- `util/git/index`: `process.env.GIT_SSH_COMMAND` deliberately overrides
the default that the two env-inheritance tests assert on. Clear it.
- `logger/pretty-stdout`: the colorized level strings are built at import
time, so the `beforeEach` setting `FORCE_COLOR` could never affect them
-- it was dead code, and the test silently assumed a terminal without
colour support. Derive the expectation the same way the module does.
* test(datasource/nuget): use a throwaway cache dir for nupkg downloads
The `determine source URL from nupkg` tests do not mock the filesystem,
so they really wrote downloaded packages into a fixed `/tmp/cache` path.
That fails wherever the system tmpdir is not writable, and two checkouts
running the suite at once would share the same directory.
Use `tmp.dir({ unsafeCleanup: true })` and clean it up afterwards, which
is the pattern the other specs that touch disk already use.
* feat(lint): add renovate/prefer-stub-env rule
Enforces in specs what the migration to `vi.stubEnv` established:
- assigning to or deleting `process.env.X` inside `beforeEach`, `afterEach`,
`it` or `test` is reported, with a fixer that rewrites it to `vi.stubEnv()`
- module scope, `describe` bodies and `beforeAll` / `afterAll` are left alone:
`unstubEnvs` reverts stubs *before* each test, so a stub created there is
gone by the time the first test runs, and those must keep assigning directly
- replacing the `process.env` object is reported everywhere and never fixed.
`vi.stubEnv()` captures the original object when the worker starts, so after
a reassignment its deletes silently target the old object
The rule found 28 further violations in specs the earlier sweep missed, since
that one was anchored on `delete process.env`. Twelve were auto-fixed; the
rest replaced `process.env` wholesale and are now `clearEnv()` from test/util.
`util/exec/index.spec.ts` opts out with a comment: it asserts on the whole
child env, and `vi.stubEnv()` cannot express that because it maps `PROD`,
`DEV` and `SSR` onto '' instead of deleting them.
As noted in #43489, it's not currently possible to install the nightly
variant of Rust.
Although Containerbase does have support for `install-tool rust nightly`
it never gets that far, as the version lookup occurs through the
`docker` datasource.
We can instead make sure we move to the `rust-version` datasource and
its associated `rust-release-channel` versioning.
The result from the `rust-release-channel` returns a version like
`nightly-2026-06-19`, which needs extra handling in Containerbase.
We can also update **??** what supported Rust versions we'd accept.
* docs(constraints): list the possible `constraints` and `installTools`
Now we have a stronger type for what a valid tool constraint is, and
what additional constraints are, we should be very explicit with what
each option is, and specify them in the `constraints` as well as our
`installTools` docs.
* test: add coverage
Co-authored-by: Claude Sonnet 4.6 <jamie.tanna+claude-code@mend.io>
---------
Co-authored-by: Claude Sonnet 4.6 <jamie.tanna+claude-code@mend.io>
* chore(types): add a strong type for `ToolName`s
Until #41849, we can use our local definitions of the tool names that
Containerbase supports by extracting out a `ToolName` type.
* fixup! chore(types): add a strong type for `ToolName`s
Co-authored-by: Claude Opus 4.6 <jamie.tanna+claude-code@mend.io>
Co-authored-by: gpt-4.1 <jamie.tanna+github-copilot@mend.io>
* fixup! fixup! chore(types): add a strong type for `ToolName`s
* no-never
* simplify
* coverage
* sq
* warn
* sq
* fixup! sq
* sq
* fixup! sq
---------
Co-authored-by: Claude Opus 4.6 <jamie.tanna+claude-code@mend.io>
Co-authored-by: gpt-4.1 <jamie.tanna+github-copilot@mend.io>
After recent changes to remove `shell: true`, it's clear from #40231
that this breaks Containerbase's installing of tools, as we use the
`hash` shell built-in.
As we no longer need the `hash` call, we can remove it.
Michael Kriese