<details>
<summary>Automated notes</summary>
Use Node.js 24.22.1 for the `setup-node` jobs affected by the eMudhra
intermediate CA compromise. [Node.js security
release](https://nodejs.org/en/blog/vulnerability/october-2026-security-releases)
Both workflow jobs read their Node.js version from `.nvmrc`.
</details>
Adds `require-checksums`, which enforces the "intrinsic or explicitly
provided checksum" idea. See #1032.
<details>
<summary>Automated notes</summary>
Add the opt-in input `require-checksums`. When enabled, setup-uv
verifies the archive against the workflow's `checksum` input or the
digest bundled for the resolved uv version, architecture, and platform.
If neither is available, setup fails before downloading the archive. By
default, setup-uv continues to accept manifest checksums as it does
today.
The setting also applies to custom manifests, which can still supply
download URLs. On persistent runners, strict mode keys tool-cache
entries by the expected archive digest so a previous manifest-verified
installation cannot bypass the check.
Closes#1032. Builds on #1033.
## Validation
- `npm ci --ignore-scripts` and `npm run all` passed under Node 24 (182
tests).
- The packaged-action regression test installs a custom build into a
temporary runner tool cache. It checks that strict mode rejects a
manifest-only version even with a warm cache, reuses a verified
installation for the same digest, and rejects a changed digest or a
custom build that conflicts with the bundled digest.
</details>
---------
Signed-off-by: William Woodruff <william@yossarian.net>
Pin the shared actions to commit
`67dc9ee1ddc113eccd8116e7b949fb6cdcd89889` from [release
2026.11.16.0](https://github.com/astral-sh/github-actions/releases/tag/2026.11.16.0).
Add the release version beside each YAML pin.
Parsed the changed YAML and verified that each referenced action or
reusable workflow exists at the pinned commit. The workflow also passed
actionlint and offline zizmor checks.
## Summary
- Reproduce the remaining dependency changes from nine open Dependabot
PRs against current main, without reverting unrelated changes from older
PR branches.
- Update actions/setup-node to 7.1.0 (#1127), release-drafter to 7.10.0
(#1126), and CodeQL init/analyze/autobuild to 4.38.2 (#1116, #1115,
#1114).
- Update @types/node to 26.6.2 (#1108), undici to 8.11.0 (#1107), Biome
to 2.5.14 (#1106), and Jest to 30.5.2 (#1105).
- Align the Biome schema and regenerate affected action bundles.
## Excluded
- #1029 (TypeScript 7.0.2): all 16 test suites fail before executing
tests because ts-jest 29.4.11 expects the TypeScript compiler API and
declares a >=4.3 <7 peer range. Excluded with operator approval;
TypeScript remains at 6.0.3.
- None of the ten open Dependabot PRs were already no-ops against main.
## Validation
- npm ci --ignore-scripts
- npm run all: passed (16 suites, 179 tests; no Biome fixes)
- actionlint: passed
- uvx zizmor --offline .github/workflows: no findings
- npm audit signatures: 353 verified registry signatures, 38 verified
attestations
- Workflow all-tests-passed dependency check: passed
- git diff --check: passed
npm installation reports 30 audit vulnerabilities, matching the baseline
count; unrelated audit fixes are outside this rollup.
Refs: pi-session 01a1d01c-e089-7536-b5be-fed125cdaa18
## Summary
`<=`, `==` and `>` with a partial version are valid PEP 440 specifiers,
but the resolver tries them as semver ranges first, and semver reads
them differently:
| specifier | semver (before) | PEP 440 (what uv enforces) |
|---|---|---|
| `<=0.5` | `<0.6.0`, installs 0.5.31 | `<=0.5.0`, installs 0.5.0 |
| `==0.5` | `==` is stripped and `0.5` is read as `0.5.x`, installs
0.5.31 | `==0.5.0`, installs 0.5.0 |
| `>0.5` with `resolution-strategy: lowest` | `>=0.6.0`, installs 0.6.0
| `>0.5.0`, installs 0.5.1 |
So `required-version = "<=0.5"` or `"==0.5"` in `uv.toml` /
`pyproject.toml` installs a uv that does not satisfy the project's own
`required-version`, and uv refuses to run there (`Required uv version
... does not match the running version ...`).
This is the same problem #1111 fixed for `~=`. Instead of adding more
operators to that check, specifiers that `@renovatebot/pep440` accepts
as a valid range now skip the semver evaluation, and `==` is only
stripped from a full version, so `==0.5` stays a PEP 440 pin.
Notes:
- Semver-only syntax is not valid PEP 440 and still goes through semver:
`0.4.x`, `^0.5`, `~0.5`, `>=0.4.0 <0.5.0` and a bare `0.5`.
- Inputs that are valid in both, like `>=0.4.0` from the docs, now go
through PEP 440. They mean the same in both, and the PEP 440 path uses
the same first-match lookup over the manifest.
- Full pins like `==0.9.26` still resolve as explicit versions, and
compound specifiers like `>=0.4,<=0.5` already went through PEP 440.
## Test plan
- Added cases for `<=0.9` and `==0.9` with the default strategy and
`>0.9` with `lowest` in `download-version.test.ts`. All three fail
without the change.
- `npm ci --ignore-scripts`, then `npm run build`, `npm run check` (no
fixes), `npm run package` and `npm run test:unit` on macOS with Node
24.13.0: 179 tests across 16 suites pass. The rebuilt
`dist/setup/index.cjs` only differs in the changed functions.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
## Summary
Fix two workflow issues exposed by the `ubuntu-latest` transition to
Ubuntu 26.04, in separate commits:
- Use `npx --no-install tsc` in the workflow job dependency check
instead of relying on a globally installed TypeScript compiler, matching
https://github.com/astral-sh/ruff-action/pull/455.
- Disable caching in `test-specific-version`,
`test-version-file-version`, and `test-tool-versions-python-version`.
These jobs only verify versions or outputs and do not create a
dependency cache directory.
## Background
The cache key includes the OS version, so the runner transition caused
cache misses. Previously restored caches bypassed cache saving and hid
the missing-directory problem. On a cold cache, cleanup tried to save a
directory that these tests never created. Disabling caching makes these
non-cache tests independent of existing cache entries; dedicated cache
tests remain unchanged.
## Validation
- `npm ci --ignore-scripts` and `npm run all` using npm 11.10.0 after
the compiler fix (176 tests passed)
- Workflow TypeScript compilation and job dependency check
- `actionlint` and `uvx zizmor .github/workflows/test.yml` passed after
both changes
- `git diff --check`
Refs: pi-session 01a1cfe6-d342-70b6-a474-504b84877d9e
With `cache-python: true`, a Windows runner can report a cache hit while
extracting the cache incorrectly. When `@actions/cache`'s tar extracts
uv's minor-version junction before the patch directory it points to, it
creates a file symlink. Resolving that link later raises `[WinError 5]
Access is denied`.
This PR fixes things up after restoring the Python cache, recreating the
junction if resolving the link fails with access denied and its recorded
target is a matching, accessible CPython installation in the same
directory.
This is a temporary workaround until
https://github.com/actions/toolkit/issues/2562 gets resolved and
`@actions/cache` itself becomes resilient to these situations.
I tested this by restoring the same archive on two fresh `windows-2022`
runners. The original run passed because resolving the minor-version
link raised the expected WinError 5. With the patch, `uv sync --offline
--managed-python --no-python-downloads` created a virtual environment
whose `home` in `pyvenv.cfg` resolved to the cached CPython 3.10.22
installation. [Original
run](https://github.com/astral-sh/ruff/actions/runs/41469344610/job/125358623949#step:9:1)
· [Patched
run](https://github.com/astral-sh/ruff/actions/runs/41469344610/job/125358623835#step:10:1).
## Summary
When the uv version is read from dependencies (`requirements.txt`, or
`project.dependencies` / `optional-dependencies` / `dependency-groups`
in `pyproject.toml` via `version-file`), the specifier was matched with
`/^uv([=<>~!]+\S*)/`. PEP 508 allows whitespace around operators and
after commas, and those valid forms were misread:
| dependency | before | after |
|---|---|---|
| `uv >= 0.5.0` | not found, latest uv installed | `>=0.5.0` |
| `uv == 0.11.20` | not found, latest uv installed | `==0.11.20` |
| `uv>=0.5.0, <0.6.0` | `>=0.5.0,`, fails with `No version found for
>=0.5.0,` | `>=0.5.0,<0.6.0` |
The regex now allows whitespace around operators and between
comma-separated clauses, and strips it from the result. Inputs that
worked before give the same result, including inline `--hash=` options,
trailing ` # comments`, the `\` continuation lines from `uv pip compile
--generate-hashes`, and names that only start with `uv` (`uvicorn`,
`uv-build`).
## Test plan
- Added cases for spaced operators, spaced clauses, inline `--hash` and
trailing comments. The four whitespace cases failed before the change.
- `npm run all` on Ubuntu (WSL) with Node 24.19.0: build, Biome (no
fixes), packaging, and 165 tests across 15 suites pass. The rebuilt
`dist/setup/index.cjs` only differs in the changed line.
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
## Summary
`~=` is the PEP 440 compatible release operator, but the resolver tries
`semver.validRange()` first, and semver accepts `~=` as a tilde range:
| specifier | semver | PEP 440 |
|---|---|---|
| `~=0.9` | `>=0.9.0 <0.10.0` | `>=0.9, ==0.*` |
So with the default `highest` strategy, `version: "~=0.9"` (or
`required-version = "~=0.9"` picked up through `version-file`) installs
the highest 0.9.x instead of the highest 0.x. The same happens with
`manifest-file`, where `tc.evaluateVersions` is semver-based.
This skips the semver evaluation for specifiers containing `~=`, so they
are only evaluated with `@renovatebot/pep440`. No semver range
legitimately uses `~=`.
Notes:
- Only a bare two-segment `~=X.Y` resolved incorrectly. `~=X.Y.Z` gives
the same result in both, and compound specifiers like `~=0.9, !=0.9.3`
were already rejected by semver and fell through to PEP 440.
- `minSatisfying` gets the same check for consistency. The lower bound
is the same under both interpretations, so `lowest` didn't produce wrong
results before.
- A single-segment `~=1` is invalid PEP 440. semver used to resolve it
to 1.x; it now fails with `No version found for ~=1`, same as other
invalid specifiers.
## Test plan
- Added tests for `~=0.9` with the default manifest and with
`manifest-file`. Both returned `0.9.26` instead of `0.10.2` before the
change.
- `npm run all` on Ubuntu (WSL) with Node 24.19.0: build, Biome (no
fixes), packaging, and 161 tests across 15 suites pass. The rebuilt
`dist/setup/index.cjs` only differs in the changed lines.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
## Summary
The post step logs `uv cache saved with key: ...` (and `Python cache
saved with key: ...`) even when the save failed. The log in #769 shows
it right after `Warning: Failed to save: ...`:
```
Warning: Failed to save: "C:\Program failed with error: The process 'C:\Program Files\Git\usr\bin\tar.exe' failed with exit code 2
uv cache saved with key: setup-uv-2-x86_64-pc-windows-msvc-windows-11-3.12.12-pruned-...
```
`@actions/cache`'s `saveCache` doesn't throw for these failures. It logs
them and returns `-1`, and `saveCacheToKey` ignored the return value.
This now checks for `-1` before logging, which is the same check the
`actions/cache` save step uses (`if (cacheId != -1) core.info("Cache
saved with key: ...")`). `@actions/cache` already logs the reason, so
nothing else is logged on failure, and the exit code stays the same.
This only changes the log line. It doesn't address the underlying
tar/gzip failure in #769. One limitation: with the legacy (v1) cache
service, an upload that fails after the cache was reserved still returns
the reserved id, so that one case is still reported as saved. That's on
the `@actions/cache` side and affects `actions/cache` the same way.
Related to #769
## Test plan
- Added `__tests__/cache/save-cache.test.ts`. The `-1` case fails before
the change and passes after it.
- `npm run all` on Ubuntu (WSL) with Node 24.19.0 / npm 11.17.0: build,
Biome (no fixes), packaging, and all 161 tests across 16 suites pass.
The regenerated `dist/save-cache/index.cjs` only differs in the changed
lines.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
## Summary
Roll up the remaining dependency changes from eight open Dependabot PRs
against current `main`, without reverting newer workflow settings:
- CodeQL init, analyze, and autobuild: 4.37.7 → 4.38.1 (#1097, #1096,
#1095)
- zizmor-action: 0.6.3 → 0.6.4 (#1085)
- @biomejs/biome: 2.5.9 → 2.5.11, including its schema URL (#1075)
- @types/node: 26.2.0 → 26.4.0 (#1074)
- Jest: 30.4.2 → 30.5.0 (#1073)
- js-yaml: 5.3.0 → 5.4.1 (#1072)
All nine open Dependabot PRs had remaining dependency changes; none was
a no-op. TypeScript 7.0.2 (#1029) is intentionally deferred with
maintainer approval: ts-jest 29.4.11 requires TypeScript <7, and
applying the update caused all 15 test suites to fail before running
with `Cannot read properties of undefined (reading 'readFile')`.
The lockfile was regenerated using npm install with scripts disabled. No
runtime dependency entries changed, and repackaging produced no `dist/`
changes.
## Validation
Using Node 24.19.0 and npm 11.10.0:
- `npm ci --ignore-scripts`
- `npm run all` — passed; 15 test suites, 159 tests; Biome applied no
fixes
- `npm audit signatures` — all 353 audited packages have verified
registry signatures; 38 have verified attestations
- `actionlint .github/workflows/codeql-analysis.yml
.github/workflows/test.yml` — passed
- `uvx zizmor --no-progress --offline
.github/workflows/codeql-analysis.yml .github/workflows/test.yml` — no
findings
- `git diff --check` — passed
Refs: pi-session 01a15ec2-de1a-7557-a0bb-00369b0a4a10
## Summary
- Preserve class names in the save-cache bundle so `@actions/cache`
error-name comparisons distinguish validation errors, benign
cache-reservation races, and finalization errors correctly.
- Add regression coverage that exercises the bundled `ValidationError`,
`ReserveCacheError`, and `FinalizeCacheError` constructors.
Fixes#1028
## Verification
- `npm run all` passed on Ubuntu 24.04 with Node 24.19.0: build, Biome,
packaging, and all 159 tests across 15 suites.
- `git diff --check` passed.
The save-cache bundle increases by about 77 KB raw (14 KB gzip) to
preserve the required names.
CI installs the locked npm dependency tree without running lifecycle
scripts, but it does not verify registry signatures or available
provenance attestations. Run `npm audit signatures` after installation
in both the test and checksum-update workflows, before `npm run all`
executes dependency tooling. The check follows npm's registry-signature
support and verifies attestations when present; it does not require
every dependency to publish provenance.
Related: astral-sh/ruff-action#402 adds the same signature and
provenance verification. astral-sh/setup-uv#1026 separately hardens
`npm` install defaults.
Co-authored-by: zaniebot <242828183+zaniebot@users.noreply.github.com>
Windows ARM64 jobs may need x64 Python independently of their Python
version pin. Add `python-arch` to export `UV_PYTHON_ARCH` for subsequent
steps and include the selected architecture in the uv and managed-Python
cache keys. The input takes precedence over an inherited environment
value, and unsupported uv versions fail before environment creation.
Requires a uv release with `UV_PYTHON_ARCH` support.
We move the `macos-14` test jobs to `macos-15` ahead of GitHub’s
[October 5 brownout and November 2, 2026
retirement](https://github.com/actions/runner-images/issues/13518).
We also remove the macOS 14 cache-key test case; the matrix already
covers macOS 15.