Take the brace-expansion and postcss fixes, drop a retired allowlist entry, release 2.7.3 - #71
Conversation
…ry upstream retired Three advisories were reachable in the lockfile with fixed versions already published: brace-expansion (two High, one per locked copy) and postcss. None had been taken because bun does not re-resolve a locked version that still satisfies its range, so a routine refresh could never reach them. A clean re-lock takes all three and dedupes minimatch and typescript-eslint. The brace-expansion allowlist entry is deleted rather than waived. It said the 1.x line had no fixed release, which was true when written; upstream has since backported to 1.1.17 and then 1.1.18, and minimatch@3 asks for ^1.1.7, so the fix arrives with no override and none of the API breakage that made forcing 5.x impossible. osv-scanner now reports the entry as an unused ignore, which is what a stale suppression looks like. The esbuild override stays. Re-tested by removing it: natural resolution is 0.27.7, which carries GHSA-g7r4-m6w7-qqqr, so the floor is still doing work. No manifest change, so no release is needed and no consumer has to act. Verification: lint, typecheck, knip, build and test (166) all green on a clean install, and osv-scanner v2.4.0 reports no issues. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Version bump for release. bun.lock carries no root version field, so unlike the uv side there is nothing to re-lock alongside it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
Scope grew after the original body: this branch now also bumps The published artifact is still functionally identical to 2.7.2.
Gates re-run after the bump: lint, typecheck, build all exit 0; tests 166 passed, 9 skipped. |
Summary
Three advisories were reachable in the lockfile with fixed versions already published:
brace-expansion(two High, one per locked copy) andpostcss. None had been taken because bun does not re-resolve a locked version that still satisfies its range, so a routine refresh could never reach them. A clean re-lock takes all three and dedupesminimatchandtypescript-eslint.The
brace-expansionallowlist entry is deleted rather than waived. It said the 1.x line had no fixed release, which was true when written; upstream has since backported to 1.1.17 and then 1.1.18, andminimatch@3asks for^1.1.7, so the fix arrives with no override and none of the API breakage that made forcing 5.x impossible. osv-scanner now reports that entry as an unused ignore, which is what a stale suppression looks like.The
esbuildoverride stays. Re-tested by removing it: natural resolution is 0.27.7, which carries GHSA-g7r4-m6w7-qqqr, so the floor is still doing work.Held versions are unchanged: eslint /
@eslint/json 9.x and typescript on 6.x, both deliberate cross-repo decisions rather than routine bumps.typescript-eslint8.66.0 exists but is inside the 24h release-age soak and will arrive on its own.Type of change
Public API
None. No exported type, signature, wire format or response shape changed.
package.jsonis untouched apart from nothing at all: the diff is the lockfile and the scanner allowlist, so no release is needed and no caller has to act.Test plan
Run from a clean install (
rm -rf node_modules bun.lock && bun install), not a tree left over from an experiment:bun run lintexit 0bun run typecheckexit 0bun run knipexit 0bun run buildexit 0bun run test: 166 passed, 9 skippedbun.lock: no issues foundThe override re-test was done in a scratch copy of the manifest so the working tree was never left in a half-resolved state, and the advisory status of both candidate versions was confirmed against the OSV API rather than inferred.
Checklist