chore: remove verified dead items - #244
Conversation
vladimirrott
left a comment
There was a problem hiding this comment.
Good work, and thanks for holding the diff to the four items. Windows stopped you short of clippy and nextest, so I ran the gates twice, once on your head commit and once with main merged in (bdfbe36):
cargo fmt --all -- --checkcleancargo clippy --workspace --all-features --all-targets --locked -- -D warningscleancargo doc --no-deps --workspace --lockedunderRUSTDOCFLAGS=-D warningscleanscripts/test_baseline.sh: 1759 tests run, 1759 passed, count matchingtests/evidence/workspace-tests.json
The trap in #235 is what I wanted verified most, and the workspace clippy run covers it. I saw Compiling sysknife-shell v0.8.0 (apps/sysknife-shell/src-tauri) in the output, so the paused GUI builds with the re-exports gone. Your lockfile edit is the minimal one, dropping the sysknife-proto edge while prost-build keeps prost-types as a transitive dependency, and cargo metadata --locked asks for no regeneration. On the stale allow: rustc 1.97.1 on edition 2021 emits no unused_unsafe for that block, so it was carrying nothing.
Two comment lines to fix here, both left behind by the deletions:
crates/sysknife-daemon/src/actions/resolvectl.rs:5namesDEBIAN_ONLY_ACTION_NAMES, which this PR removes from the workspace. The surviving name isDEBIAN_ONLY_ACTIONS.crates/sysknife-brain/src/prompt.rs:1143still heads that block as a set of re-exports it no longer holds.
Push those two and I will merge. The release side is mine to handle: sysknife-core and sysknife-brain both publish to crates.io, so dropping PRODUCTION_LISTEN_URI, the distro re-export, UnknownActionName and the two prompt aliases ships in 0.9.0 rather than a patch release.
I approved the workflow runs on both your PRs. A red security-audit is RUSTSEC-2026-0258 in h2, fixed on main in #245.
If you want another, #229 and #233 are open and in the same vein: deriving the story-coverage figures instead of restating them, and the clustered map of the 64 uncovered actions.
|
On the red security-audit: your run checked out a merge with main at bb06cb6, which predates the h2 bump in #245. I merged current main into your branch locally and ran |
|
Addressed both stale comment lines in 209cd9a. No code behavior changed; |
vladimirrott
left a comment
There was a problem hiding this comment.
Both comment lines are right, and the merge you pushed puts you on current main.
Re-ran the full gate set at 97dfb87 after that merge: fmt clean, clippy --workspace --all-features --all-targets --locked -- -D warnings clean, cargo doc under RUSTDOCFLAGS=-D warnings clean, and 1759/1759 under nextest with the count still matching tests/evidence/workspace-tests.json.
Your Windows note is accurate. vsock and std::os::unix::net::UnixDatagram stop the workspace before clippy finishes, so that gate was out of your reach. I ran it on both heads, and the log shows Compiling sysknife-shell v0.8.0 (apps/sysknife-shell/src-tauri), so the paused GUI still builds with the re-exports gone. That was the one real risk in #235.
Approving. The version bump is mine to handle: sysknife-core and sysknife-brain publish to crates.io, so these four removals go out in 0.9.0 rather than a patch release, and I will say so in the notes.
For a next one, #248 is the same flavour of compile-proved deletion (a dead exclusive_resource arm), and #221 is an easy audit bug if you want to move from removals into behaviour. Two clean PRs in a day is a strong start.
…erences-envelope * upstream/main: chore: remove verified dead items (lacs-project#244) chore(deps): bump the github-actions group with 4 updates (lacs-project#258) chore(deps): bump the cargo-minor-patch group with 2 updates (lacs-project#257) chore(deps-dev): bump @testing-library/jest-dom (lacs-project#255) chore(deps): bump library/rust from `77fac8b` to `0e2bcae` (lacs-project#254)
Bumps every package version, the internal path-dependency pins, and the five JSON manifests from 0.8.0 to 0.9.0, and writes the release notes. The middle digit moves because #244 removed published items: PRODUCTION_LISTEN_URI, the sysknife-core root re-export of the distro helpers, UnknownActionName, and the two prompt aliases. Cargo treats the leftmost non-zero component as the compatibility unit, so a consumer on "0.8" would have adopted a patch release carrying those removals. #243 also changed setup behaviour: a malformed .mcp.json that earlier versions overwrote now aborts the run. Two additions that came out of the release itself: - check_release_versions.sh now checks the 15 internal `sysknife-* = { path, version }` pins as well as the package versions, and fails when its manifest globs match nothing. Missing one pin would publish a crate depending on the previous release of its sibling. - docs/release.md documents how the digits are chosen while the leading zero stands, and names three conditions for 1.0.0 so the switch to MAJOR-for-breaking is a decision rather than a mood. CONTRIBUTING.md points contributors at it, next to the test-baseline artifact they need to regenerate when they add a Rust test.
Summary
prost-typesdependency and update the lockfileunused_unsafeallowPRODUCTION_LISTEN_URIconstant and tautological assertionFixes #235
Testing
cargo fmt --all -- --check— passed--lockedvsock, thenstd::os::unix::net::UnixDatagram) fail before the workspace finishes