fleetd #581: cover completion CAS removals #585

This branch's head dfeb9340b is an ancestor of origin/main, brought in by 1db79bcaa
("Merge worker/581-completionresolver-cas-sites-0542b7-6"). Gitea never recorded it as a PR merge
because the merge was done locally and pushed, not through the forge.

$ git merge-base --is-ancestor dfeb9340b origin/main && echo merged
merged
$ git rev-list --ancestry-path --reverse dfeb9340b..origin/main --merges | head -1
1db79bcaa

I also verified #581's coverage is real rather than assumed, by mutation rather than by reading.
Two builds, each mutating two of the nine 2-arg CAS sites in CompletionResolver to the 1-arg
remove(target) form:

sites :445 and :566 → 3 named failures, all three asserting a superseded turn must not evict
its successor's registration.
sites :335 and :549 → 2 named failures, both on the early-return paths
(aSupersededDoneTurnMustNotEvictItsSuccessorsRegistration:643,
aSupersededDoneFailedTurnMustNotEvictItsSuccessorsRegistration:738).

Every mutation was killed, and CompletionResolver.java was restored to its pristine
9b5f2f6d810fa45e after each build. So all nine sites are killable and none is an equivalent
mutant.

One correction to something I wrote on #581 earlier: my "7 tests, one per unguarded site" claim was
a map I had not measured. Measured, two of the seven pin the early returns, and :445/:566 were
already covered by tests that existed before this branch.

Already in `main` — closing as merged. This branch's head `dfeb9340b` is an ancestor of `origin/main`, brought in by `1db79bcaa` ("Merge worker/581-completionresolver-cas-sites-0542b7-6"). Gitea never recorded it as a PR merge because the merge was done locally and pushed, not through the forge. ``` $ git merge-base --is-ancestor dfeb9340b origin/main && echo merged merged $ git rev-list --ancestry-path --reverse dfeb9340b..origin/main --merges | head -1 1db79bcaa ``` I also verified #581's coverage is real rather than assumed, by mutation rather than by reading. Two builds, each mutating two of the nine 2-arg CAS sites in `CompletionResolver` to the 1-arg `remove(target)` form: - sites `:445` and `:566` → 3 named failures, all three asserting a superseded turn must not evict its successor's registration. - sites `:335` and `:549` → 2 named failures, both on the early-return paths (`aSupersededDoneTurnMustNotEvictItsSuccessorsRegistration:643`, `aSupersededDoneFailedTurnMustNotEvictItsSuccessorsRegistration:738`). Every mutation was killed, and `CompletionResolver.java` was restored to its pristine `9b5f2f6d810fa45e` after each build. So all nine sites are killable and none is an equivalent mutant. One correction to something I wrote on #581 earlier: my "7 tests, one per unguarded site" claim was a map I had not measured. Measured, two of the seven pin the early returns, and `:445`/`:566` were already covered by tests that existed before this branch.

ltms closed this pull request

2026-09-12 16:42:31 +02:00

CI / shell-tests (pull_request) Successful in 5s

Details

CI / build (pull_request) Failing after 1m31s

Details

CI / contract (pull_request) Successful in 1m35s

Details

Pull request closed

Please reopen this pull request to perform a merge.

Sign in to join this conversation.

2 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: fleet/fleetd#585