Cover a still-running one-shot in the rolling-swap tests #98

New issue

Open

opened 2026-10-10 05:40:58 +00:00 by ahmad · 0 comments

ahmad commented

2026-10-10 05:40:58 +00:00

Owner

Context

From the review of #95. The behaviour suite never exercises the branch where a one-shot is still running when the wait starts: the docker stub always answers exited. So a mutant that never waits passes. Covering it needs date stubbed, which the harness does not do yet. Also, a migration with restart: on-failure is not waited for, because only policy no is; the README does not say so.

Scope

In: a time-controllable date stub in the harness, a case for a one-shot that is running and then exits, and a README word on on-failure.
Out: changing which policies count as one-shots.

Acceptance criteria

Given a mutant that skips the wait loop, when the behaviour suite runs, then a case fails.
Given a reader of the rolling section, when their migration uses restart: on-failure, then the README tells them it is not waited for.

### Context From the review of #95. The behaviour suite never exercises the branch where a one-shot is still running when the wait starts: the docker stub always answers `exited`. So a mutant that never waits passes. Covering it needs `date` stubbed, which the harness does not do yet. Also, a migration with `restart: on-failure` is not waited for, because only policy `no` is; the README does not say so. ### Scope - In: a time-controllable `date` stub in the harness, a case for a one-shot that is running and then exits, and a README word on `on-failure`. - Out: changing which policies count as one-shots. ### Acceptance criteria - [ ] Given a mutant that skips the wait loop, when the behaviour suite runs, then a case fails. - [ ] Given a reader of the rolling section, when their migration uses `restart: on-failure`, then the README tells them it is not waited for.

ahmad added the

labels

2026-10-10 05:40:58 +00:00