syntaqxA run that is still going offers Cancel run, and not the re-runs, on every page of it. Which buttons a run's page shows is worked out in one place (runControls, lib/run-controls.ts): the run is going while its latest attempt is, as the attempt list's last entry says, whichever attempt the page shows, and on the latest attempt a job that has not finished (waiting for the jobs it needs, held for an environment's approval, queued for a runner, running) keeps it going whatever the run's own status says. So an earlier attempt's page, and its ?job= links, offer Cancel run while a later attempt runs, where they offered nothing, and keep refreshing until it ends; the re-runs show only on the latest attempt of a finished run. A page left open on a finished run asked for nothing and heard nothing when the run was re-run from somewhere else (the API, the MCP tool, another tab): it kept showing the attempt that had ended, with its Re-run buttons and no Cancel run, until a job of the new attempt started or finished. A re-run now tells pages open on the run, as a job starting or finishing does. The run view's status for its current attempt is read by one function (shown_status), with tests that a job queued for a self-hosted runner leaves its run in progress and a job held at an environment makes it wait. The Actions guide says when Cancel run shows.73a1fc9History
- ..
- 0001_init.sql
- 0002_reusable_workflows.sql
- 0003_settings_access.sql
- 0004_self_hosted_runners.sql
- 0005_actions_cache.sql
- 0006_artifacts.sql
- 0007_run_protection.sql
- 0008_actions_access.sql
- 0009_run_attempts.sql
- 0010_jobs_self_hosted_lower.sql
- 0011_schedule_rules.sql
- 0012_runner_task_runs.sql
- 0013_action_registry.sql
- 0014_managed_workflows.sql
- 0015_runner_isolation.sql
- 0016_log_chunks.sql
- 0017_workspace_managed_workflows.sql
- 0018_desktop_runners.sql