mirror of
https://github.com/woodpecker-ci/woodpecker.git
synced 2026-05-04 12:32:36 +00:00
a7739a2bb7
When the agent started before the server was available, it retried the connection as expected. However, once the server came up and a workflow was picked up, the pipeline would immediately fail without running any steps — the agent logs showed `workflow context done` firing instantly after `received execution`. The root cause was a package-level `shutdownCtx` shared across retry iterations. On each failed attempt, `stopAgentFunc` stamped it with a 5-second timeout — starting the clock immediately. By the time the agent successfully connected and received a workflow, `workflowCtx` was derived from this already-expired context, so execution failed before Docker even started a container. The fix removes the global mutable shutdown context and the `stopAgentFunc` indirection. Instead, `runner.Run()` no longer accepts a `shutdownCtx` parameter — it creates a fresh one locally only when needed for the `Done()` fallback call. The healthcheck server shutdown does the same. This makes the lifetime of each shutdown window explicit and local.