mirror of
https://github.com/docker/compose.git
synced 2026-11-20 06:59:53 +00:00
docker-agent review: moving the WaitTimeout deadline to wrap ctx before executePlan closed the in-plan hang, but opened a narrow new one -- executePlan can return nil right as the deadline fires, leaving the subsequent getContainers call (for the final --wait verification) to run against an already-expired ctx and propagate a raw context.DeadlineExceeded instead of the existing friendly "application not healthy after %s" message. Same translation the executePlan and waitDependencies error paths already apply. Also, a second docker-agent finding: execRenameContainer's containersByService refresh silently no-ops if the container isn't found at its expected slot (can't happen today -- the DAG guarantees execCreateContainer's append runs first -- but nothing would surface it if a future change broke that). Logs a warning instead of staying silent, so a regression here is visible rather than just quietly reintroducing the stale-name bug this PR fixed. Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com> |
||
|---|---|---|
| .. | ||
| api | ||
| bridge | ||
| compose | ||
| dryrun | ||
| e2e | ||
| mocks | ||
| remote | ||
| utils | ||
| watch | ||