test:add ci for cleanup fifos

[ningmingxiao]

add some ci for #4316

[fahedouch]

is it possible to fix on the same PR #4316 ?

[ningmingxiao]

is it possible to fix on the same PR #4316 ?
pr 4316 long time noby answer, you can mege it and then merge my pr.
and i think we should call io.Close() after @fahedouch

case status := <-exitCh:

[fahedouch]

Some suggestions:

• Close IO even when task is already Stopped:

  • In Stop, for the early return case case containerd.Stopped, consider closing IO before returning to prevent leaving FIFOs up until another path cleans them:
    • Example change:
      • Before return nil for containerd.Stopped, do:
        • if io := task.IO(); io != nil { if err := io.Close(); err != nil { log.Warnf(...) } }
  • Some shims/plugins may leave IO present until an explicit close; proactively closing avoids FIFO linger in this rare path.

• Log levels/messages:

  • The warning on IO close failure is good. Optionally include more context (namespace, task pid if available) for easier troubleshooting.

[fahedouch]

CI is failing

[ningmingxiao]

done

[fahedouch]

is it possible to avoid goroutine and make sequential check to avoid synchronisation issues and then flaky tests ?

[fahedouch]

The test compares global counts in /run/containerd/fifo/. If other tests run in parallel and create/cleanup tasks, the global count may fluctuate, making the test flaky:

option: Prefer filtering by container/task ID in the FIFO names, if possible. Otherwise, add a short stabilization poll or mark the test as non-parallel and document the global dependency.

[ningmingxiao]

if don't use t.Parallel() the test cases will execute serially. if this testcase is running other test case with t.Parallel() will not run.

[fahedouch]

Thanks