completion: container diff: no such pod

[edsantiago]

Seen when running system tests in parallel:

# [12:02:39.744447467] $ bin/podman __completeNoDesc  container diff
# [12:02:40.521000319] [Debug] [Error] no such pod
# :4
# Completion ended with directive: ShellCompDirectiveNoFileComp
# #/vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
# #|     FAIL: container diff: actual container listed in suggestions
# #| expected: '.*-ctrt111_bcqa22xf
# ' (using expr)
# #|   actual: '[Debug] [Error] no such pod'
# #|         > ':4'
# #|         > 'Completion ended with directive: ShellCompDirectiveNoFileComp'
# #\^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

I've been unable to reproduce

[Luap99]

This looks very strange, I cannot see where the pod comes from here.

[Luap99]

Do you have the actual test log? Is this sqlite or boltdb?

[edsantiago]

Seems to be both sqlite and boltdb. Here's f40 rootless in CI

[edsantiago]

If you want a reproducer, pull #23275 and run hack/bats --rootless --tag=para (I absolutely need to come up with a better tag name than that!) It fails about half the time on my laptop. You will need to instrument podman somehow to dump debug info when this happens, because no useful state is preserved on failure.

[Luap99]

What I don't get though is why doe sit only print no such pod, This is a valid error message but I don't see any call where such message would not be at wrapped by more context.

[Luap99]

as I expected this is not related to shell completion at all as this just lists things, running

while :; do bin/podman ps -a --pod  || break ;done

in parallel to the system tests reproduces this as well. The important bit is --pod which is always used be the shell completion code but not for most normal podman ps calls so this is why you did not notice it elsewhere so far