APIv2 /containers/{id}/wait doesn't work with correct condition enumeration

[matejvasek]

Currently the used enumeration is here:
podman/libpod/define/containerstate.go

const (
	ContainerStateUnknown ContainerStatus = iota
	ContainerStateConfigured ContainerStatus = iota
	ContainerStateCreated ContainerStatus = iota
	ContainerStateRunning ContainerStatus = iota
	ContainerStateStopped ContainerStatus = iota
	ContainerStatePaused ContainerStatus = iota
	ContainerStateExited ContainerStatus = iota
	ContainerStateRemoving ContainerStatus = iota
)

But docker uses different values:

const (
	WaitConditionNotRunning WaitCondition = "not-running"
	WaitConditionNextExit   WaitCondition = "next-exit"
	WaitConditionRemoved    WaitCondition = "removed"
)

[matejvasek]

Calls like this fails.

POST /v1.38/containers/ctrId/wait?condition=next-exit

[matejvasek]

I could work on this if nobody else wants to.

[rhatdan]

You got it.

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[rhatdan]

@matejvasek Any progress?

[matejvasek]

@rhatdan I should get to this soon. I think I could get PR sometimes next week.

Unfortunately, before starting working directly on this I spent (or wasted?) some time on related issue.
I really didn't like the idea of checking state of a container every 250ns. I was trying do add some kind of notification mechanism to lock.Locker, something like condition variable. Long story short: it didn't work.

There is a problem when waiting process is forcefully killed, like with SIGKILL (however SIGTERM or SIGINT can be handled).
First I tried to use standard posix cond var. The issues is that posix cond var is not robust.
Then I tried to write my own cond var using semaphores,
but there would be a small memory leak anytime waiting process dies.
Last but not least, this would work only with SHM lock, not FS locks.