Autostart is not working correctly

[Sebastian Rietzscher]

Hey,
I just noticed that autostart for containers is not working correctly in the current release northstar-runtime 0.5.0.
Here is a bit more detailed description:

When configuring a container to autostart, Northstar should start the container automatically
currently those containers are not started anymore.
Here is an examplified config:

name: hello-world
version: 0.0.1
init: /hello-world
uid: 1000
gid: 1000
autostart: relaxed
env:
  HELLO: northstar
io:
  stdout: pipe
  stderr: pipe
mounts:
  /dev:
    type: dev
  /proc:
    type: proc
  /lib:
    type: bind
    host: /lib
  /lib64:
    type: bind
    host: /lib64
  /system:
    type: bind
    host: /system

IMO the handling of containers that should be auto started is wrong in northstar-runtime/src/runtime/state.rs:274

        if !to_mount.is_empty() {
            self.mount_all(&to_mount).await;

            for (container, autostart) in autostarts {
                info!("Autostarting {} ({:?})", container, autostart);
                if let Err(e) = self
                    .start(&container, &[], &HashMap::with_capacity(0))
                    .await
                {
                    Self::warn_autostart_failure(&container, &autostart, e)?
                }
            }
        }

With this implementation only containers with a resource will be auto started.

Possible solution mount resources then start

        if !to_mount.is_empty() {
            self.mount_all(&to_mount).await;
        }
        for (container, autostart) in autostarts {
             info!("Autostarting {} ({:?})", container, autostart);
             if let Err(e) = self
                   .start(&container, &[], &HashMap::with_capacity(0))
                   .await
              {
                    Self::warn_autostart_failure(&container, &autostart, e)?
              }
         }

Cheers
Sebs