Deploying to a host

rho deploy puts a built program on a host that runs an operating system — a
Linux board, a gateway machine — as opposed to a part that is written by a probe.
What the host is, and how it is reached, is a property of the target: a spec
carries a deploy block, and this command reads it through rho target show and
carries it out.

rho deploy -target uno-q -run build/gateway

The deploy block says:

"deploy": {
  "method": "rsync",            // rsync (default) or scp
  "host": "uno-q.local",
  "port": 22,
  "user": "arduino",
  "path": "/home/arduino/apps",
  "run": "cd {artifact} && ./gateway"
}

A docker block in the same place deploys a container image instead:

"deploy": {
  "host": "uno-q.local", "user": "arduino", "path": "/home/arduino/apps",
  "docker": {
    "file": "Dockerfile", "image": "robohub-gateway", "tag": "v1",
    "run": ["--net", "host", "-p", "8080:8080", "{image}"]
  }
}

The image is built here, saved, carried, loaded and started on the target.

-n prints every step and runs nothing. The address, login, path, port and
method can be redirected from the environment — RH_DEPLOY_HOST,
RH_DEPLOY_USER, RH_DEPLOY_PATH, RH_DEPLOY_PORT, RH_DEPLOY_METHOD — so a
spec is not a fact about one bench.