gitlab-runner register drops some arguments based on order

Summary

When running gitlab-runner register with command line arguments, some arguments are ignored based on where they appear in the list.

Steps to reproduce

Example of full parsing of arguments

gitlab-runner register --non-interactive \
      --url $GITLAB_URL \
      --executor kubernetes \
      --kubernetes-namespace gitlab \
      --kubernetes-cpu-limit "1000m" \
      --kubernetes-memory-limit "128Mi" \
      --kubernetes-cpu-request "100m" \
      --kubernetes-poll-interval 3 \
      --kubernetes-poll-timeout 300 \
      --kubernetes-memory-request "128Mi" \
      --kubernetes-service-cpu-limit "1000m" \
      --kubernetes-service-memory-limit "128Mi" \
      --kubernetes-service-cpu-request "100m" \
      --kubernetes-service-memory-request "128Mi" \
      --kubernetes-helper-cpu-limit "1000m" \
      --kubernetes-helper-memory-limit "128Mi" \
      --kubernetes-helper-cpu-request "100m" \
      --kubernetes-helper-memory-request "128Mi" \
      --kubernetes-image "ubuntu:16.04" \
      --kubernetes-privileged "true"

Example of incomplete parsing of arguments

gitlab-runner register --non-interactive \
      --url $GITLAB_URL \
      --executor kubernetes \
      --kubernetes-image "ubuntu:16.04" \
      --kubernetes-privileged "true" \
      --kubernetes-namespace gitlab \
      --kubernetes-cpu-limit "1000m" \
      --kubernetes-memory-limit "128Mi" \
      --kubernetes-cpu-request "100m" \
      --kubernetes-poll-interval 3 \
      --kubernetes-poll-timeout 300 \
      --kubernetes-memory-request "128Mi" \
      --kubernetes-service-cpu-limit "1000m" \
      --kubernetes-service-memory-limit "128Mi" \
      --kubernetes-service-cpu-request "100m" \
      --kubernetes-service-memory-request "128Mi" \
      --kubernetes-helper-cpu-limit "1000m" \
      --kubernetes-helper-memory-limit "128Mi" \
      --kubernetes-helper-cpu-request "100m" \
      --kubernetes-helper-memory-request "128Mi"

Note that --kubernetes-image and --kubernetes-privileged are located higher in the list

Actual behavior

When running the problematic invocation, /etc/gitlab-runner/config.toml has

[[runners]]
  name = "redacted"
  url = "redacted"
  token = "redacted"
  executor = "kubernetes"
  [runners.cache]
  [runners.kubernetes]
    host = ""
    image = "ubuntu:16.04"
    namespace = ""
    namespace_overwrite_allowed = ""
    privileged = true
    service_account_overwrite_allowed = ""
    [runners.kubernetes.volumes]

Expected behavior

Both should result in the same output, which happens on the non-problematic invocation

[[runners]]
  name = "redacted"
  url = "redacted"
  token = "redacted"
  executor = "kubernetes"
  [runners.cache]
  [runners.kubernetes]
    host = ""
    image = "ubuntu:16.04"
    namespace = "gitlab"
    namespace_overwrite_allowed = ""
    privileged = true
    cpu_limit = "1000m"
    memory_limit = "128Mi"
    service_cpu_limit = "1000m"
    service_memory_limit = "128Mi"
    helper_cpu_limit = "1000m"
    helper_memory_limit = "128Mi"
    cpu_request = "100m"
    memory_request = "128Mi"
    service_cpu_request = "100m"
    service_memory_request = "128Mi"
    helper_cpu_request = "100m"
    helper_memory_request = "128Mi"
    poll_interval = 3
    poll_timeout = 300
    service_account_overwrite_allowed = ""
    [runners.kubernetes.volumes]

Relevant logs and/or screenshots

I believe the above section is sufficient. There are no errors provided to me on the problematic invocation.

Environment description

Custom integration of gitlab in a kubernetes environment.

Chart is a bit of a fork gitlab-ce chart and gitlab-runner fork .. runner hasn't been updated with new code because of this issue.

If it means anything, Kubernetes 1.6 w/ rbac

Used GitLab Runner version

I was using gitlab/gitlab-runner:v9.5.0 docker image inside of a pod in kubernetes.