Sign in to a server and share swarms through it

[evoggy]

Signing in

cfcli auth login signs in through the browser, so there's no key to copy:

1. cfcli listens on 127.0.0.1 and opens the server's sign-in page.
2. The user signs in there and allows "cfcli on ".
3. The browser brings a one-time code back to cfcli, which exchanges it for an API key.

This is OAuth 2.0 authorization code with PKCE (RFC 7636) and a loopback redirect (RFC 8252). The key is stored in credentials.json next to the cfcli config, readable only by the user, and the server lists it on its API keys page, where it can be revoked.

• cfcli auth status: who cfcli is signed in as. Exits with code 20 when it isn't signed in.
• cfcli auth logout: revokes the key and forgets it.
• --no-browser: only prints the link.

Shared swarms

When cfcli is signed in, swarms on the server sit next to the local ones, named <organization>/<swarm>. Local swarms stay <swarm>, and every swarm command takes either kind of ID.

• Listing: swarm config list shows both, with where each is stored.
• Creating, changing and deleting: create and delete with an <org>/<swarm> ID act on the server. add, remove, rename and rechannel change the shared swarm.
• Moving: swarm config move lab org/lab shares a local swarm, and move org/lab lab takes it back, deleting it on the server after asking. move also renames a swarm, and the selected swarm follows the move.
• Sync (cfcli settings sync on|off):
  • On, the default: commands check the server first, which is a cheap "not modified" answer when nothing changed, and upload changes right away. If someone else uploaded in between, the change is made again on their version.
  • Off: commands use this computer's copies, and swarm config pull and push sync them. push stops when someone else changed the swarm since the copy; pull --force or push --force chooses which version wins.
• Without the server: commands use the copies and keep changes for later, so working with the Crazyflies never waits for the network.

The copies are kept in synced/ next to the config, outside the swarms folder, so Swarmkeeper and import never take them for local swarms. Shell completion lists shared swarms from the copies and never touches the network.

Testing

• Unit tests: PKCE, the browser callback listener, and parsing of shared IDs.
• End to end against a local server instance, with a second user changing swarms in between:
  • Sign-in, status, logout, and cancelling in the browser.
  • Creating, adding and renaming, with the other user's changes kept.
  • Moving a swarm both ways.
  • Sync off: push conflicts, pull keeping unpushed changes, and --force.
  • Offline: falling back to the copies, then uploading by itself when the server came back.
  • Deleting the selected swarm, completion without the network, and errors when signed out.

The server side is deployed at https://arc.bitcraze.io, the default server, so cfcli auth login works against it as it is.