Make node start explain its failures (v0.15.1)

[gmpassos]

Born from a real debugging session: a node that authenticated and was then refused registration retried forever in total silence. It took a live packet trace to find the cause — a grant with role server/gpu instead of node.

The root cause

The signal was never missing. The omnyhub node runtime already receives the Hub's rejection — NodeErrorMessage(code: forbidden, message: "principal server-node may not register node sites-menuici") — and logs it. But its logger defaults to a NoopLogger that OmnyServer never replaced (node_runtime.dart:287). Every diagnostic — the rejection reason, connection-refused, bad-certificate — went to a silent sink.

Changes

1. Wire a real logger into the runtime (ConsoleNodeLogger). Problems-only by default; full lifecycle under node start --verbose. Repeats within a short window are collapsed, so a node that keeps retrying a rejection it can't fix says why once, not once per backoff. (Chosen behaviour: stay up and recover automatically when the grant is fixed — no restart.)
2. Connecting to <url> … before it blocks, and a terminal auth failure now exits with a clean error: <reason> instead of Unhandled exception: + stack trace.
3. The Hub logs a refused registration on its own side (refused registration: … needs the "node" role), so journalctl shows it, not just the audit trail.
4. Capability probes get a timeout. CommandDetector/GPU detectors ran Process.run with no deadline, and the scanner awaits every probe before registration — so one wedged command (a hung nvidia-smi) would freeze the node silently. A timed-out probe is killed and treated as "not detected."

Before / after

# before: nothing, until Ctrl-C
$ omnyserver node start --hub wss://hub:8081 --id web-01 --principal p --token t
^C

# after
$ omnyserver node start --hub wss://hub:8081 --id web-01 --principal p --token t
Connecting to wss://hub:8081/node …
hub rejected the connection (forbidden): principal p may not register node web-01
  (the node.register action needs the "node" role)

Verification

• 285 tests (up from 273). The headline test/integration/node_registration_logging_test.dart reproduces the field failure exactly — a non-node-role grant logs may not register and keeps retrying — plus unit tests for the logger's formatting/dedup and the probe timeout (a sleep 30 killed at a 300ms deadline).
• Driven against the live Hub: a bad token → Connecting to… then error: Invalid token (no stack trace); an unreachable host → Node connection failed: … Connection refused instead of silence.
• dart analyze --fatal-infos --fatal-warnings, dart format, dependency_validator clean.

Follow-up (separate)

An omnyhub 1.7.0 change — map a terminal-coded NodeErrorMessage to a typed exception so registration fails fast instead of waiting out a 10s timeout — is planned next. This omnyserver release already delivers the full user-visible fix against omnyhub ^1.6.0.

🤖 Generated with Claude Code