How do I fail a PR build when someone introduces a bean cycle?

[ddsha441981]

We keep accumulating dependency cycles that nobody notices until refactoring time. Can WireDoctor gate this in CI?

[ddsha441981]

Yes — this is the architectural regression guard. Commit a baseline like a lockfile for your architecture, and the PR that introduces a new cycle fails its build the same day.

1. Capture the baseline once, commit it:

./mvnw spring-boot:run \
  -Dspring-boot.run.arguments="--wiredoctor.baseline=wiredoctor-baseline.json --wiredoctor.baseline-write=true"
git add wiredoctor-baseline.json && git commit -m "chore: WireDoctor baseline"

2. Arm the gate in CI:

# application-ci.properties
wiredoctor.baseline=wiredoctor-baseline.json
wiredoctor.fail-on=new-cycle

A tripped gate fails startup with a precise message (WireDoctorRegressionException), writes wiredoctor-gate.status (PASS/FAIL) and wiredoctor-diff.json for CI scripting — and the HTML report is still written even on failure, so you can see exactly which cycle was introduced.

Other available gates in the same fail-on list: condition-changed (Upgrade Guard), startup-time and slow-bean (performance gates, dual-threshold and noise-tolerant). Empty fail-on = report-only diff.

Cycle detection itself is Tarjan SCC on the resolved graph — and the report's lazySuggestions ranks which @Lazy breaks the most cycles with the smallest blast radius, so you get fix advice, not just a red build. Full workflow: CI gating guide.