Skip to content

0011-start-ci-quality-checks-as-advisory-reports.md

Latest commit

聽

History

History
61 lines (45 loc) 路 2.26 KB

0011-start-ci-quality-checks-as-advisory-reports.md

File metadata and controls

61 lines (45 loc) 路 2.26 KB
Raw

Start CI quality checks as advisory reports

  • Status: superseded by ADR-0012
  • Date: 2026-07-06
  • Deciders: Eric Bouchut

Context and Problem Statement

CI adds linting (#47) and coverage (#48) to a codebase that has never been linted and has no coverage baseline. Should these checks block merges from day one, and should coverage go to an external service? A blocking linter on an unlinted codebase fails every PR until a large style-only cleanup lands, which stalls feature work.

Decision Drivers

  • Do not block feature PRs on pre-existing style debt
  • Zero new config files and no external service accounts (solo project, certification deadline)
  • Keep a clear path to tighten later

Considered Options

  • Blocking checks from day one (Checkstyle check goal, coverage threshold via jacoco:check, Codecov)
  • Advisory reports first: Checkstyle report-only with the bundled google_checks.xml, JaCoCo report bound to the test phase, both published as workflow artifacts, no external service

Decision Outcome

Chosen: "Advisory reports first", because it makes quality visible immediately without gating merges on historical debt, adds zero config files (bundled Google ruleset, plugin versions pinned in pom.xml: Checkstyle 3.6.0, JaCoCo 0.8.15), and keeps secrets and accounts out of scope. The lint job also sets continue-on-error: true so even setup errors stay advisory.

Consequences

  • Good: Every PR publishes a Checkstyle XML report and a JaCoCo HTML/XML report as workflow artifacts.
  • Good: mvn test locally now produces target/site/jacoco/ with no extra flags.
  • Trade-off: Nothing enforces quality yet; violations can accumulate until the gate is switched on.
  • Follow-up: To tighten, switch the lint job to checkstyle:check with failOnViolation=true, drop continue-on-error, and add a jacoco:check minimum threshold.

Pros and Cons of the Options

Blocking from day one

  • 馃憤 Debt cannot grow
  • 馃憥 Every PR is red until a big-bang style cleanup; discourages small PRs
  • 馃憥 Codecov adds an account, a token, and an external dependency

Advisory reports first

  • 馃憤 Immediate visibility, zero friction, easy to tighten incrementally
  • 馃憥 Relies on discipline until the gate exists