Fairy Fox

Plan — coverage maximization + full-layer test pass (2026-07-21)

Goal (owner mandate). Codecov ≥90% (aim near 100%); real tests on every feature at every testable layer — permissions, data backends, commands, management, config, high-load, performance; Scorecard as high as a single-collaborator repo can get.

Baseline (measured). Kover line coverage 44.2% (535/1211), branch 20.7%. Zero/near-zero classes: RemoveMaterials 0/61, DespawnProcess 0/54, DespawnBlockIntoAir$Companion 0/49, DespawnEffect 0/36, RecycleAction 0/19, DespawnItemIntoEntity 1/87, DespawnIntoCooker 1/61, DespiActions 4/127, DespawnIntoStorage 1/23, StorageFactory 3/41, CommandFeedback 1/9.

Phases

  1. Command + permission layer — MockBukkit Brigadier dispatch for every /despi subcommand + /recycle; permission matrix (none / despi.use / despi.elevated / recycle.use), player-vs-console gating, invalid-input paths; DespiActions direct tests; CommandFeedback incl. targetBlock failure path.
  2. Despawn pipeline — DespawnProcess (placement, partial, contraband, no-locations, unloaded world, exhaustion), DespawnEffect (loops, disable paths, force-destroy), all five into/* strategies against real MockBukkit worlds/blocks/entities, OnItemDespawnEvent incl. cancelled-event skip.
  3. Management — RemoveMaterials: owner-filtered/global walks, both-null error, empty-store, progress messages, per-sender exclusivity, force-destroy.
  4. Data backends — StorageFactory selection (yaml/aliases/sqlite/unknown), YAML→SQLite migration, sqlite pool behaviour; LocationManager debounced flush, reload/shutdown/saveNow paths. MySQL connect is N/A without a server (Testcontainers deferred — CI-only Docker dependency).
  5. Config + packaging — config defaults/invalid/reload; plugin.yml validity test (main class, api-version, permissions match code constants).
  6. High-load + performance — queue cap with both drop policies, per-tick budget, max-concurrent bound, flood (10k+) guardrails, existing store benches kept.
  7. Coverage gate — Kover verify rule: min 90% line coverage, gating check/build.
  8. Scorecard — repo-side re-audit: token permissions on all workflows, pinned actions, dangerous-workflow scan; note externals that stay owner-side.
  9. Release — green build → notes/changelog/VERSION → commit dev → PR to main.

Checklist enumeration (notes/plans/testing.md items this pass touches)

Each item's outcome is recorded individually in testing.md at completion (S8/S9 apply):

  • §1–7 boundary/equivalence/negative/parameterized expansion — commands + config
  • §3 parameterized: reward materials, command matrix, permission matrix
  • §5 duplication invariant (items in = out + consumed) — storage strategy tests
  • §8 lifecycle disable; PDC (/recycle progress); scheduler cancel-on-disable
  • §9–10 ItemDespawnEvent handling + ignoreCancelled
  • §11–12 plugin.yml validity; config invalid-value warnings
  • §13–15 migration (YAML→DB); pool behaviour
  • §23 command matrix; §59 permission matrix at dispatch level
  • §59–64 authorization, input validation (bad material names), DoS flood test
  • §65–73 performance guardrails (kept + extended)
  • §47–52 jar-content — deferred to a Gradle task follow-up if time runs out (disclose)
  • §87 regression pins — kept; no new bugs fixed this pass unless found

Not in scope (disclosed up front)

Real-server headless smoke in CI, Mineflayer in-game bot, Testcontainers MySQL, JMH, Pitest — all remain ⏳/⛔ in testing.md with reasons. External badge activations (Codecov app/token, Sonar, CodeFactor, SCORECARD_TOKEN PAT) remain owner-side.