Skip to content

TESTING_PROTOCOL.md

Latest commit

 

History

History
178 lines (150 loc) · 7.4 KB

TESTING_PROTOCOL.md

File metadata and controls

178 lines (150 loc) · 7.4 KB
Raw

MC-ELMA Rope Testing Protocol

Use this checklist before tagging or publishing a release.

Automated Local Checks

Run with JDK 21:

./gradlew build

The build runs verifyGameplayInvariants and verifyReleaseJar, which check:

  • no mixin configuration is introduced
  • teleport rope correction does not return
  • the mod remains server-safe/client-optional in metadata
  • critical gameplay defaults stay pinned
  • visual sync keeps client packets optional and avoids active-link work while idle
  • the release jar exists in fabric-mod-dev/release/
  • fabric.mod.json exists in the jar
  • mod id and version match Gradle metadata
  • Minecraft and Fabric API dependencies match the target versions
  • required server and optional client classes are present
  • the Fabric GameTest entrypoint and smoke test class are present

Run the headless Fabric GameTest suite with Java 21:

scripts/run-headless-gametests.sh

The script writes an XML report to build/gametest-results/mc_elma_rope-gametest.xml.

Dedicated Server Boot

  • Start a Fabric dedicated server for Minecraft 1.21.10.
  • Install Fabric API 0.138.4+1.21.10.
  • Install mc_elma_rope-0.4.0-beta.3.jar on the server.
  • Confirm startup creates or reads config/mc_elma_rope.json.
  • Confirm no client-only classloading crash occurs.

Latest local dev-runtime smoke evidence is recorded in SERVER_SMOKE_REPORT.md.

Fabric GameTest Smoke Coverage

The release jar declares com.mcelma.rope.test.McElmaRopeGameTests under the fabric-gametest entrypoint. These smoke tests cover:

  • default config values
  • default fence and chain anchor support
  • manager self-link rejection
  • anchor add, length clamping, and cleanup
  • persisted link restore duplicate protection
  • tied target disconnect penalty queue and reconnect effect application
  • controller disconnect cleanup without punishing the tied target
  • tied target disconnect lead refund to the survival-mode controller
  • anchored target disconnect persistence without lead refund
  • anchored target unavailable before disconnect callback persistence
  • anchored target reconnect restore and reconnect penalty application
  • broken anchor cleanup for offline anchored rope records
  • invalid offline anchor restore rejection
  • command-created ropes not queuing reconnect penalties by default
  • controller release and anchored release permission behavior
  • broken anchor block cleanup for matching anchored ropes
  • dead and spectator endpoint cleanup
  • one-way taut rope physics and loose rope no-op behavior
  • timed bind with lead consumption
  • timed controller release and slower third-party rescue
  • third-party rescue disabled by config
  • tied-player rescue denial
  • spawn-protected bind denial
  • timed anchor tying
  • default taut self-escape denial
  • config value sanitization and malformed config fallback
  • custom anchor id/tag reload
  • protected player name reload
  • protected player UUID reload
  • physics preset override behavior
  • holder damage type filter reload
  • persisted player-player rope save/load restore
  • persisted anchored rope save/load restore
  • pending restore behavior for missing player endpoints
  • max held duration expiration
  • active-link performance smoke scales: 0, 50, 100, 256
  • manager tick active-link invariant checks

Run these tests with Fabric's GameTest server tooling when the MC-ELMA test environment is available. They are smoke tests, not a replacement for the multiplayer gameplay checklist below.

Optional Client Renderer

  • Join once without the client mod and confirm gameplay still works.
  • Join once with the client mod and confirm visible rope rendering uses the default vanilla-like braided style, not the legacy three-line style.
  • Confirm visual packets do not block gameplay when the client mod is absent.

Gameplay Smoke Tests

  • Survival player binds another player with a lead after 3s.
  • Survival bind consumes one lead.
  • Controller releases after 3s with empty main hand and receives one lead.
  • Third-party rescue works after 12s when enabled.
  • Third-party rescue fails when enableThirdPartyRelease=false.
  • Tied players cannot bind another player.
  • Tied players cannot release another player's rope.
  • Anchor binding works on fences, fence gates, walls, chains, iron bars, lightning rods, end rods, and bells.
  • Anchor release returns one lead on manual release when eligible.
  • Breaking the exact anchor block releases the anchored rope and does not refund a lead.
  • Breaking a different anchor block does not release unrelated anchored ropes.
  • Anchor a tied player, disconnect that tied player, reconnect, and confirm the rope restores to the same anchor when the anchor block still exists.
  • Confirm anchored disconnect does not return a lead to the controller.

Self Escape

  • Tied player starts escape with empty main hand.
  • Escape progress pauses while the holder is inside selfEscapeGuardRadius when selfEscapeGuardProgressMultiplier=0.0.
  • Escape cancels while taut when selfEscapeCancelWhenTaut=true.
  • Escape progresses slowly while taut when selfEscapeCancelWhenTaut=false.
  • Escape cancels when the tied player moves too far.
  • Escape cancels when the tied player takes damage.
  • Escape completion respects selfEscapeSuccessDenominator.
  • Cooldown prevents immediate repeated attempts.

Holder Damage Drop

  • Holder damage drop respects holderDamageDropDenominator.
  • Denied damage type ids never trigger drop.
  • Non-empty allowed damage type ids restrict the drop roll to that list.
  • Anchor ropes do not drop from holder damage.

Lifecycle

  • Held player-player disconnect clears active ropes.
  • Held tied target disconnect returns one lead to the controller when refundLeadToControllerOnTargetDisconnect=true.
  • Anchored tied target disconnect stores offline anchor state when persistAnchoredRopesOnDisconnect=true.
  • Anchored tied target reconnect restores the anchor rope and applies the configured disconnect penalty.
  • If the tied player endpoint disappears before the disconnect callback, the anchor rope is still written to offline state.
  • If login temporarily places the player in another dimension, the offline anchor record remains pending and retries after the player reaches the anchor world.
  • An active anchored rope is captured during an orderly server shutdown and restores when the tied target reconnects after restart.
  • Breaking an anchor while the tied player is offline clears the offline anchored rope record.
  • Command-created ropes do not queue reconnect penalties when disconnectPenaltyOnlyLeadCreatedRopes=true.
  • Controller disconnect does not punish the tied target.
  • Tied target reconnect receives Mining Fatigue I and Slowness I for 2 minutes with default penalty settings.
  • enableDisconnectPenalty=false prevents reconnect penalty application.
  • persistDisconnectPenalties=true preserves pending penalties through server restart.
  • Death clears active ropes and does not refund leads.
  • Spectator mode clears active ropes and does not refund leads.
  • Dimension or portal mismatch clears active ropes and does not refund leads.
  • maxHeldDurationTicks clears old ropes when enabled.
  • persistRopes=false keeps restart behavior non-persistent.
  • persistRopes=true saves world-local state and restores when endpoint players rejoin.

Performance

  • Confirm idle server cost with 0 active ropes.
  • Confirm only active links tick with 50, 100, and 256 active ropes.
  • Confirm /mcelmarope list remains readable and bounded.