Files
OEVG 403ba232e6 Release Puppeteer 1.2.1 with resurrection icon
Improve aura refresh, frame rendering, roster updates, heal prediction, and spell cost display. Add outline dragging, a resurrection indicator, and Lua 5.0 regression coverage.
2026-11-18 12:01:41 -04:00

14 KiB

Puppeteer 1.2.1 validation

Run from the repository root with Lua 5.0:

sh tests/run_all.sh /path/to/lua5.0

The suites load the real PTUnit, HealPredict, Pool and aura event-handler modules with mock WoW APIs. They extract the timer, aura renderer/comparator and spell-lookup functions from their source files to avoid constructing the whole game UI. They do not copy those implementations. parse.lua parses files without executing them. Tests are not listed in the addon TOC and do not run in game.

Outline dragging and cleanup checklist (pending)

Install Puppeteer-1.2.1-outline-fix.zip (corrects the reported upward drop offset); Puppeteer-1.2.1-before-outline-cleanup.zip restores stage 5 code. Version stays 1.2.1. The prior stage 4 SavedVariables backup guidance still applies.

  1. First drag a group a short distance vertically, then horizontally, then release without moving; repeat several times to check for accumulating offsets. Test Move All too. Unlock a group and drag its title: only the outline should move until release, then the real frames should land at the outlined position. Check normal scale and a changed UI scale, different layout directions/saved anchors, screen-edge placement and persistence after reload.
  2. Enable Move All and drag with several visible groups. Check relative spacing and exclude locked/hidden groups. Hold the configured alternate movement modifier to invert single/all behavior. Repeat with Move All disabled.
  3. Hide titles and use Shift+Ctrl+Alt + left-drag on a unit frame. Confirm it moves the group without casting. Confirm normal clicks/casts and the header right-click menu still work.
  4. During single and bulk dragging, hide a participating group or toggle Alt+Z. No outline or moving state should remain stuck. Restore the UI and drag again. Check raid/party transitions while dragging if practical.
  5. Repeat long drags with GentleGC reporting. Compare responsiveness/freezes with stage 5; mocks cannot establish native engine improvements.
  6. Switch targets repeatedly with timed auras, remove/re-add focus units, and verify other units' timers remain correct. Check incoming-heal text colors with direct heals and HoTs.
  7. If you use experimental enemy tracking, check enemies disappearing from slots after the midpoint of the list. If you use experimental cast icons, test success, interruption, overlapping casts and fade/disposal. These features remain experimental and are not enabled by this update.

All six regression suites and all-file parsing passed with real Lua 5.0 on 2026-11-13. Added tests use production drag code with mocked frames to verify single/bulk motion, mixed scales, anchors, locks/hidden exclusions, modifier inversion and hide cleanup. Cleanup tests exercise owned-versus-borrowed aura timing, trailing enemy slots and actual cast-icon callbacks/disposal/fade timing. Native mouse delivery, UI scale behavior and rendering require the checks above.

Stage 5 in-game checklist (pending)

Install Puppeteer-1.2.1-stage5.zip; Puppeteer-1.2.1-before-stage5.zip restores stage 4 code. Version remains 1.2.1. Stage 4 has not yet been explicitly reported tested; stage 3 remains the last confirmed working build. This ZIP includes stage 4, so retain the SavedVariables backup described below.

  1. With Use ClassicAPI enabled and ClassicAPI 1.20.2+, hover bindings for an explicit lower rank and the highest known rank. Compare displayed costs with the game spell tooltip and actual resource spending. Check Life Tap/free spells and a deliberately unknown rank/name.
  2. Apply/remove an available cost modifier while the binding tooltip is open. Verify cost text and casts-left change; also check a cost change that leaves casts-left equal. Check ordinary mana regeneration/spending and equipment/maximum-mana changes.
  3. If available, test rage/energy spells and druid form switches. ClassicAPI rage is already in displayed units. Per-second funnel spells deliberately retain the existing cost provider; this stage does not add a channel-duration affordability model.
  4. Disable Use ClassicAPI and reload. Verify the usual Nampower or tooltip-scan fallback; re-enable afterward. Older or unversioned ClassicAPI builds deliberately keep fallbacks. ClassicAPI-only testing, if practical, checks independence from Nampower/SuperWoW.
  5. Keep ShaguTweaks enabled for the normal-stack test. Check ordinary frame clicks, heals and aura displays; report full Lua errors and triggers. No settings reset is required.

Automated coverage loads actual provider, renderer and event code: version/capability gates, ranked/highest selection, slot-to-ID conversion, free/unknown distinction, Nampower/tooltip fallbacks, rage units, cached repeated reads, cost/resource/max-power display changes with unchanged casts-left, and event invalidation. All five suites and Lua 5.0 parsing pass. These mocks do not validate native engine costs or measure in-game CPU savings.

Implementation reference: ClassicAPI spell-cost source and its scalar spell-info API. The displayed-unit correction is included from release 1.20.2 onward.

Stage 4 in-game checklist (pending)

Install Puppeteer-1.2.1-stage4.zip; the tested stage 3 code rollback is Puppeteer-1.2.1-before-stage4.zip. Version stays 1.2.1. Before the first run, with the game closed, back up WTF/Account/<account>/SavedVariables/Puppeteer.lua if you want to restore pruned learned estimates too; reverting addon code alone does not restore those entries.

  1. With the usual mods and ShaguTweaks, check direct heals, interrupted/overlapping casts and repeated HoT ticks. Prediction values should remain plausible after reload; generic estimates remain available for pruned healers. No settings reset is needed.
  2. Hover an out-of-range frame, leave it, switch frames and toggle the arrow option. Check direction/display and repeat after hiding/showing groups and Alt+Z. With no hover, PTArrowUpdater:GetScript("OnUpdate") should be nil; hovering with the option and position provider enabled should restart it.
  3. With SuperWoW, add/remove focus units, remove the last one, and add one again. Check current health/auras and units disappearing/returning. PTCustomUnitUpdater:GetScript("OnUpdate") should be nil only when no custom units remain (other enabled custom-unit types count too).
  4. Repeat a basic smoke check with Use ClassicAPI disabled. If practical, test without SuperWoW: HealComm behavior should remain unchanged, and the optional heal-learning/custom-unit subsystem stays disabled.
  5. Compare idle CPU and a long party/raid session against stage 3. Record addon CPU/memory and GentleGC observations; automated checks do not establish an in-game speedup.

Stage 4 tests cover startup age/count pruning, legacy migration, realm separation, generic fallback, direct/equal HoT learning, modified-target exclusion, stale record rejection/reclamation, actual cast-event wakeup, arrow hover/option lifecycle, and custom-unit empty/add/remove/re-add behavior with older/current SuperWoW. All four suites and Lua 5.0 parsing passed on 2026-11-12.

Stage 3 in-game checklist (reported working by OEVG)

Version remains 1.2.1. Install Puppeteer-1.2.1-stage3.zip; Puppeteer-1.2.1-before-stage3.zip restores the stage 2 build reported working. No settings reset is needed.

  1. In combat with the usual mod stack and ShaguTweaks, check health/power responsiveness, overlapping/cancelled incoming heals, resurrection text, range/LOS fade and PvP state. Click bindings must act on the correct unit.
  2. Refresh several auras, add/remove others and switch targets quickly. Check icons, timers, stacks and enemy ordering for stale values.
  3. Join/leave parties and raids, move subgroups, assign roles, dismiss/resummon same-name pets, and check offline/reconnecting members. Frames and raid mana headers should stay current.
  4. Hide/show groups and toggle Alt+Z during healing/aura updates. Toggle heal predictions off/on, change profiles/hide options, and enter/leave /pt testui. Restored frames must catch up. Try /pt check for a forced refresh.
  5. Disable Use ClassicAPI and reload for a smoke check. If practical, also test without SuperWoW to exercise no-GUID roster refreshes. Restore the usual stack afterward.
  6. Compare the same scenario against stage 2 with /console scriptProfile 1 and reload, using addon CPU/memory totals, not GetFrameCPUUsage. Report full Lua errors and their triggers.

Automated stage 3 coverage uses the actual queue/event handlers and extracted frame/roster methods: coalescing, aura union/full priority, re-entry, stale-work cancellation, heal equality/direct totals, hidden/show catch-up, ClassicAPI/SuperWoW/no-GUID rosters, shared caches, subgroup/role changes and pet replacement. A 2,000-request mock burst produces 40 health and 40 power redraws across 40 units. These are behavior/call-count checks, not native client or performance tests.

Stage 2 in-game checklist (reported working by OEVG)

Version remains 1.2.1 by request. Use Puppeteer-1.2.1-stage2.zip; Puppeteer-1.2.1-before-stage2.zip is the tested stage 1 rollback. Stage 1 was reported working by OEVG on 2026-11-11.

  1. With the usual mod stack and Use ClassicAPI enabled, refresh a visible tracked buff (for example Renew from another player) and increase/decrease a stacking aura. Check timer text, sweep, stack count and player-cast borders. Other icons should stay in place.
  2. Apply and remove buffs/debuffs, and switch between targets. Confirm the complete icon list and hover tooltips still match the unit. Test duplicate copies of an aura from different casters if available; these deliberately retain full scans.
  3. As a warlock, refresh DoTs on an enemy so their remaining durations change order. Confirm the displayed debuffs reorder, with your effects retaining their existing ownership priority. Friendly buff priority order should stay unchanged.
  4. Hide/show frames and use Alt+Z while auras refresh or expire. Confirm current timers/stacks on return. Also check an out-of-view party member returning to view.
  5. Turn Use ClassicAPI off and reload. Verify aura icons, tooltips and SuperWoW/Nampower timers still work through their existing path. Re-enable it afterward. ClassicAPI-only testing, if practical, confirms native timers do not depend on SuperWoW.
  6. Check one ordinary party/raid session for Lua errors or stale auras. No settings reset is needed. Keep ShaguTweaks enabled to exercise the normal combined stack.

Automated stage 2 coverage includes changed-ID and reusable-buffer reads, no layout calls during refresh, unchanged-icon/sweep preservation, stack font thresholds, actual enemy comparator ordering, timing-loss fallback, missing/duplicate/replaced identities, unexpected metadata, full/add/remove event routing, partial-delta failure, hidden cache updates, native/legacy timers and unavailable/disabled ClassicAPI. A 40-unit mock burst makes exactly 40 instance reads and no full slot enumerations; this does not predict in-game CPU savings.

Automated coverage

  • Target/self changes in both directions; raid-slot reassignment; independent LOS updates; GUID caches preserve immutable self identity.
  • PvP nil, false, true and 1 results, plus instance/visibility changes, without repeated unchanged-state notifications.
  • Hidden and Alt+Z-style timer expiration, reshow before expiration, refresh/reuse after expiration, multiple timers and idle shutdown.
  • Exact ranks across spell tabs, highest rank, unranked and unknown spells, with no temporary lookup table.
  • Normal removal, overlapping healers, multi-target timeout, fresh casts surviving another caster's timeout, HoT overwrite grace, normal/timeout HoT removal, missing receiver buckets, empty target lists, callbacks starting new casts and 100 changing receivers.
  • Every repository Lua file parses under the selected interpreter.

Passed with real Lua 5.0 on 2026-11-11. These tests do not measure game performance or validate the native UI/client mods.

Stage 1 in-game checklist (reported working by OEVG)

  1. Install the supplied Puppeteer folder manually, keeping the previous build for rollback. Confirm version 1.2.1 and normal settings/frame positions after reload.
  2. On the usual mod stack, verify health/power, range fade, LOS, click bindings, aura stacks, timer sweeps and flashing countdown text. Repeat a basic smoke check with Use ClassicAPI off and reload.
  3. With ClassicAPI but without SuperWoW, target yourself and reload. Select a distant party member: range/LOS should update. Target yourself again. If practical, repeat after changing your raid subgroup/slot. Turning off Use ClassicAPI alone does not disable SuperWoW, so it does not test this cache path.
  4. Hide a group with timed auras, let them expire, then show the group. Repeat with Alt+Z, including restoring the UI before an aura expires. Text should be current, expired timers absent, and newly applied/refreshed timers normal. With ClassicAPI, /run DEFAULT_CHAT_FRAME:AddMessage(tostring(PTAuraTimerDriver:GetScript("OnUpdate"))) should print nil after all registered timers have expired (assuming no other aura timer is still running).
  5. Toggle PvP and verify the intended icon/state, including entering an instance. No visible changes should occur while the state is unchanged.
  6. With a healer, check successful and interrupted heals, two healers on one target, group heals and HoTs. Cancelling/completing one heal must preserve the other healer's prediction. Timeout and missing-event cases are covered by automated mocks; they need not be deliberately induced in game.
  7. If testing without Nampower 3, verify ranked and highest-rank spell bindings still display expected tooltip costs and cast normally. Spell-cost calculation itself has not changed.

Keep ShaguTweaks enabled for the usual-stack smoke test. Puppeteer's noCooldownCount compatibility flag remains intact. ShaguTweaks' optional Hide Combat Tooltip can still hide the shared tooltip in combat even if Puppeteer is set to Always; that settings interaction is unchanged.

Report Lua errors with the full message and what triggered them. No measured speedup is claimed for this correctness release.