# Class in-game macro test checklist **Unperformed plan, not gameplay evidence.** This copy is for **Rogue**. Every new case begins `NOT_RUN`. Reading this checklist, ticking a reminder or editing a record changes no canonical macro, build list or library status. The generated case list copies each current canonical ID, version, exact body, body SHA-256, requirements, targeting order, modifiers and known limitations. It also retains historical records separately. Command-derived prompts describe the written conditions; they are not a simulation or certification of the game parser. Source statements, provisional tuning and unresolved restrictions remain questions to verify. No observed client build, date or result is prefilled. Use the blank [class test record](templates/CLASS_TEST_RECORD.md) and the [validation matrix](VALIDATION_MATRIX.md). Paladin retains its separately authored [plan](PALADIN_CLIENT_TEST_PLAN.md) and per-macro cases; this extension does not replace those records or downloads. ## 1. Choose and freeze a small current batch - [ ] Back up existing macros and bindings. Check actual free slots; catalogue alternatives do not all have to fit one character at once. - [ ] Choose actions you actually learned and can exercise. Keep deprecated actions out of routine current acceptance; investigate acquisition separately. - [ ] Record each canonical ID/version/name/body/hash and read back the native `/macro` editor text. Document newline comparison, truncation or extra lines. A personal edit is a different test subject, not a pass for the original. ## 2. Capture the actual client and character - [ ] Record actual date/time/timezone, client version/build/interface, locale, installation/ruleset and evidence of beta, release or accepted-RC channel. Never copy a research export build into an unobserved test session. - [ ] Capture level, race/faction, actual legal talents/Legacy, learned names and ranks, weapon/shield, pet family/training where relevant, gear and reagents. - [ ] Record addons, action-bar owner, real binding/modifier conflicts, self/focus-cast overrides and button-down/up settings affecting the test. - [ ] If used, preserve Macro Lab's `/aslab scan` after `/reload` or normal exit. It describes environment/visible availability, not successful macro execution. Importer `/asm status` and an import receipt also do not establish a cast. ## 3. Prepare controlled recipients and ordinary spell controls - [ ] Arrange distinct mouseover A and current target B, including applicable friendly, hostile, dead, absent, pet, party or creature-type setups. Use a controlled location and cooperating helpers for control/threat/transfer cases. - [ ] Record starting combat, form/stance, stealth, aura, health/resource, cooldown/GCD, range, line of sight, facing, equipment and pet states. - [ ] Put the ordinary learned spell on a separate button. Compare it under the same conditions. A failed control identifies a setup/ability question, not a positive macro pass. Use single deliberate hardware presses; no automation. ## 4. Run the common baseline and literal targeting cases - [ ] Confirm exact text, tooltip/icon behavior and actual rank where observable. Observe an intended action with valid requirements and the correct recipient. - [ ] Exercise each written modifier chord, combined modifiers, bracket group and semicolon choice in order. Do not add a focus, self, form, target or readiness branch that the body does not contain. - [ ] Test relevant friendly/hostile/dead/missing and conflicting recipients. Record whether the selected target changes instead of assuming it does. - [ ] Distinguish an invalid condition from a failed cast: a dead mouseover can fail `nodead` and allow the next condition, but a matching living out-of-range, obstructed or ineligible mouseover does not become a retry on another unit. - [ ] Test relevant unavailable-spell, resource, cooldown/GCD, range, line-of-sight, equipment and eligibility cases. A failed selected action is not proof that a different action should be selected. Record the actual error and recipient. ## 5. Run command side effects and source-guided special cases - [ ] Observe conditional versus unconditional `/stopcasting` independently of the following cast. Cancellation may happen before an unavailable attempt. - [ ] Observe `/stopattack` and pet orders even when a later cast cannot work. Existing periodic damage, missiles, allies and pet behavior remain separate. - [ ] Test aura cancellation with the own aura present/absent and modifier guards held/released. Removing an aura does not prove training, restore cooldowns or remove another unit's aura or a separate restriction. - [ ] Exercise broad versus spell-specific channel guards with legitimate available channels. Player and pet channel states are different questions. - [ ] Apply the record's exact requirements and limitations: actual forms, stances, stealth, weapon/position, reactive trigger, rank, pet, party, reagent, creature, corpse and ground-placement conditions where applicable. - [ ] For melee/queued/auto attacks, compare native and exact-macro behavior from attacks stopped, after target changes and on success/failure. Missing an explicit attack-start command does not prove what a native ability does. - [ ] Verify class-specific questions appended below. Do not substitute Retail, Classic, another class or a similar spell's mechanics for a missing result. ## 6. Reset and observe each required case twice - [ ] Write the expectation before the press. Restore its stated conditions and obtain two consistent observations. This is this checklist's practical maintainer evidence rule, not a statistical guarantee or validator feature. - [ ] Link each observation to recipient/action/rank, errors, side effects and original evidence IDs/timecodes. A screenshot of the editor proves text only. - [ ] Use separate sessions for long cooldowns if needed, recording each actual build/setup. Missing prerequisites or an unavailable repeat are incomplete coverage; do not turn them into passes or N/A to finish the form. ## 7. Check persistence and save a private evidence bundle - [ ] After `/reload`, read back the exact text and repeat meaningful positive and negative cases. Record a normal-exit/restart readback if performed. Persistence or successful import alone does not demonstrate gameplay behavior. - [ ] Preserve originals privately, for example in ignored `research/captures/private/`. SavedVariables can identify players/realms; treat Lua as data, never execute it, and do not put captures in public output. - [ ] Prepare a reviewed redacted report with non-identifying evidence IDs, hashes and timecodes. Preserve originals for inspection; no public upload or automatic report submission is part of these tools. | Outcome | Meaning | | --- | --- | | `PASS` | An observed positive case matches its stated expectation. | | `EXPECTED_FAILURE` | A predeclared negative case fails as intended; it does not prove its positive branch works. | | `FAIL` | A recipient, action, side effect or other result differs. | | `BLOCKED` | A required spell, talent, pet, recipient, reagent, channel or other setup is unavailable. | | `NOT_RUN` | No observation was made. | | `NA` | Truly irrelevant to this exact body, with a written reason; missing prerequisites are blocked coverage. | ## 8. Obtain a separate maintainer decision before any status change - [ ] Review exact body/version/hash, actual build/channel/setup, every applicable claimed branch, negative and special case, reset repeats and evidence mapping. Resolve failures, explain exclusions and retain supported scope/limitations. - [ ] Keep `researched` while required coverage is partial, blocked, unknown or unresolved. Do not add a partially tested new build to whole-macro `tested_builds`. No generic `tested` status exists. - [ ] `beta-tested` requires completed applicable testing on an evidenced actual beta build. It cannot stand in for release testing. `release-ready` requires complete applicable retesting on release or an explicitly accepted release candidate plus maintainer/release gates; do not invent a beta history. - [ ] Historical records remain `deprecated` until a separate acquisition/design decision. A surviving alternative does not validate an unsupported whole key. - [ ] A later approved canonical edit uses existing fields, actual fully covered builds, scoped limitations, a reviewed report reference and appropriate version/changelog. Do not invent source IDs or evidence fields. A local/private capture path is not a public HTTPS source URL. - [ ] Run relevant release checks after a separately reviewed content change. Existing validators do not authenticate observations, repeat coverage or review decisions. Website/addon tests cannot substitute for client evidence. Changes to game build, spell/talent behavior, body, requirements, locale or bindings trigger scoped regression review. An old observation is not automatically current. **Every generated case remains NOT_RUN until a real session records observations; this tool itself never promotes status.** ## Rogue source-guided questions to verify - [ ] Exercise the literal stealth, nostealth and nocombat conditions, actual weapons, position, combo points and reactive requirements. Do not add an unstated Stealth requirement to Premeditation or assume a macro enters Stealth for an opener. - [ ] Compare the ordinary spell with each exact opener or melee key from attacks stopped, after target changes and on failures. Stopattack does not clear poisons, bleeds, projectiles or other damage that may break control. - [ ] Vanish's attack/cast cancellation happens separately from whether Vanish succeeds. Check actual reagent, incoming-effect and combat behavior; similarly, test Riposte's real activation instead of inferring it from the button. - [ ] Cold Blood, Mutilate, Blade Flurry, Adrenaline Rush and Shift-only Preparation are manual choices. Verify actual talent prerequisites, learned ranks and cooldown effects; no buff stacking, proc choice or reset list is calculated by the macro. - [ ] Venom is passive in the retained current source. Keep the historical Venom / Slice and Dice selector quarantined; direct Slice and Dice success does not restore its unsupported default cast. Current snapshot: 25 current records and 1 historical records. All session cases begin NOT_RUN.