On a PvE realm, zone names were still coloured friendly, hostile or contested.
In 1.12.1, `GetZonePVPInfo` (`0x48d540`) answers a type only when its display gate is open (`0x48d5c2`/`0x48d5cc`). The gate opens on the realm's PvP flag or in a zone with Flags `0x10`, which is the six capitals. The realm flag is read from `Cfg_Configs.dbc` for the selected realm's type and stored once per world entry (`0x4015de`). A type with no row skips the store. Otherwise the type is nil, so the stock zone splash shows cream and the minimap zone text gold.
benilla now reads `Cfg_Configs.dbc` instead of a transcription that stopped at type 8, so realm type 10 lists as PvP. It latches the realm flag at world entry and applies the gate.
Tests cover the gate, the six capitals and every shipped Cfg_Configs row; each failed before the fix.
Fixes#816
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- spell text: $g/$G take the active player's gender and the chosen form is space-trimmed, so a female character reads "her" (0x508180, 0x508214)
- spell text: trim the $g docs, drop the colon-less $g pin
- spell tooltips: $g reads the view's player store; nets cover gender
- item tooltips: re-substitute $-texts when the gender byte moves
- tests: share one real-spells loader beside Spells
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
Two ways the emote-state base pick differed from 1.12.1.
- A blacksmith kept hammering while you used his vendor window. The resolver `0x5fd770` passes over a state whose `Emotes.dbc` flags have `0x2000` (work, mining, ready, dance) while the unit is the interact NPC (`0x5fd7e2`-`0x5fd7fd`). `SetInteractNPC` re-picks on open (`0x493198`) and on close (`0x493219`, before the name clears at `0x49334b`), never for a player. So the NPC stands, and the loop resumes at its next re-pick.
- A set state outranks the idle fallbacks: `0x5fd770` runs before `0x5fd830`, and a row with AnimID 0 claims Stand. After `/stand` (state 26), a swimmer at rest shows Stand until it moves.
Tests cover the open, the close hold, players, a pose under a one-shot, and the precedence order.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Three ways `DoEmote` and the chat talk gesture differed from 1.12.1:
- `/wave Bob` ignored the name and targeted the selection. 1.12.1 resolves a name that does not start with `%` as a player, at any range (`0x49fdb1`-`0x49fdf0`, `0x493aa0`), and sends the emote untargeted on a miss. `DoEmote(token, name)` from Lua does the same.
- A polymorphed, stunned or sleeping speaker still played the talk gesture. `UNIT_FIELD_FLAGS 0x20000000` suppresses it when the sender's name is cached (`0x49d7f7`).
- `DoEmote` sent while our unit was possessed (`0x5ef57e`). A chat-only emote now reads `Emotes.dbc` row 0's flags, so it is refused while asleep or dead (`0x5ef5b1`, `0x47db8e`).
Tests cover the name, miss, `%t` and self cases, the gesture gate, and both `DoEmote` gates against the install's row 0.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Every text emote's voice played for every performer, so every `/cheer` in a city was heard. 1.12.1's composer tail (`0x623c80`) plays it only for you or a member of the four party slots (`0x623cc2`-`0x623cce`, `0x5efea0`), and drops an ignored sender before that (`0x49dc36`).
What changed:
- The voice plays at attachment `0x11` and stops the unit's previous emote voice once a new one opens (`0x623c10`).
- Race and sex come from the display override before the descriptor (`0x60c6a0`/`0x60c6c0`).
- `SMSG_EMOTE` no longer plays `EventSoundID`. The one-shot path never reads it (`0x5e66b0` → `0x5fcd20`); only the state emote's `$ESD` does.
Tests cover each relation class, an ignored performer and the per-unit voice slot.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A crowd of arriving players was composited in one frame, on the main thread, under the archive lock: 40 geared looks took 76–115 ms.
In 1.12.1 an unforced full composite never waits on its section loads (`0x44b430`), and a unit whose first composite is pending is not drawn: ShouldRender `0x607da0` feeds the draw flag `0x710c50`, and the flag goes to the mount model when mounted. It appears dressed once the composite is done.
Now an arriving body composites on the async pool, one task per look, reading outside the lock. The body, and its mount, are built once the atlas lands. Arriving now costs the main thread 0.26–0.45 ms for 40 looks. The appear fade runs from arrival, as `FadeTo` is armed at create.
A re-dress, a rig heal and the glue and dressing-room previews stay in the frame. Baked NPC skins are unchanged.
Fixes#609🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A ghost heard no ambience. The ghost bed was looked up as a SoundEntries row named "Ghost", which does not exist, so it fell back to kit 0. The underwater bed was a hard-coded row.
At init, 1.12.1's `0x4609b0` walks SoundEntries from the last row down. It keeps the rows whose names match `"Ghost (DONOTRENAME)"` (`0x836444`) and `"Underwater (DONOTRENAME)"` (`0x836428`) exactly and case-sensitively (`0x64a480`, a `strncmp` through the NUL). In the install those are rows 4160 (`GhostState.wav`) and 4209.
The catalog now resolves both beds once, the same way, and the ambience selector takes them from it, with order and fades unchanged.
Tests cover the install's two rows, the exact-name match, the walk order, a ghost selecting its bed over water, and an unresolved bed staying silent.
Fixes#721🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
An ambience change read and decoded the whole bed on the main thread, about 40 ms a swap. 1.12.1 opens the bed as a looping FMOD stream that reads the archive through its file callbacks (`0x460b54` → `0x7a54d0`, `FSOUND_Stream_Open` at `0x7a5525`).
Now:
- MPQ files open for in-place reads, one sector at a time.
- The bed streams from the archive, opened on the IO pool. A newer swap drops an open still in flight.
- kira moves to 0.12.4 on symphonia 0.6. 0.5 sized PCM frames from `nBlockAlign`, which 98 stereo WAVs store as 2, so a streamed bed fell silent halfway.
- A fork patch stops a stream at its last packet instead of spinning.
- A sector that inflates short is refused.
Tests cover a bed's true length, a seamless loop, an off-thread open, the stop at the last packet and the refusal.
Fixes#700Fixes#8🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Spell descriptions printed `$b`, `$f`, `$F`, `$i`, `$q`, `$u` and `$v` raw. The reference's expander resolves each from Spell.dbc:
| token | column |
|---|---|
| `$b` | EffectPointsPerComboPoint |
| `$f`/`$F` | DmgMultiplier |
| `$i` | MaxAffectedTargets |
| `$q` | EffectMiscValue |
| `$u` | StackAmount |
| `$v` | MaxTargetLevel |
`$b` chops toward zero. `$f` prints one decimal, and `$F` rounds half away from zero, both after the scale and in double precision. The integer arms print a signed `%d`.
This change reads the new columns and adds the arms. `$f` and `$F` keep keying the `$l` plural, under the recorded deviation. Tests pin each arm, the schema's float columns and every shipped use of the six letters. Chain Heal now reads "Each jump is 50% as effective".
Fixes#593
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
A warrior chasing a running mob was refused Hamstring at 6-7 yd with "Out of range.", where 1.12.1 lets it go out.
After both arms, `GetMinMaxRange 0x6e3480` adds 2.6667 yd (`[0x811914]`) to the max when all of these hold (`0x6e3648`-`0x6e36a2`):
- the caster and the target unit both move, with a bit of `0x200d` set;
- both run faster than twice their walk speed (`0x5fc350`);
- the row is melee, or the target is a player.
On the melee row, the auto-attack target stands in when no unit is passed.
`min_max_range` now takes each unit's motion and adds the bonus. Every caller passes what its reference call site passes. The ambiguity ceiling rises by the 11 pairs this adds, with the reason beside it.
Test: Hamstring goes out at 7.5 yd with both units running and is refused standing still.
Fixes#678
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
`UnitCreatureType` read only a creature's template type, and fell back to "Humanoid" for a player. The form stage was not built. So a druid in cat, bear, travel or aquatic form and a shaman in Ghost Wolf all answered "Humanoid", where 1.12.1 answers "Beast".
`UnitCreatureType 0x51a280` names the type the resolver `0x605570` returns, and 0 names nil (`0x51a2c3`). The resolver tries three stages in order:
1. **Form:** the form byte (`UNIT_FIELD_BYTES_1` byte 2) indexes `SpellShapeshiftForm.dbc` column 12, which applies only above 0 (`0x60559a`).
2. **Template:** else the cached creature record (`0x6055a8`).
3. **Race:** else the race's `ChrRaces.dbc` type (`0x6055ca`).
A token naming no unit answers nil (`0x51a2b8`). The unit tooltip's level line takes the type only for a non-player (`0x52a4a6`).
Changes:
- `ui_unit::snapshot` sets `creature_type_name` for every unit through the shared `CreatureTypeSources::of`: players, creatures, pets and group tokens alike. `enrich_unit` no longer sets it.
- `UnitCreatureType` reads that field, with no fallback.
- The feeds at Bevy's parameter limit carry the sources in a `SnapshotTables` param.
- A player's tooltip level line is unchanged, since it reads race and class.
Behaviour that now matches the reference beyond the issue:
- a held pet, `partypetN` and `raidpetN` answer their template's type rather than nil;
- a party member in a form reads the form's type;
- a roster member with no object, and the pre-arrival player seat, answer nil.
Tests:
- `the_shipped_forms_name_a_beast_or_fall_through_to_the_race` reads the real `SpellShapeshiftForm.dbc`.
- `unit_creature_type_names_the_form_before_the_race_and_a_creature_its_template` runs `feed_units` and asks the Lua binding.
- Mouseover, party, raid and pet feeders each have a test.
- `a_players_level_line_ignores_the_snapshots_creature_type`.
Fixes#641🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
`min_max_range` took one `target_reach` that served two roles: the melee arm's fallback to the auto-attack target and the ranged arm's explicit target. The ranged arm padded whenever it was `Some`. The tooltip passed the auto-attack target's reach, so while you swung, every ranged spell's tooltip read its row plus both reaches: Fireball read "38 yd range" where 1.12.1 reads 35.
`GetMinMaxRange 0x6e3480` counts its target (`[ebx+0x10]`) only when it is a unit (`0x6e34e1`-`0x6e34f9`).
- **Melee arm** (`0x6e3552`-`0x6e3594`): with no unit target, it falls back to the caster's auto-attack target (`0x6e356a`), then to the caster's own reach.
- **Ranged arm:** pads only for a non-null target (`0x6e35d8`, `0x6e35ec`). The max always takes the pad, and the min only when nonzero.
What each caller passes:
| Caller | Call site | Target |
|---|---|---|
| Tooltip | `0x52e9c2` | none |
| `ActionHasRange` | `0x4e5852` | none |
| The cast check and `IsActionInRange` | through `CanTargetUnit 0x6e4440` → `0x6e47b0` | the bound or selected target |
Changes:
- `min_max_range` takes `RangeTargets { target, attack_target }`.
- The melee arm reads the target, else the attack target, else the caster. The ranged arm pads for the target alone.
- Each caller passes what its reference call site passes. The action button's two verbs are now two calls: `has_range` with no target and `in_range` with the selected one.
- `min_max_range_reads_reach` narrows to melee rows. The tooltip safety net then counts 813 reach-dependent views instead of 10,298, so starting to swing no longer rebuilds every ranged tooltip.
Tests:
- `ranges.rs`: `the_ranged_arm_pads_only_for_the_explicit_target`, `the_melee_arm_falls_back_from_the_target_to_the_attack_target_to_the_caster`, and the narrowed exactness test.
- `range_cell_on_real_data` reads Fireball as "35 yd range" and Charge as "8-25 yd range" with an auto-attack target present.
- `cast_range_refusal_follows_the_two_compares` covers a bound target's pad.
- `a_ranged_button_pads_its_range_by_the_selected_target_alone`.
Fixes#643🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Every spell-tooltip invalidation rebuilt every view the feed had ever pushed. The remembered set only grows: the book, every talent rank and trainer service, and every aura ever seen. So in a long session each of these frame-hitched for 1.5 to 4 ms:
- a crit or dodge change;
- a worn-slot change;
- a target swap while auto-attacking;
- a spell-modifier packet;
- a reagent cast.
The reference builds a tooltip at the hover (`0x52e610`) and keeps nothing. benilla pushes views ahead of the hover, so it has to know what each view was built from.
Each pushed view now records, as it is built, the inputs it read: the `$z` home area, the form line, the worn slots the equipped-item search reads, and so on. Each record sits at the line that reads the input.
The inputs:
- the `$z` home area;
- the form line (a spell with `Stances`);
- the worn slots the equipped-item search reads, each slot's item, template and broken state, and the disarm flag;
- the chance line's percentage (`Effect[0]` 20, 22, 23, 78);
- the combat reach the range cell reads;
- the skill lines the `$`-tokens scale by, the spell's own and those of each spell it cross-references;
- the modifier family-flag bits of the spell and of those it cross-references;
- the reagents it lists;
- the player's level, base mana, max health and max power where the cost or cast cell reads them;
- the ranged attack time the cooldown cell pads by.
The feed diffs those inputs frame to frame and rebuilds only the views whose recorded reads intersect what moved. A class-family change still rebuilds everything, and the pet's own inputs still rebuild the pet views.
The feed moves out of `ui_tooltip/mod.rs` into `spell_feed.rs` (the builder and the system) and `spell_deps.rs` (the records and the diff). `TokenContext::home_area` becomes a closure, so a `$z` read is recorded exactly.
Measured in release on the real data with 2,000 remembered ids, per forced change:
- crit or dodge: 3.3 ms → 26-38 µs;
- one worn slot: 3.5 ms → ~100 µs;
- one modifier cell: 3.5 ms → ~60 µs;
- all 64 modifier bits: 3.5 ms → ~0.6 ms;
- reach: 3.3 ms → 1.3 ms, still high because the range cell also pads ranged rows by reach (#643);
- a quiet frame: unchanged.
The bench is `ui_tooltip/bench.rs`, `#[ignore]`d.
Tests:
- `deps_tests.rs`, the safety net, builds all 22,357 real views. For each input it moves the input alone and fails on any view that changed without having recorded it, naming the spell and the input.
- It also moves everything each view did not record at once, from both sides.
- It pokes every player-store field (and every pet-store field for the pet views) to discover reads mechanically.
- It enumerates the equipment slot states the item search distinguishes.
- `feed_tests.rs` drives `feed_spell_tooltips` itself. A dodge change re-pushes only the dodge chance-line view, and a modifier cell only its family's views.
Fixes#535🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The spell target binder went straight to the relation checks and tested none of `BindTarget`'s unit gates. So:
- a heal with a dead friendly player selected was sent at the corpse and refused by the server;
- Hibernate, Banish or Turn Undead at the wrong creature type was sent and refused, instead of refusing locally with "Invalid target";
- a self-excluding spell with yourself selected bound you.
`BindTarget 0x6e5b40` refuses a unit silently, before any relation arm, when any of these holds:
- the unit is the caster and the spell has `AttributesEx & 0x80000` (`6e5bd8`-`6e5bf7`);
- the unit has `UNIT_FIELD_FLAGS & 0x10000` (`6e5c1d`);
- the spell's `TargetCreatureType` mask excludes the unit's type (`6e5c26`-`6e5c53`). A zero mask admits every unit. Otherwise a unit of type 0 is refused and a typed one needs bit `type - 1`;
- the unit is dead (signed `UNIT_FIELD_HEALTH` not above 0), unless `UNIT_DYNAMIC_FLAGS & 0x20`, a word in `0x8600` or `AttributesEx2 & 1` admits it (`6e5c5f`-`6e5c8b`);
- the unit is alive and the word has `0x400` (`6e5ca2`).
A refused selection falls to the player behind `autoSelfCast`, then to the enemy-word refusal or the cursor. `SpellCanTargetUnit 0x6e6460` runs the same five gates (`6e6507`-`6e65ab`).
- `spell/bind_gates.rs` holds the five gates. `cast_target::unit_binds` is the one predicate for the selection, the `autoSelfCast` fallback, the cursor's click, `SpellTargetUnit` and the hover verdict. The cursor's separate `excludes_caster` checks fold into it.
- The dead test now matches the reference, which catches a corpse whose create block sent no health. The `0x400` bit is now cleared by a dead unit's bind (`6e5d89`, `6e5ddb`), where before any non-self candidate cleared it.
- `creature_type.rs` is the one `0x605570`, reading these in order:
- the form's `SpellShapeshiftForm` creature type;
- the creature template's type, by `OBJECT_FIELD_ENTRY`;
- `ChrRaces`'s type.
The minimap's tracking dots used their own copy, keyed by the guid's entry bits and missing the fall-through past the form. They now call the shared resolver.
- `SpellDisplay` reads `TargetCreatureType` (column 14).
Tests:
- `bind_gates.rs` has a truth table per gate. Its `ladder` module drives `resolve_cast_target` through:
- a dead heal target, with `autoSelfCast` on and off;
- Hibernate at the wrong type;
- a self-excluding spell;
- the `0x10000` flag;
- the `0x402` word.
- `the_press_the_click_and_the_hover_bind_the_same_units` runs one table through all three paths.
- `real_spell_catalog_reads_the_bind_gate_columns` pins the real rows for Hibernate, Banish, Turn Undead, Mind Control, Bloodlust and Revive.
Fixes#449🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A pet-bar spell press armed no global cooldown on the pet's list. Claw or Firebolt sent `CMSG_PET_ACTION`, and no slot on the pet bar swept, so the bar looked ready while the pet could not cast. Only a spell with its own recovery timer, such as Bite, showed a sweep.
In the reference, the pet bar's spell arm (`0x4bd240` in `0x4bd1d0`) first handles two cases that skip the cooldown:
- A spell that is a live aura on the pet sends the cancel and returns.
- A possessed unit, or a spell with `AttributesEx4 & 0x20` (`0x4bd355`, `0x4bd35e`), takes the generic cast entry `0x6e4b60`.
Every other press calls `StartGlobalCooldown 0x6e2de0(spellId, 1)` (`0x4bd367`-`0x4bd36e`) before the `0x175` send. That call applies the spell's own `StartRecoveryCategory`/`StartRecoveryTime` pair, under spell-mod op 21, into the pet's list (`0xcecaec + 0x18`). This press is the only place the pet's GCD starts, because the `SMSG_SPELL_GO` pet leg inserts none (`0x6e85f7`).
Changes:
- `SpellDisplay` reads `AttributesEx4` (column 10).
- `PetPress::press` arms the pet list's GCD through the existing `Cooldowns::start_gcd`, with the player's modifier tables. It skips the arm for the cases above and for an unresolved record or pet object (`0x4bd2fe`, `0x4bd34f`).
- The feed's cooldown diff then fires `PET_BAR_UPDATE_COOLDOWN`.
In the shipped `Spell.dbc` the pair is 133/1500 for most pet spells (later Claw ranks, Bite, Lash of Pain, Torment, Suffering) and 133/1000 for Firebolt and Claw rank 1. A real-data test pins both values.
Tests:
- `ui_pet/press_tests.rs` presses through `PetPress::press_slot` and covers:
- the GCD across the slots sharing the category, read back through `GetPetActionCooldown`;
- no pair;
- possessed;
- `AttributesEx4 & 0x20`;
- the aura cancel;
- no pet object;
- op 21.
- `real_pet_spells_carry_the_gcd_pair_and_ex4_reads_column_10` reads the real `Spell.dbc`.
Fixes#589🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
`CharCreateCatalog::load`'s misparse guard demanded each race's `CharBaseInfo` classes equal the shipped set exactly. A patch that adds a race/class row therefore failed the load, and without the catalog the create screen is disabled and the glue preview stands no body. 1.12.1's class-list builder (`0x4706b0`) lists every `CharBaseInfo` row for the selected race, skipping only Dwarf+Mage (`0x4706e3`-`0x4706ec`), so the patched class appears there.
The guard, now `shipped_combos_present`, requires every shipped pair to be present and lets extra pairs through, so a misparse that loses a row still refuses. `a_patched_combo_passes_the_guard_and_a_lost_one_does_not` fails under the old exact-match rule. Built and gated on macOS.
Fixes#617🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- fix: apply per-level scaling to spell descriptions
- The description's per-level terms scale by the player's skill in the spell's line, not the character level
- The skill value reads its bonus word as the reference does: the high half unsigned over a positive value, the low half signed
- The $d and $o durations take the row's per-level term at the spell's own skill level
- $d reads a duration at or below 0, or no duration row, as until cancelled
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
The reference's $a, $d, $t, $e, $c, $p, $f, $F and $z arms never write the plural count
[0xbe0b84], so 1.12.1's Blizzard and Consecration read "every 1 seconds". benilla keys on the
nearest token and prints the grammar the text means; a test pins it.
Built and tested on macOS.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- fix: spell modifiers
- fix: round scaled spell tooltip values to client tenths
- fix: refresh only modifier-dependent item tooltips
- The pet leg of the SPELL_GO cooldown insert takes op 11, and the item leg takes none
- The tooltip's cast cell is GetCastTime's unclamped result
- The Improved Fire Blast tooltip test reads Fire Blast's real recovery columns
- The effect-points aura ops follow the reference's jump table
- The template filler rounds %f and %g ties half up, as the client's CRT does
- The $s, $S, $m, $M and $o tokens compute and print their values as the reference does
- The $t, $r, $e and $a tokens follow the expander's arms
- The unclamped cast-cell test skips without an install, as its GlobalStrings come from one
- start_gcd's doc names the reference's gate: category or time nonzero, then op 21, and nothing armed for a zero result
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
- Sinister Strike, Heroic Strike, Judgement and the other combat-starting spells pressed with no target select the nearest enemy, as the Attack button does
- A finisher pressed with no combo points refuses before the attack pick, and a self-cast press always Tabs past the caster
- The plain cast send skips the attack pick because no 1.12 item, form, recipe or chained spell carries predicate 0x6e5200, and a real-catalog test pins it
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
The fourteen 1.12.1 functions #336 lists are defined, each derived from its reference address, and benilla's layer drops the two places it stood in for the stock bag code.
- **Keyring:** `KeyRingButtonIDToInvSlotID` (`0x4c8150`) exists, and the keyring slots answer as inventory slots 82-113. The stock `ContainerFrame.lua:616-619` hover path now runs unchanged, and the layer's keyring `OnEnter` wrapper is deleted.
- **Bags:** the layer no longer replaces `CloseAllBags`. The stock one (`ContainerFrame.lua:702-707`) closes the backpack only when it was not open before the vendor or mailbox opened it (#541). The deviation that opens every bag stays, and its comment now names its only callers.
- **Inventory:** `ShowInventorySellCursor`, `SetInventoryPortaitTexture` and `UpdateInventoryAlertStatus` exist.
- **Character and pet:** `GetDamageBonusStat` (ChrClasses field 2), `GetPetTimeRemaining` (armed from `SMSG_PET_SPELLS`) and `GetActionAutocast` exist.
- **Mail:** `ClearSendMail`, `GetNumPackages`, `GetPackageInfo` and `SelectPackage` exist, over a new `Package.dbc` loader. A selected package rides in the mail send. The compose reset now also clears the stationery, the package and the attachment, as `0x4acdc0` does.
- **Tabard:** `GetTabardInfo` opens the tabard designer through the registrar (`0x5e00e0`).
- **World map:** `GetWorldLocMapPosition` projects a map position onto the displayed map (`0x4a7360`).
- **Loot:** `SetLootPortrait` binds a unit loot source's portrait (`0x4c2b40`).
Verified:
- A test per function for its shapes, raises and effects, failing on main.
- A keyring hover test through the stock path, failing on main.
- A backpack-over-vendor test, failing on main.
- `scripts/gates.sh` and `scripts/smoke.sh` are green.
- A live A/B run: hovering a key asked for `player:82` and showed the key's tooltip. After a vendor, opening the Help panel left the backpack open and closed bag 1; on main the backpack closed.
Built on macOS; no platform seam touched.
Fixes#336Fixes#541🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Lua VM keeps no surface the 1.12.1 client lacks.
- **Removed:** the nine Era-only globals: `UnitAura`, `CancelUnitBuff`, `GetCursorInfo`, `GetInventoryItemID`, `GetPlayerFacing`, `GetTradePartnerName`, `IsGossipOptionCoded`, `UnitIsAFK`, `UnitIsDND`.
- **Hooks off Lua:** `GameTooltip:BenillaSetItemById` and `:BenillaGetTooltipOwner` are replaced by Rust-side calls. The uncalled `Benilla*` hooks are deleted.
- **Frame strata:** `SetFrameStrata` knows the reference's eight strata and raises its message for any other (`0x774450`).
- **`SetDrawLayer`** reads the layer name alone (`0x79a780`).
- **`Set<State>Texture`** reads a number as a path (`0x781b23`); there is no colour form.
- **`PlaySound`** names a kit and returns nothing (`0x4586d0`). `PlaySoundFile` returns one value (`0x458780`); benilla answers 1, a recorded deviation, since it plays after the call.
- **`OnClick`** sets only `arg1` (`0x779540`).
- **Quest getters:** `GetQuestItemInfo` and the quest-log reward and choice getters answer the reference's five values and its miss behaviour (`0x501f80`, `0x4e0b00`, `0x4e0cf0`).
- **`PickupContainerItem`** returns nothing (`0x4f9b30`).
- **Surface gates:** the bare VM's globals and every widget class's methods are held to `reference/1.12-globals.tsv` with one commented allowlist. A new gate on the full production load (core plus layer) allows beyond that only the layer's `Benilla`-prefixed globals.
Verified:
- A test per removal and per corrected form; a scripted undo of each fails its test.
- Each gate fails on a re-added name.
- `scripts/gates.sh` and `scripts/smoke.sh` are green.
- A live run showed the same warnings and 0 script errors before and after.
- In the 219-addon corpus, AckisRecipeList's bundled AceGUI-3.0 and Roid-Macros' Grid extension now behave as they do on 1.12.1.
Built on macOS; no platform seam touched.
Fixes#171🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A dev build finds its project folder, where its `WoW` link, `benilla-config/` and `.probe-identity` live, from its launcher, not from a benilla crate's own `CARGO_MANIFEST_DIR`. For a crate built on top of benilla, where benilla is a git dependency, that manifest dir is cargo's checkout of benilla.
- `benilla_buildstamp::emit()` stamps `BENILLA_PROJECT_DIR`, the launcher's workspace root (the nearest folder up holding `Cargo.lock`), only when the launcher's `dev` feature is on. A player build stamps it empty and carries no source path. The shims pass it in as `BuildId::project_dir`.
- `benilla_formats::project_folder` is the one home for the folder. Each entry point records the stamped folder once, before anything resolves from it. The install's dev rung, the state folder (which keeps its step from a linked worktree to the primary checkout) and the probe identity all read it. With nothing recorded (a test, an example, a tool) it is benilla's checkout, as before, and in a player build it is `None`.
- Live `assets/ui` and the addon corpus stay on benilla's own source.
- The example launcher's header now says a dev build of a crate on top keeps these in its own folder.
- The comment above the stamp's resource names the build banner, the stamp's reader beside the panel footer.
Fixes#494Fixes#493🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- fix(characters): preserve unmanaged geosets above 1700
- The geoset selection's comments state the 1.12 fact
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
With realms in more than one category, the realm list shows the stock category tabs, so every realm can be reached. Before, it listed only the first category's realms and had no way to the rest.
- **The Region** is `Wow.ini`'s `[WoW Config] Region`, read through the archive chain as `0x40249d` loads it (the stock install ships it in `interface.MPQ`), with the section and key compared case-insensitively (`0x445356`, `0x4453cb`) and the value parsed as `0x445470` does. It is 0 when the file or the key is missing.
- **The categories** are the `Cfg_Categories.dbc` rows whose Region column equals it, in file order (`0x46e300`): id from column 0, name from the row.
- **Placement** (`0x46e510`): a realm's category byte 0 takes the first category's id (`0x46e5b8`), then joins every category with that id (`0x46e60b`). A byte matching no category puts the realm on no tab, as in 1.12.1. Tab ordinals count the non-empty categories only (`0x46edf0`), and the load-band statistics still run over every realm.
- **The tabs** are built from `RealmList.xml`: hidden for one category, `RealmListTab1`..`N` with the category names for two to eight, the front one raised and disabled. A click fronts the tab and re-lists (`RealmListTab_OnClick`), the highlight falls back to the current realm or none, and the scroll clamps to the new tab. Change Realm and `OnShow` keep the stored category (`0x46f188`, `0x46f400`).
- **No Region rows**: every realm lists on one untabbed page. The reference reads a NULL category there (`0x46f26a`, `0x46e5b8`), and benilla does not copy the fault.
- **Logging in** by name, the remembered or current realm, still searches every realm, as the reference resolves by name.
- A dev capture, `WOW_CAPTURE=glue-realmlist`, shows two tabs over five invented realms.
Checked: tests for the Region read, the category rows for a US and an EU Region from the install's DBC, placement (byte 0 folded, an unmatched byte on no tab), the strip hidden for one category and named for two, a tab click through the real click path, the highlight and scroll on a switch, the stored category across Change Realm, the statistics over every realm, and the no-rows fallback. Each fails with its piece undone. Gates and the smoke run green on macOS; no platform seam touched.
Fixes#362🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
With Resurrection, Redemption or Rebirth armed and nothing selected, the targeting cursor takes a released party member's corpse: it lights over the corpse in range, and a click casts on it. Before, the cursor stayed grey over a corpse and a click bound nothing.
- **The pick** (`0x480816`, `0x6e6260`). While targeting, a corpse is a candidate only under a word in `0x8600`: under `0x8000` a corpse that is not bones and passes `0x6067d0`, and under `0x200`, tried only without `0x8000`, one that is not bones and fails it. Outside targeting every corpse stays a candidate, as before.
- **The reaction gate `0x6067d0`**: the caster's faction template toward the corpse owner's race template (`ChrRaces`, as `0x5d7120` reads it), friendly at 4 and up. The faction catalog now carries each race's template, and the PvP team loader reads it from there.
- **The cursor**: `0x6e6460`'s corpse leg (`0x6e6719`-`0x6e676d`) with its range tail. `GetMinMaxRange 0x6e3480` pads a corpse like a unit, reading the caster's own reach twice (`0x6e3605`-`0x6e361e`).
- **The click**: `BindTarget`'s corpse arm (`0x6e5f89`-`0x6e6020`), the `0x8000` arm then the `0x200` arm, through the same range merge as a unit (`0x6e602c`). Out of range raises the error and keeps the cursor armed, as the unit arm already does.
- **The wire**: `CMSG_CAST_SPELL` with the corpse bit and the corpse's packed guid, which the client writes for any mask bit in `0x8a02` (`0x7e4e10`) and vmangos reads for either corpse bit (`SpellCastTargetsInfo.cpp:156-157`); `CMSG_USE_ITEM` the same, for Goblin Jumper Cables.
Checked: byte-exact goldens for both packets; unit tests for the pick filter, the cursor verdict (bones, a hostile corpse, out of range, the reach pad), the bind arms from a world click to the sent command, and the reaction gate on the shipped DBCs. Each fails with its piece undone. There is no live run: the probes cannot aim or click in the 3D world. Gates green on macOS; no platform seam touched.
Fixes#450🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Targeting an NPC whose faction you are Exalted with no longer opens a script error. `UnitReaction` answered 8 there, and stock `TargetFrame_CheckFaction` indexes `UnitReactionColor[reaction]`, which has seven entries (`TargetFrame.lua:6-14`, `:183`).
- **The cap.** `UnitReaction 0x6061e0` takes `0x606530`'s answer through `min(rank, 6)` at `0x606439` (`cmp esi,6` / `jl` / `mov esi,6`), so Revered and Exalted both read 6, and the Lua binding adds one (`0x51683e`), so Lua sees at most 7. `ring_reaction` now caps its answer there.
- **The ladder stays whole.** `reputation_rank` is untouched: the reputation pane and the item requirement lines still read Exalted.
- **Every caller** of `ring_reaction` models `0x6061e0` and tests at 1, 3, 4 or 5, so none changes but the four Lua unit feeds (target, targettarget, npc, mouseover), which now share one `unit_reaction` helper. The vendor discount's Exalted row reads as Revered, as in the reference, and still takes 10%.
- **Comments** that gave the scale as 0..7 or 1..8 now say 0..=6 and 1..=7.
Checked: `the_reputation_rank_is_capped_at_revered` walks both sides of the Honored, Revered and Exalted thresholds; `an_exalted_npc_target_reads_revered_and_its_plate_is_green` runs stock `TargetFrame` on an Exalted Stormwind unit. With the cap loosened to 7 both fail, the second with the issue's `TargetFrame.lua:183: attempt to index field '?'`; restored, both pass. Gates green on macOS; no platform seam touched.
Fixes#440🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A friendly or resurrection spell that binds nothing now enters the unit targeting cursor, as `TryCast` does for any residual word without the enemy bit (`0x6e507b`–`0x6e50ab`). `autoSelfCast` boots at its reference default of `"0"`, and its `Deviates` row is gone.
- **Binding.** The cursor binds a unit through a world click, and through `SpellTargetUnit` from the unit frames (`BindTarget 0x6e5b40`). The selection never moves.
- **Relation checks.** Heals and buffs ask `CanAssist 0x6066f0` instead of reaction alone, so a friendly NPC without the PvP flag is no longer assistable. That is part of #50.
- **Range.** `SpellCanTargetUnit` and the cursor answer per unit through `0x6e6460`'s unit leg, including the spell's min/max range (`6e677c`–`6e6802`). A click on a unit out of range raises "Out of range." or "Target too close" and keeps the cursor armed. The reference's abort there also strands the cursor, and that is not copied.
- **Your own model** joins the world pick only for a word `0x6e61a0` admits (`0x850e`, no `AttributesEx 0x80000`), and never in first person.
Fixes#136
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
- A repair the purse cannot cover is refused as in 1.12.1: the not-enough-money error and its voice line, no repair sound, nothing sent; Repair All tests the worn gear alone
- Tidy three comments: the gift-wrap getter's reader, the repair-reset test's claim, the merchant module doc's width
Fixes#412
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- Repair costs take the vendor's price discount, 10% off at Honored and up to 10% more by honor rank at a PvP-flagged vendor, rounded as 1.12.1 rounds them
- Flight map fares take the flight master's price discount, as 1.12.1 shows them, while an express flight still sends the full fare
Fixes#411Fixes#434
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- fix: repair mode
- repair mode: a right-click puts a held payload back, then repairs the item
- repair clicks sound ITEM_REPAIR as they are sent, under the kit's no-duplicates gate
- the repair-all button is no longer driven from Rust
- repair mode: the click bindings' order, for every cursor payload
- worn-item repair clicks drain beside the bag's, in no system of their own
Fixes#147Fixes#148
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
The comment and string passes' last leftovers, and one lookup made to match the reference.
**Comments that stated a wrong reference fact**, now stating the right one:
- A cast failure clears the GCD, and a failed cooldown-on-event spell also drops its parked record (`0x6e73cc`).
- `SetUnit` is `0x505d70`; `0x476cb0` is the component clone it reaches.
- Interior sound follows the character, where the listener sits, not the camera.
- Template inheritance re-runs `LoadXML` on the one object against each template (`0x76985c`), and reads `name` and `id` from the instance alone (`0x769770`).
- On the ghost-to-alive edge the reference clears its corpse cache and sends nothing (`0x5ee990`, `0x491f50`).
- The name cache has no high-bit miss. An empty-name answer evicts the record in the reference (`0x55f6f0`).
- The combat-log deferred queue (`0xc4e208`) waits only on an item name. A unit name never waits.
- A flying spline's segments are paced by curve length (`0x454320`, `0x453760`), not by chord length.
- A stop within 3 yd of its stop point keeps the arrow nocked (`0x6019e4`).
- A missile's sound id is at `CMissile+0x40`.
- A nil or unresolved `AssistUnit` token is a gap against `0xb8` ERR_GENERIC_NO_TARGET, not a deviation.
**Strings:**
- Test, log and tool messages lose leftover history wording and placeholder labels.
- The extractor's scan summaries describe the current tint and sequence-clock paths.
- The smoke script's failure text carries no private numbers.
**Line continuations:** twelve log, panic and assert strings get back the `\` continuations they had lost. Each printed a run of spaces mid-message. Fixes#106.
**One lookup:** a region's `inherits=` is looked up as one verbatim name, matched case-insensitively (`0x6ee6f0`, `_strnicmp` at `0x6ee747`), as `framexml::expand` already did for frames. The gate before it split on commas and matched case-sensitively, so a texture or font string whose `inherits=` spelled its template in another case never received the template. `a_region_takes_a_mis_cased_template` fails before this change and passes after.
**Tests:** the quest tests stop loading three FrameXML files that no quest file uses.
**In game:** nothing changes for the stock UI. Log lines read as one sentence, and an addon texture that names its template in another case now gets it.
Verified with `scripts/check.sh` on the changed crates, plus the loader and quest tests with the install present.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Part of #92, for `benilla-ui`, `benilla-world` and `benilla-formats`.
- 88 string literals drop private record and bug numbers, the dates of private incidents, and the workspace's name for the maintainer. They are assert and panic texts, log and printout lines, the `--` comments of the Lua prelude and of test fixtures, and one Lua assert. Each keeps its fact in its own words.
- Two test names lose their numbers: `burning_steppes_lava_builds_a_magma_surface_at_the_reported_pin` and `join_channel_by_name_answers_the_row_custom_and_nil_legs`.
- A test-only event, `B271_PROBE`, is now `ERROR_PROBE` in the two crates that register and fire it. A fixture's error text in `benilla-app`'s `ui_script` tests drops its bug number, together with the assertion that reads it.
- Doc comments that named a client function by a disassembler's label (`FUN_0068a830`) now give its address, in `benilla-formats`, `benilla-assets` and one `benilla-app` file.
- Every literal change keeps each literal's line count, proven with comments stripped and every literal masked. The two renamed tests differ only in their names.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Part of #92, for the crates whose comment pass has landed and whose literals are few: `benilla-assets`, `benilla-blp`, `benilla-protocol` and `benilla-srp`.
- Assert messages, printed labels and a probe's error text drop their private record and bug numbers and the workspace's name for the maintainer. Each keeps its fact in its own words. One assert names a client function by its address, `0x68a830`, instead of a disassembler's label for it.
- Two terrain tests named after a bug number are renamed for what they check: `the_pin_stands_one_chunk_west_of_an_impassable_band` in `benilla-formats`, and `a_body_walking_east_from_the_pin_is_stopped_at_the_wall` in `benilla-world`. Those two crates' other literals follow in their own pass.
- Every change is to string-literal text, proven with comments stripped and every literal masked, except the two renamed tests.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Fixes#32. Every vmangos `File.cpp:N` citation in the crates whose comments have been rewritten now reads in one revision, vmangos core `6b34a36`, the one the comments are written against, a few commits from the server benilla is run against. Some citations had been taken from a newer vmangos checkout, where the lines have moved. #32's own targets came from that newer checkout too, so where this PR and #32 differ, this PR follows `6b34a36`.
- 53 citations in 31 files are re-pointed: most in benilla-protocol, a few in benilla-ui and benilla-app. Each new line was read and is the construct its comment names.
- #32's cases:
- The root-ack kick is `MovementHandler.cpp:715-722`; three comments cited the falling deferral after it.
- Movement is discarded before `CMSG_SET_ACTIVE_MOVER` lands at `:291-293`, through `Player::GetConfirmedMover`.
- The released mover's parting stop is `:955-964`.
- The mount-special echo is sent with `self` false at `:972`. Whether the sender gets its own echo depends on the server's broadcaster (`Object.cpp:2273-2280`); that is stated once, where the packet is parsed, and the opcode comment points there.
- Ten more citations were a line or two off in every revision. One comment also stated a wrong fact: vmangos's ignoring-you party refusal names the invitee, and the empty name is the raid-convert confirmation's.
- Comment text only. Every file strips to identical code, and `cargo fmt --check` is clean.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Part of #41: the comments of `benilla-formats` (the library, its extract tools, examples and tests) are rewritten to the rule in `docs/METHOD.md`, which is what the code does and the reference fact behind it, in a line. 19,654 comment lines become 7,264.
Fixes#105: the WMO collision comments now name walking's box-query leaf `0x6bca50` and the camera and line-of-sight segment-query leaf `0x6bc700`, in `models/collision.rs`, `examples/wmo_bsp.rs` and `tests/chain.rs`.
- **The mechanical strip.** The citations that were only a private record number go first, and each edit is shown to be a pure removal.
- **Wrong facts corrected.** The rewrite corrects the comments that stated wrong facts. Among them:
- A placed doodad's material alpha reads file slot 0 on its spawn clock, while the reference reads bone 0's armed sequence, re-rolled every play window. The old text said doodads are armed once at load.
- The fog end is stored ×36 and scaled once at load (`0x6d6090`). It is not raw.
- The MCSH bit picks a map doodad's sun scale, 1.0 lit or 0.5 shadowed. 2.5 belongs to an entity's light node.
- `SpellRec` is the row as stored, column × 4, not a translated copy.
- A global sequence's clock is anchored at the instance's attach time (`0x714352`).
- Particle twinkle is a gated size multiplier (`0x7b2adc`).
- The blob shadow sizes from the Stand box.
- Data claims are checked against the install: the weapon-trail kits' parameters and alphas, the beam rows' noise, and the maps' `field_16`.
- Dozens more turned up on the way: renamed or deleted items, docs sitting on the wrong item, stale counts and links.
- **Deviation labels.** "Deviation:" now marks only a choice with a reason. An unbuilt behaviour says so plainly.
- **Verification.** Every file strips to identical code, and `cargo fmt --check` is clean. Four second readers compared every file old against new, restored the facts the rewrite had dropped, and checked each new citation against the reference, the code and the shipped data.
String literals are untouched; they are #92.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Comments that named a chat reporter, cited a chat channel or message id, or quoted a report word for word now state the symptom in their own words. The test fixtures that used real character names use invented ones.
Verified: 52 of the 56 files strip to identical code with their comments removed; the other 4 differ only in the fixture names, and with those mapped back they strip identical too. `cargo fmt --check` is clean, and the `ui_party`, `unit_frame_tests` and `capture` tests pass with the install present and no skips. Record numbers still in these comments go with the comment pass (#41).
Fixes#34🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Comments that pointed into the maintainer's private reverse-engineering notes (a note by name, a
path, a section or finding id, the analysis's own name for a client function, a process word) now
carry the fact itself, cited by the reference client's address, by vmangos file and line, or as
"the reference". Nobody outside could follow those pointers.
Comment text only, plus the test-assert and log strings that carried the same pointers. No claim
changes: a comment the notes have since corrected keeps its wording here, and correcting it is
separate work.
The two reference generators, `scripts/gen-reference-globals.py` and
`scripts/gen-reference-verb-events.py`, named the private notes' layout to find their inputs. They
now take the input files as arguments (`--capture`, `--framexml`, `--names`, `--extents`,
`--disasm`). Run on the same inputs, both write the committed tables row for row; only the two
header lines that named the old environment variable change.
Checked per file:
- the code is identical: each file with its comments stripped compares equal before and after; a
changed string literal is compared under a mask that proves only literal text moved and every
literal kept its line count;
- no client address the old text cited was dropped;
- a second reviewer checked every added address against its derivation;
- a detector for the notes' names, paths, ids, function names and process words finds none left;
- `scripts/gates.sh` is green.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
For blend modes 5 and 6 the reference zeroes the tint·M2Color term and
forces the primary colour to the blend identity (0x70c507/0x70c5b8),
keeping only the alpha. benilla multiplied the batch's M2Color (baked
into its vertex colours, or animated through the material tint) and the
body tint into the texel. Mod and Mod2x batches now bake no M2Color, and
the shader draws them from the bare texel; the alpha still reaches them
through the identity lerp.
Of the install's 615 Mod and Mod2x batches, four carry a non-white
M2Color and none an animated one. The Scholomance testing jars (empty,
ghoul and skeleton) lose the (0.93, 1.00, 0.54) cast on their Mod layer,
and Offhand_Naxxramas_D_01's Mod2x sheen loses its (1.00, 0.95, 0.85)
warmth.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
One line per dependency where the reason is not obvious, measurements
kept as the number and the unit. Parsed with tomllib before and after,
every manifest is identical.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
465 tests loaded the stock interface off the install without declaring
it, and 25 addon tests assumed the install beside the corpus; each now
opens with `wow_data_or_skip!`. The gates run the workspace tests a
second time with `WOW_DATA=`, so a test that reads the install
undeclared fails at land instead of on a contributor's clone. The skip
log writes each line in one call so parallel tests do not interleave.
Verified: the workspace tests with `WOW_DATA=` in a checkout that has
the addon corpus pass 6972, fail 0, and log 1555 skips; on a fresh
clone of main before this change they failed 465.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
What a clone lacked: the gates went green with a thousand skipped tests and said nothing, every
login script assumed a worktree-pool slot and its probe account, the client defaulted to the
maintainer's account, a docker path and a home directory were in the code, the temp files broke
on GNU coreutils, and two generators could only run against a sibling repo.
Now the test gates print how many data-gated tests skipped and why; a scripted run logs in as
the account its checkout declares in `.probe-identity` or as the three variables name, never a
default, and the client refuses a scripted login on any other account; the event and message
tables regenerate from vendored inputs, byte for byte; the docs state the real prerequisites.
Out of the tree, into the maintainer's own tooling: the worktree pool, its hooks, the perf leg
runner, the cross-platform compile and the Vulkan container leg — a way of working, and setups
not every machine has. One hook stays: no writes into the install.
Two tests that could never run are gone with their subject: the quest-timer geometry diff
compared a copy of the window that no longer exists (and joined a chain path onto a folder, so
it always skipped), and the FrameXML flag gate now reads the stock files off the player's own
patch chain instead of an extracted folder only one machine had.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Until now this repo was an export of a private tree, pushed as squashed snapshots. From here on
it is the working tree itself, and the private repo is archived.
The root holds AGENTS.md and the Rust and GitHub files only. Everything an agent needs to orient
is in docs/: METHOD.md (the rules and how the project is built, 123 lines, project method only),
MAP.md (generated from the code at every land) and CONTRIBUTING.md (what gets in and how a
change is judged). One skill, `probe`, holds the procedure for running the client unattended
against the local server.
Brought over because the project needs it: reference/ (the 1.12 name catalogues 34 source
files read; the export failed its own tests without them), the gate chain and the pool
(scripts/), the hooks that enforce the install rule and the worktree rule, the full .gitignore,
the `cargo play` alias, and the original wording of eleven source files the export had reworded.
Left out: the decision records (archived with the private repo), the export tooling, the
operator's Discord and triage loop, the lab-laptop scripts and the instruments nothing
references, which live outside the repo now. What read the records is gone from the tooling:
no corpus lint, no records index, no `reserve` verb, no number-collision check.
Three hooks are gone: fmt-at-commit and the file-size nudge were redundant with the gates, and
the assets/ui guard is now a test (`assets_ui_does_not_grow`) that fails for everyone, not only
for a Claude Code session. The remaining pool hooks stand down on a machine without a pool.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- a Vanish, a stealth or a unit leaving range clears your target at once, not two seconds later, and TAB reaches a mob behind you (2345)
- a fully painted terrain layer covers fully: roads and rock no longer let about 6% of the grass underneath show through (2345)
- companion pets show their names, the saved camera pitch survives a login, and a missed resurrect offer no longer blocks later ones (2345)
- ESC on a quest giver closes it, browsing the auction house by category searches the right items, and abandon removes the quest you chose (2345)
- a profession window open at logout no longer reopens for the next character (2345)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- the trainer window re-derives every service's state when your money, level, spells or talents change, instead of asking the server for a second list after a purchase (2333, 2336)
- the backpack button reads the stock bag scan: with Bagnon holding the bags it stays unlit, as the reference leaves it (2332)
- three unused dev switches retire, and fifteen others are named as instruments (2335)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- the cast ladder moves under the spell: try, target bind, the usable walk and the refusal rungs live in one place, and the cast reply sends a chained cast itself (2330)
- every server packet decodes through one exhaustive match, and an opcode-coverage script joins the reference's 387 handled opcodes against ours (2331)
- a landing that touches a platform seam compiles Linux and Windows first, and the seven WMO portal-cull audits join the tests (2331)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- the spell owns its lifecycle state: the in-flight cast, the local cancel, the cooldowns and the spell-modifier tables live in one module (2328)
- a malformed MPQ, DBC, M2 or WDT file is refused instead of crashing the client, nine read sites bounded
- a test that needs the install or the addon corpus fails instead of skipping when the data is there; the Bagnon render suite runs again (2329)
- the dev crash injector is armed on every platform, and no instrument writes a path relative to the working directory
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- inside a battleground the battle map draws its terrain, overlays, arrow and blips: the world map's selection carries the instance maps as the reference's third cell (2298)
- talent spell modifiers land: the 45 packets a level-60 login threw away fill the reference's two tables, and a spell's power cost reads them, so a cast you can afford is no longer refused (2298)
- a ghost's Accept on the spirit-heal dialog can no longer be eaten by frame order, and a latched spirit guide no longer follows you into the next character's session (2298)
- the battleground probe refuses GM mode, exits when done, and reads the scoreboard before toggling it (2298)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- the character-create and main-menu backdrop runs its material animation: the clouds drift on their 20, 30 and 40 second loops, which were seeded once and never sampled (2300)
- Arathi Basin and Alterac Valley are entered and censused beside Warsong Gulch, the flag is picked up, and a match runs to its end (2296)
- the animation variation roll is one clock-seeded stream, as the reference's C runtime is, not four streams seeded 0 (2299, 2301)
- a cancelled channel never turns the bar red and a finished one closes on the bar's own clock, now pinned by tests that drive the real packet sequence (2284)
- the field watch fires one edge per moved descriptor dword, from the merge that holds both sides (2297)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
