- Map GuildCommandError codes to human-readable strings instead of showing raw
error numbers (e.g. \"error 4\" → \"No player named X is online.\")
- Handle errorCode==0 for QUIT command: show \"You have left the guild.\" and
clear guild state (name, ranks, roster) — previously silent
- Handle errorCode==0 for CREATE and INVITE commands with appropriate messages
- Substitute %s-style error messages with the player name from data.name
- Suppress repeated \"Guild: <Name>\" chat message on every SMSG_GUILD_QUERY_RESPONSE;
only announce once when the guild name is first learned at login
Previously "You are now in a group with N members." was shown on every
GROUP_LIST packet, which fires for each party stat update. Now only show
a message on actual state transitions: joining, leaving the group.
- SMSG_SUMMON_REQUEST: fall back to playerNameCache when entity not in
range; include zone name from getAreaName() in the summon message
(e.g. "Bob is summoning you to Stormwind.")
- SMSG_TRADE_STATUS BEGIN_TRADE: fall back to playerNameCache when the
trade initiator's entity is not visible
- SMSG_QUESTGIVER_QUEST_FAILED: look up quest title from questLog_ and
include it in the failure message (same pattern as QUESTUPDATE_FAILED
fix from previous session)
- SMSG_ACHIEVEMENT_EARNED: fall back to playerNameCache for non-visible
players before showing a raw hex GUID in the achievement message
- SMSG_DURABILITY_DAMAGE_DEATH: use the actual pct field from the packet
instead of hardcoding "10%"
- SMSG_PET_CAST_FAILED: read reason as uint8 (not uint32), look up spell
name and show human-readable failure reason to player
- Trade status 9 (REJECTED): show "Trade declined." instead of "Trade
cancelled." to distinguish explicit decline from cancellation
SMSG_QUESTUPDATE_FAILED and SMSG_QUESTUPDATE_FAILEDTIMER were emitting
generic "Quest 12345 failed!" messages. Now looks up the title from
questLog_ and shows e.g. "\"Report to Gryan Stoutmantle\" failed!" for
a much more readable notification. Falls back to the generic form if
the title is not cached.
SMSG_SPELLLOGEXECUTE POWER_DRAIN reads drainPower but was not passing it
to addCombatText, so drained-resource returns showed as blue (mana) even
for rage or energy. Now correctly colored following the energize palette
added in the earlier commit.
The dispel-failed handler was showing the failure notification for every
dispel attempt in the party/raid, regardless of who cast it. Now checks
casterGuid == playerGuid before showing "X failed to dispel." so only
the player's own failed dispels surface in chat.
SMSG_CHANNEL_NOTIFY carries many event types that were silently dropped
in the default case: wrong password, muted, banned, throttled, kicked,
not owner, not moderator, password changed, owner changed, invalid name,
not in area, not in LFG. These are now surfaced as system chat messages
matching WoW-standard phrasing.
SMSG_DUEL_WINNER type=1 means the loser fled the duel zone rather than
being defeated; was previously treated the same as a normal win. Now
shows "X has fled from the duel. Y wins!" for the flee case vs the
standard "X has defeated Y in a duel!" for a normal outcome.
Previously a spell failure like "Not in range" gave no context about
which spell failed. Now the message reads e.g. "Fireball: Not in range"
using the spell name from the DBC cache. Falls back to the bare reason
string if the spell name is not yet cached.
SMSG_REMOVED_SPELL and SMSG_SEND_UNLEARN_SPELLS both erased spells from
knownSpells but left stale references on the action bar. After a respec
or forced spell removal, action bar buttons would show removed talents
and spells as still present. Now both handlers clear matching slots and
persist the updated bar layout.
SMSG_SPELLLOGMISS contains miss events for both directions: spells the
player cast that missed, and enemy spells that missed the player. The
victim side (dodge/parry/block/immune/absorb/resist) was silently
discarded. Now both caster==player and victim==player generate the
appropriate combat text floater.
Mana (0)=blue, Rage (1)=red, Focus (2)=orange, Energy (3)=yellow,
Runic Power (6)=teal. Previously all energize events showed as blue
regardless of resource type, making it impossible to distinguish
e.g. a Warrior's Rage generation from a Mage's Mana return.
Power type is now captured from SMSG_SPELLENERGIZELOG (uint8) and
SMSG_PERIODICAURALOG OBS_MOD_POWER/PERIODIC_ENERGIZE (uint32 cast
to uint8) and stored in CombatTextEntry::powerType.
When buying a higher spell rank from a trainer, SMSG_TRAINER_BUY_SUCCEEDED
already announces "You have learned X", and SMSG_SUPERCEDED_SPELLS would
then also print "Upgraded to X" — two messages for one action.
Fix: check if the new spell ID was already in knownSpells before inserting
it in handleSupercededSpell. If so, the trainer handler already announced
it and we skip the redundant "Upgraded to" message. Non-trainer supersedes
(quest rewards, etc.) where the spell wasn't pre-inserted still show it.
SMSG_INITIAL_SPELLS delivers active cooldowns which were stored in
spellCooldowns but never propagated to the action bar slot
cooldownRemaining/cooldownTotal fields. This meant that spells with
remaining cooldowns at login time showed no countdown overlay on the
action bar. Sync the action bar slots from spellCooldowns after
loadCharacterConfig() to restore the correct timers.
Replace the generic "Party command failed (error N)" message with
WoW-standard error strings for each PartyResult code, matching what
the original client displays (e.g. "Your party is full.", "%s is
already in a group.", "%s is ignoring you.", etc.).
SMSG_TRAINER_BUY_SUCCEEDED pre-inserts the spell into knownSpells and
shows "You have learned X." The subsequent SMSG_LEARNED_SPELL packet
would then show a second "You have learned a new spell: X." message.
Fix: check if the spell was already in knownSpells before inserting in
handleLearnedSpell. If it was pre-inserted by the trainer handler, skip
the chat notification to avoid the duplicate.
After automatically upgrading action bar slots to the new spell rank
in handleSupercededSpell, save the character config so the upgraded
slot IDs persist across sessions.
When a spell is superceded (e.g. Fireball Rank 1 -> Rank 2 after
training), update any action bar slots referencing the old spell ID
to point to the new rank. This matches WoW client behaviour where
training a new rank automatically upgrades your action bars so you
don't have to manually re-place the spell.
Same-map teleports (dungeon teleporters, etc.) clear casting state but
were not clearing lastInteractedGoGuid_. If a gather cast was in progress
when the teleport happened, the stale GO guid could theoretically trigger
a spurious CMSG_LOOT on the destination map.
Also clears lastInteractedGoGuid_ in handleNewWorld alongside the rest of
the casting-state teardown for consistency with other reset paths.
SMSG_LOOT_START_ROLL was not calling queryItemInfo(), so the roll popup
would display item IDs instead of names when the item had not been
previously cached (e.g. first time seeing that item in the session).
Also update renderLootRollPopup to prefer the live ItemQueryResponseData
name/quality over the snapshot captured at parse time, so the popup
shows the correct name once SMSG_ITEM_QUERY_SINGLE_RESPONSE arrives.
wasInTimedCast checked casting == true but not whether the completing spell
was actually the gather cast. A triggered/proc spell (SMSG_SPELL_GO with
a different spellId) could arrive while a gather cast is active (casting==true),
satisfying the old guard and firing lootTarget prematurely.
Require data.spellId == currentCastSpellId so only the spell that started
the cast bar triggers the post-gather CMSG_LOOT dispatch.
When the server sends SMSG_GAMEOBJECT_CUSTOM_ANIM with animId=0 for a GO
of type 17 (FISHINGNODE), a fish has been hooked and the player needs to
click the bobber quickly. Add a system chat message and a UI sound to
alert the player — previously there was no visual/audio feedback beyond
the bobber animation itself.
Mailboxes, doors, buttons, and other non-lootable GOs set shouldSendLoot=false
so no CMSG_LOOT is dispatched — but lastInteractedGoGuid_ was still set.
Without SMSG_LOOT_RESPONSE to clear it, a subsequent timed cast completion
(e.g. player buffs at the mailbox) would fire a spurious CMSG_LOOT for the
mailbox GUID.
SMSG_CAST_FAILED is a direct rejection (e.g. insufficient range, no mana)
before the cast starts. Missing this path meant a stale gather-node guid
could survive into the next timed cast if SMSG_CAST_FAILED fired instead
of SMSG_SPELL_FAILURE.
If a gather cast was interrupted by SMSG_SPELL_FAILURE (e.g. player took
damage during mining), lastInteractedGoGuid_ was left set. A subsequent
timed cast completion would then fire CMSG_LOOT for the stale node even
though the gather never completed.
Clear lastInteractedGoGuid_ in all cast-termination paths:
- SMSG_SPELL_FAILURE (cast interrupted by server)
- SMSG_CAST_RESULT non-zero (cast rejected before it started)
- cancelCast() (player or system cancelled the cast)
- World reset / logout block (state-clear boundary)
handleSpellGo fired lootTarget(lastInteractedGoGuid_) on ANY player spell
completion, including instant casts and proc/triggered spells that arrive
while the gather cast is still in flight. Save the casting flag before
clearing it and only dispatch CMSG_LOOT when wasInTimedCast is true — this
ensures only the gather cast completion triggers the post-gather loot send,
not unrelated instant spells that also produce SMSG_SPELL_GO.
Two bugs fixed:
1. Retry logic (for Classic) re-sent CMSG_GAMEOBJ_USE at 0.15s while the
gather cast was in-flight, causing SPELL_FAILED_BAD_TARGETS. Now clears
pendingGameObjectLootRetries_ as soon as SMSG_SPELL_START shows the player
started a cast (gather accepted).
2. CMSG_LOOT was sent immediately before the gather cast completed, then
never sent again — so the loot window never opened. Now tracks the last
interacted GO and sends CMSG_LOOT in handleSpellGo once the gather spell
completes, matching how the real client behaves.
Action bar changes (dragging spells/items) were only saved locally.
Now notifies the server via CMSG_SET_ACTION_BUTTON so the layout
persists across relogs. Supports Classic (5-byte) and TBC/WotLK
(packed uint32) wire formats.
Classic 1.12 does not send SMSG_DEATH_RELEASE_LOC, leaving corpseMapId_=0
and preventing the 'Resurrect from Corpse' button from appearing.
- When health reaches 0 via VALUES update, immediately cache movementInfo
as corpse position (canonical->server axis swap applied correctly)
- Do the same on UNIT_DYNFLAG_DEAD set path
- Clear corpseMapId_ when ghost flag is removed (corpse reclaimed)
- Clear corpseMapId_ in same-map spirit-healer resurrection path
The CORPSE object detection (UPDATE_OBJECT) and SMSG_DEATH_RELEASE_LOC
(TBC/WotLK) will still override with exact server coordinates when received.
- Remove NoDecoration flag to allow ImGui drag/resize
- Store questTrackerRightOffset_ instead of absolute X so tracker
stays pinned to the right edge when the window is resized
- Persist position (right offset + Y) and size in settings.cfg
- Clamp to screen bounds after drag
gameObjectDisplayIdWmoCache_ was not cleared on world unload/transition,
causing stale WMO model IDs (e.g. 40006, 40003) to be looked up after
the renderer cleared its model list, resulting in "Cannot create instance
of unloaded WMO model" errors on zone re-entry.
Changes:
- Clear gameObjectDisplayIdWmoCache_ alongside other GO caches on world reset
- Add WMORenderer::isModelLoaded() for cache-hit validation
- Inline GO WMO path now verifies cached model is still renderer-resident
before using it; evicts stale entries and falls back to reload
When the M2 model cache is full (>6000 entries), loadModel() returns
false and the model is never added to the GPU cache. The WMO instance
doodad path was calling createInstanceWithMatrix() unconditionally,
generating hundreds of "Cannot create instance: model X not loaded"
warnings on zone entry. Add the same guard already present in the
terrain doodad path.
SMSG_TALENTS_INFO wire format sends 0-indexed ranks (0=has rank 1). Both
handlers were storing raw 0-indexed values, but handleSpellLearnedServer
correctly stored rank+1 (1-indexed). This caused:
- getTalentRank() returning 0 for both "not learned" and "has rank 1",
making pointsInTree always wrong and blocking tier access
- Prereq check `prereqRank < DBC_prereqRank` always met when not learned
(0 < 0 = false), incorrectly unlocking talents
- Click handler sending wrong desiredRank to server
Fixes:
- Both SMSG_TALENTS_INFO handlers: store rank+1u (1-indexed)
- talent_screen.cpp prereq check: change < to <= (DBC is 0-indexed,
storage is 1-indexed; must use > for "met", <= for "not met")
- talent_screen.cpp click handler: send currentRank directly (1-indexed
value equals what CMSG_LEARN_TALENT requestedRank expects)
- Tooltip: display prereqRank+1 so "Requires 1 point" shows correctly
Vendors that open directly (without gossip menu) never triggered the
armorer gossip path, so canRepair was always false and the Repair button
was hidden. Now also check the NPC's unit flags for NPC_FLAG_REPAIR when
the vendor list arrives, fixing armorers accessed directly.
TBC races like Draenei use version-264 M2 files with no embedded skin;
indices come from a separate .skin file loaded after M2::load().
The premature isValid() check (which requires non-empty indices) always
failed for WotLK-format character models, making Draenei (and Blood Elf)
players invisible.
Fix: only check vertices.empty() right after load(), then validate fully
with isValid() after the skin file is loaded.
Two issues in the WotLK SMSG_ATTACKERSTATEUPDATE parser:
1. subDamageCount could read a school-mask byte when a packed GUID is
off by one byte, producing values like 32/40/44/48 (shadow/frost/etc
school masks) as the count. The parser then tried to read 32-48
sub-damages before hitting EOF. Fix: silently clamp subDamageCount to
floor(remaining/20) so we only attempt entries that actually fit.
2. After sub-damages, AzerothCore sends victimState(4)+unk1(4)+unk2(4)+
overkill(4) (16 bytes), not the 8-byte victimState+overkill the
parser was reading. Fix: consume unk1 and unk2 before reading overkill.
Also handle the hitInfo-conditional HITINFO_BLOCK/RAGE_GAIN/FAKE_DAMAGE
fields at the end of the packet.
When SPLINEFLAG_ANIMATION (0x00400000) is set, AzerothCore inserts 5 bytes
(uint8 animationType + int32 animTime) between durationModNext and
verticalAccel in the SMSG_UPDATE_OBJECT MoveSpline block. The parser was
not accounting for these bytes, causing verticalAccel, effectStartTime,
and pointCount to be read from the wrong offset.
This produced garbage pointCount values (e.g. 3322451254) triggering the
"Spline pointCount invalid (legacy+compact)" fallback path and breaking
UPDATE_OBJECT parsing for animated-spline entities, causing all subsequent
update blocks in the same packet to be dropped.
When unloadTile() was called for a tile still in finalizingTiles_
(mid-incremental-finalization), terrain chunks already uploaded to the
GPU (terrainMeshDone=true) were not being cleaned up. The early-return
path correctly removed water and M2/WMO instances but missed calling
terrainRenderer->removeTile(), causing descriptor sets to leak.
After ~20 minutes of play the VkDescriptorPool (MAX_MATERIAL_SETS=16384)
filled up, causing all subsequent terrain material allocations to fail
and the log to flood with "failed to allocate material descriptor set".
Fix: check fit->terrainMeshDone before the early return and call
terrainRenderer->removeTile() to free those descriptor sets.
wmo_renderer: when portal BFS starts from a group with no portal refs
(utility/transition group), the rest of the WMO becomes invisible because
BFS only adds the starting group. Fix: if cameraGroup has portalCount==0,
fall back to marking all groups visible (same as camera-outside behavior).
renderer: minimap player-orientation arrow was pointing the wrong direction.
The shader convention is arrowRotation=0→North, positive→clockwise (West),
negative→East. The correct mapping from canonical yaw is arrowRotation =
-canonical_yaw. Fixed both render paths (cameraController and gameHandler)
in both the FXAA and non-FXAA minimap render calls.
World-map W-key: already corrected in prior commit (13c096f); users with
stale ~/.wowee/settings.cfg should rebind via Settings > Controls.
The failed-model cache introduced in f855327 would persist across map
changes, permanently suppressing models that failed on one map but might
be valid assets on another (or after a client update). Clear it in the
world reset path alongside the existing gameObjectDisplayIdModelCache_
clear, so model loads get a fresh attempt on each zone change.
Two follow-up fixes for the ribbon emitter implementation and the
transport-doodad stall fix:
1. loadModel() rejected any M2 with no vertices AND no particles, but
ribbon-only spell-effect models (e.g. weapon trail or aura ribbons)
have neither. These models were silently invisible even though the
ribbon rendering pipeline added in 1108aa9 is fully capable of
rendering them. Extended the guard to also accept models that have
ribbon emitters, matching the particle-emitter precedent.
2. processPendingTransportDoodads() ignored the bool return of
loadModel(), calling createInstance() even when the model was
rejected, generating spurious "Cannot create instance: model X not
loaded" warnings for every failed doodad path. Check the return
value and continue to the next doodad on failure.
Three root causes identified from wowee.log crash at frame 134368:
1. processPendingTransportDoodads() was doing N separate synchronous
GPU uploads (vkQueueSubmit + vkWaitForFences per texture per doodad).
With 30+ doodads × multiple textures, this caused the 489ms stall in
the 'gameobject/transport queues' update stage. Fixed by wrapping the
entire batch in beginUploadBatch()/endUploadBatch() so all texture
layout transitions are submitted in a single async command buffer.
2. Game objects whose M2 model has no geometry/particles (empty or
unsupported format) were retried every frame because loadModel()
returns false without adding to gameObjectDisplayIdModelCache_.
Added gameObjectDisplayIdFailedCache_ to permanently skip these
display IDs after the first failure, stopping the per-frame spam.
3. renderM2Ribbons() only checked ribbonPipeline_ != null, not
ribbonAdditivePipeline_. If additive pipeline creation failed, any
ribbon with additive blending would call vkCmdBindPipeline with
VK_NULL_HANDLE, causing VK_ERROR_DEVICE_LOST on the GPU side.
Extended the early-return guard to cover both ribbon pipelines.
Servers send a 9-byte packet (guid+lootType) with lootType=LOOT_NONE when
loot is unavailable (locked chest, another player looting, needs a key).
The previous parser required ≥14 bytes (guid+lootType+gold+itemCount) and
logged a spurious WARNING for every such failure response.
Now:
- Accept the 9-byte form; return false so the caller skips opening the
loot window (correct behaviour for a failure/empty response).
- Log at DEBUG level instead of WARNING for the short form.
- Keep the original WARNING for genuinely malformed packets < 9 bytes.
- TOGGLE_QUEST_LOG: change default from Q to None — Q conflicts with
strafe-left in camera_controller; quest log already accessible via
TOGGLE_QUESTS (L, the standard WoW binding)
- Equipment Set Manager: remove hardcoded SDL_SCANCODE_GRAVE shortcut
(~` should not be used for this)
- World map M key: remove duplicate SDL_SCANCODE_M self-handler from
world_map.cpp::render() that was desync-ing with game_screen's
TOGGLE_WORLD_MAP binding; game_screen now owns open/close, render()
handles initial zone load and ESC-close signalling via isOpen()
Parse M2RibbonEmitter data (WotLK format) from M2 files — bone index,
position, color/alpha/height tracks, edgesPerSecond, edgeLifetime,
gravity. Add CPU-side trail simulation per instance (edge birth at bone
world position, lifetime expiry, gravity droop). New m2_ribbon.vert/frag
shaders render a triangle-strip quad per emitter using the existing
particleTexLayout_ descriptor set. Supports both alpha-blend and additive
pipeline variants based on material blend mode. Fixes invisible spell
trail effects (~5-10%% of spell visuals) that were silently skipped.
- canReclaimCorpse() and getCorpseDistance() compared canonical movementInfo
(x=north=server_y, y=west=server_x) against raw server corpseX_/Y_ causing
the proximity check to always report wrong distance even when standing on corpse
- Fix: use corpseY_ for canonical north and corpseX_ for canonical west
- Also detect OBJECT_TYPE_CORPSE update blocks owned by the player to set
corpse coordinates at login-as-ghost (before SMSG_DEATH_RELEASE_LOC arrives)
