When spline parsing consumes the wrong number of bytes, the subsequent
blockCount read lands on garbage data (e.g. 71 instead of ~5 for UNIT).
Previously the parser logged a warning but continued, reading garbage
mask/field data until hitting truncation. Now it returns false for
CREATE_OBJECT blocks with suspicious counts, letting the block loop
skip cleanly to the next entity.
Also downgrade ~44 diagnostic LOG_WARNING messages to LOG_DEBUG across
17 files (equipment, transport, DBC, heartbeat, chat, GO raypick, etc.)
to reduce log noise and make real warnings visible.
MAX_CULL_INSTANCES was 16384 but game object instances were allocated
at indices 20000+, beyond the cull buffer. Increased to 65536.
Also compute fallback boundRadius from vertex extents when M2 header
reports 0, and seed bones for global-sequence-only animated models.
M2 GPU instancing
- M2InstanceGPU SSBO (96 B/entry, double-buffered, 16384 max)
- Group opaque instances by (modelId, LOD); single vkCmdDrawIndexed per group
- boneBase field indexes into mega bone SSBO via gl_InstanceIndex
Indirect terrain drawing
- 24 MB mega index buffer (6M uint32) + 64 MB mega vertex buffer
- CPU builds VkDrawIndexedIndirectCommand per visible chunk
- Single VB/IB bind per frame; shadow pass reuses mega buffers
- Replaced vkCmdDrawIndexedIndirect with direct vkCmdDrawIndexed to fix
host-mapped buffer race condition that caused terrain flickering
GPU frustum culling (compute shader)
- m2_cull.comp.glsl: 64-thread workgroups, sphere-vs-6-planes + distance cull
- CullInstanceGPU SSBO input, uint visibility[] output, double-buffered
- dispatchCullCompute() runs before main pass via render graph node
Consolidated bone matrix SSBOs
- 16 MB double-buffered mega bone SSBO (2048 instances × 128 bones)
- Eliminated per-instance descriptor sets; one megaBoneSet_ per frame
- prepareRender() packs bone matrices consecutively into current frame slot
Render graph / frame graph
- RenderGraph: RGResource handles, RGPass nodes, Kahn topological sort
- Automatic VkImageMemoryBarrier/VkBufferMemoryBarrier between passes
- Passes: minimap_composite, worldmap_composite, preview_composite,
shadow_pass, reflection_pass, compute_cull
- beginFrame() uses buildFrameGraph() + renderGraph_->execute(cmd)
Pipeline derivatives
- PipelineBuilder::setFlags/setBasePipeline for VK_PIPELINE_CREATE_DERIVATIVE_BIT
- M2 opaque = base; alphaTest/alpha/additive are derivatives
- Applied to terrain (wireframe) and WMO (alpha-test) renderers
Rendering bug fixes:
- fix(shadow): compute lightSpaceMatrix before updatePerFrameUBO to eliminate
one-frame lag that caused shadow trails and flicker on moving objects
- fix(shadow): scale depth bias with shadowDistance_ instead of hardcoded 0.8f
to prevent acne at close range and gaps at far range
- fix(visibility): WMO group distance threshold 500u → 1200u to match terrain
view distance; buildings were disappearing on the horizon
- fix(precision): camera near plane 0.05 → 0.5 (ratio 600K:1 → 60K:1),
eliminating Z-fighting and improving frustum plane extraction stability
- fix(streaming): terrain load radius 4 → 6 tiles (~2133u → ~3200u) to exceed
M2 render distance (2800u) and eliminate pop-in when camera turns;
unload radius 7 → 9; spawn radius 3 → 4
- fix(visibility): ground-detail M2 distance multiplier 0.75 → 0.9 to reduce
early pop of grass and debris
buildDefaultPlayerGeosets() was inserting all submeshIds 0-99 into
activeGeosets, showing every hair variation simultaneously. Now uses
the hairGeosetMap_ (from CharHairGeosets.dbc) to select only the
correct hair scalp geoset for the player's race/sex/style, matching
the existing NPC geoset filtering logic in EntitySpawner.
Revert static JSON layout changes (15-22 back to 14-21) since WotLK
loads the Classic 23-field DBC. Add getItemDisplayInfoTextureFields()
helper that detects field count at runtime and adjusts the texture
base index accordingly (14 for 23-field, 15 for 25-field).
Instance was created with Vulkan 1.1 but depthResolveSupported_ was gated
on the physical device's API version (1.2+ on RADV). This caused
vkCreateRenderPass2 (core 1.2) to dispatch through a null function pointer
when MSAA was enabled. Now requests 1.2 instance with 1.1 minimum fallback
and gates depth resolve on the actual instance API version. Also removes
all diagnostic crash-phase instrumentation from the previous investigation.
Add signal-safe render-phase markers throughout GameScreen::render() and
Application::render() so the crash handler can report which render call was
active when a SIGSEGV occurs. The AMD RADV crash backtrace only shows 2
frames due to missing frame pointers, making it impossible to identify the
actual crash site.
Changes:
- Add volatile g_crashRenderPhase marker updated before each major render call
- Upgrade Linux signal handler to sigaction with SA_SIGINFO for faulting address
- Set ImGui CheckVkResultFn to log silent Vulkan errors in ImGui backend
- Enable -fno-omit-frame-pointer in all build configs (not just Debug/RelWithDebInfo)
The warmup loop waited up to 20 seconds for getHeightAt() to return a
terrain height within 15 units of spawn Z before accepting the ground
as ready. In practice, the terrain was loaded and the character was
visibly standing on it, but the height sample didn't match closely
enough (terrain LOD, chunk boundary, or server Z vs client height
mismatch).
Reduce the tile-count fallback timeout from 20s to 5s: if at least 4
tiles are loaded after 5 seconds, accept the ground as ready. The
exact height check still runs in the first 5 seconds for fast-path
cases where it does match.
Remove all OpenGL/GLEW code and dependencies. The Vulkan renderer has
been the sole active backend for months; these files were dead code.
Deleted (8 files, 641 lines):
- rendering/mesh.cpp+hpp: OpenGL VAO/VBO/EBO wrapper (never instantiated)
- rendering/shader.cpp+hpp: OpenGL GLSL compiler (replaced by VkShaderModule)
- rendering/scene.cpp+hpp: Scene graph holding Mesh objects (created but
never populated — all rendering uses Vulkan sub-renderers directly)
- rendering/video_player.cpp+hpp: FFmpeg+GL texture uploader (never
included by any other file — login video feature can be re-implemented
with VkTexture when needed)
Cleaned up:
- renderer.hpp: remove Scene forward-decl, getScene() accessor, scene member
- renderer.cpp: remove scene.hpp/shader.hpp includes, Scene create/destroy
- application.cpp: remove stale "GL/glew.h removed" comment
- CMakeLists.txt: remove find_package(OpenGL/GLEW), source/header entries,
and target_link_libraries for OpenGL::GL and GLEW::GLEW
- PKGBUILD: remove glew dependency
- BUILD_INSTRUCTIONS.md: remove glew from all platform install commands
- input: fix undefined behavior in SDL mouse button loop — SDL_BUTTON(0)
computes (1 << -1) which is UB. Start loop at 1 since SDL button
indices are 1-based (SDL_BUTTON_LEFT=1, RIGHT=3, MIDDLE=2)
- big_num: guard BN_bn2hex/BN_bn2dec against nullptr return on
OpenSSL allocation failure — previously constructed std::string
from nullptr which is undefined behavior
- asset_manager: add size guard before fsPath.substr(size-4) in
tryLoadPngOverride — resolveFile could theoretically return a
path shorter than the extension
- wmo_loader: name kDoodadNameIndexMask (0x00FFFFFF) with why-comment
explaining the 24-bit name index / 8-bit flags packing and MODN
string table reference
- window: add why-comment on LOG_WARNING usage during shutdown —
intentionally elevated so teardown progress is visible at default
log levels for crash diagnosis
- Dockerfile: fix LLVM apt repo codename (jammy → noble) for ubuntu:24.04
- build-linux.sh: add missing mkdir -p /wowee-build-src before tar extraction
- Dockerfile: remove dead ENV OSXCROSS_VERSION=1.5 and its unset
- CMakeLists: scope -undefined dynamic_lookup to wowee target only
- GameServices: remove redundant game:: qualifier inside namespace game
- application.cpp: zero out gameServices_ after gameHandler reset in shutdown
Introduce `GameServices` struct — an explicit dependency bundle that
`Application` populates and passes to `GameHandler` at construction time.
Eliminates all 47 hidden `Application::getInstance()` calls in
`src/game/*.cpp`, completing SOLID-D (dependency-inversion) cleanup.
Changes:
- New `include/game/game_services.hpp` — `struct GameServices` carrying
pointers to `Renderer`, `AssetManager`, `ExpansionRegistry`, and two
taxi-mount display IDs
- `GameHandler(GameServices&)` replaces default constructor; exposes
`services() const` accessor for domain handlers
- `Application` holds `game::GameServices gameServices_`; populates it
after all subsystems are created, then constructs `GameHandler`
(fixes latent init-order bug: `GameHandler` was previously created
before `AssetManager` / `ExpansionRegistry`)
- `game_handler.cpp`: duplicate `isActiveExpansion` / `isClassicLikeExpansion` /
`isPreWotlk` anonymous-namespace helpers removed; `game_utils.hpp`
included instead
- All domain handlers (`InventoryHandler`, `SpellHandler`, `MovementHandler`,
`CombatHandler`, `QuestHandler`, `SocialHandler`, `WardenHandler`) replace
`Application::getInstance().getXxx()` with `owner_.services().xxx`
SDL2 requires video/window functions to be called from the main thread
(the one that called SDL_Init). The watchdog thread was calling
SDL_SetRelativeMouseMode, SDL_ShowCursor, and SDL_SetWindowGrab directly
on stall detection — undefined behavior on macOS (Cocoa requires main-
thread UI calls) and unsafe on other platforms.
Now the watchdog sets an atomic flag, and the main loop checks it at the
top of each iteration, executing the SDL calls on the correct thread.
Final sweep across mpq_manager, application, auth_screen, wmo_renderer,
character_renderer, and terrain_manager. All now use the unsigned char
cast pattern. No remaining bare ::tolower/::toupper or std::tolower(c)
calls on signed char in the codebase.
std::tolower(int) has undefined behavior when passed a negative value,
which signed char produces for bytes > 127. The rest of the codebase
correctly casts to unsigned char first; these 3 sites were missed.
The per-tick GPU upload budget check ran before consuming async futures,
so after 1 upload ALL remaining ready results were deferred — including
permanent failures and cache hits that need zero GPU work. Moved the
budget gate after failure/cache-hit processing so only actual uploads
count. Re-queues over-budget results as pending spawns for next frame.
Loop iterated 20 hair geoset lookups for Human Male but the if-body
was empty — the LOG statement that was presumably there was removed
but the loop skeleton was left behind.
Stormwind WMO collision takes 25+ seconds to fully load. The warmup
ground check couldn't detect the WMO floor because collision data
wasn't finalized yet. Player spawned and immediately fell through
the unloaded WMO floor into the terrain below (Dun Morogh).
New approach: suspendGravityFor(10s) after world entry. Gravity is
disabled (Z position frozen) until either:
1. A floor is detected by the collision system (gravity resumes instantly)
2. The 10-second timer expires (gravity resumes as fallback)
This handles the case where WMO collision loads during the first few
seconds of gameplay — the player hovers at spawn Z until the floor
appears, then lands normally.
Also fixes faction language for chat (ORCISH for Horde, COMMON for
Alliance) and adds SMSG_MESSAGECHAT diagnostic logging.
loadOnlineWorldTerrain() was called directly from the worldEntryCallback
inside the packet handler, running the 20s warmup loop synchronously.
This blocked ALL packet processing and froze the game for 20-41 seconds.
Now defers the world reload to pendingWorldEntry_ which is processed on
the next frame, outside the packet handler. Position and camera snap
immediately so the player doesn't drift at the old location.
The /y respawn report was actually a server-initiated teleport (possibly
anti-spam or area trigger) that hit this 41-second blocking path.
Equipment: the first emitOtherPlayerEquipment call fired before any item
queries returned, sending all-zero displayIds that stripped players naked.
Now skips the callback when resolved=0 (waiting for queries). Equipment
only applies once at least one item resolves, preventing the naked flash.
BG announcer: broadened filter to match ALL chat types (not just SYSTEM),
and added more patterns: "BGAnnouncer", "[H: N, A: N]" with spaces.
Also added diagnostic logging in setOnlinePlayerEquipment to trace
displayId counts reaching the renderer.
Action bar hearthstone: the slot was type SPELL (spell 8690) not ITEM.
castSpell sends CMSG_CAST_SPELL which the server rejects for item-use
spells. Now detects item-use spells via getItemIdForSpell() and routes
through useItemById() instead, sending CMSG_USE_ITEM correctly.
Far same-map teleport: hearthstone on the same continent (e.g., Westfall
→ Stormwind on Azeroth) skipped the loading screen, so the player fell
through unloaded terrain. Now triggers a full world reload with loading
screen for teleports > 500 units, with the warmup ground check ensuring
WMO floors are loaded before spawning.
Stormwind players stand on WMO floors ~95m above terrain. The previous
check only tested if terrain existed at the spawn XY (it did — far below).
Now checks WMO floor first, then terrain, requiring the ground to be within
15 units of spawn Z. Falls back to tile count after 10s.
Also adds diagnostic logging for useItemBySlot (hearthstone debug).
Stormwind players stand on WMO floors, not terrain. The terrain-only
check passed immediately (terrain exists below the city) but the WMO
floor hadn't loaded yet, so the player fell through.
Now checks three ground sources in order:
1. Terrain height at spawn point
2. WMO floor height at spawn point (for cities/buildings)
3. After 8s, accepts if 4+ terrain tiles are loaded (fallback)
Won't exit warmup until at least one ground source returns valid height,
or the 25s hard cap is reached.
The terrain readiness check was using getCharacterPosition() which is
(0,0,0) during warmup — always returned a valid height and exited
immediately, causing the player to spawn before terrain loaded.
Now uses the server-provided spawn coordinates (x,y,z from world entry)
converted to render coords for the terrain query. Also logs when terrain
isn't ready after 5 seconds to show warmup progress.
Player spawn callbacks and equipment re-emit chain confirmed working.
Added terrain readiness check to the warmup exit condition: the loading
screen won't drop until getHeightAt(playerPos) returns a valid height,
ensuring the ground exists under the player's feet before spawning.
Also increased warmup hard cap from 15s to 25s to give terrain more time
to load in cities like Stormwind with dense WMO/M2 assets.
Equipment re-emit chain confirmed working: items resolve 3-4 seconds
after spawn and equipment is re-applied with valid displayIds.
NPC and other-player melee swing callback was hardcoded to animation 16
(unarmed attack). Now tries 17 (1H weapon), 18 (2H weapon) first with
hasAnimation() check, falling back to 16 if neither exists on the model.
Inspect: CMSG_INSPECT was writing full uint64 GUID instead of packed GUID.
Server silently rejected the malformed packet. Fixed both InspectPacket and
QueryInspectAchievementsPacket to use writePackedGuid().
Follow: was a no-op (only stored GUID). Added client-side auto-follow system:
camera controller walks toward followed entity, faces target, cancels on
WASD/mouse input, stops within 3 units, cancels at 40+ units distance.
Party commands:
- /lootmethod (ffa/roundrobin/master/group/nbg) sends CMSG_LOOT_METHOD
- /lootthreshold (0-5 or quality name) sets minimum loot quality
- /raidconvert converts party to raid (leader only)
Equipment diagnostic logging still active for debugging naked players.
Shoulder pieces are M2 model attachments (like helmets), not body geosets.
Load left shoulder at attachment point 5, right shoulder at point 6.
Models resolved from ItemDisplayInfo.dbc LeftModel/RightModel fields,
with race/gender suffix variants tried first. Applied to both online
player and NPC equipment paths.
Other players previously appeared partially naked — only chest, legs, feet,
hands, cape, and tabard rendered. Now renders full equipment:
- Helmet M2 model: loads from ItemDisplayInfo.dbc with race/gender suffix,
attaches at head bone (point 0/11), hides hair geoset under helm
- Weapons: mainhand (attachment 1) and offhand (attachment 2) M2 models
loaded from ItemDisplayInfo, with Weapon/Shield path fallback
- Wrist/bracer geoset (group 8): applies when no chest sleeve overrides
- Belt/waist geoset (group 18): reads GeosetGroup1 from ItemDisplayInfo
- Shoulder M2 attachments deferred (separate bone attachment system)
Also applied same wrist/waist geosets to NPC and character preview paths.
Minimap: batch 9 individual vkUpdateDescriptorSets into single call.
