The master server dropped a transfer request it started no instance
for, so the world server never heard back and the player waited. It
now answers the request with an empty address and the reason in place
of the port: no room for another instance (max_instances), a zone that
is not in the ZoneTable, or another reason. A request for a private
zone that does not exist is answered the same way.
The world server passes this on as a TRANSFER_TO_WORLD without an
address, which the client shows as a failed map change, and leaves the
character where it is. The auth server fails the login when there is
no character select to send the player to.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The key was a 32 bit number, sent to the client as the MD5 hash of its
decimal text. The client only stores the key and sends it back, so it
is now 16 bytes from the system entropy source written as 32 lowercase
hex characters, which fills the same 33 character field of the login
response.
The auth, master and world servers pass the key itself between them as
a 33 character string in place of the number: SET_SESSION_KEY,
SESSION_KEY_RESPONSE and NEW_SESSION_ALERT keep their IDs and carry the
string where the number was. The master server keeps one key for each
username, and a world server compares the key a client sends with the
one from the master server in constant time. A key stays valid until
the next login of its account replaces it, as before.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Several game messages read a 32 bit length and then read that many
characters without checking it, so the loop ran for the whole length
even after the message had ended. The length now has to fit in the
data the message has left. This covers the activity state, cinematic
path, death type, removed item info and inventory group messages and
the shared wide string reader.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
An sd0 buffer could hold any number of chunks, and each chunk grew the
inflated data by a full chunk whatever it held, so the data of a chunk
that was not full was followed by a gap. A buffer now holds at most
4096 chunks and inflates to at most 64 MiB, and every chunk is inflated
right after the one before it.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The AMF3 reader read an array inside an array to any depth and kept
reading values after the data had ended. It now refuses arrays nested
more than 64 deep and throws when the data ends before a value does.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The key came from the shared game random engine, which is seeded from
the start time. It now comes from std::random_device through a small
helper. The key stays a 32 bit number, so the packets do not change.
The master and world servers no longer write the key or its hash to
the log.
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
When a brick by brick save is split into models, each model was centred on
the middle of its bones' origins, on the lowest origin. A brick's origin is at
a corner of it, not its middle, so a single 2x2 came back offset by half a
brick, and the model's floor was taken from the origins rather than from the
bricks.
Now each model is placed as the live server placed it: at the bottom middle
of the box around its bricks. Each brick's own box comes from the client's
brickdb by its design ID and is turned and moved by its bone (all 8 corners),
and the model's box holds every brick's. Its x and z middle and its lowest y
are the new origin, still snapped to the 0.8 grid on x and z, and every bone
moves by that; rigids and rotations are untouched. A design the brickdb has no
box for counts its bones' origins, as before, so no brick is left out.
Lxfml::NormalizePosition and Split take the boxes by design ID, and the brick
by brick save passes the ones read from the client (read on the first save,
then kept).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
BrickBounds reads the <Bounding><AABB/> of every Primitives/<designID>.xml in
the client's res/brickdb.zip (through the asset manager, so packed and
unpacked clients both work), by design ID. A design listed only as another
primitive's alias gets that primitive's box. Get() reads the file the first
time it is called and keeps the boxes for every later call.
The zip is read by a small reader in dCommon (Zip::ForEachEntry) that handles
stored and deflated entries, with a raw inflate added to ZCompression.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Lxfml::NormalizePosition moved only the first bone of each part, so after a
save a flexible part's other bones kept their world positions while the rest
of the model was moved to its own origin. Now every Bone moves by the same
amount, and the box that places the model holds every bone. Split maps every
bone of a part too, so a rigid system holding only a later bone keeps its
brick.
Only bones move, as the live server did: the Rigid transformations are sent
back as the client wrote them, and rotations are not touched.
Also:
- a model with no readable bone is kept as it is at the origin instead of
getting a center 10000 up
- the math is done in doubles and numbers are written as the shortest text
that reads back as the same float (-0.4, not -0.400002; 12.1678, not
12.1677)
- bricks and rigid systems come out in the file's order, so the same model
always splits into the same bytes
- parts over 5 MB are normalized like any other (the input is capped at
10 MB), and the loop's safety limit no longer drops every later model
The pivot is still snapped to the 0.8 grid (the bricks don't move in the
world; the model's position stays on the grid), which can put it half a
stud from the middle.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
LwoNameValue::FromString (0x010fa220) splits a config at every comma
and line break and leaves out entries without a '='; the world split
only at line breaks. Objects and sounds are now split the client's way.
ReadLvlObjectData (0x0103ba20) also adjusts an object's config as it
loads it, and now so does the world:
- before version 47 a respawn time was in milliseconds: a float over
100, or any u32, becomes seconds (a float);
- a model spawner (LOT 176 spawning template 14) gets its blueprint
from its subkey unless it has one, behaviors off unless it says
otherwise, and no render wrapping;
- before version 44 an object that overrides its scene goes to layer 0
of the scene it names.
Check in game: every world spawns its objects as before; spawner
respawn times are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Before version 46 the u16 after a particle's effect names is the names'
wide null terminator, 0 in every shipped scene file. It is skipped
instead of kept as an unknown value.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Files from before chunks now read their sounds and particles after the
objects, and chunked files their particle chunk:
- sounds (SceneLoader::LoadSounds, version 5 on, only before chunks):
position and config (before 35 a 260 character name, kept as
soundName);
- particles (SceneLoader::LoadParticles, version 5 on): priority (43
on), position, rotation, effect names and config (46 on; before, an
unused u16); the client stops at one with no effect, and before 44 a
file may end before them.
Every scene file of every client on disk reads each part to the next
part's start (up to the client's 16 byte padding): 5408 referenced by
zone files and 5125 on disk, versions 31-43 before chunks and 44-49
chunked. One client file, a LUP world, ends before its sounds; its
objects are kept.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The world read only a scene's objects and skipped the rest with
file.ignore. Level now reads a .lvl the way SceneLoader::LoadLvlFile
does (1.10.64) and keeps what it reads.
Chunked files: the file info chunk gives where the environment, object
and particle chunks start (the chunks were walked in order before); the
environment chunk gives where the lighting, skydome and editor settings
start, and the client reads all three only when there is lighting.
Files from before chunks: the no-editor-settings byte only from version
32, the revision from 37, a file newer than 48 is not read, a file
older than 3 has no objects; then lighting, skydome, editor settings
and objects in turn.
Kept, by version:
- lighting: blend time (45 on, default 10), ambient, specular, upper
hemisphere, position, min and max draw distances (31 on; before 39
fog near and far only), cull settings (40 on), fog colour (31 on),
directional colour (36 on), and before 42 the zone's spawn point
(before 33 the fixed one the client uses);
- the skydome and its five other files (34 on; they were skipped from
33);
- the editor's colours (37 on, the first 32);
- each object's node type (38 on), glom ID (32 on) and render
techniques (7 on: a 64 byte name, then per technique a 64 byte name,
u32, u8 and 16 floats); before, a count of techniques was read and
the techniques themselves were not, so an object with any misread
the rest of the file.
Object config is split at line breaks as before. A file that ends early
keeps the objects read before its end.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
LuzFile::ReadLUZFile copies the path chunk by its size and ReadLUZPaths
reads it with LevelPath::FromBuffer (1.10.64). Reading as they do:
- the paths are read from their chunk, never past its end;
- a chunk version of 2 or more, 10000 paths or more, or a path the
client refuses (version 19 or more, 10000 waypoints or more, more
than 99 name/value pairs on a waypoint) leaves the zone with no
paths, logged; the rest of the zone file still reads;
- before version 3 a path's type is a name ("platform" or a movement
path) and every waypoint has a rotation, the moving platform values
and a name/value list, whatever its type;
- property headers are there from version 3, spawner and camera
headers from version 4;
- a camera waypoint's FOV before version 11 is divided by 1.25;
- a property path's reputation multiplier outside 0-1000 becomes 0.
Every path chunk of every client's zone files on disk (858) reads to
its end.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ZoneBoundary's destinationZone and destinationSceneID and the scene's NiColor, so main's scene tools and tests read
the zone this branch's reader loads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The dashboard's copy of the terrain reader in dCommon is gone; main's in dZoneManager reads for everyone, and keeps
what the copy had over it: colour and texture maps from before version 32 become RGBA, cut to the resolution the
client keeps. The scene map (ZoneScenes) and the dCommon tests compile against that reader.
A scene ID takes its layer as the zone file numbers it (0 General, 1 Audio) as well as the enum, so main's scene tools
and tests build; /getscene prints the layer as a number.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Main's model behaviors (every rotation, DoDamage, OnTimer, OnStartup, MoveBackToStart, moves along the model's
local axes), raw terrain reading with scene lookups and scene tools, property model and brick by brick fixes and the
chat server's player count fix, merged onto this branch's structures:
- game messages stay structs; the angular velocity messages join them and the handler sets a message's target
- strips own their speed again and the block names, moves and rotations stay in the shared action table
(PropertyBehaviorActions, which the dashboard's behavior player reads; it now lists rotations too)
- the zone keeps reading its file through ZoneFile and takes the boundaries from it, then loads the raw terrain as main
does; the old dNavigation terrain reader is gone
- the show all response stays a struct and logs a player count mismatch as main does
- the property management component tells models who sent a chat message
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat: raw terrain parsing for scene data
Replace old dNavigation/dTerrain raw parser with new Raw module in
dZoneManager. Parse heightmaps, color maps, and scene maps from .raw
files to determine which scene a position belongs to. Build scene
adjacency graph from terrain data and scene transitions.
Adds NiColor type, SceneColor lookup table, eSceneType enum, terrain
mesh generation with OBJ export, and debug slash commands for scene
visualization.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* refactor: deduplicate terrain grid lookup logic, remove unnecessary cast
Add IsValidForSceneLookup(), GetSceneIDAtGrid(), and GridToWorldPos()
to Raw::Chunk. Use them in GenerateTerrainMesh, GetSceneIDFromPosition,
SpawnScenePoints, and SpawnAllScenePoints instead of duplicated inline
math. Remove unnecessary static_cast in LoadSceneTransitionInfo.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: add version < 32 support to raw terrain parser
The parser was only correct for v32 files. For v<32:
- colorMapResolution is width-1, not width
- Color map reads width*width*4 bytes (BGRA per-pixel), not colorMapRes^2*4
- No DDS lightmap
- No texture settings byte or blend DDS
- Scene map: v31 reads (colorMapRes+1)^2 cells with border skip,
v<31 skips 1 byte
- No mesh data
Verified against all 309 .raw files across multiple client versions.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: log and reject legacy raw terrain format, clean up ReadRaw API
Log a clear message and throw if the zone uses a legacy raw format
(LUZ version < 30) since no known maps use it. Remove ReadLegacyRaw
and the mapVersion parameter from ReadRaw.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: revert v33 scene field renames to unknown1/unknown2
These fields are skipped by the client in every known build (ptr += 12
and ptr += 4). No client ever reads them into a struct. Keep the
original unknown1/unknown2 names from main.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix: load the zone without scene lookups when its raw terrain is missing or unreadable
Log at debug when it parses; build the terrain mesh only for the OBJ export; read the zone name and description only for version 31 and later, as the client does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix: look up the scene at a position as the client does, in constant time
Chunk bounds span (width - 1) * scale; the nearest scene map cell is used and the far edge is kept one cell inside. Chunks are indexed by grid cell when they form a grid. Skipping reads instead of seeking, so a cut-off file can't seek past the end. Terrain mesh indices follow the vertex order.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat: scene point commands take a step and use brick LOTs from the BrickIDTable
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore: rename NiColor::ToXYZ to Luminance
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test: raw terrain reading, scene lookup, scene graph and scene commands
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test: close the OBJ file before removing it so the test passes on Windows
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix: keep the zone boundary lines and read the luz scene count as the client does
Each boundary line's destination is read as a map ID and an instance ID into a LWOZONEID. The scene count is a byte before version 37 and a u32 from it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix: name the fourth flair colour byte padding
The client never sets or reads it, so it holds leftover memory.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix: keep the scene map of version 31 and older raws
These files have a (resolution + 1)^2 scene map, kept in sceneMapLegacy. Version 31 uses its first resolution rows and columns; older versions are all scene 0, as in the client.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* refactor: simplify the raw heightmap size check
The dimensions are capped at kMaxResolution, so the product can't overflow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore: remove the scene point spawn commands and send /getadjacentscenes as one message
The spawn commands were only for testing. /getscene and /getadjacentscenes stay.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* add rotation behavior logic (not working correctly for multiple axes)
* Fix FetchContent_Declare speed
* remove manual quaternion math
* feat: working rotations
* remove flag
* feat: make directional movements based on model rotation and not world axes
* poggers rotation on the strip
* Add OnStartup and OnTimer
* go home works
* Add DoDamage behavior and dont try and start the same node multiple times
* fix: pause bbb models
* interrupt rotations and save interrupt count across resets
Appended message ID 50 (dashboard -> master): the update check's state and a
one-line summary. Master logs it whenever the summary changes, so the first
one after master starts says when an update is out.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
wchar_t is 16-bit on Windows and 32-bit elsewhere; the game's wide strings
are UTF-16 and already split through the char16_t overload.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CMake refuses to configure when pointers aren't 8 bytes, and dCommonVars.h
asserts it, so sent and stored sizes can't silently change width.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MATCH_TRANSFER is appended after GUILD_DISBAND (71); the chat ids from
CREATE_TEAM on are pinned.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Appended at the end of MessageType::Master (47-49); the pin test is
updated. A world sends its zone file list to master once it is ready.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ZoneFileLog keeps the .luz, .lvl, .lutriggers, terrain .raw and navmesh a
world read while loading its zone: absolute path of the loose file (or
the name in the pack), size and an FNV-1a hash of the bytes read.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Worlds, the UGC server and the dashboard open the pair named in
resServer/cdclient-current (CDServer.sqlite without an fdb when there is
none). Worlds switch to a new pair on CDCLIENT_RELOAD between frames.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Appended at the end of MessageType::Master (46); the pin test is updated.
With names it tells a server which fdb copy and CDServer.sqlite to switch
to; without them it asks master to check the client's fdb now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
So a worker can make a new CDServer.sqlite without touching the shared
CDClient connection or the logger.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
FdbSnapshot names copies cdclient-<hash>.fdb / CDServer-<hash>.sqlite, writes
them under temporary names and renames them in, keeps a pointer file naming
the current pair, removes old copies, and decides when a changed file has
settled enough to hash. Nothing in it logs, so it can run on a worker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each address's requests are counted per signed-in account (the dashboard's
session or API key, as the auth middleware found it), so several people
behind one address stay apart, with the account's user name; a WebSocket
upgrade counts under its account.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The dashboard's own loop is framed too, with a scope per module update. Frame time and stacked phase charts per server, the servers' loop summary, longest frames, packet handling times, the last 50 slow frames with a nested timeline, and profiling sessions (profiling_run, GM 8) drawn as a flame graph with folded stacks to download. PerfHistory keeps it in memory and is unit tested; the layouts are tested with node. Task 96.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An optional section after marker 3 carries the report's frame timing; older readers stop before it and reports without it still read. Phase times go with their count so readers with fewer or more phases read them. PROFILE_REQUEST and PROFILE_RESULT are appended to the master messages (44, 45). Task 96.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Profiler.h: frames, named scopes and phases, recorded on the main thread only (other threads' scopes do nothing). Per second: frames, total and longest frame time, a mergeable frame time histogram (TrafficStats::Histogram) and time per phase. Per report: the packet types that took longest, the worst frames and the frames over a threshold with their heaviest scopes. Profiling sessions merge every frame's scope tree for a while into one (folded stacks for flame graphs). Optional Tracy client (DLU_TRACY, off by default) gets the same frames and scopes. Task 96.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Locale::LoadFromFile can keep the other locales' phrases too, and
GetPhrase(id, locale) answers in one of them, falling back to the default.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
FdbMappedFile maps a file read-only (CreateFileMapping/MapViewOfFile on
Windows, mmap elsewhere) and falls back to reading it into memory when
mapping fails. FdbReader reads the table and column headers from it and
looks rows up by their first column through the fdb's own buckets,
decoding every integer as little-endian with bounds checks, so the rows
never get copied out of the file.
Tests write small fdb files (collisions, text, int64, nulls) and read
them both mapped and from memory.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
client_sysinfo (GM 5, grantable) shows the client system info; the Log pruning
task deletes rows not seen for log_client_sysinfo_days (90).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The login request's memoryStats is a run of short parts with no separators
(process working set and private bytes, memory load, physical memory, commit
limit, the 32-bit client's own address space, peaks). ParseMemoryStats splits
it into numbers; a text cut short keeps the parts before the cut.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The traffic topic now carries each server's rates with its split by
peer (null for servers that report none), link statistics and gauges.
New routes: /api/diagnostics/network, /network/server (message types
and a 10 minute series) and /network/connections (remote ends grouped
by address). Addresses need the new network_ips permission; without it
each is a salted token. They stay in memory from the last report only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every server now splits its packet counts by peer: its own connections
(players on auth and worlds), its master link, or other servers (the
worlds on chat, every server on master, a world's chat link, which is
now counted too). HTTP requests carrying X-Darkflame-Server count as
another server's, the dashboard counts its own requests to the UGC
server, and each HTTP client address is counted. The report also gets
each RakNet connection's statistics (worlds name the player on it),
trimmed to the 32 busiest with the rest summed. Counting stays on the
main loop, except the dashboard's UGC fetches, which only touch the
locked recorder.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
chat_dms (GM 8) reads whispers; chat_private is now team and guild chat only.
Adds chat_flag and chat_flag_review (GM 3). Both can be granted per account or
character with the permission grants.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The client's render component sets the object's own position and
rotation on the root node it loads, over the stored ones, so a root's
turn never shows in game (its scale stays). NifFile now reads roots the
same way. LU Toolbox's .nif (root turned 90 degrees about X, the
NiLODNodes turned back) now stands up in icons and the dashboard's
views; every game .nif has an unturned root and reads as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
What makes a model's files becomes a processing option like ray_backend
and denoise: native (the UGC server) or toolbox-blender (LU Toolbox in
Blender). Old stored options parse as before; the retired hidden-face
word toolbox is still skipped and is not toolbox-blender.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every server logs BuildInfo's build string as its version; the About page
also shows the build kind and commit, and log bundles name the build.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
cmake/BuildInfo.cmake runs on every build and rewrites the generated
BuildInfo.cpp only when something changed, so a new commit recompiles one
file and relinks. The build kind comes from DLU_BUILD_KIND
(release/ci/local) or CI=true, defaulting to local.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The third ray backend, so every machine has a library for it: Embree on x86
CPUs (embree), HIPRT on AMD and NVIDIA GPUs (hiprt), Embree through SYCL on
Intel Arc and Xe GPUs (embree-gpu).
- CMake option DLU_EMBREE_SYCL (off). dUgcServer/EmbreeSycl is a project of
its own built by a SYCL compiler (DLU_SYCL_CXX; icpx through ONEAPI_ROOT or
the path, or the open source DPC++'s clang++ through DPCPP_ROOT) as an
external project: Embree 4.4 with EMBREE_SYCL_SUPPORT, linked statically and
bound inside (-Bsymbolic, only its C functions exported, so it never meets
the servers' own Embree), and the GPU kernels (nearest hit skipping a ray's
triangle, any hit), into libdlu_embree_sycl next to the servers
- UgcRaysEmbreeGpu loads it the first time embree-gpu is asked for; one GPU for
the process (embree_gpu_device picks it), the workers take turns, the
occlusion rays in batches as for hiprt
- without the build, the library or a supported Intel GPU it falls back to
embree and says why (the UGC server's log at start, --make-model on stderr)
- the option names, the settings page, the dashboard's picker, /reprocessproperty
Verified here: the default build and ctest; the SYCL build with the open source
DPC++ 7.1.0 (compiles, links against oneAPI's libsycl.so.9, exports only its C
functions); on this machine (no Intel GPU) it loads, finds no GPU and falls
back to embree. Not verified: tracing on an Intel GPU (none here).
Check: on a machine with an Intel Arc or Xe GPU and oneAPI, configure with
-DDLU_EMBREE_SYCL=ON and run UgcServer --make-model x.lxfml out embree-gpu;
the UGC tests then compare it with Embree.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Embree 4 (always built) replaces the UGC server's own bounding volume
hierarchies (the nearest-hit one and the occlusion rays' any-hit one), with no
fallback to them. ray_backend is embree (default) or hiprt (optional build;
Embree when it can't be used).
Settings, stored options and stats that say builtin still read: it is embree
(UgcRays::Parse, UgcProcessOptions::Parse). The dashboard's picker,
/reprocessproperty and --make-model offer embree and hiprt.
Tests: the backends are compared with Embree (hiprt when built), Embree against
rays whose hits are known, and the clutter's occlusion against what the old
hierarchy worked out (296 vertices summing to 114.5, 26 open, 183 dark); the
pinned model hashes are unchanged with Embree.
Check: ray_backend=builtin in an ini still starts and uses embree; the
settings page offers Embree and HIPRT.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The path tracer that imitated LU Toolbox's Remove Hidden Faces took 17 to 100
times as long as the renders from 42 directions around the model, too heavy to
use. It goes with its settings (hsr_method, hsr_samples, hsr_bounces,
hsr_sample_spacing, hsr_min_points), its side by side tracing for GPUs and its
tests. The renders (UgcRender::VisibleFromAround) remove hidden faces as before
that method existed; their size is hsr_resolution (1024), and the memory
estimate counts their buffers again.
The hidden-face method is no longer a processing option: the dashboard's picker,
/reprocessproperty and --make-model take only the ray backend (the occlusion's)
and denoising. Options stored before (made_options, process_options,
ugc_process_runs, --make-model arguments) that name toolbox or fast still
parse; the word is skipped.
Check: the settings page has no hsr_method or path settings and has
hsr_resolution; /reprocessproperty embree toolbox still works (toolbox
ignored); models made with the defaults keep their hashes (UGC tests).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
