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>
After serviceType the reply now sends major, minor, patch, a flags byte
(build kind in bits 0-1, dirty in bit 2), the first 32 bits of the commit
hash and a u16 length-prefixed build string, instead of the stale fixed
ASCII "0.1.3". unknown stays "DLU3". The 1.10.64 client reads only
netVersion and serviceType and never checks the length, so the extra
bytes are ignored. The build string is optional on read.
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 client reads RebuildComponent.activityID over a level's activityID
unless the level sets actIDovrd, but live followed the level's: the AM
Center draw bridge (LOT 12047) has activityID 12047 without actIDovrd and
dropped no loot in 4 of 4 live completions, where the template's activity 6
would have dropped quickbuild rewards. DLU already does what live did.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Live GF Large Crates (LOT 1859, path CrateMast) and FV small white shrines
(LOT 3141, path gate_statue_quickbuild) dropped 1-point powerups outside
their template matrices, from the path's smashable_loot_matrix=1:29 with
smashable_loot_matrix_set=7:1. DLU already carries a spawner path's
waypoint config into the spawned entity's settings; this pins it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A quickbuild's HasItem preconditions took their items only when the build
completed and never gave them back. Live took them when the build started
and gave them back when it was cancelled: the FV Stone Warrior pedestal
(LOT 8551, precondition 99: 5 of LOT 6194) took the items at Building 5
times and added them back with the loot source Quickbuild on each of the 3
cancels in the live captures.
Preconditions now report their item costs instead of removing items while
checking. A quickbuild takes them at the start, gives them back on cancel
or a reset during the build, keeps them on completion, and gives them back
when the builder leaves the world mid-build.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Live sent AddItemToInventoryClientSync with loot_type_source Pickup for the
items of opened packages (token bags, surprise packs, the imaginite bag;
all 98 items added after a live UseNonEquipmentItem). DLU used Consumption.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Live sent quickbuild completion loot (about 190 samples) and the dragon,
BONS, spider queen and Frakjaw chest and wishing well rewards with the
player who earned them as the DropClientLoot source and owner,
use_position true and the spawn position at the object. DLU used the object
as the source, so use_position was false.
DropActivityLoot now uses the player as the source by default. The growing
flowers and the VE mission console have no live samples and keep the
object as the source.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ActivityRewards coins were always read from CurrencyTable npcminlevel 1.
Live used the reward row's ChallengeRating as the npcminlevel, and level 1
when the currency index has no row for that level:
- FV foot races (ChallengeRating 4, index 1) gave 36 and 48 coins, which
only level 4 (30-50) fits; level 1 is 3-5.
- Frakjaw's chest (activity 58, ChallengeRating 6) gave 250 each to a team
of 2: its currency indices 123-126 only have a level 6 row (500), so DLU
gave nothing.
- Quickbuilds, wishing wells and chests have ChallengeRating 1; survival
and the shooting galleries have ratings with no row and keep level 1.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The booty dig chest (LOT 14590) is the one live drop that sent its items
before its 75 coins (5 of 5 live samples).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DLU sent a DropClientLoot for 0 coins with every drop whose coin range was
0, such as item-only smashables and scripted drops. None of the 3163 live
currency drops was for 0 coins.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Live sent a drop's currency DropClientLoot before its item drops in 1498 of
1502 live drop groups that had both. DLU sent the coins last.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
docs/UgcServer.md: embree-gpu in the processing options, the build option
DLU_EMBREE_SYCL (a SYCL compiler: oneAPI's icpx or the open source DPC++), the
library it builds and loads, the run time it needs, the GPUs it supports and
embree_gpu_device; which backend covers which hardware.
Check: the section against the settings page.
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>
docs/UgcServer.md: "Hidden faces" describes the renders from 42 directions
(hsr_resolution) and what they remove that LU Toolbox keeps, and why the path
traced version was removed; the parity table says so. "Processing options" has
ray_backend (embree, hiprt) and denoise, how options stored with builtin,
toolbox or fast are read, and what the tests compare the backends with.
Check: the sections match the settings page and the UGC page.
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>
docs/UgcServer.md gets "Processing options": the three settings and their
choices, how they compare with builtin/toolbox/off, the optional builds
(DLU_OIDN, DLU_HIPRT) and what they need, what each make records and where the
dashboard shows it; the migration, /reprocessproperty's options, the command
line's and the UGC page's buttons.
Check: the section reads right against the settings page and the UGC page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
HIPRT (MIT) behind the CMake option DLU_HIPRT (off): its headers come from its
SDK (HIPRT_ROOT, else ROCm's /opt/rocm) and are copied next to the servers; its
library is loaded when first used (hiprtew), as HIP or CUDA are by Orochi
(MIT, fetched pinned by hash; CUDA when its toolkit is found). The trace
kernels (nearest hit skipping the triangle a ray leaves, any hit) are compiled
the first time and kept in cache/hiprt. One GPU context for the process
(hiprt_device picks the GPU); the workers take turns on it. When HIPRT, the
GPU or a scene's upload fails, Embree is used instead, and the UGC server logs
why at start.
For a GPU the rays go in batches (UgcRays::Scene gets batch queries; the CPU
backends answer them a ray at a time):
- hidden faces: with a batch backend the paths are traced side by side, a
bounce at a time (the path code split into Start, Scatter and Bounce, the one
by one tracing unchanged); the same paths with the same random numbers, so
the same triangles are decided (tested with builtin side by side)
- the occlusion bake and the denoised icons' traced occlusion always ask in
batches (the same rays, the same results)
Its symbols are hidden: the servers export theirs (-rdynamic), and HIPRT's
library, which has an Orochi of its own, would otherwise call ours.
Check: configure with -DDLU_HIPRT=ON on a machine with ROCm (or HIPRT's SDK)
and an AMD RDNA or NVIDIA GPU; UgcServer --make-model x.lxfml out hiprt; the
UGC tests (hits, hidden faces and occlusion against builtin); a build without
it leaves everything as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Intel Open Image Denoise 2 (Apache-2.0) behind the CMake option DLU_OIDN (off):
an installed OIDN is used when found, else Intel's release package (pinned by
hash) is downloaded and its libraries copied next to the servers.
A denoiser only removes noise that differs from pixel to pixel; the icons'
occlusion comes from the bake, per vertex, which it leaves as it is (checked:
white noise 0.17 -> 0.006 relative spread, per-vertex blocks unchanged). So
with denoise=oidn a model's icon is drawn from model.noao.nif (its colors
before the bake) with its occlusion traced per pixel of the supersampled image
(denoise_samples rays, default 4, with the bake's distance and strength and the
ray backend), box filtered and denoised at the icon's size, guided by the colors
and normals. The model keeps its baked occlusion. Icons drawn again from stored
files use the stored model.noao.nif the same way.
OIDN works on a thread of its own; its time is charged to the job's thread
(UgcThrottle::Charge), so the CPU budget and the recorded CPU time include it.
Check: configure with -DDLU_OIDN=ON; UgcServer --make-model x.lxfml out oidn
and compare its icon.png with one made with off; an OFF build leaves icons as
they were.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The UGC page's Make again buttons (a model's, the failed ones, everything) use
the ray tracer, hidden-face method and denoising picked next to them (the
settings' when left alone). The List view's new Options column shows what made
each model and what its next make will use; "Processing options compared"
shows every combination the made models used with its makes, models and the
average time, CPU time, hidden faces', occlusion's and icon's time and share of
triangles removed per make.
Check: the three selects next to Make again (models only); make a model again
with embree fast and see its Options column and the comparison table.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Staff can make models again with other processing options than the UGC
settings' (ray backend, hidden-face method, denoising) and every make records
what made it, for comparing the options.
- migration mysql 99 / sqlite 82 (ugc_process_options): ugc.process_options
(picked for the next make, cleared once made), ugc.made_options (what made
the current files) and ugc_process_runs (every successful make: options,
wall and CPU time, hidden faces', occlusion's and icon's time, bricks,
triangles before and after)
- IUgc: ResetUgcModelProcessing and ResetPropertyUgcModelProcessing take the
options; PendingModel carries them; RecordUgcModelRun, GetUgcRunSummaries;
list entries have madeOptions and processOptions
- the UGC server applies a job's options over its settings and records the run
- /api/ugc/reprocess takes options ("embree fast oidn"); /api/ugc/options
lists the choices, the settings' defaults and averages per combination
- /reprocessproperty [builtin|embree|hiprt] [toolbox|fast] [off|oidn], any
order, all optional
Check: run the migration on MySQL and SQLite; /reprocessproperty embree fast
on a property, then the models' made_options and ugc_process_runs rows;
/api/ugc/options.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
New UGC settings (ugcconfig.ini and the dashboard's settings page):
- ray_backend: builtin (default), embree or hiprt (falls back to embree when
the build or machine can't); the hidden faces' paths and the occlusion rays
- hsr_method: toolbox (default, LU Toolbox's paths) or fast (the renders from
around the model), and hsr_fast_resolution (1024) for the fast one
- denoise: off (default) or oidn (icons; off until a build has it)
UgcProcessOptions (dCommon/UgcKeys.h) names the choices for everything that
passes them on ("embree fast oidn", any order, left out: the setting's).
UgcJobs::ApplyOptions puts a choice over the settings and MadeWith says what a
make used after fallbacks; a made model's stats.json records it (rays,
hsrMethod, denoise). UgcServer --make-model and --make-modular take the
choices after the folder and print the CPU time and what made it.
Check: the defaults make the same files as before; the settings page shows the
four settings under UGC; UgcServer --make-model model.lxfml out embree fast.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
UgcHsr::Options::method picks how hidden faces are found: toolbox (LU Toolbox's
paths, the default, unchanged) or fast, the test the UGC server used before
"hidden faces removed as LU Toolbox's Remove Hidden Faces decides them",
restored unchanged as UgcRender::VisibleFromAround: the opaque mesh rendered
from 42 directions around the model (Options::fastResolution pixels square,
1024 as before) and the triangles that show in none removed. It is much
faster, but it also removes faces seen only by bounced light (interiors,
recesses), which LU Toolbox keeps. Nothing sets it yet.
Check: the UGC tests (the fast method removes a box seen only through a
chimney, which the paths keep; its files are the same every time).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
UgcRays::Scene answers the two ray queries the UGC server makes: the nearest
hit for the hidden faces' paths (never the triangle a path leaves) and any hit
for the ambient occlusion rays. Two backends:
- builtin: the two hierarchies the queries had before, moved unchanged (each
built the first time it is asked), so the files made are the same bytes
- embree: Embree 4 on the job's thread (a device per worker thread, no threads
of its own, so its time counts in the CPU budget), watertight, the skipped
triangle filtered out
UgcHsr::Options::rays and UgcRender::AoOptions::rays pick the backend; both
default to builtin, and nothing sets them yet.
Check: the UGC tests (builtin's files keep their hashes; UgcRays tests compare
embree's hits, hidden faces and occlusion with builtin's).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Embree 4.4.0 (Apache-2.0) is fetched like glm and curl and built as a static
library: its own task scheduler (no TBB), triangles only, single rays, no ISPC,
SYCL, tutorials or tests; SSE2, AVX and AVX2 kernels picked at run time. Its
Debug build keeps its optimizations but not its assertions. Nothing uses it yet.
Check: a clean configure fetches embree and the build links (Linux and Windows);
the extra build time.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Item::RemoveFromInventory ended with `delete this`, so every caller that still used the item afterwards (SetCount(0)
returning into its caller, loops that remove several items, proxies purged while their parent is handled) touched
freed memory. The inventory now takes the removed item and frees it at the inventory component's next update, or
with the inventory.
Check in game: sell, drop, delete, trade, mail and use up stacks (including the last of a stack); unequip and
remove an item set piece and a proxy-bearing item (rocket, modular car); donate items; nothing crashes and the
inventory shows the right counts after relogging.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The client's TacArcBehavior::DoHit (0x00fb10c0) writes the closest max-targets ids from a set, so ascending and
each once, then the action data per id in that same order; DoUnserializeBS (0x00fb26a0) reads them the same way and
skips empty ids. The server handled the targets in list order and skipped ids whose object it could not find
without reading their action data, so every later target read the wrong bits (several pirates under a Doom Slicer).
Handle now reads the action for every listed id in ascending order, and the server's own casts write ids and actions
in ascending order after picking the closest targets.
TacArcBehavior::Cast (0x00fb2d10): a picked target that passes the filter gets the action with no TacArc data (the
server's own cast now calculates that action instead of handling it); otherwise the target is dropped, and an arc
measured from the target's position writes nothing.
Check in game: Doom Slicer and multi-target katanas on groups of pirates/admirals damage each of them; apes still
take damage during their stun; enemies with arc attacks (apes, Maelstrom horsemen) still hit players.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The live boss script's rain of fire takes every target of the first ROF target group and ROFImpactCnt (2) different
random targets of each other group. The server took one per group, so the rain was sparser than live. The impacts are
now picked like the live script (without repeats inside a group).
Check in game: AG Spider Queen stage 3; the rain of fire lands on the centre ring and on two spots in each outer ring.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The live boss script tracks which arena zone volume (Zone1Vol..Zone8Vol, TeleVol for the default Zone3Vol) each
player last entered, and its rapid fire picks a random player, takes the three RFS target groups around that zone,
sorts each by the targets' CWOrder (CWOrder2 when the sweep crosses between zones 8 and 1), clockwise or
counter-clockwise at random, drops the first and last target of the middle group (shared with its neighbours),
turns with skill 1480 at the fourth target and fires 1394 at every target in turn, playing attack-shoot-right or
attack-shoot-left. The server shot a single random target with the single-shot animation, and the AG property zone
never subscribed the boss to the zone volumes. The zone now registers the volumes (retrying until they are spawned)
and the boss builds the sweep like the live script.
Check in game: AG Spider Queen stage 2; the rapid fire is an arc of many shots sweeping across the arena near the
player, left or right, and follows the player to other parts of the arena; after teleporting it starts from the
default side.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every hit started both the rapid fire shooter and the rain of fire timers, from stage 1 on, so the two specials
ran in every stage and on top of each other, each turning the boss's AI off and on again under the other's
animations. They also wrote the "stoppedFlag" that the no-players-around attack stop uses, which could leave her
stopped for good. As the live script: after she comes back down, a skill manager fires the rapid fire shooter in
stage 2 and the rain of fire in stage 3 only, again 10 to 15 s after each ends; for 3.1 s after her melee smash
(skill 322) a due special waits and fires when the smash ends. The rain of fire keeps her from attacking until
its last impact.
Check in game: Spider Queen fight: no specials before the first spiderling wave; stage 2 only rapid fire, stage 3
only rain of fire; she never freezes in the smash animation and keeps attacking after each special.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The zone script never handed the boss her landing target and scream emitter: ZoneAgProperty::ProcessGroupObjects
was empty and nothing answered the boss client script's "QueryZoneScript" event. As the live scripts: the boss asks
the zone ("RetrieveZoneData"), the zone stores the first object of Land_Target and Spider_Scream on her as
LandingTarget and ScreamEmitter (looking again every 0.3 s until they are spawned), and each spiderling death sends
NotifyClientObject "EmitScream" with the emitter, which the boss's client script plays as the scream. The landing
skill and camera shake now come from the landing target, not the boss.
Check in game: AG Spider Queen (property or instance): kill a spiderling: the scream plays from the mountain; when
she comes back down, the landing blast hits around the landing spot and the camera shakes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The client's AreaOfEffectBehavior::Cast (0x004ec590) writes every target id, then runs and writes the action once
per unique target in ascending id order. The server ran it for every listed target in list order, so a target
listed twice (the caster with a magnet, Everlasting items' refill) was handled twice, and later targets' data was
read against the wrong target. Handle and the server's own Calculate now go through the unique ids in ascending
order. Check: Thumpin' Bass / Flowin' MC refill once; Shinobi charge with a magnet gives imagination once; area
attacks on several enemies still hit each of them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
WBL_Frog_Bridge had no script. As the live script: when a player who finished mission 946 comes within 25 units,
the four FrogBridge pieces move out to form the bridge; after 10 seconds they go back one a second from the tip,
and 7 seconds later the frog looks for players again. Check: in Portabello, with mission 946 done, walk up to the
frog: the tongue bridge comes out and goes back; without the mission it doesn't. Depends on moving platforms
following their paths.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
WBL_Starflies had no script. As the live script: when the mission given at that object is handed in, every object of
the Starflies group starts its path. Check: in Portabello, hand in the starflies mission: the starflies fly off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
L_ACT_HORSEMEN_1 (one path spawner of Maelstrom Cavalry, LOT 7816, in Forbidden Valley) had no script. It does what
L_FV_MAELSTROM_CAVALRY does (tell the group's turret it spawned, report a Brick Fury kill to the horsemen trigger),
so it uses that port. Check: the cavalry on that path wakes the ninja turret and Brick Fury kills count.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
L_TURRET (LOTs 12286 and 13093, e.g. the Crux Prime auto turrets) had no script, so these turrets could be stunned,
knocked back and pulled. They now push the same immunities the live script does. Check: stun and knock back a
Crux Prime auto turret: nothing happens.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The turret (LOT 9303-9305, 12618) had no script: it fought as soon as it was spawned and lasted the generic 60
seconds. As TURRET.lua: its combat AI is off until it is built, it can't be stunned, interrupted, knocked back or
pulled, and it dies 30 seconds after it was placed once nobody is building it. Kill credit going to the builder is
not done. Check: summon the turret with Engineer gear: it doesn't shoot until built, then fights and goes away
after about 30 seconds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Master stops with a message naming client_location, its value and the folder it resolves to when that folder
doesn't exist, instead of failing later on missing client files; a missing dump_folder gets a warning. Check: set
client_location to a folder that doesn't exist: master says so and stops.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The WSL steps used software-properties-common, lsb-release and the deprecated apt-key. They now follow Kitware's
current instructions: the key as a keyring referenced with signed-by, the release codename from /etc/os-release.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
RepositionPlayer added the offsets in world X/Z, set the height to 0 and built the keyboard turn around the position
vector. As l_ns_concert_instrument_qb.lua: the offset is in the instrument's own frame, the instrument's height is
kept, the player faces the instrument's way, and at the keyboard turns -0.8 further. Check: build and play all four
instruments on the Nimbus Station stage: the player stands at each one facing it, not below the stage.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
