Resolve private effects only from the level-wide Game compartment and validate
their resident declarations with the same parser used by the mod merge.
Release surviving native services during world changes, check decal realm
ownership before cleanup, defer ordinary off-thread releases, and prune expired
VM handles so they return their quota. Size the temporary decal allowance from
the live count while preserving external settings changes.
Gate contact geometry behind events.physics.geometry, document the manifest
migration, and test Blood impacts through the plugin host using its own manifest.
Add regression coverage for residency, lifetime, quota and draw-budget handling.
Validation: Release builds of the runtime and plugin/effect test targets succeed.
The plugins, effect_geometry and effect_lifetime tests pass. Existing MSBuild
asset-directory dependency warnings remain; native visuals need an in-game check.
grind_exit gave way to shape_launch upstream, so Enums.GrindSide becomes
Enums.Launch, the words shape_launch checks; Delete joins Enums.Key.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
skater.pose's places are worked out with the pose's scale in them (keep_children, or a joint placed
relative to another joint or the world under a scaled one), but the others were sent only rotation
and position, so such a joint landed up to a thousand times its distance away on their screens. A
place may also be far off in the parent's frame, which drew a skater's arms and head across the map
for everyone. The sent pose now carries each joint's scale (0.001 to 10), and a joint's offset from
its parent is cut to 4 m in what is sent. What the player sees locally is unchanged.
Since the declarations were read from the applied catalogue, a file that could not be read or
parsed threw out of the whole refresh, so no mod's declared inputs applied (from launch, or after a
live apply). Each file is read on its own again: a broken one is logged with its mod and skipped,
and only a failed read of item ownership keeps the previous set.
Plugins: support declared colour inputs on mod-owned board materials
Conflicts with #363 resolved by keeping both: reconcile_variant_additions beside
active_mod_sources in mod_catalog.h, and both test targets in cmake/Apps.cmake.
Add permission-gated effect handles, native surface attachments, and world lifetime cleanup. Extend physics_step with optional contact geometry and body poses. Document the generic API and validate it with self-contained plugin and geometry fixtures.
So far a plugin could only add to what the game's animation does with a joint. Now it can give the
joint its own transform, which the animation's is replaced by: its own animations, joints reused for
body parts the skater doesn't have, joints moved by a simulation in Luau, joints pinned to a point.
A joint's entry takes place = {position, rotation | angles, relative_to} and keep_children = true
(its children stay where they were); a pose takes up to 128 joints. Each frame the place is turned
into the joint's value under its real parent, the joints posed in an order that puts each after its
posed parents and the joint it is placed relative to; a pose without such an order is refused with
the reason.
Extension/Skater/pose_place.h/.cpp hold the maths, with pose_place_tests; skater_animation applies it
after a joint's offsets; plugin_api_world reads it and sorts the joints.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2.1.2 left such an actor alive and forgotten. When the level went, the game released its parent
through the same pointer, took one off the byte below the parent's vtable pointer and called
through it (Skate.exe+0x2b8092f and +0x2b7fcfb, every report after the "link ... is damaged"
warning with the same bus). A link exactly one less than the parent the bus also holds at +0x48 is
now repaired before the destroy; other damage is still left alive. What shortens the link is not
known yet.
The skater sound state sent the packed material words of the surfaces under a skater as they were.
A Studio map's surfaces are appended to the root grid from slot 116 on, numbered by each install's
own set of map mods, so the same word named another surface on the other side, or a slot past the
end of its grid, and the game's material lookups (0x7111c0, 0x45e120) read past MaterialIndexMap
(Skate.exe+0x41da7b and +0x6fc93b, 240 installs in a day on 55a7342, on Tokyo 2026, Community
Center, Olay Way and other maps with their own surfaces). Those words now cross as the grid's
default, when sent and when received, so older clients sending their own numbers are covered too.
reskate.outfit.worn() is meant to list full-body costumes: it has a "Costume" slot for them, between
the head and the feet. But no costume ever reached it. With a costume on, the list came back without
it, so a plugin that goes with a costume mod could not tell whether its costume was worn. Now it is
listed like any other item: {slot = "Costume", item = "Own_Hobkin", title = "Hobkin"}.
Why: the list leaves out the body's own parts (head shape, nose, skin...) by their catalog categories,
which start with "cust" (custhead, custnoses, custbodies...) or name the body. A full-body costume's
category is "cust_fullBodyCostume", so it matched both and was dropped before its slot was looked up.
A category that names a costume is now kept; everything else is decided as before.
The rule (category to slot) moves out of worn_items.cpp into outfit_kind.h as it was, so it can be
tested without the game: outfit_kind_tests (with the plugin tests) checks clothes by their slots, the
costume, the body's own parts, and an unknown category.
Checked in the game: with a costume mod's costume on, outfit.worn() returns
Costume | Own_Hobkin | Hobkin, where it returned nothing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
skater.pose changes the skater relative to the game's animation. Some of those changes were applied
twice every third frame, so a posed skater shook in a three-frame rhythm, and their board with them.
The more a plugin changed, the worse it got: an avatar with the hips scaled and offsets on most of
the body needed workarounds in its plugin to hold still. Now every pose is applied exactly once, and
a plugin can give skater.pose the values it means.
Why: ReSkate draws a pose by writing it into the game's bone buffer just before the render, and the
game evaluates the next frame from that same buffer without overwriting all of it. Two of three
frames it resets the buffer itself; in the third the pose drawn was still there and the next pose
went on top of it. Now the game's own values go back as soon as the renderer has its copy, so the
game, the body physics and the next animation update only ever see the game's pose.
This replaces own_scale, the earlier fix for the same cause (the Filmer's forearm at 0.001 shrinking
to nothing). It covered only scale, and put it back after the game had already run on the posed
values, so rotations and offsets still doubled. The new way covers every part of every joint, for
pose, carry and reach, with nothing to track per joint and nothing to clean up after a pose ends.
Checked in the game: the avatar with the hips at 0.667 and offsets on 48 joints holds still with
only skater.pose (shoulder distance 0.117 m every frame, where it was 0.117 0.117 0.067 ...). The
Filmer 1.0.0's camcorder preview looks and measures as on main, and the skater is back to normal
after it.
Details: the renderer copies the bones in submit_render_pose (0x4c7e7a0), which the render pose
caller (0x2c20d99) calls right after render_pose with the same render data; found with a guard page
on the bone buffer that logged every access with its call stack. native_skater hooks it and, for the
render data the pose override just wrote, calls the new set_pose_override_drawn_listener.
skater_animation keeps what on_render wrote over and puts it back in on_drawn.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Installs, updates and uninstalls are jobs run one after another in the order asked for: asking for
one while another runs queues it instead of being refused. A job waits while the mods are applied or
a session's map downloads, and works out what its mod needs when it starts, from the Mods folder as
it is then. Buttons say ADD TO QUEUE while a job runs and QUEUED once in line; the foot of the page
says how many wait, and CANCEL ALL empties the queue.
The carousel's middle card: the mod's picture as its backdrop, darkened, and big as a tipped sticker
beside the text; a FEATURED and INSTALLED / UPDATE AVAILABLE tag, the facts as chips, as much of the
description as fits above the buttons, and the buttons under the text with the pad's prompts beside
them.
The front page stopped after four category rows. Now every category has one, the biggest first,
after Most popular, Recently updated and New; mod packs, libraries, tools and AI generated stay off
the front page (still found by search and the CATEGORIES panel).
The map grids (Mod Options' MAPS, the ReSkate menu's world page) found a custom map's icon among the
mods merged at launch, so a map installed or switched on since had none. The live apply now keeps
which mod each level comes from (live_mods::level_mod_directory), and map_icons asks it, falling
back to the launch's list before any live apply.
The reload after a live apply was a console "load", which in a session either ended it (a guest's
map going away is "your local map changed"), was refused (a guest without map rights), or moved
everyone (a server admin's load changes the server's map; a lobby host's reload sent every guest
through it too).
Now client_tick tells the session first (multiplayer::begin_local_reload) and loads the same map
locally with no multiplayer rule (queue_local_load). The session keeps the reload as travelling in
the same world: a guest keeps telling the host or server it is loading (world_ready, so it does not
time out) and that it is back; a host says it is not in the world for that long, so its guests wait
instead of moving, and nothing is a host map change. A reload refused by the loader is taken back
(cancel_local_reload); one that never starts ends after 20 s; in no state to reload (changing maps,
joining) the mods show at the next level load. The dedicated server already handles a player
reloading on their own.
A third ReSkate tab in the game's pause menu (native_menu.cpp slot 2, the Store's tab art), drawn by
the overlay as the other two are (mod_manager_page.inl, included into hub_page.cpp):
- BROWSE: search, categories in a panel of their own (shown only, hidden or either, kept in the
profile), sort, a featured carousel and rows (most popular, recently updated, new, the biggest
categories), a grid for results, and a mod's own page with its README, what it brings with it,
and INSTALL, switch on or off and uninstall. Works from a controller: the carousel's card is
one stop that left and right turn, and B goes back from a mod's page (the page's Back is held
meanwhile).
- MY MODS: the Mods folder with the states the ReSkate menu shows; UPDATES: Update all.
- The foot of the page shows the install's progress and the apply's, and Apply (View on a pad).
Leaving the menu with changes not applied asks first (a two-answer card in message_card.cpp,
answered by the runtime as the message is); LATER closes the menu.
Behind it, Extension/Assets/mod_store: Thunderstore's listing (cached on disk), the installed mods,
and a queue that downloads and installs packages with their dependencies, updates and uninstalls
them. An update or uninstall of a mod the game holds open waits in Mods/.reskate-pending and the
launcher makes it before the next start (launcher_mods::apply_pending). Package icons load into
their own SRV slots (store_icons.cpp), sharing the launcher's icon cache.
Tests: mod_store_view (search, categories, the front page's rows, install plans), the pending
updates and uninstalls in mod_manager_tests, and the third tab in native_menu_view_tests.
From the first 6k Sentry reports of 55a7342 (2026-10-24), whose top sites were all new since 2.1.1:
- Trainer: scale_underflip_bands wrote the underflip rate bands' MinScale and MaxScale every tick on
every copy it had ever found, checking only that the address was readable. A level's copies go back
to the game's heap with it, so 1.0 and 0.6 landed in the upper half of whatever the game made there
next (Skate.exe+0x157f930, +0x12a1549, +0x10084de, +0x11ef49e, +0x3db81ea, +0x12c21ca: about a
third of the crashes, every dump with a pointer whose high dword is one of those floats). A copy is
now written only while it still is that band and holds the game's band or the last one written;
any other is forgotten.
- Throwdowns: models the page released were destroyed 3 s later by the menu tick, which does not run
while a level loads, so the game's own flush of the unloaded level's types destroyed them after
their assets were freed (Skate.exe+0x3dc5eaa, 18%). Before a level unload they are destroyed at
once, as the native menu's own are.
- Textures: the guard's "default for the id" is the entry itself, so for a dead entry resolve gave a
null texture, and the texture view cache asked it for its key (Skate.exe+0x3e6ddc1, 17%, every
report with TheVlixOne-Nike_Pack). A stub at the cache insert hands back the empty view the cache
already built when the texture is null; the guard is on only when that patch applies.
- Physics: the wake guard widens every pass because ReSkate raises the owner pool to 160000; its log
line now says so once instead of claiming a prevented crash thousands of times a session.
Built; not yet checked in the game.
DingoSDKDebugInitialize answers the launcher with a bare TRUE or FALSE. On
FALSE the launcher threw "DingoSDKDebugInitialize returned FALSE", which the
STATUS tile turned into "ReSkate could not attach to Skate" with anti-virus
advice, although the attach had worked and the runtime had logged its reason.
ReSkate.dll now exports DingoSDKStartupFailure, a 1 KB buffer that every
FALSE path fills with its reason (the first one stays). The launcher reads it
out of Skate before closing it. With an older ReSkate.dll it reads the ERROR
lines that Skate process wrote to ReSkate.log during the call instead, and
with neither it says it was not told why. problem.h explains these messages
before the attach rules: a save in use or a folder another game holds says
Skate is already running, hook and overlay failures keep the attach advice,
and anything else leads with the runtime's own reason. Bootstrap and the
local profile hand their reason back, and the profile checks that used to
return false without a word now log one.
Fixes#90, fixes#346.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A server lists Thunderstore packages in plugins.required ("Owner-Name", or "Owner-Name-1.2.3"
for at least that version, at most 16). It checks each on Thunderstore as it starts and leaves
out one not listed as a plugin, and sends the rest with its map offer (protocol 55).
A joining player's game checks them before the server's map loads. Those missing, switched off
or too old go on one card with the map when that is missing too, so the player is asked once.
GET ALL downloads and installs each, switches on any that were off, runs one live merge when
any brings game data, and joins; CANCEL leaves. The game fetches a package only when
Thunderstore lists it as a plugin and it installs as one (a reskate-plugin.json naming Luau
code), whatever a server asks; anything else is taken out again, so a cosmetic can never be
required.
map_download handles a list of packages instead of one map, and the Thunderstore chooser takes
the category it needs. A map alone keeps its card as it was.
The mod list now notices a reskate-plugin.json that names an entry, reskate-objects.json and
reskate-music.json, and mods::mod_contents lists what a mod holds, most telling first. The
launcher's Type column shows the first of them (with "+" when there are more, all of them on
hover), and the in-game Mods page joins them. A reskate-plugin.json with only variants or
resident assets is not a plugin.
The UPDATE button moved from its own spot before the Type column to right after the blue
old -> new version, cutting a long name short rather than pushing into Type.
Hide and Seek, tag, spectating and ghost runs need one player to be
gone for some others, which no plugin could do.
reskate.server.hide(target, viewers, {skater, name, voice, sound}) and
reskate.server.unhide(target, viewers?) do it per viewer; viewers is a
player, a list of players or "all" (everyone on now but the target).
Wholly (the default) the target leaves those viewers' rosters, so their
games take the skater, its name tag and its objects away as for a player
who left, and the server sends them nothing of the target: poses, sound,
voice, outfit, effects and objects. No protocol change: the roster per
viewer is the one the cards and polls use, and the server's introduce
already holds a player back until shown. Shown again, the target is
introduced anew (outfit, then whole poses, then all its objects).
Without the skater (skater = false), only the voice and/or the sound is
held back and the player stays. Name and skater go together: a game has
no skater without its player, and hiding one without the other would
need a protocol change.
Refused while the target shares a party or a throwdown with one of the
viewers, whose game would lose a party member or a player mid-way; while
hidden, the two cannot come into one party. A plugin that stops shows
everyone it hid; a player who leaves is dropped from every list.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SYabE7FCPgj3SaWeF92EML
A server copy on a dedicated server had no chat, an empty session.world() and no way to change the
map or the time of day, though the server does each with its own commands. Now:
- chat (chat.read): each player's line as the server passes it on, the server's own lines to
everyone, and a /command the server does not know itself, which goes to the plugins instead of
"Unknown command" while one hears chat, so a game mode takes its own commands;
- session.world(): the server's world and the name of its map;
- travel.levels() and travel.load(): the server's maps, and a map change as the console's map does
(checked, saved and logged);
- world.set_time(): the time of day for everyone, as tod does, when the server has its world
layers and layer-sync is on.
The names custom chat commands may not take now include every server command: lock, unlock, locks,
throwdown-locks, words, bone-reach and poll-run were missing, so a custom command could hide them.
Plugins: server cards and polls, mods.content on dedicated servers, world_to_screen bounds
Conflicts in the plugin and server plugin tests, where #348's role tests and this PR's tests were
added in the same places: both kept.
"bind reskatemenu" runs the menu command, which went with the drone's commands when the drone
became a plugin, so every press said "Unknown command" and a player without an Insert key had
no way into the menu. At startup the log now warns about a bind whose command is not registered.
The server took a player as away unless their skater moved or they chatted, talked or changed
objects, so a pilot of the FPV drone plugin, whose skater stands still, was removed by the away
limit. The game now notes when the player works the controls (keys, mouse, a gamepad's buttons,
sticks and triggers, a joystick a plugin reads) and tells the server every 20 s while they do,
with a new presence packet (protocol 54). The server counts it as activity, once in 10 s at most.
A server plugin could only reach players through chat lines (say, tell).
A mode that asks players something (which track next, join the next
round) had to read chat commands, and could not show a card.
reskate.server.card(who, text, seconds) and reskate.server.poll(who,
question, answers, {seconds}) put a card or a poll on the screens of a
player, a list of players or "all", in the slot the roster already has
for the server's announcement and vote, so games need nothing new and
the protocol is unchanged. Each player's is their own and shows over the
server's until it ends, then the server's shows again. A poll has one id
for everyone it asks; a player's answer ("/2", as the game sends it)
closes it for them and comes back to the plugin that asked as a
poll_answer event; a poll that runs out, is replaced by a newer one for
that player, is closed with reskate.server.poll_close, or whose player
leaves comes back as poll_closed with why. The plugin counts the answers.
Both need the "server" permission and a server copy on a dedicated
server; a lobby's host has no roster slot for them. Answering a plugin's
poll is not a server vote, so it does not ask the player's role for
"votes". A plugin that stops takes its cards and polls with it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SYabE7FCPgj3SaWeF92EML
A server plugin with mods.read could call reskate.mods.content, but the
server's game had no content of its own, so it got nil and "not
available". A mode that keeps its data in mods (a race reading start
and finish from a park mod) could only read it in the players' games.
The server now answers from every mod folder in its Mods, by name, as it
finds its maps there; it has no load order or mods.json. A hidden folder
(.reskate_plugin_data) and names a plugin could not ask for are left out,
and a mod's title is its manifest.json's "name".
The listing and reading moved from plugin_game_runtime.cpp into
plugin_content.cpp, which the game and the server share, so both keep
the same limits: the folders the manifest's "content" names, 1000 files,
four folders deep, 1 MB a file, nothing outside the mod's folder. The
game's side is unchanged apart from calling it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SYabE7FCPgj3SaWeF92EML
A point just in front of the camera and off to the side projects to
millions of pixels. world_to_screen handed that back, and drawing it
(d:poly, d:line, ...) raised "must be a finite position", so a client
plugin that drew a world marker near the camera was switched off after
five errors. world_to_screen now returns nil when x, y or the depth is
not finite or beyond 1e6, the same bound the drawing functions use, as
it already did for a point behind the camera.
The draw test now also checks that its handler ran to the end; before,
a failing assert inside it went unnoticed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SYabE7FCPgj3SaWeF92EML
The FakeGame test for player.role ran its console command straight after apply(), before the
host had taken the session's players (Host::follow_session runs from tick), so the server copy
indexed a nil player. It ticks each host once first, as plugins_server_copies_are_the_authority
does; the player's copy and the plugin without server are then refused for the reasons they test.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Hs7xxJDf6SDcQUaXnzmtL5
A server plugin could not see the roles of a dedicated server (data/roles.json), so a plugin that
wanted only moderators to use some of its commands had to keep its own list of SteamIDs. In a
server copy with the server permission, a player table now has role: a read-only {name, rank,
staff}, the role the player has now (the one given them, or the server's default), read from the
server each time. staff is whether the role gives any of the server's commands. It is nil for a
player with no role, for the server itself and on a lobby's host. Each read counts as a server call.
dingosdk_server_plugin_tests reads a given role, the default role and a role given while the
player is on, and checks the table is read-only and a plugin without server is refused;
dingosdk_plugin_tests checks the API on a server copy, and that a player's copy and a plugin
without server cannot read it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Hs7xxJDf6SDcQUaXnzmtL5
zeex64