An array record's hash is one per field type in a document (every
array the base game's files hold follows this, and each such hash is a
key of the document's reflection group table, now read). A new array
takes the hash of an existing array of the same holding type and field;
callers can ask the writer to refuse when there is none, and the mod
merge keeps the old fallback. Checked on the game's own
level documents.
Each array record carries a hash the original files make specific to
the field or element type; the writer used to give every array the hash
of the first record of the same element type, so a rewrite changed
hundreds of them in bam's level root. Arrays read from a document now
keep their record's hash; only arrays with no source record borrow as
before.
A rewrite copied a boxed value's bytes, pointers and all, but wrote the
arrays and strings those pointers reach only when a field also reached
them. Rewritten documents lost that data and the boxed pointers aimed
at whatever moved into its place; the mod merge rewrites real documents
this way. Read documents now keep their EBXD bytes, and the writer
carries every array, string, instance and boxed value a boxed pointer
reaches, repointed. Found by a round-trip check on bam's level root (one string and
about 1,120 bytes of arrays lost).
Players reported the game hanging for about a second while skating. A
profile showed the memory line in the log walking every heap
(HeapSummary) on the client thread, 0.5 to 2 s every two minutes. An
audit of everything that runs on the game's update and render threads
found more of the same, and Sentry's top crashes had causes of ours.
Freezes:
- Memory log line: no heap walk or thread snapshot; the GPU part and the
write happen on a thread of their own.
- Profile database: synchronous=NORMAL in WAL mode. A save is still kept
when the game crashes; every play event no longer waits for the disk
inside the game's hooks.
- ReSkate's own preferences and values (menu clicks, session sliders,
Discord) are saved by a background writer in order; reads see a value
until it is saved.
- Input: the foreground window and the keyboard are sampled on a thread
of their own (GetAncestor and GetAsyncKeyState stalled the client tick
40-50 ms when another program held Windows' lock); flight input uses
the sample; ClipCursor at most twice a second while a menu has the
pointer.
- Hooks installed while the game runs go in one transaction per place
(HookBatch): throwdowns, remote players, the party, Hall of Meat.
- Off the update thread: the host's map package lookup (one table read in
the background, not the Mods folder per offer), the trainer's map file
and settings save, Board Break's settings save, mod music playlists and
their covers, the Spotify station covers.
- Off the render thread: the Modding page's and the deck card's Mods
folder reads. The overlay's three startup waits share five seconds.
- Less work per frame: Steam names reused while one a frame is asked
again, 2 friends a tick, steam_api64.dll checked once per process,
cosmetic names read in place rather than a system call per character,
an outfit load no longer copies the inventory per slot, the throwdown
type scan runs at background priority over resident memory, the Board
Break deck label is kept a minute.
Crashes:
- Render-resource pool: the metadata fields are written to the state
owner only. The other renderer is 0x2daa00 bytes, and +0x11aae28 is far
past its end (store_field crash; heap corruption in every release).
- Other players' skateboard effects: a stale skateboard component is no
longer followed, and the joint lookup is walked with a bound first (an
endless loop in reused memory ended in GameStall).
- Launcher: Oodle decodes into the buffer it asks for with corrupt data
possible, and its assertions no longer stop the process (heap corruption
in the store-copy scan; a breakpoint in oo2core).
Timings (Send performance timings):
- A frame of half a second or more in a level is sent as game.freeze:
where the time went (ReSkate's tick, the game's frame, or outside it),
whether the renderer kept presenting, and ReSkate's busiest zones.
- Closing the game sends the level session it was in.
Music after a connection drop:
- Spotify device: an account Spotify could not be reached with is kept and
tried again (3 s doubling to a minute); a session that drops is taken
up again. Only an account Spotify refuses is asked for anew.
- YouTube playlist pictures that could not be fetched are tried again,
from 30 s to 10 minutes apart.
The renderer gives each frame 131,072 shader-visible descriptors past a persistent 120,000, in
heaps of 333,333. Dense custom maps (the Los Santos City sublevels, San Andreas) draw more than
that in a frame, and the renderer wrote past the end of the heap: shaders read garbage
descriptors, which faulted or hung the GPU (page faults at nonsense addresses, in draws, texture
fetches and ray tracing alike). ReSkate now patches the device setup at boot so each heap is
D3D12's 1,000,000 and a frame may use 524,288; heaps are 644,288 descriptors. LS-B2, LS-B3 and
San Andreas (210,000 a frame at its peak) now hold.
Also in this change:
- Memory: perf memory owners/heap/gpu/report, a ledger of who commits the game's RAM; the
render-resource pool no longer allocates its 115 MB twice; the EadpErrorsData arena is
capped at 32 MiB (EA's error reporting is blocked anyway).
- Crash reports: tagged with the map, destination, graphics card and driver version (protocol
v3), so Sentry can group GPU hangs by map and card.
- GPU diagnostics (--gpu-diagnostics): a fuller DRED report, the device-removal and video
memory watches, a large-draw watch, a record of which game code made and freed each GPU heap
and resource, checks of every GPU address bound, and a meter of each descriptor heap's use.
- World pools: peak use of entity pages, render instances and spatial blocks in the log.
- dingosdk_mod_mesh_check: checks every MeshSet a mod ships against its geometry; the MeshSet
reader now knows static meshes' layout.
Shader tables
- Reverts a61f82c ("Shader crash fix"). Since 2.0.8 a mod's own shader
programs were copied into the merged table, and the tables that cosmetic
bundles carry were merged across mods. Each such bundle then held other
mods' programs, and the game built them as the bundle loaded, before the
files with their compiled variants were there. That was 2.0.8's most
reported crash (Skate.exe+0x3f86da6, some forty machines in ten minutes).
2.0.9's guard passed such a program over, and the same players then
crashed a little later in cosmetic loading (+0x3f7bb65, +0x16fe1a6,
+0x3dc07ed).
- Whatever crash a61f82c fixed comes back with this for the mods that
needed it.
- The guard stays.
Music
- The game's read of one song for its scorer (0x14f23a0) is replaced by the
same read with the check it lacks: a place past the end of the playlist
asset's songs has no song, which the game already answers with 0. 2.0.9's
check in music_select_next_hook looked at what the queue held before the
game was asked and did not see these; the game asked for places such as
0xd570 and crashed reading them (7 reports on 6 machines).
- Said in the log the first eight times.
Not confirmed in game: neither change has been run by a player who was
crashing.
Player list. A player on the developer, staff or homie list sees each
player's Steam ID in the overlay's PLAYERS list for the session they are
in, after "Host" / "Guest" and the distance, on their own row too. A click
copies it. The lists are the backend's, the same ones the tags come from;
every client in a session already knew the IDs, so this changes who is
shown them and nothing else.
Custom scripts. Nothing in scripts/Custom is read or run any more, whatever
the launcher's "Loose files" setting says: nothing used them, and a mod was
using the folder to load a DLL into the game (package.loadlib from its
init.lua). The loader is kept behind one switch (custom_scripts_enabled in
custom_script_loader.cpp) for the day they come back.
For that day, the loader now also takes Lua's ways of running native code
away before the first custom script runs: package.loadlib, require through
the two C searchers and package.cpath, os.execute and io.popen. If that
does not hold, no custom script runs.
The overlay's Mods page no longer has the INSTALLED / SCRIPTS tabs; it is
the installed mods. Editing the game's own Lua and config through loose
files is unchanged.
As in skate. 3: when the local skater bails, the bones it hurt show over
the world through the body, yellow where a hit bruised one and red where
one broke it, with the bail's Meat card in the bottom left corner in the
place of skate.'s own HUD there: its time, hits (head hits double,
vehicles half again), broken bones, road rash, airtime, fall, top speed
and rotations, each scoring, and the map's best. A bone breaking pulses
the screen's edges red and, in single player, slows the game down a
moment. Off until switched on, as in skate.: in skate.'s menu (Custom
Stuff > Player) or with `hallofmeat 0|1`, saved with the profile.
Everything it draws is skate.'s own, read from the installed game: its
skeleton mesh (the Dem Bones costume) posed with the renderer's own
skinning matrices, its scoring icons, the THRASHER wordmark and its UI
brush strokes. Nothing is shipped.
How it works: No Bail already hooks the skeleton response, which runs
once a physics step right after the body contacts; it now hands that step
on, with the local skater it follows, but only while Hall of Meat is on.
Each step reads what every ragdoll body hit and how hard, and whether the
skater is a ragdoll; the bail runs on the game's own clock so slow motion
measures what the body did. The feature's own hooks, on the renderer's
skinned draw packets and main view and on the UI model's write, are
prepared at startup and switched on only with the feature, and off again
once the HUD corner and the game's speed are given back: switched off, the
game never calls them and the physics step pays nothing. The skeleton mesh
is read the first time it is on; only the images are read at startup,
since the overlay builds its atlas once. The renderer's picture of the
skater is read only while a bail shows. The menu's switch reads the
overlay model like the others. Every address is in the build table with a
fingerprint checked before use. Extension/HallOfMeat/README.md describes
the parts; its tests build with DINGOSDK_BUILD_HALL_OF_MEAT_TESTS.
The slow motion changes SimulationTime's rate, frame cap and time scale as
a session override and steps the local skater along, since skate. sets a
skater's step only when it builds its core. It runs the same at every
break, but never faster than the game ran: with the trainer's game speed
below 1 it only takes the part that is slower still. Afterwards it writes
back what it found, each setting only while it still holds what the slow
motion wrote, so the trainer's speed comes back as it was; a setting
changed from elsewhere meanwhile is left as it is and ends that slow
motion. named_setting_value reads a setting's value for it.
Tuned in play: a bruise from 7 m/s and a break from 12 m/s along the
contact normal (raised twice: a hard fall on flat ground broke bones), a
break worth ten of the lightest hits (a heavy bail's hits and breaks
outweighed the rest of the card) and a fall half of skate. 3's rate.
Fix: the overlay now draws in the back buffer's pixels, not the window's,
so it lines up on displays Windows scales (1920x1080 borderless on 4K).
The skater's render hooks, its skeleton and its body maps were first
found by gBGYo's ReSkate fork (github.com/gBGYo/ReSkate).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each built its text with path::string(), which throws for a character outside the ANSI code page, so the real error or message was replaced by a code-page error.
Merge: carry added assets with a changed list, combine in-place values, list repeats once
Three merge bugs found by auditing 104 installed mods:
- A changed list (the music playlist) reached every bundle copy that holds
it, but the songs a mod added beside it only reached copies of the bundle
the mod added them to. The game keeps the playlist in eight bundles, so
the other seven named songs they did not hold. Added assets now follow
the change into whichever bundle copy takes it.
- An entry naming an asset that lost its name to another mod's asset is
now dropped from the list, both in the merged bundle and in the list
other mods' copies are given.
- Two mods changing values in place in one asset (lighting) kept one mod's
copy whole and silently dropped the other's. Plain values are now
combined, the higher-priority mod winning where both changed one. An
edit that also moved a reference or resized a list is taken whole or not
at all, and the merge notes whose changes are missing.
- A reference two mods both append to a list is listed once.
- Read Dev / Creator / Homie Steam IDs from api.reskate.dev instead of the hardcoded list
- Creator (red) and Homie (gold) tags and hoodie/board colours; Admin tag is now pink
- Global bans: lobby hosts and dedicated servers turn banned players away
("global_bans": false in ReSkateServer.json opts a server out)
- SPECIAL menu page for listed players to hide their own tag and items
- Fix hoodie and board keeping the animated colour after the animation turns off
- Multiplayer protocol 39 -> 40
With several cosmetic mods installed the game crashed as soon as an
equipped custom item came into view, and so on every launch after it
was equipped.
Cosmetic presets (Engine/Resource/bundle_ref_table.*)
- The character bundle-reference table says which bundle holds each
preset, under the preset's full path and under its leaf name, one row
per hash. Mods number their presets alike in different folders
(cust_tops/1_ap, cust_bottoms/1_ap), so two mods can share a leaf
name. Adding the second preset threw "collides with an existing
lookup", the merge fell back to the highest-priority mod's copy of
the table, and every other cosmetic mod's presets were unregistered.
The game then built the skater with a preset it could not find and
called through a bad pointer.
- A preset whose leaf name is taken is now added under its full path
only; the leaf name stays with the preset that had it. Only a clash
of full paths still fails. MergedTable::shadowed lists each such
preset with the one that holds its name.
Merge log (Engine/Vfs/mod_merge_combine.cpp)
- Each mod with shadowed presets gets a line naming how many and an
example pair, so an item that shows the wrong preset can be traced to
the mods that clash.
Checked against 39 installed mods, 15 of them cosmetic: the table
combines 274 presets where it used to keep one mod's, all reachable by
full path and 267 by leaf name, with the game's own 6264 rows unchanged.
The seven left without a leaf row are a naming clash between the mods
themselves (two Carhartt packs, a Zero tee and deck); whether the game
finds those by full path is not known.
bundle_ref_table_tests takes an optional second mod and checks that two
mods' presets are all added, in either order, and that a shared leaf
name is reported with the preset that holds it.
Dave Johnson