254ce7c156
Lane PAR localised the AI's RNG variance to this function and stopped. This
reads it from the instruction stream (0x006ad700..0x006ae61a, swept to the next
function start; the padding confirms Ghidra's size is right on this one) and
measures it live with a sub-bracket on VM145.
THE MODEL. Nine live draw sites plus two provably dead. Only three of the nine
can cost more than one word, and only through NextInt's rejection loop; the four
cl_Chance probabilities are all strictly inside (0,1), so neither zero-word
early-out is reachable anywhere in the function. The loop-carried draw fires
once per SMALL STANDARD WEAPON BANK selected by
f = 1.00 / 0.75 / 0.50 by request flags, hull size and one 0.3 coin
M = (int)(N * f); D' = max(1, (N+1)/M)
L = #{ bank j : PointDefence section, or j mod D' == 0 }
which is not monotone in N -- at f=0.75, N=4 costs 4 words and N=5 costs 3.
VERIFIED. Client 32's seven turn-1 words decompose as TWO composer calls: a
costOnly=1 price query (3 words) and a costOnly=0 build (4). That was predicted
from two push literals at 0x006cda9a/0x006cdb17 and committed before the probe
existed. Two pinned runs in fresh processes agree row for row. The probe's bank
counts 3/1/2 match the Tarkas section catalog's bank counts exactly -- a hooked
pointer walk and a parsed data file agreeing from opposite directions. The
detour is behaviour-neutral: unpinned it reproduces the published
d59bb9f2fd0eb535.
NOT VERIFIED, and this is the part worth reading. The loop-carried draw has
NEVER FIRED -- six composer calls across four runs, zero. On turn 1 the weapon
lookups gate it; on turn 15 every composer call is a price query that returns
nine steps earlier. Five of the nine sites have never fired and two of nine
exits have ever been taken. The formulas for those are read, not measured, and
section 6 says so.
CORRECTS roll-parity.md: its site table omits a live draw site (0x00691e9b,
which the same lane measured live at 0x00691ea0), so the AI turn has 22 live
sites and not 21; and 'six of seven turn-1 words come from the composer' is
five from its body plus two from a helper it calls -- all seven are in its
subtree.
NEW: on turn 15 the composer is 10 of 16 words, not 16 of 16. The task
system's coin at 0x0069086a fires four times and never fired on turn 1.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ARBgSooAfokKUy6wKUKEyZ
|
||
|---|---|---|
| campaign | ||
| findings | ||
| ghidra | ||
| guides | ||
| notes | ||
| objects | ||
| scripts | ||
| tools | ||
| verify | ||
| .gitignore | ||
| README.md | ||
sots-re
Reverse-engineering worklog for Sword of the Stars (2006, SOTS1) — the 32-bit DX9 original + expansions. The nitty-gritty: static/dynamic analysis notes, Ghidra & ReVa scripts, function/struct maps, D3D9 call traces, decomp progress, findings.
Where this runs
Analysis lab on spicy (PVE, 192.168.3.201):
- CT111
sots-re— Linux workspace: Ghidra + headless ReVa server, radare2/rizin/cutter, binwalk. Hosts the Samba share and this repo's working tree at/srv/re-lab/notes. - VM140
sots-re-win10— Win10 runtime + dynamic analysis (x64dbg, Cheat Engine, RenderDoc/apitrace, DXVK→CPU-Vulkan for GPU-less rendering).
Infra (guests, storage, network, share, ReVa endpoint) is documented from the
system-maintainer POV in trikilli → services/re-lab.md. This repo is everything else.
Layout
findings/— the running findings log (append-only), one file per subsystem.ghidra/— exported scripts, data-type archives, struct definitions.traces/— D3D9 / Win32 API call captures + analysis.scripts/— helper tooling (loaders, extractors, parsers).notes/— session notes, scratch, hypotheses.
Ownership / legality
Game binaries come from the owner's own GOG/Steam copy. RE is for personal
interoperability, bug-fixing, and preservation. Binaries themselves are not committed
here (see .gitignore) — they live on the lab's Samba share /srv/re-lab/samples.
Campaign (how this repo is run)
A 4-agent crew (defined in ~/.claude/agents/re-*.md) runs the exploration:
re-quartermaster (backlog + board) → re-analyst (maps via ReVa) →
re-verifier (proves vs real data; old-vs-new differential once reimpl starts) →
re-scribe (files the note, links it, commits).
campaign/board.md— live status board (start here).- Live tracking = Forgejo issues on
alex/sots-re(labelsstatus/*are the kanban columns;type/*,conf/*).campaign/board.mdis the editable mirror — publish withscripts/forgejo_campaign.py bootstrap(needsFORGEJO_TOKEN). campaign/backlog.md— prioritized target queue.campaign/open-questions.md— unresolved threads.findings/_template.md— the record format every finding follows.findings/{objects,control-flow,subsystems}/— the growing engine map.verify/{parsers,traces,harness,results}/— validation: struct parsers now, golden-trace replay + shim compare-mode for reimplementation.ghidra/— exported scripts + datatype archives.
Approach & north star: findings/00-strategy.md. Binary facts: findings/01-fingerprint.md.
Sibling repo
alex/sots-engine — the from-scratch engine source (clean-room, public-capable). This repo keeps the
evidence + planning for both; binary facts cross over only via ghidra/addresses.json → tools/gen_addresses.py.
guides/re-windows-2000s-howto.md— annotated bibliography + how-to for RE of mid-2000s MSVC/DX9 Windows games, with our-experience call-outs.