rrt/docs

RT3 Reverse-Engineering Handbook

This handbook is the project bootstrap for reverse-engineering and rewriting Railroad Tycoon 3. It is written for future us first: enough structure to resume work quickly, without pretending the project is already mature.

Canonical Target

• Canonical executable: rt3_wineprefix/drive_c/rt3/RT3.exe (patch 1.06)
• Reference executable: rt3_wineprefix/drive_c/rt3_105/RT3.exe (patch 1.05)
• Canonical SHA-256: 01b0d2496cddefd80e7e8678930e00b13eb8607dd4960096f527564f02af36d4
• Reference SHA-256: 9e96b0695cb722a700f99c8dce498d34da7235e562b1e275bcc1764f8c9b7eb1

Documents

• Stable docs

• setup-workstation.md: toolchain baseline and local environment setup.

• re-workflow.md: how to analyze the binary, record findings, and export reusable artifacts.

• function-map.md: canonical schema and conventions for function-by-function mapping.

• control-loop-atlas.md: compatibility index for the split atlas, preserving legacy anchors.

• control-loop-atlas/: canonical section files for the atlas narrative.

• subsystem-views/: curated cross-cut subsystem views over the atlas.

• runtime-rehost-plan.md: bottom-up runtime replacement plan and milestone breakdown.

• Active queue and history

• rehost-queue.md: active implementation and research queue.

• rehost-queue/: archived queue snapshots and related preserved worklog material.

• history/progress-history.md: preserved high-detail milestone narrative from the former root README.

• Artifact indexes

• ../artifacts/exports/rt3-1.06/README.md: canonical 1.06 export manifest.

• ../artifacts/captures/README.md: committed capture inventory and policy.

Repo Conventions

• docs/: stable project guidance and durable design notes.
• tools/py/: committed Python helpers for analysis and validation.
• artifacts/exports/: committed derived outputs that can be regenerated.
• artifacts/captures/: committed logs, sample binaries, and retained capture evidence.
• artifacts/tmp/: scratch-only local working area; do not commit new files here.
• Local-only state stays untracked: .venv/, Ghidra projects, Rizin databases, crash dumps, and other bulky/generated working files.

Current Baseline

The current technical milestone is a repeatable loop-mapping workflow for the 1.06 executable. Before injection work or deep file-format work, we capture:

• executable hashes and PE metadata
• section layout, imports, and notable strings
• a starter subsystem inventory plus a control-loop atlas
• focused address and string context exports for branch-deepening passes
• a reusable CLI RE kit for branch dossiers where the atlas needs deeper grounding
• a stable curated function ledger in artifacts/exports/rt3-1.06/function-map.csv

Current coverage is broad enough to support future sessions without rediscovery, especially in:

• CRT startup and bootstrap handoff
• shell frame, layout, presentation, deferred-message, and frontend overlay flow
• Multiplayer.win UI, chat, session-event, and transport ownership
• map/scenario load and text-export paths
• shared support layers such as intrusive queues, vectors, hashed stores, and tracked heaps

README maintenance rule:

• Keep this section at subsystem level only.
• Do not mirror per-pass function additions here.
• Detailed mapping progress belongs in artifacts/exports/rt3-1.06/function-map.csv and the derived branch artifacts under artifacts/exports/rt3-1.06/.

Current local tool status:

• Ghidra is installed at ~/software/ghidra
• ~/software/ghidra/ghidraRun launches successfully in an interactive shell
• Rizin is installed and available on PATH
• winedbg works with rt3_wineprefix
• RT3 launches under /opt/wine-stable/bin/wine when started from rt3_wineprefix/drive_c/rt3

Next Focus

The atlas milestone is broad enough that the next implementation focus has already shifted downward into runtime rehosting. The current runtime baseline now includes deterministic stepping, periodic trigger dispatch, normalized runtime effects, staged event-record mutation, fixture execution, state-diff tooling, tracked save-slice documents for captured-runtime inputs, overlay import documents that combine captured snapshots with save-derived state, and a packed-event persistence bridge that now reaches per-record summaries and selective executable import.

The highest-value next passes are now:

• preserve the atlas and function map as the source of subsystem boundaries while continuing to avoid shell-first implementation bets
• keep using overlay imports as the context bridge when selectively executable packed rows still need runtime context that current save slices and raw save inspection do not yet persist
• treat broader real grouped-descriptor recovery as the active packed-event frontier now that the first company-scoped batch already parses, summarizes, and executes through the ordinary runtime path when overlay context resolves its symbolic company scope: descriptor 2 Company Cash, descriptor 13 Deactivate Company, and descriptor 16 Company Track Pieces Buildable
• descriptors 1 Player Cash and 14 Deactivate Player now join that executable real batch through the same ordinary runtime path, backed by the minimal player runtime and overlay-import context
• the first chairman-targeted real grouped rows now execute too through that same path when the hidden grouped target-subject lane resolves to grounded chairman scope ordinals 0..3: condition_true_chairman, selected_chairman, human_chairmen, and ai_chairmen; wider chairman ordinals stay parity-only under blocked_chairman_target_scope
• chairman runtime ownership is broader now too: selected-chairman condition rows for chairman cash, holdings value, net worth, and purchasing power import through the same service path, and the first grounded company governance issue batch now executes too via book-value-per-share, investor-confidence, and management-attitude thresholds; wider chairman target ordinals remain frontier
• checked-in save-slice documents can now carry explicit company rosters and chairman profile tables too, so the current company-targeted and chairman-targeted descriptor/condition batches can execute from standalone save-slice fixtures without overlay snapshots when that context is present; raw .gms inspection/export now reconstructs those company/chairman collections directly from save-side tagged record families, while the fixed save-side 0x32c8 world block still provides selected company/chairman ids plus the grounded campaign-override, issue-0x37 value/multiplier, and chairman-slot / role-gate analysis bytes; the current raw-save frontier is narrower now: company/chairman identity, active flags, links, chairman cash, chairman holdings, chairman purchasing power, company debt, company track-laying capacity, and collection counts are grounded, while broader company finance/governance scalar lanes plus controller-kind reconstruction still remain conservative defaults until their raw offsets are pinned more strongly; the offline analysis command runtime inspect-save-company-chairman <save.gms> now dumps those remaining raw record candidates directly from the rehosted parser, including fixed-world chairman slot / role-gate context, the grounded fixed-world issue-0x37 pair, the fixed-dword world finance neighborhood rooted at [world+0x0d] with the saved calendar tuple and absolute counter, the separate six-float economic tuning band, derived holdings-at-share-price and cached purchasing-power totals, context, company dword candidate windows, and richer chairman qword cache views; the current rehosted company-side owner state now also includes a typed market/cache map carrying saved outstanding-shares, support/share-price/cache words, salary lanes, calendar words, connection latches, and the first grounded stat-band root windows at [company+0x0cfb], [company+0x0d7f], and [company+0x1c47] for each live company, so later finance/stat-family readers can attach to owned runtime data instead of one more guessed save offset; the first runtime-side 0x2329 stat-family reader seam is now also rehosted for slots 0x0d and 0x1d, and the saved stat-band windows themselves now carry 32 dwords per root; the matching world-side issue reader seam is now rehosted for the grounded 0x37 lane, and the adjacent raw issue-byte strip 0x37..0x3a now also flows through save-slice/runtime restore state for later credit / prime / management readers; selected-company summaries now expose the unassigned share pool derived from outstanding shares minus chairman-held shares for later annual finance logic; that same owned company market state now also backs a bundled annual-finance reader seam for assigned shares, public float, and rounded cached share price; the fixed-world finance neighborhood is now widened to 17 dwords rooted at [world+0x0d] so later issue-family readers can target a broader owned restore-state window, and the saved absolute counter now flows through normal runtime restore state instead of staying on shell-context metadata; that same world owner surface now also carries the packed year word and partial-year progress lane behind annual-finance recent-history weighting; the same annual-finance state now also feeds a shared company market reader for stock-capital, salary, bonus, and the full two-word current/prior issue-calendar tuples, and now derives elapsed years since founding, last dividend, and last bankruptcy for later annual finance-policy rehosting; live bond-slot count now travels through that same owned annual-finance state for the stock-capital branch gate, and the grounded bond table now also contributes both the largest live bond principal and the chosen highest-coupon live bond principal into that same owner-state surface, plus the highest live coupon rate for the stock-capital approval ladder; the same fixed-world save block now also carries the raw stock, bond, bankruptcy, and dividend finance-policy bytes, and the first annual creditor-pressure branch now executes as a pure runtime reader over that owner state instead of remaining atlas-only; the later deep-distress bankruptcy fallback now runs on that same save-native cash and trailing- profit seam; the annual bond, stock-repurchase, and stock-capital issue branches now do too, and periodic boundary service now also commits the bond-issue branch through that owned stat-post and live-bond seam instead of leaving it as a reader-only finance lane; the rehosted company market reader now carries the first bundled annual net-profit surface too; the annual dividend-adjustment branch now does as well through the shared year-or-control-transfer reader and board-approved dividend ceiling helper; the same owner seam now also carries the fixed-world building-density growth setting plus the linked chairman personality byte, which is enough to run the annual stock-repurchase gate headlessly as another pure reader; periodic boundary service now also chooses one annual-finance action per active company and already commits the shellless creditor-pressure-bankruptcy, deep-distress-bankruptcy, dividend-adjustment, stock-repurchase, stock-issue, and annual bond-restructure branches against owned runtime state, with bankruptcy now following the grounded “halve live bond debt and stamp the year” path rather than a liquidation shortcut; the same live bond-slot owner surface now also carries save-native maturity years into annual bond policy summaries, derives the current live coupon burden directly from owned bond slots, and now also commits the shellless “repay matured live bonds, compact the table, then issue the exact staged count” path during periodic service; the same service surface now also carries the per-cycle retired-versus-issued principal totals that feed the later debt-news family plus the issued-share and repurchased-share counts behind the later equity-offering and buyback news tails; runtime summaries now also expose the grounded retired-versus-issued relation directly, and annual finance service now maps that same relation onto the exact debt headline selectors 2882..2886 while persisting the last emitted annual-finance news events as structured runtime-owned records carrying company id, selector label, action label, and the grounded debt/share payload totals
• shellless calendar advance now also drives that annual seam directly: StepCount and AdvanceTo invoke periodic-boundary service automatically on year rollover instead of requiring a second manual service pass to make the annual finance stack run
• that stepped world-time path now also refreshes the derived selected-year gap scalar owner lane [world+0x4ca2], so later selected-year and periodic-boundary world work can build on runtime state instead of a frozen load-time scalar
• that same save-native world restore surface now also carries the grounded locomotive-policy bytes and cached available-locomotive rating from the fixed world block, so the All Steam/Diesel/Electric Locos Avail. descriptor strip now writes through owner state instead of living only as mirrored world flags
• the selected-year seam now follows the same rule: the checked-in 0x00433bd0 year ladder now drives a derived selected-year bucket scalar in runtime restore state, and the economic-tuning mirror [world+0x0bde] now rebuilds from tuning lane 0 instead of freezing one stale load-time word
• that same selected-year owner family now also rebuilds the direct bucket trio [world+0x65/+0x69/+0x6d], the complement trio [world+0x71/+0x75/+0x79], and the scaled companion trio [world+0x7d/+0x81/+0x85] from the bucket scalar instead of preserving stale save-time residue
• the same save-native company direct-record seam now also carries the full outer periodic-company side-latch trio rooted at 0x0d17/0x0d18/0x0d56, including the preferred-locomotive engine-type chooser byte beside the city-connection and linked-transit finance gates
• that same seam now also resolves the base world route-preference byte at [world+0x4c74], the effective electric-only override fed by 0x0d17, and the matching 1.4x versus 1.8x route-quality multiplier as a normal runtime reader instead of a pure atlas note
• that same seam now also owns the first route-preference mutation path directly: beginning the electric-only periodic-company override rewrites the world route-preference byte to the effective company preference, ending it restores the base world byte, and runtime service state now carries both the active and last applied override
• save inspection now also separates the shared 0x5209/0x520a/0x520b save family correctly: the smaller direct 0x1d5 collection is the live train family and now exposes a live-entry directory rooted at metadata dword 16, while the actual region collection is the larger non-direct Marker09 family; the tagged placed-structure header 0x36b1/0x36b2/0x36b3 remains grounded alongside them, so the remaining city-connection / linked-transit blocker is record-body reconstruction rather than missing save-side collection identity
• the project rule on the remaining closure work is now explicit too: when one runtime-facing field is still ambiguous, prefer rehosting the owning source state or real reader/setter family first instead of guessing another derived leaf field from neighboring raw offsets; the checked-in docs/rehost-queue.md file is the control surface for that work loop, and after each commit the next queue item should run unless the queue is empty, a real blocker remains that cannot be advanced by any further non-hook work without guessing, or approval is needed; final responses are stop-only there too, so active work should continue under commentary updates rather than placeholder status replies
• rrt-runtime now also exposes higher-layer probe commands for the current blocked frontier: runtime inspect-periodic-company-service-trace <save.gms>, runtime inspect-region-service-trace <save.gms>, and runtime inspect-infrastructure-asset-trace <save.gms>. These reports make the current shellless city-connection / linked-transit blockers explicit as missing owner seams rather than generic “still unresolved” residue.
• a checked-in EventEffects export now exists at artifacts/exports/rt3-1.06/event-effects-table.json, and a checked-in semantic closure layer now exists at artifacts/exports/rt3-1.06/event-effects-semantic-catalog.json
• recovered descriptor rows now land on explicit semantic frontier buckets such as blocked_shell_owned_descriptor, blocked_evidence_blocked_descriptor, and blocked_variant_or_scope_blocked_descriptor instead of generic unmapped-descriptor residue
• the first recovered governance descriptor tranche now executes through the generic company-governance scalar effect surface: descriptor 56 Credit Rating and descriptor 57 Prime Rate
• adjacent recovered finance/control-transfer descriptors such as 55 Stock Prices and 58 Merger Premium now land on explicit shell-owned descriptor parity instead of generic unmapped descriptor residue, with tracked fixtures now pinning finance, scenario-outcome, and control-transfer shell rows explicitly
• the recovered whole-game scalar economy/performance strip 59..104 now has a bounded runtime landing surface too: representative rows execute into RuntimeState.world_scalar_overrides through stable normalized keys such as world.build_stations_cost and world.track_maintenance_cost
• the runtime-variable strip 39..54 now executes too through bounded event-owned scalar maps on world/company/player/territory state, and the matching ordinary-condition strip now gates imported records through those same runtime-owned variable maps; these variables are still runtime-owned only in the current model and are not yet reconstructed from raw saves
• the grounded aggregate cargo-economics descriptors now execute too: descriptor 105 All Cargo Prices and descriptors 177..179 All Cargo Production / All Factory Production / All Farm/Mine Production land on bounded event-owned cargo override state, and the grounded named cargo-production strip 180..229 now lands on named cargo production overrides too
• the named cargo-price strip 106..176 now lands on named cargo price overrides too; the checked-in static selector reconstruction is now explicit because the broader 1.06 CargoTypes corpus has 51 names, the Cargo106 cargoSkin corpus has 70, and the rehosted offline selector builder now closes the 71-row price strip as cargoSkin plus the core Rock carry-over; a new checked-in offline cargo-source report at artifacts/exports/rt3-1.06/economy-cargo-sources.json now parses both CargoTypes and the Cargo106.PK4 cargoSkin descriptors through rehosted code, normalizes localized ~####Name tokens into visible names, builds a merged live cargo registry, and now derives exact named cargo-production and named cargo-price selectors from the checked-in bindings; dedicated CLI inspector commands now expose both grounded selectors directly, while the same report still shows the nine excluded CargoTypes-only industrial names that sit outside the 71-row price strip
• the add-building strip 503..519 is now explicitly classified as recovered shell-owned parity, with tracked fixture coverage, instead of generic unresolved descriptor residue
• widen real packed-event executable coverage descriptor by descriptor after identity, target mask, and normalized effect semantics are all grounded, not just after row framing is parsed
• the first grounded condition-side unlock now exists for negative-sentinel raw_condition_id = -1 company scopes, and the first ordinary nonnegative condition batch now executes too: numeric thresholds for company finance, company track, aggregate territory track, and company-territory track
• exact named-territory binding now executes too, while named-territory no-match cases remain the explicit binding blocker frontier
• real descriptors 8 Economic Status, 9 Confiscate All, and 15 Retire Train now join the executable batch through the same ordinary runtime path, backed by the opaque economic-status lane and the minimal event-owned train roster
• descriptor 3 Territory - Allow All now executes as company-to-territory access rights through the same ordinary runtime path; shell purchase-flow parity remains out of scope, and mixed supported/unsupported real rows still stay parity-only
• whole-game ordinary-condition execution now exists too: special-condition thresholds, candidate-availability thresholds, and economic-status-code thresholds now gate imported runtime records, and the packed-event frontier now reports explicit unmapped world-condition and world-descriptor buckets
• that whole-game condition batch is now metadata-driven too: special-condition label ids, economic-status, and the generic %1 Avail. candidate-availability template plus candidate-name side strings all decode through checked-in world-condition metadata instead of fixture-only ids
• the first real whole-game grouped-descriptor batch is now metadata-driven too: checked-in descriptor metadata covers special-condition and candidate-availability setters, and descriptor 110 Disable Stock Buying and Selling now executes too through the checked-in keyed runtime flag world.disable_stock_buying_and_selling
• that world-toggle path now covers a broader recovered boolean scenario-rule band too: descriptors 111..138 now decode through checked-in metadata into either keyed world_flags or the bounded world_restore.limited_track_building_amount scalar for finance/trading, construction, and governance restrictions
• the late recovered world-toggle band now executes too where current evidence is equally strong: Use Bio-Accelerator Cars, Disable Cargo Economy, Disable Train Crashes, Disable Train Crashes AND Breakdowns, and AI Ignore Territories At Startup
• whole-game ordinary-condition coverage is broader now too: checked-in world-flag condition ids can lower into world_flag_equals gates for boolean equality/inequality forms, so real packed rows can gate whole-game effects on existing world_flags
• the tracked parity save-slice now keeps its remaining non-imported residue as structured real_packed_v1 parity records, with the first captured leftover now pinned to the unresolved upper locomotives-page availability tail and moved onto explicit descriptor parity
• the next recovered locomotives-page descriptor batch is partially executable too: descriptors 454..456 (All Steam/Diesel/Electric Locos Avail.) now lower through checked-in metadata into keyed world_flags, while the wider locomotive availability/cost scalar bands now split cleanly between the grounded lower executable prefix and the explicit unresolved tails
• raw .smp inspection/export now reconstructs the persisted save-side named locomotive table and derives a minimal locomotive catalog from its row order, so save-slice documents can carry both RuntimeState.named_locomotive_availability and the catalog context needed for descriptor lowering
• recovered scalar locomotive availability and locomotive-cost descriptors now import through that save-native or embedded RuntimeState.locomotive_catalog context into the ordinary named_locomotive_availability and named_locomotive_cost runtime maps
• the grounded executable lower locomotive prefix now extends through save-backed locomotive id 61 (Zephyr); the unresolved lower tail and upper locomotive bands stay on explicit parity
• cargo-production 230..240 and territory-access-cost 453 now execute too through minimal world-side scalar landing surfaces: slot-indexed cargo_production_overrides and world_restore.territory_access_cost
• recipe-book probing now derives a save-native cargo_catalog too, so save-slice documents carry stable cargo slot labels and token-stem evidence into runtime state without requiring a separate cargo simulation layer
• world-scalar ordinary-condition coverage now matches those runtime surfaces too: checked-in metadata lowers named locomotive availability, named locomotive cost, named cargo-production slot thresholds, aggregate cargo production, factory/farm-mine/other cargo production, limited-track-building-amount, and territory-access-cost rows into explicit runtime condition gates
• cargo slot classification is now checked in and save-native too, so the remaining cargo frontier is broader descriptor/condition breadth rather than classification or save/import plumbing
• the company/chairman frontier has moved too: checked-in save-slice documents can now carry that context natively, so the next work on that axis is broader recovery and eventual raw save reconstruction rather than overlay-only ownership
• keep in mind that the current local .gms corpus still exports with no packed event collection, so real descriptor mapping needs to stay plumbing-first until better captures exist
• use rrt-hook primarily as optional capture or integration tooling, not as the first execution environment
• keep docs/runtime-rehost-plan.md current as the runtime baseline and next implementation slice change

Regenerate the initial exports with:

python3 tools/py/collect_pe_artifacts.py \
  rt3_wineprefix/drive_c/rt3/RT3.exe \
  artifacts/exports/rt3-1.06

Regenerate the startup-focused Ghidra exports with:

python3 tools/py/export_startup_map.py \
  rt3_wineprefix/drive_c/rt3/RT3.exe \
  artifacts/exports/rt3-1.06

Regenerate the checked-in EventEffects table export with:

python3 tools/py/extract_event_effects.py \
  rt3_wineprefix/drive_c/rt3/RT3.exe \
  rt3_wineprefix/drive_c/rt3/Data/Language/RT3.lng \
  artifacts/exports/rt3-1.06/event-effects-table.json

Regenerate the checked-in EventEffects semantic catalog with:

python3 tools/py/build_event_effect_semantic_catalog.py \
  artifacts/exports/rt3-1.06/event-effects-table.json \
  artifacts/exports/rt3-1.06/event-effects-semantic-catalog.json

That default export now walks two roots:

• entry:0x005a313b
• bootstrap:0x00484440

For a focused branch-deepening pass, regenerate the analysis context exports with:

python3 tools/py/export_analysis_context.py \
  rt3_wineprefix/drive_c/rt3/RT3.exe \
  artifacts/exports/rt3-1.06 \
  --addr 0x00444dd0 \
  --addr 0x00508730 \
  --addr 0x00508880 \
  --string gpdLabelDB \
  --string gpdCityDB \
  --string 2DLabel.imb \
  --string 2DCity.imb \
  --string "Geographic Labels"

For the pending-template dispatch-store branch, regenerate the new branch dossier with:

python3 tools/py/rt3_rekit.py \
  pending-template-store \
  rt3_wineprefix/drive_c/rt3/RT3.exe \
  artifacts/exports/rt3-1.06

That dossier is now a targeted follow-up tool, not the default first pass.