# MECHSUB — reconstruction notes (phase 2 kickoff, 2026-07-19)

**Status: base-class ANALYSIS done; the reconstruction must be done FRESH against
the 4.10 structure, NOT backdated from BT411 (its base class + damage model
diverge). Key findings below de-risk the actual code.**

MechSubsystem is the base of the ~30-class subsystem roster (sensor, gyro, torso,
hud, generator, myomers, weapons…) the Mech ctor's segment-table walk
instantiates. The staged MECHSUB.HPP is interface-only (one `Mech *owner`
member, no bodies). Reconstructing the base is phase-2 step 1.

## Dependencies (all survive in the 4.10 archive)

- `Subsystem` (CODE/RP/MUNGA/SUBSYSTM.HPP) — the base.
- `DamageZone` (CODE/RP/MUNGA/DAMAGE.HPP/.CPP) — the engine damage zone.
- `Mech__DamageZone : public DamageZone` (CODE/BT/BT/MECHDMG.HPP) + the BT damage
  model (MechCriticalSubsystem, redirect tables, critical hits).
- `AlarmIndicator` (just reconstructed, MUNGA/ALARM.HPP) — statusAlarm.

## Finding 1 — the base Subsystem DECLARES the zone but does NOT create it (RESOLVED)

SUBSYSTM.HPP:181 declares `DamageZone *damageZone;` in the **base** Subsystem and
its TakeDamage (SUBSYSTM.HPP:177-178) delegates `damageZone->TakeDamage(damage)`.
The base ctors ARE reconstructed — **staged source410/MUNGA/SUBSYSTM.CPP** (built,
in munga.lib; 38 Subsystem:: syms in the map). Both base ctors set
`damageZone = NULL` (resource ctor line 63) — the base does **not** allocate the
zone. So the DERIVED MechSubsystem creates the `Mech__DamageZone` (BT411's
zone-creation is structurally right here) — but the ARMOUR it seeds must follow
Finding 2 (per-type streamed), not BT411's per-facing hand-seed.

## Finding 2 — CRITICAL: the damage model diverged

The 4.10 base DamageZone (DAMAGE.HPP/CPP) is a **per-damage-TYPE** model:
- members: `defaultArmorPoints` (Scalar) + `damageScale[Damage::DamageTypeCount]`
  where DamageTypeCount = 5 = {Collision, Ballistic, Explosive, Laser, Energy}.
- TakeDamage: `damageLevel += damage.damageAmount * damageScale[damage.damageType]`
  (DAMAGE.CPP:413); the zone STREAMS its own armour (`*stream >> defaultArmorPoints;
  *stream >> damageScale[ii]`, DAMAGE.CPP:335,340).

BT411's reconstruction is a **per-FACING** model: `armorByFacing[5]` (front/back/
left/right/top) + `structureReference`, **manually seeded** by MechSubsystem's
resource ctor with an armour→absorption-coefficient inversion. Those field names
(`armour[5]`, `structureReference`) exist only in BT411's `ReconDamageZone`
proxy — NOT in the surviving 4.10 DamageZone/Mech__DamageZone. The staged
`MechSubsystem__SubsystemResource` (armorByFacing[5]+structureReference) was
copied from BT411 and is likewise suspect for 4.10.

**Consequence:** BT411's MechSubsystem ctor (heat.cpp/mechsub.cpp) CANNOT be
backdated verbatim — it would install a damage model the 4.10 engine's DamageZone
doesn't implement, corrupting every subsystem's damage behaviour. The 4.10
MechSubsystem must let the base Subsystem + DamageZone stream ctor handle armour
(per-type), not hand-seed per-facing values.

## Corrected reconstruction approach (fully scoped — ready to write)

1. ~~Pin the base Subsystem ctor~~ **DONE (Finding 1)** — staged SUBSYSTM.CPP sets
   `damageZone = NULL`; MechSubsystem creates the `Mech__DamageZone`.
2. Reconstruct MechSubsystem: chain the base Subsystem, `new Mech__DamageZone(this,
   index)`, add the mech-specific members (statusAlarm = AlarmIndicator, vital
   flag, alarmModel, criticalReference, collisionCriticalHitWeight,
   printSimulationState, configureActivePress) and the TakeDamage override, using
   the 4.10 **per-type** damage model — let the zone stream its own armour
   (defaultArmorPoints + damageScale[5]); do NOT hand-seed BT411's per-facing
   armorByFacing[5]/structureReference.
3. Re-derive `MechSubsystem__SubsystemResource` against the 4.10 DamageZone stream
   format (the staged struct's per-facing fields are BT411-copied and suspect).
4. Then the roster (GAUSS/PPC/SENSOR source + .TCP partials + decomp).

Analysis complete; step 2 is the next code to write. This worksheet is the
deliverable that keeps the subsystem foundation on the 4.10 damage model rather
than BT411's divergent one.

## Progress + roster hierarchy findings (2026-07-19)

**DONE:** `MechSubsystem` base (MECHSUB.HPP/.CPP) and `HeatableSubsystem`
(HEAT.HPP/.CPP) reconstructed + compile-verified. Verified vertical slice so far:
`Subsystem → MechSubsystem → HeatableSubsystem`.

**The subsystem hierarchy is DEEPER than the staged headers show.** From BT411:

```
MechSubsystem            (done)
  └ HeatableSubsystem    (done; ctor chains MechSubsystem + ResetToInitialState
                          -> currentTemperature=300, heatLoad=0)
     └ HeatSink          (MISSING from the staged tree -- not in source410 nor
                          CODE; VDATA HeatSinkClassID=0xBC5. Must be added.)
        └ PoweredSubsystem (staged POWERSUB.HPP wrongly derives it from
                          HeatableSubsystem -- real base is HeatSink; FIX the
                          staged header. Also its ctor has embedded AlarmIndicators
                          (electricalStateAlarm 5-level, modeAlarm 3-level) and a
                          voltageSource SlotOf.)
           └ Sensor       (surviving header CODE/BT/BT/SENSOR.HPP; leaf)
           └ Generator, ... (other leaves)
```

**PoweredSubsystem has inter-subsystem wiring** — its ctor resolves a voltage
source (generator) by indexing the OWNER MECH's subsystem roster
(`owner->GetSubsystem(res->voltageSourceIndex)`), not the skeleton. So powered
subsystems must be created AFTER the generator in the segment-walk order, and the
Mech must expose GetSubsystemCount()/GetSubsystem(). This ordering constraint is
a segment-walk (phase-4 ctor) concern; note it there.

## Complete subsystem family tree (mapped 2026-07-19) + build status

```
MechSubsystem                         DONE (MECHSUB.HPP/.CPP)
├─ HeatableSubsystem                  DONE (HEAT)
│   └─ HeatSink                       DONE (HEAT)
│       ├─ PoweredSubsystem           DONE (POWERSUB)
│       │   ├─ Sensor                 DONE (SENSOR; interface survives in CODE)
│       │   └─ Myomers                todo (myomers.hpp)
│       ├─ Generator                  DONE (POWERSUB; source, GNRATOR.TCP partial)
│       └─ Condenser                  todo (heat.hpp)
└─ HeatWatcher  (: MechSubsystem)     todo  -- sibling branch to HeatableSubsystem
    └─ PowerWatcher (: HeatWatcher)   todo
        ├─ Torso                      todo (torso.hpp)   -- Mech sinkSourceSubsystem
        ├─ HUD                        todo (hud.hpp)     -- Mech hudSubsystem
        └─ Gyroscope                  todo (gyro.hpp)    -- Mech gyroSubsystem
```

Plus the WEAPON family (separate subtree): MechWeapon → {PPC, Gauss, ProjectileWeapon
→ Missile/…}. PPC/GAUSS source survives in CODE/BT/BT; AMMOBIN/EMITTER/MISSILE/
PROJTILE/PROJWEAP have .TCP partials; the rest decomp-only.

**Pattern established** (6 classes verified): each = statics
(ClassDerivations(parent, "Name") + DefaultData reusing Subsystem::MessageHandlers/
AttributeIndex/StateCount), ctor chaining the parent + member init from the
resource, dtor, TestClass/TestInstance/ResetToInitialState real, per-frame
Simulation method + CreateStreamedSubsystem staged with Fail. Embedded alarms =
AlarmIndicator(N); filters = AverageOf<Scalar>; owner->GetInstance() !=
ReplicantInstance gates the master performance install.

## ROSTER CORE COMPLETE (2026-07-20) — 16 classes verified

All committed + compile-verified: MechSubsystem, HeatableSubsystem, HeatSink,
HeatWatcher, PowerWatcher, PoweredSubsystem, Generator, Condenser, Sensor,
Myomers, Gyroscope, HUD, Torso, MechWeapon, Emitter (+ surviving PPC/GaussRifle),
ProjectileWeapon, MissileLauncher, AmmoBin. Tree links clean.

## Phase-4 blueprint: the Mech-ctor segment-walk factory (mech.cpp:1146-1330)

The Mech ctor allocates `subsystemArray[subsystemCount]` (+GetSubsystem/Count),
then loops segments (from id 2) switching on `seg->classID` to `new <Class>(this,
id, seg)`, caching the key ones. **The VDATA ClassID enum names are MISLABELED vs
the real class** (BT411's mapping):

| ClassID | enum name | REAL class | built? |
|---|---|---|---|
| 0xBBD | CockpitClassID | Condenser | ✅ |
| 0xBBE | SensorClassID | HeatSink bank | ~ (HeatSink ✅; bank variant TODO) |
| 0xBC0 | CondenserClassID | **Reservoir** | ❌ TODO |
| 0xBC1 | GeneratorClassID | Generator | ✅ |
| 0xBC2 | PoweredSubsystemClassID | PoweredSubsystem | ✅ |
| 0xBC3 | MyomersClassID | Sensor | ✅ |
| 0xBC4 | GyroClassID | Gyroscope | ✅ (cache gyroSubsystem) |
| 0xBC5 | SinkSourceClassID | Torso | ✅ (cache sinkSourceSubsystem) |
| 0xBC6 | ActuatorClassID | Myomers | ✅ |
| 0xBC8 | WeaponEmitterClassID | Emitter | ✅ (weaponCount++) |
| 0xBCB | JumpJetClassID | AmmoBin | ✅ |
| 0xBCD | MechWeaponClassID | ProjectileWeapon | ✅ |
| 0xBCE | MissileWeaponClassID | GaussRifle | ✅ |
| 0xBD0 | BallisticWeaponClassID | MissileLauncher | ✅ |
| 0xBD3 | SubsystemMessageManagerID | **SubsystemMessageManager** | ❌ TODO |
| 0xBD4 | GaussWeaponClassID | Emitter (PPC) | ✅ |
| 0xBD6 | MechTechClassID | HUD | ✅ (cache hudSubsystem) |
| 0xBD8 | LegSubsystemClassID | **Searchlight** (: PowerWatcher) | ❌ TODO |
| 0xBDC | HeatableClassID | **MechTech** | ❌ TODO |
| 0xBDE | DisplayClassID | **ThermalSight** (: PowerWatcher) | ❌ TODO |

Remaining subsystem classes for a full walk: Reservoir, HeatSink-bank variant,
Searchlight, MechTech, ThermalSight, SubsystemMessageManager, Actuator (fallback).
Then the Mech member layout (subsystemArray/subsystemCount + the capability chains
controllable/heatable/weaponRoster/damageable + the alarms/reticle/state-indicators
/SequenceControllers/NameFilter/CStrings embedded members) must be named in
MECH.HPP for the ctor to set them.

## 5.3.17 — the subsystem side of the damage cascade

- The subsystem's PRIVATE zone is now ARMED (binary @4abf28): owned by the
  SUBSYSTEM (`new DamageZone(this, 0)` — Joint.cpp:512 proves subsystems are
  valid zone owners), named, `defaultArmorPoints = structureReference`,
  `damageScale[5] = armorByFacing[]` normalized with the hull-zone rule
  (1/(raw×points); guard: structureReference<=0 leaves the zone inert).
  This makes crits MEASURABLE.
- `ApplyDamageAndMeasure` (@4ac07c): TakeDamage + return the zone-level
  delta (the crit accountant charges it).
- `ForceCriticalFailure` now also sets `SetSimulationState(DestroyedState)` —
  the subsystem SIMULATION state enum (Destroyed=1/Exploding=2, from the
  binary PrintState string pool) added to MECHSUB.HPP; both weapon FSMs'
  `GetSimulationState() == 1` hard gate is now live (destroyed weapons stop
  firing — fight-verified).
- `IsVitalSubsystem()` accessor (vital crit death = mech kill in
  SendSubsystemDamage).