vwemirror/BT411
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

scoring: BIND the scenario role -- one commented-out line zeroed the whole chart

Browse Source 
BTPlayer::scenarioRole was never assigned.  The lookup sat commented out with
"the BT role registry (BTMission::GetRoleRegistry()->Lookup) has no WinTesla
analog, so the scenarioRole set by the base Player ctor stands" -- and the base
ctor sets it to NULL (PLAYER.cpp:680).  So it stood NULL forever.

Every scoring value the game has hangs off that pointer, and the shipped
content authors them correctly.  New ungated receipt in the ScenarioRole ctor
prints what a real mission loads:

  [role] 'Role::Default' model='dfltrole' killBonus=500 deathPenalty=500
         dmgInf=1 dmgRcv=0 bias=1 ff=1 return=1000

That IS the original manual's scoring chart -- +500 a kill, -500 a special-case
death, +1 per damage point.  With the pointer NULL every award multiplied
against zero: kills scored the damage tally alone (4.88), the eject charge read
0 (the field log's "PUNCH-OUT: charge=0 (role killBonus)" = #134's missing
penalty), and the death-cost block was skipped.

The analog DOES exist: Mission::GetScenarioRole(name) (MISSION.h:162) walks
scenarioRoleChain -- the same dictionary BTL4Mission fills via AddScenarioRole()
when it parses the role pages, whose own comment says the WinTesla base exposes
it.  Same lookup, same key.  Falls back to Role::Default when a creation
message names an unknown role (shipped content authors exactly one page), and
logs BOUND/NULL so this cannot fail silently again.

Benched cross-node:
  role binding    player 2:1 BOUND, player 3:1 BOUND
  KILL AWARD      505.88  (was 4.88)   <- chart's +500, verified
  death cost      victim total -500.00 <- chart's -500
  inflicted       still tracking, killer total 1017.32 kills=1

The -500 on an ORDINARY combat death is AUTHENTIC, not a bug: the binary's gate
is advancedDamageOn alone (@004c05c4 tail: `if (player+0x264 != 0) { -role+0x20 }`),
verified in the decomp.  It only shows now because the role finally binds.  It
also reconciles the chart's two death rows: an EJECT costs -500 (death) plus its
self-kill negating its own ~500 award = -1000, and an ammo death costs -500.

CORRECTION to my own earlier note: returnFromDeath=1000 is NOT the chart's
"+1000 starting the game" -- role+0x28 is a lives/return gate (`if (< 1)` ->
mission review, else respawn).  The 1000 is coincidence.  That row is still
unlocated and is most likely console-side.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NCJQkvq6G2JNrpVbA75tVZ
This commit is contained in:
Joe DiPrima
2026-08-07 10:21:15 -05:00
co-authored by Claude Opus 5
parent b2498ca39a
commit a4bfb64ace
3 changed files with 67 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
engine/MUNGA/SCNROLE.cpp
+23
View File
@@ -55,9 +55,32 @@ ScenarioRole::ScenarioRole(const CString &role_name, const CString &model_file)
damageBias = player_data->damageBias;
friendlyFirePenalty = player_data->friendlyFirePenalty;
player_res_des->Unlock();
// SCORE PROVENANCE (ungated, 2026-08-07). Every unmatched row of the
// original manual's scoring chart -- +500 a kill, -1000 an eject, -500
// an ammo death -- is sourced HERE, from the role's GameModel record
// (type 0xf, 7 dwords; the binary's ctor @00429a9c copies rec[0..6]).
// A field log showed `charge=0 (role killBonus)`, and the two ways that
// happens -- record authored zero, or resource lookup missed -- are
// indistinguishable in release because the miss path's Warn/Tell compile
// out (DEBUGOFF.h). One line per role settles it.
DEBUG_STREAM << "[role] '" << (const char *)role_name
<< "' model='" << (const char *)model_file
<< "' killBonus=" << (float)killBonus
<< " deathPenalty=" << (float)specialCaseDeathPenalty
<< " dmgInf=" << (float)damageInflictedModifier
<< " dmgRcv=" << (float)damageReceivedModifier
<< " bias=" << (float)damageBias
<< " ff=" << (float)friendlyFirePenalty
<< " return=" << (int)returnFromDeath
<< "\n" << std::flush;
}
else
{
DEBUG_STREAM << "[role] '" << (const char *)role_name
<< "' model='" << (const char *)model_file
<< "' -- RESOURCE NOT FOUND, all score values default to 0"
<< "\n" << std::flush;
Tell(role_name);
Warn(" does not exists in resource! ");
damageReceivedModifier = 0.0f;
game/reconstructed/btplayer.cpp
+43 -7
View File
@@ -1952,15 +1952,51 @@ BTPlayer::BTPlayer(
}
//
// Look the scoring role up in the role registry (keyed by the role name
// in the creation message, +0x90) and stash it as our scenarioRole. The
// BT role registry (BTMission::GetRoleRegistry()->Lookup) has no WinTesla
// analog, so the scenarioRole set by the base Player ctor stands.
// CROSS-FAMILY: needs BTMission role-registry access. BEST-EFFORT.
// Look the scoring role up in the role registry (keyed by the role name in
// the creation message, +0x90) and stash it as our scenarioRole.
//
// WIRED 2026-08-07. This was commented out with "the BT role registry
// (BTMission::GetRoleRegistry()->Lookup) has no WinTesla analog, so the
// scenarioRole set by the base Player ctor stands". The base ctor sets it
// to NULL (PLAYER.cpp:680), so it stood NULL forever -- and EVERY scoring
// value the game has lives on that pointer:
//
// killBonus -> the kill award basis (authored 500)
// specialCaseDeathPenalty-> the death cost (authored 500)
// damageInflictedModifier-> the per-hit multiplier (authored 1)
// returnFromDeath -> the entry credit (authored 1000)
//
// With it NULL: kill_bonus reads 0 (a kill scored the damage tally alone --
// benched 4.88 instead of ~500), the eject charge is 0 (field log:
// "PUNCH-OUT: charge=0 (role killBonus)" = #134's missing penalty), and the
// death cost block is skipped entirely. That is four rows of the original
// manual's scoring chart, all from one commented-out line.
//
// The analog DOES exist: Mission::GetScenarioRole(name) (MISSION.h:162)
// walks scenarioRoleChain -- the very dictionary BTL4Mission fills via
// AddScenarioRole() when it parses the role pages (btl4mssn.cpp), whose own
// comment already says the WinTesla base exposes it. Same lookup, same
// key, no cross-family gap.
//
CString role_key(creation_message->roleName); // make+0x90
(void)role_key;
// scenarioRole = playerMission->GetRoleRegistry()->Lookup(&role_key); // this[0x7e]+0x50, this[0x82]
if (playerMission != 0)
{
ScenarioRole *found = playerMission->GetScenarioRole(role_key); // this[0x7e]+0x50
if (found == 0)
{
// The shipped content authors ONE role page, "Role::Default"
// (model dfltrole). A creation message naming anything else -- or
// naming nothing -- must still score, so fall back to it rather
// than leave the pointer NULL and silently zero every award.
found = playerMission->GetScenarioRole(CString("Role::Default"));
}
scenarioRole = found; // this[0x82]
DEBUG_STREAM << "[role] player " << BTMatchHostOf(GetEntityID())
<< ":" << (int)GetEntityID() << " key='" << (const char *)role_key
<< "' -> " << (scenarioRole != 0 ? "BOUND" : "NULL (scores will be 0)")
<< (scenarioRole != 0 ? "" : " !!")
<< "\n" << std::flush;
}
if ((simulationFlags & 0xc) == 4)
{
scratchpad/night13/scoreverify.sh
+1 -1
View File
@@ -50,7 +50,7 @@ sleep 2
sleep 5
python ../tools/btconsole.py SV.EGG 127.0.0.1:1501 127.0.0.1:1601 > sv_relay.log 2>&1 &
RELAY=$!
sleep 260
sleep 400
kill $RELAY 2>/dev/null
sleep 3
bt_kill_ours