Entry 307

Outbreak's foes that took no damage: Exdefense, and two area rules

2026-09-30battlevolumes

Filescrates/piney-battle/src/damage.rscrates/piney-game/src/session/tests/survey.rscrates/piney-game/src/session/tests/side_events.rscrates/piney-game/src/session/tests/shrine.rscrates/piney-game/src/session/tests/data_bug.rscrates/piney-game/src/session.rsdocs/engine/battle.mdplans/outbreak-side-events.mdBUGS.md

Worklog 306 left three side events of Outbreak open, each on a foe that took no damage: MARLOWE-2 (256) in room 0-2 of field 84's dungeon, TERASHIMA-1 (262) in room 2-6 of field 90's, and SERVER-3 (263) at Black Death. None of them was a fault in the port. The game bars these foes from one kind of hit, and the autopilot kept using that kind.

Where the damage stopped

The survey got a log of each foe's HP and last affect (kind, damage, skill, attacker) as they change (PINEY_SURVEY_HITS). Every hit landed and every affect reached the foe. CalcBattleDamage returned 0:

• 256. Gaia Turtle (OUT row 114): pDef 700, mDef 1,220, Exdefense 0x1. Every ATTACK sent 0; Juk Kruz (skill 245) took 194 now and then.
• 262. Deadly Present (row 243): pDef 1,920, Exdefense 0x1. Juk Kruz took 383 a cast until the walk gave the last two up. With the doors shut, Kite then went at them with ATTACK for 0.
• 263. Black Death (row 176): pDef 4,500, mDef 2,400, Exdefense 0x2. Kite cast Juk Kruz for 0 on every try. Its 9,990 is the physical PP defence, the protect gauge's, not pDef.

The game's rule

ccChar::CalcBattleDamage (INF gcmn 0x0056d910) reads the row's exdefense (+0x64). Bit 0x1 bars a physical skill while the target's real pDef is not below the row's. Bit 0x2 does the same for magic and mDef, and bits 0x4-0x80 for a skill of that element. A barred hit does 0 and fills no protect gauge. Infection keeps the whole value while any matching defence holds (0x0056dc64). From Mutation on, only the bits still held count (MUT 0x005933f8, OUT 0x0058f5c0).

tools/battle.py damage runs Outbreak's own code. Kite at 75 does 0 to the Gaia Turtle with ATTACK, and 31-35 a blow to Black Death. So Black Death is beaten with blows (about 300 of them) or with spells once a debuff has lowered its mDef.

test_battle_rs.py agreed: damage, exdefense and affect gave 0 mismatches in 300-500 cases on INF, MUT, OUT and QUA, and protect on INF and OUT. A new test pins the rule on Outbreak's rows: exdefense_bars_a_kind_while_its_defence_holds.

The autopilot

StoryPilot::choose chose the kind by the higher defence alone. That meant spells at Black Death, and blows (with Skills! to the members) at the Gaia Turtle. Now a sure hit (CalcBattleDamage at 100 on a copy of the foe) tells it whether a kind is barred:

• A foe barred from one kind gets the other.
• Skills that would do 0 are left out, both Kite's and the members'.
• choose also skipped a foe the walk had given up (hopeless), even while the walk still went at it behind shut doors. The walk now records its foe (Walker::foe), and that foe is fought with skills.

SIGN-7 (265) then met its Data Bug (row 261, level 68) at another moment. The bug felled a Kite of 75 (1,395 HP) with two blows of 703 in one frame, at 47,391. Before this change, the run had finished at 131,678 by luck. The side events now raise a lone Kite to 90 (a harness aid, as god is). The story keeps its 75.

Now 25 of the 26 side events finish; only GOB3-5 (254) is open. The new times are 256 at 66,308, 262 at 41,348, 263 at 20,134 and 265 at 62,918. server_3_black_death_falls_to_blows fails with the old choice (open after 28,000 frames) and passes with the new one.

Two area rules the harness found

test_battle_rs.py area_damage showed two differences on the later discs. Both are ported:

• From Mutation on. ccSkillDamage gives up on a downed target only in its single-target branch (MUT 0x00599988). An area skill aimed at a downed foe still hits the foes around it. Infection returns at once. The port had Infection's rule everywhere. Test: an_area_skill_at_a_downed_foe_hits_those_around_it_from_mutation_on.
• From Outbreak on. All three area rules set the types to 0xe0 once they walk the foes' list: ccSkillDamage OUT 0x00595dac, its point variant 0x005962f8, ccSkillDamage2 0x0059664c, and QUA 0x004886ec, 0x00488c38, 0x00488f8c. An area skill aimed at an enemy then also hits a Data Bug or a boss in reach. Test: an_area_skill_takes_in_every_foe_type_from_outbreak_on.

Both tests fail without their fix. With both fixes, area_damage and skill_damage show 0 mismatches in 1,000 cases on INF, MUT and OUT. Before, the misses were 26 of 300 and 5 of 500.

Infection's drained forms

Reported as "too tanky" (BUGS.md). Two paths were tested:

• A drained Gremlin (row 129, 50 HP) is a target after its 60-frame grace and falls to the attack button: a_drained_form_falls_to_kites_blows.
• A drained Data Bug's base form (row 113) has no virusFlag, and a hit of 100 takes 100: a_drained_data_bug_s_base_form_takes_whole_hits. affectEnemy reads the flag from the enemy itself (INF 0x00433c0c: byte 0x250, bit 6).

Every drained form on Infection is a level-0, 50-HP row with no Exdefense, so the report stays open. 186 of Infection's 303 rows do carry an Exdefense bit, which leaves a barred kind of hit at 0. That may be what the report saw.

Checks

• Suites: piney-battle 117 + 5, piney-world 88 + 3 + 12, piney-game 171.
• Clippy (the three crates, all targets) and fmt are clean.
• Enemy AI (15 checks) and enemy motion (11) match on OUT at 200 cases.
• test_battle_kite_rs.py on OUT: main_run differs in 10 of 200 cases and main_run_ai in 31, each in one float of an AI record. Infection and Mutation match. This was there before these changes.
• outbreak_whole_story ends at 408,000, from 463,200 (203 at 63,300, from 77,700). mutation_whole_story ends at 506,100, from 613,800.