Entry 296

Gorre and its two brothers ported: patterns and tables solid, per-brother camera math approximated

2026-09-28battledecomp

Filescrates/piney-battle/src/boss/gorre.rscrates/piney-battle/src/boss/gorre/brother.rscrates/piney-battle/src/boss/gorre/tests.rscrates/piney-battle/src/boss.rscrates/piney-battle/src/boss/kyvia.rscrates/piney-gen/src/manifest.rscrates/piney-data/src/pack.rscrates/piney-world/src/combat/boss.rscrates/piney-game/src/session/tests/survey.rscrates/piney-game/src/fx.rsdocs/engine/boss-gorre.md

Gorre (ccBoss05, boss05.cpp, bossTbl row 4, file x51) is Outbreak's second phase boss, event 218 in field 5. Read from OUT gcmn.prg 0x00497010-0x0049e590 via tools/disasm.py (INF's clean listing plus tools/voldiff.py diff ... --volume OUT --overlay gcmn for the codegen delta) and tools/dwarf1.py fn for locals and source line ranges. 33 functions on ccBoss05 (~21 KB) plus 15 on ccBoss05Brother (~9 KB), as the task named them; INF's syms.txt (work/infection/syms.txt) gave the class list and confirms the byte counts almost exactly (33 methods summing near 21 KB, 15 summing to 8,956 bytes).

What Gorre turned out to be

Unlike Fidchell (ccBoss04, a single character), Gorre fights with two ccBoss05Brothers, each its own ccChar beside it. ccBoss05::Main (OUT 0x00498110) is its own override, not the base's: it inlines the centre's distance and heading itself, calls its own CalcRealEx(0) (the base's per-character CalcReal plus a protect-gauge broadcast - effProtect/SetProtect - to both brothers every frame), and drives Think and Slave through the same indirect-vtable-call pattern the base's generic Main uses (confirmed by matching the vtable slot offsets against CalcTargetInfo/Move/PreDrawAnm's known positions in ccBoss::Main's own disassembly).

The two brothers are not simulated inline the way Innis's three slaves are (innis::slave_main writing directly into their characters from a single call); each is a full Boss of its own, exactly like Kyvia's core and gomoras (docs/engine/boss-kyvia.md's Tree). crates/piney-battle/src/boss/kyvia.rs's Part<T>, plain_boss, new_char, set_base_param and head are all reused directly rather than re-copied into gorre.rs; Part<T> gained Clone, Default, PartialEq, Eq derives (previously only Debug) so Gorre's own struct, which holds [Part<Brother>; 2] as scratch for the span of one frame, can derive the same set the rest of Boss does. ccBoss05::ExecPatternIndex (0x004978d0) turned out to share Fidchell's exact shape for the -1 wrap, the move_spd/move_vector reset and patterns 9-11's target selection; unlike Fidchell, though, its fallback (patterns 2-8, 12) could call Boss::base_exec_pattern_index directly rather than duplicating it, because Gorre's own act module happens to assign the same numbers to WANDER (8), DASH_CENTER (7), EPITAPH/EPITAPH_NEUTRAL (12), ESCAPE (16) and CHASE (18) that the base module does - checked by hand against every pattern the base fallback can reach.

ccBoss05::Affect (0x00498100) is 8 bytes: jr $ra and nothing else. Gorre itself never takes a direct hit; the fight's win condition is "both brothers' HP at 0", read here by Gorre's own frame polling cx.scene.chars[brother_me].hp each frame rather than the brothers' SendMessage(this, msg) callback the game uses (SendMessage, 0x0049b680-0x0049c228, 2984 bytes - the largest function in the class, and the one this session had least time to read in full: only msg 1 and 4's outcomes, a hit and a death, were confirmed from its branch structure). Once both are down Gorre goes to its Epitaph table (OnThinkDrain); since nothing in what was read ever sets Gorre's own act to DEAD from there (its Affect is empty, so no hit can do it either), a grace period was added (Gorre::epitaph_frames, EPITAPH_GRACE = 300) after which Gorre is taken as finished and exits, the way Fidchell's own dead effect (begin_dead_effect, EffKind::Dead) already does. This is a guess standing in for whatever SendMessage's msg 21 (which touches boss05SuperActTbl and a stat reset, per its touches: line) actually does; the first survey run without it looped in field 5's menus for the full 150,000-frame budget without the story ever completing, and with it outbreak_story_survey reaches "story 218: done" (and, since nothing else in Outbreak was blocked, the staff roll) in 72.99 s wall time.

Tables

tools/disasm.py range ... --data over boss05NormalActTbl (0x005ec060, 39 words to its -1), boss05SuperActTbl (0x005ec100, 47 words) and boss05EpitaphActTbl (0x005ec1c0) gave the real pattern sequences; @1261 (0x005ec228, 12 bytes: 157, 158, 160) and @1777 (0x005ec250, 16 bytes: 157, 158, 159, 160) are OnThinkSkill's and OnThinkMagic's spells. @1672/@2074 (200, 212, 232, 264, each 16 bytes) were found but not placed in any function read closely enough to say what they time; they are not in the manifest. crates/piney-gen/src/manifest.rs gained gorre_anims, gorre_normal, gorre_super, gorre_epitaph, gorre_skills, gorre_magic_skills; cargo run --release -p piney-gen -- gen --group combat regenerated crates/piney-data/src/tables/combat.rs clean, and DATA_VERSION (crates/piney-data/src/pack.rs) moved 17 to 18.

What is not ported

Brother's own OnThinkWave, OnThinkEpitaphWave and OnThinkTornade (1668, 1476, 1356 bytes - real per-brother position and camera math past their locals, which is as far as dwarf1.py fn was read for them) are approximated as holding the brother's formation place, facing Gorre's target. Order's special on values beyond a plain ChangeAction, and all but two of SendMessage's message codes, are not ported. ccBossEffFinalPhotonFlashCreate (OnThinkKerse) is a new EffKind with a placeholder 45-frame life and no picture (crates/piney-game/src/fx.rs returns None for it, as Innis's and Kyvia's own unfinished effects do). None of this was checked against the game running in eemu - the task's harness (tools/test_battle_fidchell_rs.py and battle_probe's fidchell example) was not built for Gorre in this session; docs/engine/boss-gorre.md is filed partial rather than solid for exactly this reason.

What was verified

cargo test -p piney-battle --lib: 106 passed (105 before this entry), including four new boss::gorre::tests (stands_behind_kite_with_two_brothers_beside_it, runs_its_first_pattern_without_panicking, both_brothers_down_drains_gorre, both_brothers_down_ends_the_fight) built on the same Fight harness Fidchell's own tests use, against the real combat.bin tables this environment has for Outbreak (they do not skip). cargo clippy and cargo fmt --check are clean on piney-battle, piney-world, piney-game, piney-gen and piney-data. PINEY_SURVEY_ONLY=218 PINEY_SURVEY_GOD=1 PINEY_SURVEY_FRAMES=150000 cargo test --release -p piney-game outbreak_story_survey -- --ignored --nocapture reports story 218: done. The pilot aids levels_for_boss and approach_roamer (crates/piney-game/src/session/tests/survey.rs) were extended to Gorre the same way they already covered Fidchell (GORRE_LEVEL = 75; Gorre added to approach_roamer's boss codes) - without the level aid the first survey attempt could not be told apart from a genuine stall, since both look like a fight that does not end.