The Gen 3 extractor emitted eggCycles/eggGroups but never egg moves, so the
gen3_dexnav hidden-mon roll was inert on FireRed: a hidden mon kept its
level-up moves and could never gain an egg move.
Decode the ROM's own gEggMoves table (pokefirered/src/data/pokemon/
egg_moves.h) into pokemon/egg_moves.lua and expose it the same way as
learnsets/tmhm:
- Versions: EGG_MOVES = 0x25EF0C plus the 20000 species offset, the 0xFFFF
run terminator and a defensive per-species cap. The table is one flat u16
stream of `{ species + 20000, move..., 0xFFFF }` runs with no final
terminator, so the first word that is neither a header nor a plausible move
id ends the scan.
- PokemonExtract: extract_egg_moves + write_egg_moves_lua, written to the
cache by run(), required by ready() so a stale cache re-extracts, and
returned as pack.eggMoves. FORMAT_VERSION 3 -> 4.
- Pokemon.eggMoves(species): runtime accessor with the same species coercion
as Pokemon.learnset; nil for a species with no egg move.
- Schemas: eggMoves on monTables/monRecord/writeMon and the Gen 3 species
field list, so a mod reads record.eggMoves as move ids and writes names back.
- Versions.CACHE_VERSION 98 -> 99 to force re-extraction.
The table is sparse: species without egg moves are absent rather than an
empty list, all the way from the ROM to the mod record.
Validated against a supported FireRed USA 1.0 dump: 165 species, 973 moves,
<=8 per species, Bulbasaur {113,130,219,204,80,345,320,174}, Mankey
{157,193,96,68,179,251,279,265} (no Toxic), Mew nil.
Tests: new tests/engine/game3_egg_moves.lua (31 checks ROM-free, 39 with a
ROM), gate_gen3_mod_api 734/734, engine tier 587/587, modkit tier 37/37,
lint --gate clean.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
dev added the public Encounters.tableFor export (MapCatalog resolution,
prefix stripping, route normalization) at the same point in
src/core/game3/encounters.lua where this branch inserts the cooldown
block. The two changes are purely additive, so both are kept: dev's
export sits next to table_for where it belongs, followed by the
cooldown/rate-test block.
scripts/lint.sh auto-merged cleanly, keeping both dev's PATH export for
node_modules/.bin and ~/.luarocks/bin and this branch's luacheck --version
probe.
Re-verified after the merge that dev's RNG-parity commit does not shift
the encounter goldens: 88/88 cooldown checks still pass, including the
Route 1 4000-step counts (401 with the grace period vs 858 without).
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
luacheck 1.2.0 is the newest release and predates Lua 5.5, which made
generic-for control variables read-only. Built against Lua 5.5 it dies
loading luacheck.standards -- before linting a single file -- so the T0
gate can never pass, and the failure surfaced only as a raw loader
traceback that looks like a repo defect.
Probe `luacheck --version` before the real run:
- if it cannot run, print the captured error plus the Lua 5.4 install
command and exit 2, matching the existing not-found path
- if it runs but reports Lua 5.5, warn that 1.2.0 predates it and that
results are suspect, then lint anyway
CI is unaffected: it installs luacheck against Ubuntu's Lua, not 5.5.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Completes the pret/pokefirered parity for wild encounter rate gating,
on top of the cooldown port in b84c2c15.
Ports DoWildEncounterRateTest as Encounters.encounterRate(): *16, bike
*80/100, banked failure rate +buff*16/200, flute, Cleanse Tag *2/3,
ability (Stench /2, Illuminate *2), then the 1600 clamp last -- pret's
exact order, which the modifier-order tests pin down.
Ports AddToWildEncounterRateBuff: a failed rate test (or a rate test
that passes but picks no slot) banks the area rate; an active Repel
zeroes the bank. Without this, encounters were still flat-rate rather
than ramping after near-misses.
Also ports the Mach/Acro bike *80/100 modifier.
End-to-end goldens on Route 1 (rate 21, 4000 steps) shift 397->401
with the cooldown and 843->858 without. That matches pret's steady
state analytically: a 6-step cooldown followed by p~0.112 as the bank
ramps gives a cycle of (1-0.94^6)/0.112+6 ~ 10.3 steps, ~10%/step.
The observed 401/4000 = 10.03% confirms it.
tests/game3_rng_test.lua now resets rate modifiers per replay, since
the banked rate makes sequences history-dependent; a map load resets
in-game.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Gen3 (FireRed) had no cooldown between wild battles, so every step on an
encounter tile rolled at the area's full rate -- Route 1 (~21%) landed a
battle on roughly one step in five. Gen2 already had this via
World:wildCooldownStep(); game3 never did.
Port pret/pokefirered wild_encounter.c:
- GetMapBaseEncounterCooldown -> Encounters.mapBaseCooldown: steps of
immunity derived from the area's own rate (rate >= 80 -> none, < 10 ->
8, else 8 - rate/10).
- HandleWildEncounterCooldown -> Encounters.handleCooldown: a soft floor,
not a hard gate -- once the minimum elapses a 5%/step leak lets a battle
through anyway. Includes the White/Black Flute, Cleanse Tag, Stench and
Illuminate modifiers in pret's application order (Cleanse Tag before the
ability mod, which changes the result).
- ResetEncounterRateModifiers -> Encounters.resetRateModifiers, wired to
the two places pret resets: map load (Map.load, including seamless
connection crossings) and battle start (BattleBridge.startWild). The
latter is the bug-#1229 class -- scripted battles and fishing re-arm the
grace period even though no step rolled.
Measured on Route 1 over 4000 steps: 843 encounters before, 397 after;
first encounter never lands before step 7.
Not ported (makes encounters more likely, so it does not affect the
reported symptom): the encounterRateBuff anti-frustration ramp and the
Mach/Acro bike rate modifier.
tests/engine/wild_encounter_cooldown.lua covers the cooldown golden table,
the counter/leak mechanics, all five modifiers, both re-arm paths and the
end-to-end step count.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Collision.installWarps force-marked every map-header warp cell walkable. That
is right for the outdoor MB_WARP_DOOR tiles extract can leave solid, but it also
opened cells that only carry a dead warp event. PalletTown_PlayersHouse_1F has
one at (3,9): the wall tile directly left of the door mat, mid 26, coll 1,
behavior 0x00 MB_NORMAL. pret never lets the player stand there, so the warp
never fires; the forced COLL_DOOR let the player walk out of the house through
the wall and land in PalletTown.
The repair now fires only when the behavior is unreadable (tileset attrs stopped
before that mid, so Collision.behavior returns nil) or is a behavior pret would
actually warp on. Collision.isWarpMetatileBehavior covers 0x60-0x6F plus 0x71,
matching field_control_avatar.c: the arrow warps (TryArrowWarp), the directional
stair warps (IsDirectionalStairWarpMetatileBehavior), and everything
IsWarpMetatileBehavior accepts (cave door, ladder, fall warp, regular warp,
Lavaridge 1F warp, warp door, escalators, union-room warp).
Warp indexing is unchanged: a dead warp on a wall is still registered, only the
grid write is gated.
tests/engine/firered_house_wall_warp_bug2297.lua is ROM-free: it embeds the pret
13x10 map.bin grid and the `building` metatile attributes, translates them with
the real ScriptingCollision.fromCell, binds the real Versions.WARPS table, and
asserts the wall stays solid, the four warp cells still behave, the repair still
applies for unreadable attrs and live warp behaviors, and an exhaustive
0x00-0xFF table pins the predicate. It fails 6 checks against the pre-fix file.
luajit tests/run_engine.lua: 585/585 suites passed.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
- Battle Animation Pipeline:
* Eliminate 90-frame audio polling stall in anim_vm end/waitsound opcodes,
allowing attack animations to transition instantly into target hit flicker
and HP bar drain.
* Correct visual task counting in AnimVm:visualCount to match pret gAnimVisualTaskCount.
* Signal bg task completion on restorebg (vm.args[7] = -1).
* Fix termination conditions on shake tasks (ShakeMon, ShakeMon2, ShakeBattleTerrain),
WaterSport droplet arcs, and StartSinAnimTimer.
* Handle RGB555 color args in start_blend_anim_sprite_color and route P.destroyTask
to AnimTasks.destroy.
- RNG System & Seeding Parity:
* Replace math.random in Party.giveMon with Rng.Random32() for personality values
and Rng.Random() for IV bitfield extraction (HP/Atk/Def/Spe/SpA/SpD) matching GBA format.
* Seed RNG on Title Screen A/Start press via Rng.seedNewGame() (matching SeedRngAndSetTrainerId)
and perturb RNG state on title menu inputs.
* Route NPC overworld idle/wander timers and wild encounter slot rolls through Rng.compat.
- Pokédex Habitat & Area Maps:
* Fix encounter tables extraction and map indexing to resolve Pokemon habitat
locations across Kanto/Sevii maps rather than showing Area Unknown.
* Render steady semi-transparent red overlays for Pokédex area map locations.
- Multichoice Menus & UI:
* Implement multichoice table extraction and data population for Viridian blackboard
status ailments and special dialog interactions.
* Improve choice box borders, pagination, and Town Map wall viewing.
- Overworld Events & Flags:
* Fix Route 1 entrance girl NPC spawning and script triggers after starter selection.
* Ensure consistent save reload state in Oak's Lab.
Add ROM-backed contextual L/R help with article navigation and safe pause/resume. Record recent field scenes and gameplay events for the skippable Continue recap, with playback isolated from live progress.
Integrate extraction and cache validation, preserve bundled script flags during import, and cover navigation, persistence, event recording, playback isolation, and both ROM revisions with tests.
A link battle runs under cable rules and LinkBattle keeps them: openItems
prints "Items can't be used in a link battle!" and the wire knows move,
struggle, locked, switch and RUN. Right for the Cable Club, wrong for a
mode that is not the Cable Club -- a battle royale played over
LinkState.newFromSession fights with real, damaged parties, and the
potions and X items on the ground are its whole economy; against a bot
the bag works, against a person it says no. The mode cannot fix it:
submit, resolveLockstep and the decoder are closures, and there is no
action that means "nothing" to spend the turn on.
opts.items = true on newHost/newGuest (off by default, set on both
machines like turnLimit): the bag is BattleState.openItems -- the vanilla
BagMenu against this battle, whose picker already offers the clamped
copies -- and the effect is ItemEffects.use as in any fight. What
changes is what spending the turn means: itemUsed puts { kind = "item",
item, index, move } on the wire as the turn's action instead of running
the AI's reply. Both machines and a spectator resolve it before switches
and moves: the user's side is already applied (the bag did it), the other
side applies the same effect to its own copies of that side through
src/link/LinkItems.lua -- ItemEffects.use behind a battle whose player is
the user's battler and a save whose party is the user's copies, nothing
consumed -- and prints "<name> used <ITEM>!" plus the effect's lines.
Every effect reachable in a battle is deterministic, so the per-turn
hash still agrees. BagMenu hands itemUsed the item, target and move it
used; a local battle ignores them. A ball on the cable is refused
(ItemEffects), the way the doll is in any trainer battle.
tests/engine/link_items.lua: cable rules by default; opted in, the host's
POTION heals the guest's copy before the moves, both print the lines,
one turn, hashes agree; a spectator fed the same two messages heals its
copy too. docs/modding.md and docs/rfcs/0021-link-battle-items.md.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The Android APK job fails inside android-actions/setup-android@v3, before
any step of ours runs: the action installs its default package list,
"tools platform-tools", and Google's repository no longer serves the
legacy `tools` package -- "Warning: Failed to find package 'tools'", then
sdkmanager exits 1 and the job is over. The build never needed it; the
next step installs platform-tools, the platform, build-tools and the
NDK by name. So the action is told to install platform-tools alone.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Shane McGovern