entityRenderAll() (src/dusk/entity/entityrender.c) ports ac2's
scene.c:sceneRender() camera/view-matrix/draw-loop logic as a standalone
function - find the active camera, bind SHADER_UNLIT with its projection/
view, draw every COMPONENT_TYPE_RENDERABLE entity. display.c calls it each
frame between the screen clear and screenUnbind(), the same spot the old
sceneRender() used to sit.
A single-scene scene manager (src/dusk/scene/, ported from ac2 minus its
prefab system and physics/trigger stepping - that module hasn't been
ported yet) sits alongside this but isn't wired into the engine loop yet;
only one scene can ever exist (no SCENE_COUNT_MAX array).
entitymanager.c/h removed entirely: its three arrays (entities/components/
entitiesWithComponent) are now plain globals in entity.h/.c
(ENTITIES/COMPONENTS/ENTITIES_WITH_COMPONENT), and every entity/component
function across the whole ECS (entity, component, position, camera,
renderable, physics, trigger, animation) drops its entitymanager_t *mgr
parameter to operate on them directly - there was only ever one instance of
that struct in the whole engine, so the indirection bought nothing.
New duskscript bindings (src/dusk/script/native/entity/) expose entityAdd,
entityAddComponent/entityGetComponent, entitySetName/entityGetName/
entityFindByName, entityDispose/entityDisposeDeep,
entityPositionSetLocalPosition/entityPositionLookAt,
entityRenderableSetMesh (by name - cube/plane/sphere/capsule/triprism/quad),
and COMPONENT_TYPE_*/ENTITY_ID_INVALID/COMPONENT_ID_INVALID constants.
assets/init.ds uses them to spawn a camera (via LookAt from (3,3,3) at the
origin), a cube and a plane as the engine's first real content on the new
ECS.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
rpg/ui/scene moved to archive/ earlier left a handful of otherwise-generic
systems with dead references into them:
- asset/loader/chunk existed solely to load rpg overworld map chunks
(tile_t, worldpos_t) - deleted outright, along with its
ASSET_LOADER_TYPE_CHUNK entry in assetloader.c/h.
- save/slot wrapped party_t/storyflagval_t/map-name/cutscene-path, which is
100% rpg save data - deleted outright (save.c/h, savedevice.c/h and the
Linux save backend keep their generic settings/device-discovery logic,
just lose the slot read/write pieces that only existed for it).
- locale/localemanager.c and test/CMakeLists.txt/test/save/* had a few
stray includes and test cases that only existed for the now-gone code.
Also restores assets/locale/en_US.po (copied back from archive/assets/) -
locale init still needs a real locale file to boot into, even with the
asset-moving/archiving done separately.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Remove the per-array type category so arrays can mix types like structs.
Add 13 passing and 125 expected-failure duskscript test fixtures with a
table-driven runner. Also includes the in-progress entity/physics and
archive moves from the working tree.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
assets/init.ds is compiled and packed as init.dsb like any other script, and
engineInit() runs it as the very last init step - after scriptInit() and
sceneInit(), right before the initial scene is set - so it's free to use any
engine system (display, audio, save, rpg, ui, scenes, ...) without having to
special-case startup ordering. A syntax error or runtime failure in init.ds
is fatal, same as every other errorChain()'d init step.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
scriptLoadHandler used fopen/fread directly, but every .ds a script include()s
is packed into dusk.dsk (compiled to .dsb) like every other asset now, not a
loose file next to the binary - raw filesystem reads never found them outside
a dev checkout. It now goes through assetFileInit/Open/Read/Close like the
rest of the engine's asset loaders, failing the load the same way (returning
SCRIPT_LOAD_FAILED) for a missing entry or one bigger than capacity.
test_duskscript's fixtures moved to their own assets/ subfolder and pack into
a small test.dsk with the same tools.asset.pack used for the real dusk.dsk,
and the test opens that archive straight into ASSET.zip (skipping the full
assetInit()/assetInitPlatform(), which expects a launched engine) so the
tests exercise the real compile -> pack -> asset -> runtime pipeline instead
of a filesystem shortcut around it.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
core/CMakeLists.txt and vm/CMakeLists.txt each listed every .c file across
every sub-folder in one flat target_sources() call. Split that into one
add_subdirectory() per folder (array, bytecode, call, coroutine, error, gc,
heap, number, ops, promise, runtime, scope, string, struct, text, value and
builtin/{array,core,math,string} under core/; execute, image and ops under
vm/), each with its own small CMakeLists.txt listing just that folder's
sources - matching the convention already used in src/dusk/display/.
Verified the Dusk binary and the duskscript test suite still build and pass
identically; test_battle and test_entityinteract already failed the same
way before this change (stale against unrelated rpg/battle API changes), so
left untouched here.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
tools/asset/pack now special-cases .ds files: each one is compiled with
duskscriptcc to a temp .dsb (no .map, nothing reads one yet) and that gets
packed under the same path with the extension swapped - the raw .ds source
never ends up in the archive. Every other asset type is unchanged, packed
byte for byte as before.
CMake passes --duskscriptcc "$<TARGET_FILE:duskscriptcc>" and adds a DEPENDS
on it, but only on the linux/knulli host target where duskscriptcc is
actually built (see its own gating in src/CMakeLists.txt) - a cross-compiled
console target has no host-built compiler to point at, so for now it can
only pack an assets/ with no .ds in it; the packer raises a clear error
rather than silently skipping or mis-packing one if it finds one anyway.
The manifest's cache-skip check also now folds in duskscriptcc's own mtime,
so a compiler rebuild invalidates every .ds's cached result even when none
of their own source files changed.
Verified by hand: a .ds under assets/scripts/ packs as the matching .dsb
(confirmed by unzipping the resulting dusk.dsk), an unrelated rebuild
correctly skips re-packing, touching duskscriptcc correctly forces a
re-pack, and a syntax error in the .ds fails the build loudly with
duskscriptcc's own error message rather than silently succeeding.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Folds what used to be duskscriptnative.h's includes (builtin/value/array/
struct/promise helpers a native function needs) directly into script.h,
alongside the runtime and VM includes already there. scriptnativeconsole.h
now just includes "script/script.h" instead of enumerating the same list
itself - any new native function bindings can do the same.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Scripting is now a standard part of the engine rather than an optional
feature: removes the DUSK_SCRIPT CMake option and its compile definition,
makes src/dusk/CMakeLists.txt add_subdirectory(script) unconditionally, and
drops the #ifdef DUSK_SCRIPT guards around scriptInit()/scriptDispose() in
engine.c. duskscriptcc and its test still only build on the linux/knulli
host target (it's a dev tool, not part of the runtime), now gated on that
alone instead of DUSK_SCRIPT AND that.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
duskscriptnative.h had exactly one consumer (scriptnativeconsole.h) left
after the duskscriptvm.h removal; folds its includes in directly rather than
keeping a single-consumer indirection.
Moves scriptnativeconsole.c/.h into their own native/ subdirectory, giving
it the same add_subdirectory(native) + CMakeLists.txt treatment as core/
and vm/, so future native function bindings have an obvious home.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
It had exactly one consumer (script.h) and added nothing of its own beyond
re-including duskscript.h + image/scriptvmattach.h + image/scriptvmrun.h -
a pointless indirection now that it's not a standalone embedding SDK with
multiple external consumers.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
scriptvmtypes.h, scriptcompiletypes.h and scriptnode.h each crammed several
unrelated struct definitions into one file, unlike the rest of dusk (see
src/dusk/display or src/dusk/rpg/battle), which always gives a type its own
file named after it, with other types reached by an explicit include or a
forward declaration.
- scriptvmtypes.h: scriptvmfunction_t moves into scriptvmfunction.h (next to
the functions that already work with it), scriptvmname_t into
scriptvmname.h, scriptvmhandler_t into scriptvmtable.h, and scriptvmimage_t
gets a new scriptvmimage.h. scriptvmtypes.h keeps working as a thin
re-include umbrella for the ~45 vm files that reach into several of these
at once, the same role scripttypes.h already plays for the base types.
- scriptcompiletypes.h: scriptlabel_t moves into scriptbuilderlabel.h,
scriptloopcontext_t into scriptcompileloopcontext.h, scriptincluderesolver_t
into scriptcompile.h, and the big scriptbuilder_t gets a new scriptbuilder.h.
Breaking the resulting include cycle needed scriptbuilderlabel.h and
scriptcompileloopcontext.h to forward-declare scriptbuilder_t/scriptlabel_t
rather than pull in full definitions they only ever use by pointer, and
scriptcompile.h to define scriptincluderesolver_t before including
scriptcompiletypes.h instead of after.
- scriptnode.h: scriptfuncnode_t (a function declaration, not a kind of
expression/statement node) moves into a new scriptfuncnode.h.
Also gives scriptbuilder_t and scriptlabel_t proper struct tags (previously
anonymous, which doesn't allow forward declaration) to match the project's
_s tag / _t typedef convention from the previous pass.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Struct tags throughout duskscript reused the typedef name itself (struct
foo_t { ... }), unlike the rest of dusk which always tags with foo_s and
typedefs foo_t - renames all 15 tagged structs in src/dusk/script and
src/duskscriptcc to match. Also drops a vestigial duplicate forward
declaration of scriptsource_t left over from the port.
Unifies all 144 #ifndef/#define/#endif include guards in the ported code to
#pragma once, matching the rest of the codebase (which was already a mix of
25 files doing it the house way and 144 not).
Packs scriptvmfunction_t's isAsync/isRoot/isScript and scriptfuncnode_t's
isAsync/isInline/isNoInline/isRoot/isScript into single uint8_t flags
fields with bit constants, mirroring the SCRIPT_BINARY_FLAG_ASYNC pattern
the bytecode format already uses and the flags-byte convention used
elsewhere in dusk (animation, display). While there, zero-initializes both
scriptfuncnode_t allocation sites (scriptAlloc never zeroes memory, so
several of its fields - not just the ones being packed here - were
previously read uninitialized before their first explicit write), and
removes scriptfunction_t's dead `plain` field, which was always the exact
inverse of isAsync and never actually read anywhere.
Converts the couple of struct-level doc comments that used the ported
project's Javadoc-with-@param style to dusk's plain // prose convention,
and adds missing rationale comments to scriptruntime_t and scriptstring_t,
which had none at all.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
util/memory.h had landed as the very first #include in ~30 files from the
mechanical memcpy/memcmp conversion, ahead of each file's own umbrella type
header - moves it back after, matching the rest of the codebase's ordering.
Also updates a scriptplatform.h comment left over from the int/fixed/bignum
split, and drops an unused util/string.h include in scriptcompile.c.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Brings over the duskscript VM and compiler from dusk-vm: the runtime
(core/vm) lives under src/dusk/script, gated by DUSK_SCRIPT and wired into
the engine lifecycle with print()/consolePrint() routed to the console; the
standalone duskscriptcc compiler stays a lean host tool (linux/knulli only).
Collapses int/fixed/bignum into a single float-backed SCRIPT_VALUE_NUMBER
throughout the lexer, compiler, bytecode format and VM, removing the old
int16-rollover fast paths and the type-conversion builtins they needed.
Fixes a real big-endian bug in the bytecode's 32-bit number reader (found
while adding GameCube/Wii endian support via util/endian.h) and converts
src/dusk/script and duskscriptcc to dusk's memory/string/assert utilities,
which surfaced two more latent bugs: a conservative GC stack scan reading
its own local variable's address, and an uninitialized path buffer in
scriptReadScript when the caller path has no directory component.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Save slots now track the currently-active cutscene file (defaulting to a
new placeholder new_game.jsonc intro) so loading a save resumes it;
cutsceneSystemStartCutsceneWith persists this on every cutscene start, and
a cutsceneSystemLoad ordering bug (loadedFile stamped after starting) is
fixed so the persisted path is always correct. Also adds a
battlefighterspecies_t/battlespeciesid_t stub for upcoming species data.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Replaces three independently hand-rolled battlefighter_t JSON readers/
writers (party.jsonc roster loading, BATTLE_START enemy definitions, and
the v1 save-file party member format) with one generic pair in
battlefighter.c: battleFighterParse(fighter, obj) and its inverse
battleFighterSerialize(fighter, doc). Covers team/controller/status
(optional, default ally/player/normal), the 7 base stats (required),
current health/mp (optional, default to max), and an optional abilities
name array.
- party.c: partyInit now parses each roster member via battleFighterParse
directly, dropping the old partyParseStats/partyParseStatField.
- cutscenebattlestart.c/.h: cutscenebattlestartenemy_t now embeds a full
battlefighter_t, parsed once at Load time; Start just copies it into the
enemy pool instead of re-initializing. battletest.jsonc's enemies are
now flat (dropped the "stats" wrapper) to match.
- saveslotver1.c/.h: rewritten in place to use the shared parser/
serializer, changing the v1 on-disk field names/shape (no migration -
acceptable pre-release, documented in saveslotver1.h).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
- party.jsonc's "members" is now a key/value object (e.g. "hero",
"companion") instead of an array - partyInit seeds party_t.keys
alongside members, and partyGetMemberIndexByKey/partyGetMemberKey do the
lookups. Cutscene items (PARTY_ADD_MEMBER/PARTY_REMOVE_MEMBER/
PARTY_HEAL_MEMBER) now take a "key" JSON field, resolved to a roster
index at load time, instead of a raw index.
- Battle UI shows who's acting: uibattlemenu.c draws the current fighter's
party key above the action menu, and uibattlehud.c draws each ally
fighter's key as an extra line above their HP/MP.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
- Battle selection menu can now go back a fighter to re-decide their move,
and battle win/loss no longer leaks the round's remaining queued move
cutscenes into the ending state (cutsceneSystemDiscardQueued).
- Party roster is now driven by party.jsonc: every defined member always
exists (partyInit), while party->order alone determines who's currently
active - partyAddMember/partyRemoveMember (and their cutscene items) now
activate/deactivate an existing roster member by index instead of
creating/destroying one, with removal compacting order to stay packed.
- New PARTY_ADD_MEMBER/PARTY_REMOVE_MEMBER/PARTY_HEAL_MEMBER/PARTY_HEAL_ALL
cutscene items.
- saveSlotInit/partyInit now return errorret_t and take their target
struct directly (saveslot_t*/party_t*) instead of implicitly operating
on SAVE.slot; saveSlotReadJSON resets through saveSlotInit rather than
duplicating the reset logic.
- Test coverage: stub the JSON asset loader in save test fixtures so
partyInit's real party.jsonc load runs against an in-memory doc instead
of hanging with no asset system initialized.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
cutsceneCutsceneResolve and cutsceneSystemLoad could be called with a
name/path pointer that aliases into CUTSCENE_SYSTEM.loadedItems (e.g. a
CUTSCENE item's own name field). cutsceneLoadParse then overwrites that
same buffer while loading the target cutscene, so the subsequent
stringCopy into loadedFile read stale, corrupted memory. Copy the
name/path into a local buffer before use in both functions.
- battleStateSelectionAiDecide: AI-controlled fighters now actually
queue a basic attack against the first living enemy instead of doing
nothing every round.
- battleStateEndingInit: disposes the battle and switches back to
SCENE_TYPE_OVERWORLD instead of leaving the game stuck in the battle
scene forever.
- cutsceneSystemNext now pops each finished item off the front of the
running cutscene (cutsceneRemoveFront) instead of leaving it in place
forever, so a long-running cutscene (many battle rounds splicing
items in repeatedly) no longer grows scene->items unboundedly.
- cutsceneGoTo reworked to match: since consumed items are evicted, a
marker behind the current position can't be found in the live queue
anymore, so it now re-parses the cutscene fresh from its source file
(cutsceneLoadParse, into a persistent scratch buffer, not the stack)
and replaces the live queue with marker-onward from that fresh parse.
Only supported when the running cutscene is the one
cutsceneSystemLoad/cutsceneCutsceneResolve populated.
- Fixed a real bug this surfaced: cutsceneSystemPrepare unconditionally
cleared loadedFile, which ran *after* cutsceneCutsceneResolve had
just stamped it but *before* the cutscene started - wiping it out
immediately and breaking cutsceneGoTo for any cutscene entered via a
CUTSCENE item or cutsceneCutsceneResolve (e.g. the main menu's
NEW_GAME/OPTIONS/QUIT navigation). Now only cleared when switching
away from loadedScene entirely.
- Added cutsceneLoadParse, the shared "lock asset, require loaded,
parse items, unlock" sequence, replacing four near-identical copies
across cutsceneSystemLoad/cutsceneCutsceneResolve/
cutsceneInsertResolveAndSplice/battleStateExecutingInit. Each call
site handles its own fatal-error display; cutsceneLoadParse itself
just propagates.
- Added test_cutscene.c (cutsceneRemoveFront) and a regression test for
the loadedFile bug. Writing the zero-count RemoveFront test caught a
real self-move bug (count == 0 passed dest == src into memoryMove).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
BATTLE_FIGHTER_DAMAGE/BATTLE_FIGHTER_HEAL let a move's own cutscene
apply damage/healing to "target" or "user", resolved from
BATTLE_FIGHTER_POS_TARGET/USER sentinels against the current action by
battleStateExecutingInit. fire/slash now call BATTLE_FIGHTER_DAMAGE on
their target.
BATTLE_POST_MOVE checks fighter health after each move and sets
BATTLE.result (now a real field, previously commented out) to
win/loss, transitioning into BATTLE_STATE_WON/LOST; fleeing sets
BATTLE_RESULT_FLED. All three end states now show a textbox before
moving on to BATTLE_STATE_ENDING.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Each queued action now resolves to its move's cutscene
(cutscenes/battle/moves/<name>.jsonc), spliced into the executing state
with a BATTLE_POST_MOVE checkpoint after each one. Fighter turn order is
tracked via fighterOrder/fighterOrderCount, and resolved actions/pending
counts are cleared so fighters aren't skipped in later rounds. Also fixes
a BATTLE_WAIT_SELECTION deadlock and adds an optional "blocking" flag to
MODAL cutscene items.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Adds BATTLE_START, BATTLE_SET_STATE and BATTLE_WAIT_SELECTION cutscene
items so a battle's state transitions (opening -> pre-round -> selection
-> executing) are chained through the running cutscene, with
cutsceneSystemInsertItemsNext/InsertItems splitting "splice" from
"splice and advance immediately". Adds battlestateselection's
get-current-fighter/player-attack/player-flee API backing the battle UI,
and battleMoveGetIdByName for JSON-authored ability lookups.
Fixes: off-by-one in battleMoveInit that let the first-defined move
alias BATTLE_MOVE_ID_NULL; a segfault in scenebattle.c from rendering
unfilled (NULL) fighter slots; stale battleGetCurrentFighter/
battlePlayerAttack/battlePlayerFlee references in the battle UI.
Adds assertions across the battle/scene/UI code for fighter-index
bounds and non-null invariants. Updates battletest.jsonc for the new
BATTLE_START item schema.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Splits battlemoveaction_t into its own header, adds a per-state
struct/init/update/dispose file under rpg/battle/state driven by a
dispatch table, switches BATTLE.fighters to battlefighter_t* backed by a
pool, folds current/max health and mp into battlefighterstats_t, and adds
battleFighterDamage/Heal/AbilityAdd. Removes the now-unused battle
cutscene items (START_BATTLE, BATTLE_WAIT_STATE, BATTLE_FORCE_MOVE,
REGULAR_BATTLE) and their JSON lookup helpers, and adds JSON validation
to battleMoveInit. battle.h/battle.c are mid-rewrite around the new state
machine, so some UI consumers (uibattlehud.c, uibattlemenu.c) and
test_battle.c are currently out of sync with the new API.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
cutsceneCutsceneResolve/cutsceneInsertResolveAndSplice no longer inject
a "cutscenes/%s.jsonc" prefix/suffix around an item's "name" field -
matching cutsceneSystemLoad, which already takes a full path. Fixes a
latent double-prefix bug in initial.jsonc's CUTSCENE reference (also had
a "jsonsc" typo) that this change would otherwise have exposed.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
uiLabelInit now marks a fresh label dirty itself, instead of every call
site having to remember to do it. Adds uiLabelSetWrap (NULL buffer
disables wrapping, otherwise requires a positive width) built on the
existing textWrap(), and wires the fatal error overlay's message label
up to wrap at 80% of screen width.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Implements storyInit/storyGetFlagByName/storyGetFlagValue/etc, moves flag
storage into saveslot_t (persisted keyed by name so save files survive
flag reordering), and adds STORY_SET_FLAG/STORY_WAIT_FLAG cutscene items.
Also removes the now-superseded storyflag.c/h stub.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Comments-only change - drops implementation rationale/cross-referencing
detail in favor of shorter one-liners, matching the simplified style
already applied to part of the file.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
cutsceneSystemLoad, cutsceneCutsceneResolve and
cutsceneInsertResolveAndSplice all used to steal ownership of the
asset entry's parsed yyjson_doc (nulling entry->data.json) and force
an immediate assetReapUnused() to avoid handing back a corrupted
cache entry on a later load of the same file - which doubled as a
latent crash if that reap was ever skipped (as it briefly was),
since a stale cache hit would then have a NULL doc.
Turns out none of that was needed: every cutsceneitem_t field that
could reference the doc's memory already copies its string data by
value (cutsceneJsonCopyString into fixed buffers), so the parsed
items don't depend on the doc surviving past cutsceneParseDoc. All
three call sites now parse straight out of entry->data.json while
still locked, then unlock normally - the asset entry stays a normal,
correctly-cached resource, and CUTSCENE_LOADED_DOC/cutsceneLoadedSetDoc
are gone entirely.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Adds CUTSCENE_SYSTEM.loadedFile, set by cutsceneSystemLoad and cleared
by cutsceneSystemPrepare. cutsceneRestart now re-reads and re-parses
that file via cutsceneSystemLoad instead of just rerunning whatever's
still resident in loadedScene/loadedItems, so restarting a file-loaded
cutscene always reflects its current on-disk contents - restart of a
C-authored cutscene is unchanged.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Matches the existing assertNotNull convention in cutscene.c
(cutsceneAppend/cutsceneParseDoc etc.) - these are internal-API
preconditions on trusted callers, not untrusted input.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Nothing ever wrote to it - it only ever passed its own address through
to onComplete/CUTSCENE_CALLBACK/modal-option callbacks, so every call
site now passes NULL directly and the 8KB buffer is gone.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Replaces the separate insert buffer/scene and insertStack return-frame
mechanism with cutsceneAppend/cutsceneAppendNext, so an inserted
cutscene's items become part of the running scene's own array instead
of a jump-and-return. Moves the loaded-items buffer/scene from file-scope
globals into cutscenesystem_t, and extracts shared start-cutscene setup
into cutsceneSystemPrepare.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
cutsceneAppend adds items to the end of a cutscene_t's item array;
cutsceneAppendNext inserts them at a given index instead (e.g.
CUTSCENE_SYSTEM.currentItem + 1, to run immediately after the
currently-executing item), shifting everything from that index
onward to make room. Both are capacity-bounded against the caller-
supplied itemsMax, matching cutsceneParseDoc's convention of never
assuming ownership of the backing buffer's size.
cutscene_t.items drops its const qualifier to allow this in-place
mutation; cutsceneitem_t backing arrays in the test fixtures that get
assigned into a mutable cutscene_t.items are no longer declared const
either, since a const source would otherwise silently violate the
field's new type.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Matches the existing battleability.h/battlefighter.h split - keeps
battle.h focused on the overall battle_t state machine rather than
also owning the action data shape.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Lets a move reference a cutscene to play when used, resolved by name
the same way a CUTSCENE item's "name" field is. Still just a schema
placeholder - no loader consumes moves.jsonc yet.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Groups item data under assets/items/ to match the cutscenes/ folder
convention, updating itemInit()'s asset path accordingly. Also adds
assets/battle/moves.jsonc as a starting point for move data - there's
no JSON loader consuming it yet (battleability_t is still just a bare
enum), so its field shape is a placeholder rather than a contract.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Every CUTSCENE_XXX(...) shorthand (CUTSCENE_WAIT, CUTSCENE_TEXT,
CUTSCENE_ENTITY_WALK_TO, etc.) existed to build a cutsceneitem_t literal
in C; real cutscenes are now authored exclusively as .jsonc files, and
the only other users were test fixtures. Deletes all ~55 of these
macros (and the shared CUTSCENE_ITEM(...) helper they built on) along
with their doc comments, leaving each item's struct/enum/Start/Update/
Load declarations untouched.
test_battle.c, test_cutscenecontrol.c, test_cutscenemaparea.c,
test_cutscenesystem.c and test_entityinteract.c now build their
cutsceneitem_t fixtures as plain struct literals instead.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
CUTSCENE(...)/CUTSCENE_REFERENCE/CUTSCENE_CUTSCENE/CUTSCENE_INSERT
existed to build a cutscene_t and its item array directly in C; real
cutscenes are now authored exclusively as .jsonc files parsed via
cutsceneParseDoc, so these are dead weight in production. Deletes the
demo scene that still used them (testcutscene.h, wired to NPC #3) along
with its now-dead include in npc.c. test_cutscenesystem.c - the only
other user, building cutscenes in-memory for tests - now writes the
equivalent struct literals directly, matching the convention already
used by test_battle.c/test_entityinteract.c.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Cutscenes now load through the generic ASSET_LOADER_TYPE_JSON loader
and parse straight into fixed static buffers via new cutsceneParseDoc/
cutsceneParseItem/cutsceneCountItems helpers (cutscene.c), removing the
bespoke ASSET_LOADER_TYPE_CUTSCENE pipeline entirely. Only one cutscene
(plus one INSERT target) is ever resident, so each resolve locks,
steals the parsed doc, unlocks, and force-reaps immediately - fixing a
stale-cache-reuse bug where revisiting a cutscene name after a
different one could return data from the shared buffer's previous
occupant.
ENTITY_WALK_TO and MAP_AREA_WAIT now store their waypoints/area ids in
fixed-size arrays inline on their own item struct instead of a shared
bump-allocated pool, so pool/poolOffset are gone from every item type's
Load callback. cutscene_t.dataSize (and the CUTSCENE(...) macro's SIZE
parameter) is removed along with it - it sized entries out of that same
now-deleted pool and had no other reader.
NPC #4's battletest cutscene and the main menu's cutscene now resolve
their target by name on interact instead of eagerly at init, so a
resolved-once pointer is never held past the point its backing buffer
can be reused for something else.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Replaces the build-time item.json/tools.item.py code generator with a
single item_t ITEM registry populated at runtime by parsing
assets/items.jsonc, so items/types can grow without a rebuild. Adds
itemGetDefById/itemGetTypeById/itemGetIdByName/itemTypeGetOrCreate,
updates inventory/backpack to size off the new *_DEF_COUNT_MAX caps
and look up types dynamically, and reorders engine init so rpgInit
(which loads items) runs before uiInit (which needs the item type
count to size the backpack's tabs).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Splicing a JSON-authored INSERT target by name previously resolved
through the same shared CUTSCENE_LOADED_ITEMS/POOL buffer as whatever
outer cutscene was currently running, so loading it would clobber the
outer cutscene's own items mid-playback - that's why it was reverted.
Gives INSERT its own smaller CUTSCENE_INSERT_ITEMS/POOL buffer and lets
the cutscene asset loader be told which buffer set to parse into, so an
INSERT's dynamic load no longer stomps the running cutscene.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
memoryAllocate() becomes a macro capturing __FILE__/__LINE__ into a
fixed-size tracking table when DUSK_TRACK_MEMORY is enabled (Linux only
for now), so a leak surviving to shutdown can be traced back to its call
site. Replaces engineDispose's hard assertTrue with memoryCheckUnfreed(),
which reports unfreed allocations (with call sites when tracking is on,
just a count otherwise) instead of aborting.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Unlike CUTSCENE (a one-way jump that replaces the running cutscene),
INSERT splices a referenced cutscene's items into the running one and
resumes right after itself once they run out. Backed by a small return
stack on CUTSCENE_SYSTEM so inserts can nest.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
cutsceneitemtype_t moves from cutsceneitem.h into cutsceneitembase.h,
alongside a new CUTSCENE_ITEM(TYPE, UNION_NAME, ...) shorthand that
fills in a cutsceneitem_t literal's .type and union member together.
Every CUTSCENE_* authoring macro (CUTSCENE_WAIT, CUTSCENE_TEXT,
CUTSCENE_AUDIO_PLAY, etc.) moves out of the single, ever-growing
cutscene.h into the header of the item type it actually authors,
built on top of CUTSCENE_ITEM. cutscene.h now only holds cutscene_t
itself plus the CUTSCENE/CUTSCENE_REFERENCE/CUTSCENE_CUTSCENE macros,
which aren't tied to any one item type. Every relocated macro also
gained a proper JSDoc-style comment block with @param tags.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Adds a battleability_t enum (starting with FIRE) and a fixed
abilities[8] array on battlefighter_t. Extends the START_BATTLE
cutscene item to parse an optional per-enemy "abilities" JSON array
(via a new cutsceneJsonLookupBattleAbility name lookup) and copies
abilities onto every spawned fighter, from the party roster for
allies and from the parsed config for enemies.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
party.c's functions no longer take a party_t* - they always operate on
SAVE.slot.party, since gameplay only ever has one active party. That
forced the same change on saveSlotInit/saveSlotNewGame (they called
partyInit/partyAddMember internally), so those are now parameterless
too, always resetting/seeding SAVE.slot. Callers building a save at a
specific index (new-game creation, slot deletion) now set
SAVE.slotCurrent first instead of building an independent local struct -
safe since the select-save screen is only ever reached before any
gameplay starts.
saveSlotWriteJSON/saveSlotReadJSON deliberately keep their explicit
saveslot_t* parameter - they're genuine serialization primitives used
by the save-device layer and tests against arbitrary structs, unrelated
to the single-current-party assumption. saveSlotReadJSON's internal
reset is now a plain memset instead of delegating to the now-global-only
saveSlotInit().
Updates test_save/test_savedevice/test_savedevicelinux/test_saveslot for
the new signatures. Note: test_savedevice/test_savedevicelinux/test_save
have pre-existing failures in this sandbox unrelated to this change -
savetestfixture.c swaps $HOME, but saveDeviceLinuxGetDirectory actually
derives the save path from ASSET.baseDirectory (the executable's own
location), so the fixture's sandboxing never actually applies. test_
saveslot.c (which doesn't depend on that fixture) passes 15/15.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
testBattleStart() was never actually called anywhere (its only trace
was a stray unused include in uimainmenu.c) - battletest.jsonc's
START_BATTLE/REGULAR_BATTLE flow replaces it.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
REGULAR_BATTLE previously never completed on its own, hanging any
cutscene that reached it - START_BATTLE already waits out the whole
encounter and returns to the overworld before this item ever runs, so
it's now a completes-immediately hook point for future regular-battle-
specific follow-up.
Also updates test_saveslot.c for the mapName/partyMembers schema
additions: the round-trip test now writes a valid map name for its
in-use slot, "missingFieldsUseDefaults" is replaced with a test for the
blank-slot short-circuit it was actually hitting, and a new regression
test locks in that an in-use slot with no map name is treated as corrupt.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Replaces the hand-rolled mapName missing/null/empty check in
saveSlotVer1ReadJSON with a reusable macro.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
New games now go through saveSlotNewGame (starting party + default map
name) instead of the empty saveSlotInit, and selecting a save loads its
stored map instead of the hardcoded boot-time overworld. Save loading now
requires an in-use slot to have a non-empty map name, treating a blank one
as corruption instead of silently faking a default.
Also fixes a real off-by-one in the readString save-JSON macro: it passed
its maxLength (documented as excluding the null terminator) straight
through as stringCopy's destSize, which needs +1 for the terminator - so a
string exactly at the length limit (e.g. "overworld", 9 chars) always
asserted "src is too long".
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Party moves off a global singleton onto each save slot, letting
START_BATTLE pick specific active-order members to bring into a fight,
and round-trips through save/load via saveslotver1. A corrupt slot load
now flags the cache and shows a red "Corrupted Save" warning in the
select-save UI instead of silently resetting.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Previously deleting a save left the panel in delete mode, and pressing
back while in delete mode closed the whole save-select screen instead
of just returning to the normal load view.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
File read, yyjson parse, and all item dispatch now happen in
assetCutsceneLoaderAsync - none of it touches the main thread now that
CUTSCENE items defer their nested asset resolution to start time. The
sync phase is now just a state flip, kept only so assetUpdate()'s
onLoaded dispatch (which only fires for entries finishing via the sync
path) still runs.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Load map chunks from raw JSON at runtime instead of a baked DCF/DMF
step, add a reusable perf timing module, fix a real PSP logDebug/logError
Memory Stick I/O cost bug, and switch chunk tile storage to a dense
minified grid.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Cuts the six overworld chunk JSON files from 146KB to 10KB by replacing
{"pos":[x,y,z],"type":t,"tile":u} entries with a fixed 256-slot array
(index x + y*16, value null or [type, z]) and dropping pretty-printing.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
New src/dusk/perf module: a simple push/pop tick-stack (perfPush/perfPop/
perfReport) for measuring how long a labeled span of code takes, wired
into engine.c's init/dispose. perfGetTick() is a placeholder - real time
on DUSK_LINUX (clock_gettime) and DUSK_PSP (sceKernelGetSystemTimeWide),
0 elsewhere until a proper cross-platform answer is settled on.
Used it this session to chase a reported PSP hitch when crossing a chunk
boundary. Confirmed mapPositionSet itself (unload/load sweeps, chunk
order rebuild, entity chunkIndex recompute) is consistently under ~3ms -
not the cause. Along the way, found and fixed a real, unrelated bug:
logDebug()/logError() on PSP did a fopen+write+fclose to the Memory
Stick on every single call, which was massively inflating any perf
measurement that logged from inside another measurement (the actual
source of the ~29-128ms numbers first seen). That's now gated behind a
new DUSK_PSP_LOG_FILE CMake option (default off).
Also, as a real test based on this session's findings:
- MAP_CHUNK_LOAD_CONCURRENCY dropped from 2 to 1 (8 and 4 both crashed/
were untested further on real PSP hardware; 1 is confirmed safe).
- MAP_CHUNK_LOAD_DELAY_TEST: an intentionally exaggerated (now 50ms)
artificial delay between starting successive chunk loads, gated in
mapChunkLoadNext/mapUpdate, to observe the effect on framerate.
All perf instrumentation call sites added during this investigation
(assetChunkLoaderSync, assetMeshLoaderSync, mapPositionSet, the map area
callback invocation, engine.c's frame boundary) were removed again once
they'd served their purpose - only the reusable perf module itself and
the two real fixes above remain.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Experimental: skip the offline tools/asset/chunk bake step entirely.
The chunk asset loader now parses the same JSON schema tools/asset/chunk
already consumed (tiles/meshes/entities/areas) directly at load time,
generating the terrain mesh in C instead of reading pre-baked .dcf/.dmf
bytes. The raw assetsraw/maps/*/chunks/*.json files are now the shipped
source directly under assets/, replacing the old baked chunk/mesh/model
output.
Also drops MAP_CHUNK_LOAD_CONCURRENCY to 1 for now, per PSP hardware
testing (8 crashed, 4 untested further - 1 is the known-safe value).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Bulk assets (mp3, png) are already compressed at the file level, so a
DEFLATE pass on top bought negligible space for real CPU cost. Packs
everything into a single ZIP_STORED archive instead, so every entry
(not just locale/*) gets the reliable repeated seeking/re-opening that
libzip only supports for uncompressed entries. Header shrinks from 32
to 16 bytes; version bumped to 2 so a stale pre-change dusk.dsk is
rejected and rebuilt rather than misread.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
saveLoadSettings()/saveLoadSlot() previously let a read failure (e.g. a
version mismatch from the new schema check) propagate all the way up
through saveUpdate()/engineUpdate() to main(), which just prints the
error and exits - taking down the whole game over one bad file. Both
now catch the failure, log it, and fall back to init defaults instead,
matching the existing "no file yet" convention.
mapSetMap() now keeps SAVE.slot.cachedData.mapName in sync with the
active map, so whichever map is loaded gets written out next time the
slot is actually saved.
Splits saveslot_t/savesettings_t into a stable, version-independent
in-memory struct (save/slot/, save/settings/) and a separate frozen
per-schema-version struct (saveslotver1_t/savesettingsver1_t) used only
for JSON serialization, converted between manually. Every write now
stamps a "version" field (SAVE_SLOT_CURRENT/SAVE_SETTINGS_CURRENT), and
every read requires it to match exactly, rejecting anything else -
laying the groundwork for a real migration path instead of silently
misreading old/incompatible save data.
Per-map chunk model paths (maps/<map>/models/chunks/chunk_X_Y_Z_0.json)
were hitting the old 48-char limit, crashing assetEntryInit's length
assert during initial chunk/cutscene loading.
Move chunk assets from a single flat assetsraw/chunks (and assets/chunks)
into per-map assetsraw/maps/<map>/chunks (assets/maps/<map>/chunks), so
terrain mesh/model output no longer collides across maps sharing chunk
coordinates. map_t gains a name[MAP_NAME_MAX] field plus mapSetMap()/
mapIsLoaded() (checks name[0] == '\0') in place of the old bool loaded
flag; mapInit() now just resets state and mapSetMap() does the chunk
grid load. Updated the chunk asset tool and the map editor's dev
server/client to match.
Entities weren't being detached from their chunk's entities[] slot in
two despawn paths (CUTSCENE_ENTITY_REMOVE, item pickup collection) - the
slot leaked forever until the whole chunk unloaded. Both now call
entitySetChunk(entity, CHUNK_INDEX_INVALID) before nulling the type.
entity_t.chunkIndex and entitySetChunk/mapGetChunk now use chunkindex_t
instead of uint8_t, matching mapGetChunkIndexAt's own -1-is-invalid
convention - added CHUNK_INDEX_INVALID next to the typedef in
worldpos.h rather than reusing the old 0xFF/uint8_t sentinel, which
would silently mean +255 on a signed 16-bit field instead of -1.
mapPositionSet now also recomputes every live entity's cached
chunkIndex after mapRebuildChunkOrder() reshuffles which chunk_t sits
at each chunkOrder slot - done as a direct field recompute rather than
through entitySetChunk, which would've looked up each entity's now-stale
old index, failed to find/clear its real registration, and inserted a
duplicate entry into the correct chunk on every map position change.
Also adds saveCanSave() (false while a cutscene is running) and shows
the player's live chunkIndex in the ui/debug/uiplayerpos overlay.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Picking a save slot now sets SAVE.slotCurrent and calls saveLoadSlot()
before entering the overworld, same slotCurrent/saveSlotInit/saveLoadSlot
sequence saveLoadAllSlots uses per-slot; a load failure shows the fatal
error overlay instead of proceeding with stale/garbage slot data.
Also fixes uiTransitionUpdate firing its "finished" callback every frame
after a transition completes instead of once - latent since the generic
transition system had no real caller yet.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
The SAVE_DEVICE_FOUND/SAVE_LOAD_ALL_SLOTS block had been moved to the
end of the file (after QUIT/QUIT_GAME) without moving the LOADED
marker with it. MARKER items always auto-advance to whatever follows
them in the array, so reaching LOADED after a successful save-load fell
straight through into the QUIT block's confirm dialog instead of
completing the cutscene. LOADED now sits after SAVE_LOAD_ALL_SLOTS as
the true last item, so it falls off the end and fires
uiMainMenuOpenSelectSave again as intended.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Options and Quit now route through main_menu.jsonc markers the same
way New Game does, with a new QUIT_GAME cutscene item type ending the
process directly once quit is confirmed via MODAL_OPTIONS_MARKERS
(whose "message" field is now optional, defaulting to "" like "title").
Fixes two crashes surfaced along the way:
- uiModalLocalize asserted on an empty locale key instead of treating
it as "nothing to show" - hit as soon as a modal omitted a title or
message.
- assetLocaleGetString's new empty-message-ID guard broke the PO
format's own header-read convention (messageId "") used internally
by assetLocaleLoaderAsync, crashing on every boot. Extracted the
shared lookup into assetLocaleGetStringLookup and added a dedicated
assetLocaleGetHeader for that internal read, leaving
assetLocaleGetString's validation untouched for every real caller.
Also replaced every raw assetLocaleGetString/WithVA call across the UI
with the locale/localemanager.h helpers (adding localeManagerGetString
for the plain, non-formatting case), so app code no longer bypasses
them.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
cutsceneSystemLoad(file) centralizes the lock/assetRequireLoaded/start
sequence previously duplicated in sceneInitialInit, and falls back to
the fatal error overlay instead of asserting when the asset fails to
load. uiFatalErrorOpen now accepts a NULL message, showing a generic
"contact support" message for callers with no specific error text.
Also adds a ui/debug/uicutscene overlay showing the running cutscene's
item count and current item type, and wires main_menu's post-save-picker
flow through cutsceneSystemLoad.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Each cutscene item type now owns its own JSON parsing (a Load callback
alongside its existing runtime callbacks), instead of one monolithic
switch in assetcutsceneloader.c. Shared JSON field/enum decoding helpers
move into a new rpg/cutscene/item/json/ module so per-type parsers stay
small and colocated with the behavior they configure.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Pure whitespace/line-break reformatting (braces, newlines, and line
continuations matching this codebase's existing wrap conventions) - no
logic, string content, or identifiers changed anywhere. Confirmed via
diff against the pre-change tree and by rebuilding + re-running the
affected test suites, which produce identical pass/fail results.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
The main menu no longer needs its own scene lifecycle - it's started
directly by the initial cutscene's CUTSCENE item chaining into
main_menu.jsonc. Fold scenemainmenu.c's logic into uimainmenu.c (its
only real caller) and drop SCENE_TYPE_MAIN_MENU entirely. Also arm the
cutscene's onComplete callback unconditionally in uiMainMenuStartGame
so the save-picker still opens on a fresh boot, not just on restart.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Cutscenes now parse their authored .jsonc straight into the runtime
cutsceneitem_t/pool representation via yyjson, instead of going through a
separate Python-compiled DCTS binary format - removes the whole
build/compile step and the byte-format contract between the Python
encoder and the C decoder, at the cost of a (still tiny, one-time)
parse per cutscene load.
Also fixes dusk.dsk going stale after a build: the old custom_command
depended on a CMake-configure-time file glob, which only re-detects
added/removed assets on the next configure and could miss edits
entirely. tools.asset.pack now always runs and decides for itself
(via a small manifest) whether anything actually needs repacking, so
asset changes are never missed regardless of add/edit/remove.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Previously the main menu UI opened directly with no cutscene, and only
pressing "Start Game" spun up a brand new cutscene
(main_menu_start_game.jsonc) for the save-device check/load-slots flow.
Now a single cutscene (main_menu.jsonc) starts the moment the main menu
scene becomes active, sitting idle (new CUTSCENE_ITEM_TYPE_IDLE item -
no existing item type blocked forever with no side effect) while the
menu is shown, so it can own menu-wide state going forward.
Pressing "Start Game" now jumps the already-running cutscene to a
NEW_GAME marker via cutsceneGoTo instead of starting a second cutscene -
this already-existing mechanism needed no engine changes. Both retry
options (no device / load error) also now jump straight back to
NEW_GAME instead of RESTARTing the whole cutscene, since RESTART would
otherwise strand the player on the idle block. sceneMainMenuStartGame
also handles the case where the cutscene already ran to completion once
(e.g. backing out of the save picker and pressing Start Game again) by
restarting it landing straight on NEW_GAME, avoiding a crash from
jumping a marker with no cutscene running.
Built and verified on Linux (boots/idles without crashing, cutscene
tests pass), PSP, GameCube and Wii (Docker).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Wii NAND save storage is blocked by Dolphin's IOS ticket-rights check for
homebrew (documented earlier this session as a known, non-code limitation),
so Wii now defaults DUSK_SAVE_WII_METHOD to SD instead of NAND.
Also: asset_t gains ASSET.baseDirectory, populated by each platform's asset
loader with the real directory dusk.dsk was opened from (or, on PSP, the
directory EBOOT.PBP itself lives in, since dusk.dsk is embedded inside it).
Wii SD and PSP now write their single combined save file directly into
that directory instead of a separate hardcoded path, and no longer need to
create it (it's already known to exist). Linux now uses a `saves/`
subdirectory of that same directory instead of ~/.dusk/saves, since it
keeps multiple files (settings + one per slot). Wii NAND and GameCube/Wii
memory-card storage are unaffected - neither has a real filesystem-path
concept this applies to.
Built and verified on Linux, PSP (Docker), GameCube (Docker) and Wii
(Docker) - confirmed via the compiler invocation that Wii now compiles
with DUSK_SAVE_WII_METHOD_SD.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
saveLoadSlot() and uiSelectSaveNameEntered() both asserted
SAVE.deviceCurrent != 0xFF unconditionally, crashing as soon as a player
dismissed the "no save device found" prompt and tried to start a game or
name a new save. Both now fall back to an in-memory-only slot (no device
to persist to) instead of asserting, matching the "continue without a
save device" flow the main menu cutscene already offers.
Verified fixed on Dolphin/GameCube; built clean on Linux and GameCube
(Docker).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
- Fixed PSP pause taking up to ~1.1s to actually go silent: its output
thread ran independently of stream->state, draining its whole software
ring regardless. audiostream_t.state is now volatile and the output
thread checks AUDIO_STREAM_STATE_PLAYING before each hardware chunk,
skipping output (without consuming the ring) while paused - pause now
goes silent within about one chunk (~23ms) and resume has no gap.
- Added audioMixerPause/Resume/SetPan/SetLoop/FadeTo/IsFading to the
mixer - immediate, channel-indexed primitives for cutscenes to drive.
Fade transitions (fadeFrom/To/Duration/Time/Easing) live on
audiomixerchannelstate_t and advance every frame in
audioMixerChannelApplyVolume(), reusing the same easingApply()
interpolation uifullbox_t already uses for screen fades.
- New src/dusk/rpg/cutscene/item/audio/ with 9 cutscene item types:
AUDIO_PLAY (+ AUDIO_PLAY_SIMPLE/AUDIO_PLAY_LOOPED shorthands),
AUDIO_STOP, AUDIO_PAUSE, AUDIO_RESUME, AUDIO_FADE (+ FADE_OUT/FADE_IN
shorthands), AUDIO_FADE_WAIT, AUDIO_SET_PAN, AUDIO_SET_LOOP, and the
combined AUDIO_SET - registered through the same enum/union/callback
table/macro mechanism every other item type uses.
- Wired the same 9 types into the JSON-based (offline JSONC -> binary
.cts) cutscene asset pipeline: tools/asset/cutscene/__main__.py's
encoder and assetcutsceneloader.c's decoder. Verified round-trip by
hand-encoding/decoding a test file covering all 9 types, and confirmed
the two existing real cutscene files re-encode byte-identical.
Built and verified on Linux, PSP (Docker), GameCube (Docker) and Wii
(Docker).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
- Extracted src/dusk/audio/mixer/audiomixerchannel.{h,c}: the channel enum,
command/state structs, and all per-channel logic (StopStream,
ApplyVolume, ChannelUpdateEarly/Late) now live there; audiomixer.{h,c}
just holds the channel array and the public Play/PlayLooped/Stop API.
- Removed audioMixerSetChannelVolume()/audioMixerSetMasterVolume() - volume
is written directly (SAVE.settings.audioChannelVolume[channel] = 0.3f,
SAVE.settings.audioMasterVolume = 0.3f) and only ever validated where
it's read, in audioMixerChannelApplyVolume().
- Moved master/channel volume into savesettings_t (new writeFloatArray/
readFloatArray save-JSON macros) so they persist as user preferences;
`fade` moved from an unused mixer-wide array to a real per-channel
multiplier that's now actually folded into the volume calculation.
- audioMixerUpdateEarly()/Late() now catch and log a single channel's
error instead of letting it abort every other channel's update for that
frame (and cascade into skipping the rest of that frame's engine
update), matching the existing per-stream error-isolation pattern.
- Added persistent per-channel onLoop/onEnd callbacks + a user pointer
(audiomixerchannelstate_t), wired onto every stream a channel starts via
a trampoline (audioMixerChannelOnStreamLoop/OnStreamEnd) so they survive
across multiple plays on the same channel, unlike the underlying stream
instance itself.
- General cleanup: for loops -> while loops, if/else cascades -> guard
clauses, dropped unnecessary const on locals, wrapped to 80 columns.
Built and verified on Linux, PSP (Docker), GameCube (Docker) and Wii
(Docker).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
- PSP output thread now caches the last computed leftVolume/rightVolume and
only recomputes them when stream->volume/directionality actually changed
since the previous chunk, instead of recomputing every ~23ms chunk
unconditionally.
- Mixer channels and the mixer itself now carry their own volume
(audioMixerSetChannelVolume()/audioMixerSetMasterVolume()), multiplied
with each sound's own volume every audioMixerUpdateEarly() and applied via
audioStreamSetVolume() - so changing a channel's or the master volume
affects whatever's already playing, not just future sounds. Also fixed
audioMixerUpdateEarly() to tolerate a PLAY command queued outside the
normal per-frame cycle (no loadingAsset yet from a preceding
audioMixerUpdateLate(), e.g. during engine startup) by leaving it queued
for the following frame instead of asserting.
- engine.c now plays boa.mp3 on AUDIO_MIXER_CHANNEL_BGM_0 on loop through
the mixer instead of driving a raw audiostream_t directly, exercising the
new mixer pipeline end to end.
Built and verified on Linux, PSP (Docker) and GameCube (Docker).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
audioMixerPlay()/audioMixerPlayLooped()/audioMixerStop() now just queue a
command per channel instead of taking effect immediately:
- audioMixerUpdateLate() (end of frame) locks and begins loading any newly
queued PLAY command's asset, giving it until the start of the next frame
to finish.
- audioMixerUpdateEarly() (start of the next frame) applies the queued
command - blocking (assetRequireLoaded()) if that load hasn't finished
yet - actually starting/stopping playback and setting volume,
directionality and looping (reusing this session's new
audioStreamSetLoopLimit()) on a real audiostream_t acquired from the
shared stream pool.
Previously this was an unwired stub (never initialized/updated/disposed
from audio.c, and audioMixerUpdateLate() only printf'd). Built and verified
on Linux, PSP (Docker) and GameCube (Docker).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
- audioUpdate()/audioMixerUpdate() split into Early (before rendering, per-stream buffering) and Late (end of frame, starts newly-queued mixer sounds) phases, wired into engineUpdate() around displayUpdate(); mixer is now actually initialized/disposed/updated from audio.c.
- New src/dusk/audio/stream/audiostreamtype.{h,c}: asset-loader-type <-> audio-stream-type map (with file extension), plus audioStreamAssetTypeForPath()/audioStreamTypeForAssetType()/audioStreamAssetTypeForStreamType() lookups.
- audiostream_t/audiomixerchanneldata_t volume and directionality/pan switched from integer (0-0xFF / -128..127) to float (0.0-1.0 / -1.0..1.0), removing the manual scaling each platform backend had to do.
- audiostream_t gains loopLimit/loopRestartCount so a stream can be told to stop (fire onEnd) after a fixed number of loop restarts instead of looping forever; wired into the shared restart path (Linux/Dolphin) and PSP's own read-ahead loop decision.
Built and verified on Linux, PSP (Docker) and GameCube (Docker).
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Fixes the main-thread hitch reported on Dolphin (and, less severely,
elsewhere) whenever the MP3 decode buffer needed refilling:
audioStreamMp3Read() -> audioStreamMp3DecoderDecodeFrame() does real work
synchronously (asset I/O + libmad decode), previously called directly from
each platform's once-per-frame Feed(), blocking rendering for a frame on
every refill.
Adds a background decode-ahead ring (src/duskmad/audiostreammp3ring.c):
a one-shot-per-pass thread reads via the existing audioStreamRead() into a
ring buffer ahead of need; Feed() becomes a cheap, lock-protected drain
instead of a decode call. Reuses this project's own thread_t/threadmutex_t
(mutex+condvar) primitives, mirroring PSP's own background-thread-plus-ring
precedent. Fixes a real (if narrow) race in thread.c along the way:
threadHandler() reset thread->threadId outside the mutex it also used to
signal STOPPED, which a stop-then-immediately-restart pattern (needed once
per pass: play/seek/loop-restart) could hit as a stale-threadId assertion.
Initially built Dolphin-only, then generalized: the ring doesn't touch
ASND at all, so it was straightforward to share with Linux too, moving
both it and the libmad decoder into a new top-level src/duskmad/ - a
shared-capability directory in the same vein as src/duskgl or
src/dusknetwork, pulled in by whichever DUSK_TARGET_SYSTEM values need it
(linux/knulli, wii/gamecube) rather than PSP, which keeps its own
hardware sceMp3 decoder and ring untouched.
With both platforms now sharing real code, deduped what was left:
audioStreamComputeEndFrame() (loop-segment math, byte-identical between
Linux/Dolphin) moved to the generic audio/stream/audiostream.c, and
audioStreamMp3Ring*IfNeeded()/audioStreamReadForPlayback() wrappers (in
duskmad) collapse each platform's own `if(stream->type ==
AUDIO_STREAM_TYPE_MP3)` branches at every Init/Dispose/Buffer/Feed call
site into one check each, living in the ring module itself.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Replaces the ansnd-based backend (a third-party library with a custom
DSP ucode Dolphin's HLE audio emulation doesn't recognize, requiring
-a LLE as a workaround) with libogc's own asndlib - the standard,
HLE-supported audio API most GC/Wii homebrew already uses.
Also fixes three real bugs found while getting a large (41MB) WAV
asset actually playing on Dolphin:
- audiostreamdolphin.c previously buffered a stream's entire loop
segment into one allocation up front (inherited from the ansnd
design) - fine for a short test tone, but an out-of-memory crash for
a real multi-minute track on GameCube's 24MB (or even Wii's 88MB)
total RAM. Rewritten to stream bounded windows via ASND's real
double-buffer primitives (ASND_SetVoice + ASND_AddVoice, polled from
the main thread - never from an interrupt callback), matching how
Linux/PSP already work.
- ASND_TestVoiceBufferReady() returns SND_OK (0) when ready and
SND_BUSY (1) when not, in the actual toolchain fork
(extremscorner/libogc2) this project builds against - opposite of a
plain boolean. Treating the raw result as one meant a genuinely
ready voice's SND_OK read as false, silently skipping every
ASND_AddVoice() call forever after a stream's first buffered window.
- audioStreamPcmRead()'s 16-bit fast path read WAV's little-endian
sample bytes straight into the output buffer with no byte-swap -
correct by coincidence on every little-endian platform this engine
targets (Linux, PSP), but reversed every sample's bytes on
Dolphin/GameCube/Wii's big-endian PowerPC.
libmad is GPL-licensed (see audiostreammp3decodersw.h's own note from
the prior MP3 change); libogc's asndlib is under a permissive
BSD-style license, so this doesn't add another such dependency.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
minimp3's sliding-window frame confirmation could spuriously discard real
frames near the tail of a not-yet-full window, losing ~5.5% of a real
VBR file's content even at a tuned 256KB window (~0.3% residual) -
audible as the whole track finishing early with stutters at each drop.
libmad's mad_stream/mad_frame/mad_synth API reports "need more data"
(MAD_ERROR_BUFLEN) and "genuinely bad data" separately rather than
overloading one return value, which was the actual ambiguity minimp3
couldn't resolve. Verified against the same real file with a 32KB window
(vs minimp3's 256KB): frame count and duration match the Xing header
exactly. Confirmed building for Linux, GameCube, and Wii (the latter two
via the project's real devkitPPC/libogc Docker toolchain) - libmad ships
inside libogc itself on Dolphin, no separate fetch needed there.
libmad is GPL-licensed, unlike the rest of this MIT project - a
deliberate tradeoff, noted at each site that pulls it in.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Reorganizes src/dusk/audio/: individual stream implementations
(audiostream, audiostreampcm, audiostreammp3, audiostreammp3decodersw)
move into audio/stream/, and a new audio/mixer/ holds a channel-based
playback queue (audiomixer.c/.h) for future use - not yet wired into
the engine.
Fixes needed to keep the tree buildable after the move:
- A stray duplicate of audiostreammp3decodersw.c/.h was left at the old
flat path; removed in favor of the canonical copy in stream/.
- Updated every #include "audio/audiostream*.h" and the two platform
CMakeLists.txt (dusklinux, duskdolphin) that still pointed at the old
flat location.
- audiomixer.c/.h didn't compile: audiomixerqueue_t was referenced but
never defined (audiomixerchanneldata_t has the matching fields), a
trailing comma in audioMixerPlayLooped's parameter list is illegal in
C, and `file` was declared as an array of pointers instead of a char
buffer.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
audioStreamMp3DecoderDecodeFrame was treating minimp3's "can't confirm
a frame here" verdict as genuine garbage even when the window simply
hadn't been topped up yet - discarding real frames whenever one landed
near the tail of a not-yet-full window, since minimp3 needs to also
validate the *next* frame's header to confirm a decode. On a real ~236s
VBR file this silently dropped ~5.5% of frames, heard as the stream
finishing early ("racing") with a stutter at each drop.
Now refills before ever trusting a "not found" verdict as confirmed
garbage, and grows the window from 16KB to 256KB (past the point of
diminishing returns, ~0.3% residual loss). PSP is unaffected - it uses
the hardware sceMp3 decoder, not this file.
Also removes now-unneeded debug instrumentation from the Linux feed
path.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
