SeymourBETA

Update 2.23

6/7/2026

v2.23.0BETA

Seymour

SPELL NAME in the magic viewers (PS3 Magic HD): best-effort join (owner-authorized)

The owner told me to show the name even without the byte-proven join ("agora é pra fazer caralho", 2026-06-07 — an explicit override of the no-fabricate rule). New MagicSpellNameResolver (FfxLib/Dictionaries): magic_#### (id 0..1023, 12-bit) → tries CommandCharacter → CommandMonster1 → CommandMonster2 → Item (all Dictionary<ushort,string>), returns name ~source (e.g. Firaga ~char-cmd) or magic_#### when unmapped. Wired into PS3 Magic (HD) (Ps3MagicBrowser): Ps3MagicEntry.SpellNameDisplay becomes the list title + detail header, and search matches by name. Honest: the ~source flags it as a NON-byte-proven hypothesis (the catalog says "no names proved"); the owner verifies on screen and we refine the offset/dict if any name is wrong. Build 0 errors

v2.22.0

v2.23.1BETA

Seymour

Formation Editor: pre-filled slots now SHOW the monster name (load display fix)

The owner flagged that the 8 formation slots opened with an EMPTY ComboBox for slots that already had a value (raw 10DEh/10E2h) — the name only appeared AFTER picking manually (even the empty FFFFh slots showed blank instead of "(Vazio)"). Cause: an Avalonia gotcha — the SelectedItem set on load (object initializer) is applied before the per-row ComboBoxes materialize their ItemsSource (bound via RelativeSource), so the selection doesn't resolve and stays blank. Fix: after populating Slots, re-apply each SelectedMonster (toggle null→value) on a Dispatcher.UIThread.Post(..., Background) so the ComboBox re-resolves against a populated Items list. Guarded by slotsSyncing so it does not mark dirty (load stays "Changed: false" until a real edit). Build 0 errors

v2.23.0

v2.23.2BETA

Seymour

Aurora: "Reload battle from disk" (reflects other modules' saves on the map)

The owner edited the formation in the Formation Editor and the monsters didn't appear on the Aurora map. Investigation: the pipeline is correct — Aurora binds formation slot[i] → monster-live anchor[i] with MonsterId/Model (/work/phyre_chr_anim/models/mNNN/mNNN_animated.gltf, 340 HD models exist, server-root URL right). The cause was cached state: RefreshCatalog reloads only the scene list, NOT the open battle's bytes — so a formation edit made elsewhere wouldn't reflect until re-selecting. Fix: new ReloadSelectedBattle() (re-fires the battle read = re-reads chunk2 formation + chunk3 anchors from disk) + a "🔄 Reload battle from disk" button next to "Render". (Battle Explorer already had "Refresh" + recreates fresh on every nav — A was already there for it.) The combined root cause of "Changed: False" + "no monsters on the map" was the formation save not persisting — fixed in v2.23.1 (combo-blank). Build 0 errors

v2.23.1

v2.23.3BETA

Seymour

Formation Editor: a monster ADDED to an empty slot now inherits the "live on field" flag

The owner added monsters to slots 04-07 and they weren't recognized as live (Aurora "only 4 live monsters"). On-screen proof: the original slots (00-03) carry raw 10DEh/10E2h (high nibble 0x1000); the added ones came out 0030h/0136h (high nibble 0x0000). 0x1000 is the ACTIVE-formation-monster flag — the writer stamped 0 when filling a slot that was FFFFh. Fix (FormationSlotRow): when filling an empty slot, inherit the high nibble from a SIBLING live slot of the same battle (per-battle, no guessing; default 0x1000 if the formation was fully empty). Still slot-only byte-safe (FormationSlotLab gate 858/858). Extra diagnosis (NOT a save-side bug): save and read use the SAME path (GetPathBattle); the problem was the VALUE, not the side. Honest: (1) 0x1000=live is an evidence-based hypothesis from the on-screen proof — test in-game; (2) the Aurora MAP only has the monster anchors the ARENA (chunk3) defines (e.g. 4) — adding a monster beyond that needs a new arena anchor (arena authoring = out of this fix's scope). Build 0 errors

v2.23.2

Back to updates

SeymourBETA

Update 2.23

6/7/2026

v2.23.0BETA

Seymour

SPELL NAME in the magic viewers (PS3 Magic HD): best-effort join (owner-authorized)

The owner told me to show the name even without the byte-proven join ("agora é pra fazer caralho", 2026-06-07 — an explicit override of the no-fabricate rule). New MagicSpellNameResolver (FfxLib/Dictionaries): magic_#### (id 0..1023, 12-bit) → tries CommandCharacter → CommandMonster1 → CommandMonster2 → Item (all Dictionary<ushort,string>), returns name ~source (e.g. Firaga ~char-cmd) or magic_#### when unmapped. Wired into PS3 Magic (HD) (Ps3MagicBrowser): Ps3MagicEntry.SpellNameDisplay becomes the list title + detail header, and search matches by name. Honest: the ~source flags it as a NON-byte-proven hypothesis (the catalog says "no names proved"); the owner verifies on screen and we refine the offset/dict if any name is wrong. Build 0 errors

v2.22.0

v2.23.1BETA

Seymour

Formation Editor: pre-filled slots now SHOW the monster name (load display fix)

The owner flagged that the 8 formation slots opened with an EMPTY ComboBox for slots that already had a value (raw 10DEh/10E2h) — the name only appeared AFTER picking manually (even the empty FFFFh slots showed blank instead of "(Vazio)"). Cause: an Avalonia gotcha — the SelectedItem set on load (object initializer) is applied before the per-row ComboBoxes materialize their ItemsSource (bound via RelativeSource), so the selection doesn't resolve and stays blank. Fix: after populating Slots, re-apply each SelectedMonster (toggle null→value) on a Dispatcher.UIThread.Post(..., Background) so the ComboBox re-resolves against a populated Items list. Guarded by slotsSyncing so it does not mark dirty (load stays "Changed: false" until a real edit). Build 0 errors

v2.23.0

v2.23.2BETA

Seymour

Aurora: "Reload battle from disk" (reflects other modules' saves on the map)

The owner edited the formation in the Formation Editor and the monsters didn't appear on the Aurora map. Investigation: the pipeline is correct — Aurora binds formation slot[i] → monster-live anchor[i] with MonsterId/Model (/work/phyre_chr_anim/models/mNNN/mNNN_animated.gltf, 340 HD models exist, server-root URL right). The cause was cached state: RefreshCatalog reloads only the scene list, NOT the open battle's bytes — so a formation edit made elsewhere wouldn't reflect until re-selecting. Fix: new ReloadSelectedBattle() (re-fires the battle read = re-reads chunk2 formation + chunk3 anchors from disk) + a "🔄 Reload battle from disk" button next to "Render". (Battle Explorer already had "Refresh" + recreates fresh on every nav — A was already there for it.) The combined root cause of "Changed: False" + "no monsters on the map" was the formation save not persisting — fixed in v2.23.1 (combo-blank). Build 0 errors

v2.23.1

v2.23.3BETA

Seymour

Formation Editor: a monster ADDED to an empty slot now inherits the "live on field" flag

The owner added monsters to slots 04-07 and they weren't recognized as live (Aurora "only 4 live monsters"). On-screen proof: the original slots (00-03) carry raw 10DEh/10E2h (high nibble 0x1000); the added ones came out 0030h/0136h (high nibble 0x0000). 0x1000 is the ACTIVE-formation-monster flag — the writer stamped 0 when filling a slot that was FFFFh. Fix (FormationSlotRow): when filling an empty slot, inherit the high nibble from a SIBLING live slot of the same battle (per-battle, no guessing; default 0x1000 if the formation was fully empty). Still slot-only byte-safe (FormationSlotLab gate 858/858). Extra diagnosis (NOT a save-side bug): save and read use the SAME path (GetPathBattle); the problem was the VALUE, not the side. Honest: (1) 0x1000=live is an evidence-based hypothesis from the on-screen proof — test in-game; (2) the Aurora MAP only has the monster anchors the ARENA (chunk3) defines (e.g. 4) — adding a monster beyond that needs a new arena anchor (arena authoring = out of this fix's scope). Build 0 errors

v2.23.2