SeymourBETA

Update 2.17

6/6/2026

v2.17.0BETA

Seymour

Magic Viewer FINALIZED (3-lane parallel workflow): web viewer wired into the editor + menus polished + root+88 RE closed:

Lane A — web viewer wired into the editor: FFXMagicViewerWeb (Codex-built — Spell Index, Package Graph, Texture Stack, Runtime Evidence, Effect Stage, Crosswalk + 6 indices/catalog) was orphaned (no editor button). New "Magic Viewer (Web) 🪄" nav button (Extras group, next to "PS3 Magic (HD)") + MagicViewerLauncher.cs that serves RuntimeTools/FFXMagicViewerWeb via python -m http.server (port 8766, distinct from Aurora's 8765) and opens the browser — identical recipe to Aurora/MapViewer (HTTP is mandatory: app.js fetch()es its catalog → file:// would CORS-hang). Did NOT use SpiraForgeHub's file:// path.
Lane B — 2 magic menus polished (read-only, no writer): PS2 Magic Effects got search (package/lane + kernel/lane) and an honest empty-state (actionable message vs a blank pane); the "Counts" card no longer echoes "Summary". PS3 Magic (HD): thumbnail decode moved off the UI thread (heavy folders no longer freeze — rows show "pending" and stream in via Dispatcher, with a generation token against stale passes), texture-name search, and Open Folder/Open File buttons now bind IsEnabled (no more dead-click). (Spell names by id stay BLOCKED — no name table exists; the magic.bin→magic_#### join is a blocked causal bridge; not fabricated.)
Lane C — root+88 RE (Pass12, doc-only, applied to the REAL db FFX_recon.i64): closed Pass11's two open items. (1) Who populates the runtime root: sub_7FD9A0 installs FFX_Magic_RuntimeRootTable[0x12A4080] ONCE at bootstrap (bin-load) and zeroes root+88; phases don't swap the table — they reuse the same root and mutate in-place. (2) Final root+88 writer: sub_817200 = opcode handler 0x21 in FFX_Magic_OpcodeHandlerTable (0xC48EC8); on family 0x1000 it writes root+84=root+88=same cursor in the work-copy, written back to the root at phase end. 5 renames + [pass12] comments applied/saved in the real db. Doc docs/reverse/FFX_MAGIC_ROOT88_WRITER_PASS12_2026-06-06.md. (Honest: concrete spell→opcode, engine-accurate timing, and full writer coverage stay BLOCKED.)
csproj bump (real UI: launch + polish). (Lane C is RE/doc-only but lands in the same entry.)

Back to updates

SeymourBETA

Update 2.17

6/6/2026

v2.17.0BETA

Seymour

Magic Viewer FINALIZED (3-lane parallel workflow): web viewer wired into the editor + menus polished + root+88 RE closed:

Lane A — web viewer wired into the editor: FFXMagicViewerWeb (Codex-built — Spell Index, Package Graph, Texture Stack, Runtime Evidence, Effect Stage, Crosswalk + 6 indices/catalog) was orphaned (no editor button). New "Magic Viewer (Web) 🪄" nav button (Extras group, next to "PS3 Magic (HD)") + MagicViewerLauncher.cs that serves RuntimeTools/FFXMagicViewerWeb via python -m http.server (port 8766, distinct from Aurora's 8765) and opens the browser — identical recipe to Aurora/MapViewer (HTTP is mandatory: app.js fetch()es its catalog → file:// would CORS-hang). Did NOT use SpiraForgeHub's file:// path.
Lane B — 2 magic menus polished (read-only, no writer): PS2 Magic Effects got search (package/lane + kernel/lane) and an honest empty-state (actionable message vs a blank pane); the "Counts" card no longer echoes "Summary". PS3 Magic (HD): thumbnail decode moved off the UI thread (heavy folders no longer freeze — rows show "pending" and stream in via Dispatcher, with a generation token against stale passes), texture-name search, and Open Folder/Open File buttons now bind IsEnabled (no more dead-click). (Spell names by id stay BLOCKED — no name table exists; the magic.bin→magic_#### join is a blocked causal bridge; not fabricated.)
Lane C — root+88 RE (Pass12, doc-only, applied to the REAL db FFX_recon.i64): closed Pass11's two open items. (1) Who populates the runtime root: sub_7FD9A0 installs FFX_Magic_RuntimeRootTable[0x12A4080] ONCE at bootstrap (bin-load) and zeroes root+88; phases don't swap the table — they reuse the same root and mutate in-place. (2) Final root+88 writer: sub_817200 = opcode handler 0x21 in FFX_Magic_OpcodeHandlerTable (0xC48EC8); on family 0x1000 it writes root+84=root+88=same cursor in the work-copy, written back to the root at phase end. 5 renames + [pass12] comments applied/saved in the real db. Doc docs/reverse/FFX_MAGIC_ROOT88_WRITER_PASS12_2026-06-06.md. (Honest: concrete spell→opcode, engine-accurate timing, and full writer coverage stay BLOCKED.)
csproj bump (real UI: launch + polish). (Lane C is RE/doc-only but lands in the same entry.)