conco/OpenMaidEngine
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 3 Wiki Activity

Defer read-message skip persistence design

Browse Source
This commit is contained in:
gamer147
2026-07-18 22:42:17 -04:00
parent 56f8adfa2d
commit 0e5c0d79c8
2 changed files with 25 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
docs/engine-re.md
View File

@@ -1272,10 +1272,15 @@ also needs the generic callback/input layer used by Hide Window, local literal-a
menu/text-surface operations. ReadTextDB can share the point where a message completes, but it cannot serve menu/text-surface operations. ReadTextDB can share the point where a message completes, but it cannot serve
as the backlog data model because it stores only read flags, not text, styling, names, or voice metadata. as the backlog data model because it stores only read flags, not text, styling, names, or voice metadata.
The dependency order was **Hide Window first** to establish reusable callback/coroutine input, then The original dependency order was **Hide Window first** to establish reusable callback/coroutine input, then
Read-message Skip, then History after both the input layer and message-completion seam exist. Hide Window is Read-message Skip, then History after both the input layer and message-completion seam exist. Hide Window is
now complete; the remaining order is Read-message Skip followed by History. The `/v2` image names/comments now complete. Read-message Skip's semantics remain understood, but its implementation is deferred until the
the cursor, callback dispatch, retained-history navigation/render/metadata, and history-voice opcode paths. port can choose storage for numbered saves and shared profile/global data as one architecture rather than
selecting an isolated ReadTextDB backend. History is therefore the sole remaining control-strip action under
active consideration; it can already reuse Hide Window's input layer, while any useful sharing with the
deferred ReadTextDB work should remain a seam rather than a storage dependency. The `/v2` image
names/comments the cursor, callback dispatch, retained-history navigation/render/metadata, and history-voice
opcode paths.
### ADV Hide Window implementation (2026-07-18) ### ADV Hide Window implementation (2026-07-18)

17
docs/phase-a-slice-plan.md
View File

@@ -1637,3 +1637,20 @@ Validation: engine 177/177, opcode tests/lints and vm0 RECOVER clean, zero-warni
**Next:** manually recheck x=772 with the original left-click restore gesture, then continue with the **Next:** manually recheck x=772 with the original left-click restore gesture, then continue with the
profile-wide ReadTextDB slice if the visual lifecycle now matches. profile-wide ReadTextDB slice if the visual lifecycle now matches.
### ADV Read-message Skip deferred pending persistence architecture (2026-07-18)
The native behavior and data model established above remain the implementation contract: read state is
profile-wide, keyed by packed script resource id and per-script message index, with message completion
queued and committed through the native `0x71` boundary. Implementation is deliberately deferred, however,
until the port has a fuller picture of numbered saves, shared `SAVE.DAT` state, `RT.DAT`, configuration, and
other engine-owned profile/global data.
In particular, the earlier suggestion to persist ReadTextDB in a port-owned format first is not an adopted
storage decision. The port should not choose native `RT.DAT`, a port-owned replacement, or an import/export
split in isolation from the broader save/profile backend, lifecycle, migration, and interoperability design.
When that work resumes, the already-reversed semantic model can sit behind whichever backend that wider
architecture selects.
With Auto, all-message Skip, and Hide Window implemented and Read-message Skip consciously deferred,
History is the sole remaining unimplemented ADV control-strip action to investigate as a current slice.