jqueuniet/dm2-tools
1
0
Fork 0
Code Issues 2 Pull requests Projects Releases Packages Wiki Activity Actions

Adopt the engine-exact standalone-missile termination for Beta saves #257

Merged
jqueuniet merged 0 commits from refs/pull/257/head into main 2026-07-17 08:47:27 +02:00
Conversation 0 Commits - Files changed -
jqueuniet commented 2026-07-17 08:42:53 +02:00 (Migrated from codeberg.org)
Copy link

Re-tests the #169 standalone-missile termination discrepancy now that the Beta walk is bit-exact (#256), and resolves it: the Beta compensating model is gone, retail and Beta share the engine-exact model.

Background

For standalone (non-creature-possessed) type-14 missiles, the Beta path deliberately kept a chain-mode / persistent-flag "compensating" model that contradicted the disassembled Beta binary, because engine-exact semantics desynced DATA/SKSAVE0.DAT at the time (#169, 2026-07-06). That desync was measured while the three since-fixed Beta decode bugs (overlay masks, teleporter pair fold, map cap) were corrupting downstream bit positions.

What changed

  • The decoder and encoder standalone-missile arms drop the beta conditional: single-record sub-read/sub-write with the termination flag set, flag unconditionally cleared — matching the retail engine and the unpacked Fires.exe reader/writer (0x1eaa4/0x1eac6, 0x1de41/0x1de60).
  • The two compensating-model unit tests are replaced by engine-exact Beta expectations (a type-15 after a standalone missile parses/encodes as a full cloud under set_beta(true) too); the retail-shaped tests stay as the shared-path guard.
  • docs/formats/save-record-stream.md §"Standalone-missile termination" rewritten from open question to resolved, with the history and a fixture-coverage statement.

Evidence (all under the flip, local fixtures)

  • Beta suites green: all 9 slots land inside the final flush byte and field-round-trip byte-exactly; full crate 321 passed; 26 retail --ignored round-trip guards green (retail unchanged).
  • Slot-0 engine oracle trace: PERFECT MATCH, all 15,295 CHECKCODE-entry bit positions identical (temporary trace instrumentation, removed before commit as in #256).
  • Fixture census: slots 6/7/8 contain 4/6/6 standalone missiles (16 total), every one holding exactly a nested termination — a shape that consumes identical bits under both models; no slot's CHECKCODE stream contains any type-15 full cloud, so the model-distinguishing shape (flag persistence misparsing a later cloud) is never exercised. The docs state this explicitly: the flip rests on the binary evidence; the fixtures corroborate that it regresses nothing.

Notes for review

  • CI exercises none of the Beta assertions (fixtures are local-only, skip-on-missing), same as #256. Local gates: fmt / clippy -D warnings / full workspace tests / doc -D warnings all green.

Closes #169.

Re-tests the #169 standalone-missile termination discrepancy now that the Beta walk is bit-exact (#256), and resolves it: the Beta compensating model is gone, retail and Beta share the engine-exact model. ## Background For standalone (non-creature-possessed) type-14 missiles, the Beta path deliberately kept a chain-mode / persistent-flag "compensating" model that contradicted the disassembled Beta binary, because engine-exact semantics desynced `DATA/SKSAVE0.DAT` at the time (#169, 2026-07-06). That desync was measured while the three since-fixed Beta decode bugs (overlay masks, teleporter pair fold, map cap) were corrupting downstream bit positions. ## What changed - The decoder and encoder standalone-missile arms drop the `beta` conditional: single-record sub-read/sub-write with the termination flag set, flag unconditionally cleared — matching the retail engine and the unpacked `Fires.exe` reader/writer (`0x1eaa4`/`0x1eac6`, `0x1de41`/`0x1de60`). - The two compensating-model unit tests are replaced by engine-exact Beta expectations (a type-15 after a standalone missile parses/encodes as a full cloud under `set_beta(true)` too); the retail-shaped tests stay as the shared-path guard. - `docs/formats/save-record-stream.md` §"Standalone-missile termination" rewritten from open question to resolved, with the history and a fixture-coverage statement. ## Evidence (all under the flip, local fixtures) - Beta suites green: all 9 slots land inside the final flush byte and field-round-trip byte-exactly; full crate 321 passed; 26 retail `--ignored` round-trip guards green (retail unchanged). - Slot-0 engine oracle trace: PERFECT MATCH, all 15,295 CHECKCODE-entry bit positions identical (temporary trace instrumentation, removed before commit as in #256). - Fixture census: slots 6/7/8 contain 4/6/6 standalone missiles (16 total), every one holding exactly a nested termination — a shape that consumes identical bits under both models; no slot's CHECKCODE stream contains any type-15 full cloud, so the model-distinguishing shape (flag persistence misparsing a later cloud) is never exercised. The docs state this explicitly: the flip rests on the binary evidence; the fixtures corroborate that it regresses nothing. ## Notes for review - CI exercises none of the Beta assertions (fixtures are local-only, skip-on-missing), same as #256. Local gates: fmt / clippy `-D warnings` / full workspace tests / doc `-D warnings` all green. Closes #169.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jqueuniet/dm2-tools!257
No description provided.