5cc66f6fff
BuildQueue::ProcessTurn 0x00890d50 disassembled to the next function start (Ghidra says 1230 bytes; the body is 1264 and its end lands inside the epilogue -- rule 17 again). Corrects the recorded prototype: the first stack argument is the StrategyServer S frame, not the system, so the build-completed event goes on the sim and the hull's turn stamp is Frame. THE HEADLINE IS A CORRECTION. Lane E2's open census leaf (shpt[0], one destroyer short on both reference pairs) cannot be closed by S11's build-queue sub-pass: measured over the corpus, both reference saves carry three empty BQ frames, every hbq false, and the one TurnCommands_v5 block empty. The order is created inside the turn by the AI -- PvSav shows 11,900 leaving the treasury before spine phase 0. The blocker moves to game/ai. Also: the per-class built counter has EXACTLY ONE writer in the image (image-wide byte scan); ShipRecords sized by enumeration; the ship/fleet birth chain end to end, correcting lane B5 -- FtFlg 0x400 is set by EVERY fleet born through the create path, not by retreat; one home fleet per system cached at ServerSystem+0x238; and a third indirection class vtable_map.py cannot see (the ship-borne wrapper is dispatched from a stack-built pointer table). 21 addresses in ghidra/addresses.d/lane-b6.json, validated to a scratch path against the merged set (1,077 entries, no duplicate names); two entries dropped as duplicates of lane B5's and recorded as agreements instead. |
||
|---|---|---|
| .. | ||
| control-flow | ||
| objects | ||
| subsystems | ||
| .keep.md | ||
| 00-strategy.md | ||
| 01-fingerprint.md | ||
| _template.md | ||