Plan Roman charcoal as an available coal production route

The current0.11.0 candidate's ordinary two-policy seed3 remains unfinished at200,000ticks. Player2has an active iron mine and24produced iron ore, but no coal or iron. A further normal50,000tick trace still produces no coal/iron while its geologist revisits the same survey targets. The existing policy requires a coal-mine before an iron smelter and never plans the implemented Roman charcoal maker.

A retained ordinary-player probe at200,000 queues legal charcoal-maker and iron-smelter builds. Both commands execute successfully. With the same normal AI and no further intervention, at220,002 the real ledger records22coal/8iron and40iron ore. Created-goods totals are unchanged and conservation is clean. The match remains unfinished; this does not establish automatic recovery, army growth or victory.

The existing Roman engineering catalog already models the charcoal maker as3planks/4stone construction and5delivered planks per1coal. This is a supported faction production route, not an inventory grant or changed recipe. Original AI priorities/timings remain unobserved.

Owning source: game economic policy and smelter prerequisites. Candidate200+201source hashes and the generated pre-probe checkpoint are retained locally and will be committed with provenance after that source is published. No original raster/audio.

Acceptance:

• Plan a real charcoal maker as an available Roman coal route when no usable coal mine is available, while preferring a viable known mine and respecting timber/material bootstrap budgets.
• Recognize legitimate coal producers in downstream smelter planning; avoid hard-coding coal-mine as the only supplier.
• Use only allowed player observations and ordinary build/settings/work orders. No hidden deposits, granted goods, altered costs/recipes or clocks.
• Retain public policy/session regressions for the captured missing producer, actual delivered plank-to-coal-to-iron production, matching restore and conservation.
• Cover a viable mine, existing/planned alternative producer and unavailable faction building without duplicate or unknown-building orders.
• Repeat ordinary matches and full source/build/browser/exact-main verification. Report remaining survey/tool/military bottlenecks honestly.

Related #73, #74, #76, #200, #201 and #4. This is a scoped engineering planner correction; broad AI and original Gold acceptance remain open.

Published source correction as Chris, without coauthors, in b0d1df3 by normal fast-forward to main. Local qualification: 315 passing unit tests (23087 assertions), lint/types/assets/fidelity/roadmap/build pass and all 385 fresh browser steps pass, including the real ten-minute 4x battle with save/reload. All four sampled ordinary two-policy matches finish with clean conservation and no failed commands: seeds 1/5/2/3 at 31801/73581/184761/139741.

The exact main CI is still running: https://github.com/stacksjs/the-settlers-iii-remake/actions/runs/40242723177. Keeping this issue open until that run and its original diagnostic artifact are verified. Existing rules, recipes, starting stocks, deadlines and framework cohort are unchanged. These are sampled engineering matches, not all-seed, human-input victory, native presentation or original Gold parity claims.

Qualified on main in b0d1df3, authored and committed by Chris without coauthors. Exact main CI 40242723177 succeeded. The original diagnostic artifact 12618051922 (19646 bytes) matches GitHub SHA256 a5a9df03b47549820dcac09d0c7defcfb69649b48b63a0da9c82cd415fc9e4fc; every ZIP CRC, all 385 browser steps, and all error collections were verified. This includes the real ten-minute 4x battle with save/reload.

The current AI uses the real Roman charcoal recipe after timber production and actual ore or coal-consumer demand exist, retains coal-mine preference when no supplier exists, and recognizes planned coal suppliers for smelting. Rules, recipes, created stock, clocks and existing battle deadlines are unchanged. Local qualification at this publication: 315 unit tests / 23087 assertions, lint/types/assets/fidelity/roadmap/build all pass.

All four sampled ordinary two-policy matches finish, with complete per-tick traces, no failed execution/enqueue and clean accounting: seed 1 at 31801 (aa0f0eb7), seed 5 at 73581 (ce4ce34e), seed 2 at 184761 (b034bac1) and seed 3 at 139741 (d6a0bdf8). Committed receipt.

Closing this scoped production-planning defect. Broader #73, #74, #76 and #4 remain open: sampled AI completion does not prove all seeds, human-input victory, sustainable indefinite play, native presentation or original Gold Edition parity. #203 tracks a separate keyboard harvest-site usability correction discovered during native-input play.