M4 iteration 2: portable modules and the Mockingbird re-carve

[Dibujaron]

One module document can now serve many (hull, slot) placements, a module can claim several adjacent slots, and the Mockingbird is re-carved so her rooms are slots — five standard cabins served by a single rijay.cabin.standard document, a central payload bay, a crew commons, and a modular engine room.

Nothing about the authored-overlay bet changes. Every target is still drawn by hand against one specific hull's coordinates, which is what guarantees its doors line up. What changed is file organisation: one document per concept instead of one per placement. Free placement stays refused, and the validator still does no geometry.

This restores what docs/modules.md originally specified and that iteration 1 simplified away, so the Sparrow and Finch can be authored once, against final rules.

What landed

34be4ec 18d33fc	a module document carries targets
5d46056 301373d	a target may claim several adjacent slots
6f15216	the Mockingbird re-carved into ten slots
ecdec33	one rijay.cabin.standard serving all five cabins
b068b9f 0a49201	schema and docs
8164269 0f74346	review fixes

The ten slots

cockpit (4 tiles) · cabin_fore_a (2) · cabin_fore_b (2) · crew_commons (12) · cabin_mess (2) · payload (45) · cabin_engineer (2) · engineering (6) · cabin_aft_stbd (2) · hold (74)

Fixed hull: the cockpit passage, the x8 corridor, the aft junction, both stairwells, the whole Mezzanine with its Q ports, and the skin. Where a divider now has slot tiles on both sides — the aft passenger cabin partitions, the wall between the large common and those cabins — it left the hull and moved into the payload module. Those walls are movable now.

Verification

• 337 Gleam tests, no failures; harness 29 passed / 2 deselected.
• The Mockingbird's deck is unchanged, tile for tile. server/test/fixtures/mockingbird_authored.json and default_loadout_reproduces_the_authored_deck_test were the arbiter throughout and are untouched — zero deletions in mockingbird_test.gleam.
• tools/slotmap.py renders the ten regions at exactly the planned tile counts, and the bare-hull render confirms the payload interior is genuinely open where the modules now own it.
• The new module grids are byte-for-byte transcriptions of the deleted ones at the same origins, not re-drawings.

Two forced number changes

The plan's own module masses and tag figures required these; both were verified against the unchanged part documents.

• Hull dry mass 96.0 → 73.0 — modules went 24.0 → 47.0, so the resolved total stays 128.0 and her flight numbers are unchanged.
• Reactor power 10 → 18 — modules now draw 17 with the Consol patch, 18 with the stock engine, preserving the pre-M4 headroom of 1 and 0 respectively.

Known follow-up (not a blocker)

tag_deficit:power is unreachable through shipped content now: every slot has one candidate module except hold, whose two both draw 1. The rule is still enforced and unit-tested; what is missing is a shipped demonstration, which is why sim_test's refused-refit case retargeted to tag_deficit:engine. Iteration 3's refit catalog brings alternatives that can overdraw, at which point that test can go back to power. Worth noting she sits at exactly zero headroom with the stock engine, so the first catalog addition forces a reactor decision.

Vocabulary change

module_wrong_hull and module_wrong_slot are gone, replaced by module_not_drawn_for_slot. A typo in a module document's slots reports as target_slot_not_on_hull:<module id> — a content error naming the file at fault — kept distinct from the player-facing slot_not_on_hull:<slot id> refusal.

🤖 Generated with Claude Code