jpmschweitzer/settled-reach
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(simulation): road-graph hub refinement + HQ placement boundary (T-1076, D-242) #178

Closed
jpmschweitzer wants to merge 0 commits from road-hub-refinement into main
pull from: road-hub-refinement
merge into: :main
Conversation 3 Commits - Files Changed -
jpmschweitzer commented 2026-07-16 16:25:16 +02:00
Owner
Copy Link
Copy Source

Summary

Implements T-1076 — the final D-242 story. Road-graph hub selection over the corrected settlement model, plus the placement-boundary re-tune Jeroen ruled during refinement (the shipped map had over-generalized Standalone).

Item 0 — placement boundary re-tune (data)

  • Standalone ONLY where the HQ is itself an industrial complex / company town: 11 values (extraction/mining/quarrying, heavy vehicle+shipyard, heavy energy, gate infrastructure, military-industrial). 9 values flipped to CityTenant (agriculture offices, craft/consumer manufacturing, high-tech fabs, marine farming) — the winery-as-belt-settlement failure mode is gone.
  • Counts: Standalone corps 95 -> 11; CityTenant 60 -> 144; settlements emitted 94 -> 11; city rows 423 -> 340; tenant links 58 -> 129. Fixpoint verified (runs 2-3 byte-identical across all 165 corps). Cygni NameLocked pin moved to Metropolis.
  • D-242 amendment states the boundary explicitly; Jeroen`s edge-of-settlement/cluster color recorded as CityTenant rendering detail for later phases.

Items 1-4 — road graph (Rust)

  • is_standalone_hq threaded through CityRecord/CityPlacement (corporations LEFT JOIN in both readers — the threading the T-1074 test comment predicted).
  • Hub selection: top-cap by population among non-HQ settlements; HUB_SPACING_DIAG_PX=64 (cap 8 on the current planet grid), HUB_CAP_MIN=6; Gate Corporation (909M, HQ) verified non-hub on live cascade.
  • Co-location collapse: exact (body_id,name), keep lowest id, loud warn (defensive, unit-tested with synthetic duplicate).
  • Hybrid minor attach: nearest-point-on-polyline snap within SNAP_MAX_PX=8 with Junction edge-split (from<to preserved via norm_edge, maintenance/route identity inherited), else A*-spur to nearest hub.
  • RailHeadFacing: new FoundingOrientation variant with octant bearing_degrees toward the dominant incident edge at degree>=3 junctions; skeleton_gen railhead_edge() feeds the D-234b flush-frontage machinery (coastal wins if both). D-213 amended.
  • Drive-by fix: believability reader honoured a hardcoded PopulationBudget (stale since T-1075) — now reads the baked settlement_class.

Verification

regen-db x3 (fixpoint stable), check-systems-db fresh, test-tooling PASS, cargo check clean; 11 new targeted tests, all touched suites green (road_graph 17, skeleton_gen 53, attractor_matching 11, city_context_reader 32, cascade 8, layer_proxy 9, plugin 20, believability 3); 4-body live-cascade hub sanity report in the ticket.

Known/expected

  • calluna-wellness + namsan-collective unmatched-by-design (their belt-HQ tenancy targets correctly vanished under the boundary; re-link via T-1115 content).
  • rail_facing fires on no current real body (max degree 2 today) — mechanism unit-tested, engages as density grows.
  • T-1116 filed: water-heavy bodies produce 0 routable edges (pre-existing A* drop path, now more visible).
## Summary Implements T-1076 — the final D-242 story. Road-graph hub selection over the corrected settlement model, plus the placement-boundary re-tune Jeroen ruled during refinement (the shipped map had over-generalized Standalone). ### Item 0 — placement boundary re-tune (data) - Standalone ONLY where the HQ is itself an industrial complex / company town: 11 values (extraction/mining/quarrying, heavy vehicle+shipyard, heavy energy, gate infrastructure, military-industrial). 9 values flipped to CityTenant (agriculture offices, craft/consumer manufacturing, high-tech fabs, marine farming) — the winery-as-belt-settlement failure mode is gone. - Counts: Standalone corps 95 -> 11; CityTenant 60 -> 144; settlements emitted 94 -> 11; city rows 423 -> 340; tenant links 58 -> 129. Fixpoint verified (runs 2-3 byte-identical across all 165 corps). Cygni NameLocked pin moved to Metropolis. - D-242 amendment states the boundary explicitly; Jeroen`s edge-of-settlement/cluster color recorded as CityTenant rendering detail for later phases. ### Items 1-4 — road graph (Rust) - is_standalone_hq threaded through CityRecord/CityPlacement (corporations LEFT JOIN in both readers — the threading the T-1074 test comment predicted). - Hub selection: top-cap by population among non-HQ settlements; HUB_SPACING_DIAG_PX=64 (cap 8 on the current planet grid), HUB_CAP_MIN=6; Gate Corporation (909M, HQ) verified non-hub on live cascade. - Co-location collapse: exact (body_id,name), keep lowest id, loud warn (defensive, unit-tested with synthetic duplicate). - Hybrid minor attach: nearest-point-on-polyline snap within SNAP_MAX_PX=8 with Junction edge-split (from<to preserved via norm_edge, maintenance/route identity inherited), else A*-spur to nearest hub. - RailHeadFacing: new FoundingOrientation variant with octant bearing_degrees toward the dominant incident edge at degree>=3 junctions; skeleton_gen railhead_edge() feeds the D-234b flush-frontage machinery (coastal wins if both). D-213 amended. - Drive-by fix: believability reader honoured a hardcoded PopulationBudget (stale since T-1075) — now reads the baked settlement_class. ### Verification regen-db x3 (fixpoint stable), check-systems-db fresh, test-tooling PASS, cargo check clean; 11 new targeted tests, all touched suites green (road_graph 17, skeleton_gen 53, attractor_matching 11, city_context_reader 32, cascade 8, layer_proxy 9, plugin 20, believability 3); 4-body live-cascade hub sanity report in the ticket. ### Known/expected - calluna-wellness + namsan-collective unmatched-by-design (their belt-HQ tenancy targets correctly vanished under the boundary; re-link via T-1115 content). - rail_facing fires on no current real body (max degree 2 today) — mechanism unit-tested, engages as density grows. - T-1116 filed: water-heavy bodies produce 0 routable edges (pre-existing A* drop path, now more visible).
jpmschweitzer added 4 commits 2026-07-16 16:25:16 +02:00
chore(meta): activate T-1076 road-hub batch — Si refinement + gap resolutions (incl. scope item 0: placement-map re-tune per Jeroen) f5bb71736b
feat(db): corp-HQ placement boundary re-tune — Standalone only for industrial-complex HQs (T-1076 item 0, D-242) a0ebdbe6b8 
95->11 Standalone / 60->144 CityTenant per Jeroen's boundary ruling: only extraction/mining/quarrying, heavy vehicle+shipyard, heavy energy, gate infrastructure keep their own settlements; craft/consumer manufacturing, agriculture offices, high-tech fabs flip to CityTenant. Cygni pin moved to Metropolis (its HQ settlement vanishes under the flip). Regen systems.db: 340 city rows, 129 tenant links, fixpoint stable.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
feat(simulation): road-graph hub refinement — scaled-cap hubs, co-location collapse, minor-attach edge split, RailHeadFacing (T-1076) 98991284bf 
is_standalone_hq threaded through CityRecord/CityPlacement via corporations LEFT JOIN (both readers); hubs = top-cap by population among non-HQ settlements (HUB_SPACING_DIAG_PX=64, HUB_CAP_MIN=6); exact-name collapse keeping lowest city_id with loud warn; nearest-point-on-polyline snap (SNAP_MAX_PX=8) with Junction edge-split preserving from<to via norm_edge, else A*-spur to nearest hub; FoundingOrientation::RailHeadFacing{bearing_degrees} assigned at degree>=3 junctions (octant bearing toward dominant incident edge), consumed by skeleton_gen railhead_edge() through the D-234b flush-frontage machinery; believability reader now honours baked settlement_class (stale hardcode since T-1075). 11 new targeted tests; existing suites green.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
docs(meta): D-242 placement-boundary amendment + D-213 RailHeadFacing amendment (T-1076) 5d37d205e1 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer commented 2026-07-16 16:39:10 +02:00
Author
Owner
Copy Link
Copy Source

Review: road-hub-refinement → main (type: code + content slice)

Reviewers: Hoshe (code quality), Tyre (architecture), Miri (world consistency — placement-boundary slice). All read source from the branch worktree (HEAD = branch tip, not sparse). Pre-push gate passed before push and was not duplicated. Runtime verification: 4-body live-cascade hub sanity report in the ticket; no client/merge-path code touched.

Hoshe (Code Quality): REQUEST_CHANGES

Hub/minor partition, collapse, and RailHeadFacing correctly implemented and well unit-tested; DB spot-checks match the D-242 amendment exactly (11/144 corps, 340 rows, 129 links, 0 dupes, stamp fresh). One real reachable bug + three coverage gaps in exactly the places that hide it.

# File Issue
H1 server/src/atlas/road_graph.rs:629-631 split_edge_at length split unsound for 2-point parents: l1 = length*(path1.len()-1)/total_segs gives l1=full/l2=0 whenever total_segs==1, regardless of where jpos falls. Reachable: A*-collapsed 2-point paths AND every snap-spur (vec![apos,mpos]) — so a later minor snapping onto an earlier minor's spur (the documented accreting behavior) always hits it. Proved with a temporary diagnostic test (reverted). Fix: split by the geometric segment-fraction t (available from project_onto_segment at the call site) when total_segs==1, or proportional-floor both halves — never all-or-nothing on a real interior split.
H2 server/src/atlas/road_graph.rs:1614-1671 No test exercises the chained-snap scenario (minor B onto minor A's spur) — the load-bearing accretion path where H1 surfaces. Both snap tests use one minor; the determinism test uses 5 placements (all hubs under HUB_CAP_MIN=6) so attach_minors never runs in it. Add: chained-snap test + a determinism assertion on a 2+-minor scenario that actually exercises attach_minors.
H3 server/src/atlas/cascade.rs:314-323 The Layer2→Layer3 RailHeadFacing wiring has no cascade-level assertion — the integration test executes the path but never inspects placements[..].founding_orientation, so a wiring regression (wrong field/layer/silently-skipped if-let) passes everything. Assert the mutation end-to-end.
H4 server/src/atlas/believability.rs:542-568 The settlement_class drive-by fix is correct but untested; both doc comments say "keep the two in lockstep" yet nothing catches drift mechanically. Add a synthetic-DB unit test mirroring the city_context_reader fixture, asserting settlement_class + is_standalone_hq.

Tyre (Architecture): REQUEST_CHANGES

Deterministic by construction (iteration traced to ORDER BY id, integer octant math, pure-function derivation); .or_else() composition cleanly reuses D-234b flush-frontage; pin-file contract coherent; believability drive-by ruled scope-appropriate. Two gaps:

# File Issue
T1 server/src/atlas/road_graph.rs:105-113 HUB_SPACING_DIAG_PX keys on today's fixed 512×256 grid diagonal with no cross-reference to D-243's elastic planetary seam — when grid size varies per body, this premise silently goes stale. Add a one-line [D-243] cross-ref to the doc comment (and/or a matching note in D-243).
T2 server/src/atlas/city_context_reader.rs:271-278 + believability.rs:536-552 The is_standalone_hq LEFT JOIN is duplicated verbatim across the two readers, synced only by comments. Extract a shared helper/const (single source of truth, per the codebase's own registry pattern) so a predicate change can't be forgotten in one place.

Miri (World Consistency): REQUEST_CHANGES

19 of 20 boundary judgments correct and faithful to Jeroen's ruling; all 11 keep-Standalone values verified corp-by-corp as genuine company-town identities; edge cases (geothermal_hub, military_industrial, lattice_material_source, marine_farming) all correctly sided; Cygni pin move canon-clean (Metropolis is a real pool name on Ferrath); D-242 amendment locks the boundary properly; D-213 amendment consistent.

# File Issue
M1 wiki/economics/corp_hq_placement.toml:124-131 + 5 corp pages The general_industrial flip is correct as a rule but leaves 5 real steel-complex/shipyard corps on the wrong side: stalownia-kowalski (300-year steelworks that IS its system's identity), westphalia-heavy-works, sede-chemical-works, sova-station-works, cygni-combines (the canonical Reach shipyard cluster — whose standalone settlement this PR just retired). The flip's own justification comment describes them exactly. Fix: re-tag the five corps' corp_specialization frontmatter to existing Standalone-shaped values (shipbuilding / military_industrial are the closest fits — both currently have ZERO corps, so this also closes a content hole; sede is the closest call, fuel_production-adjacent in shape) + regen. Content fix, not a boundary-rule change — the craft-majority framing stays.

Verdict: CHANGES REQUESTED

7 findings. H1 is the only correctness bug (proved reachable); H2-H4 are the tests that pin it and the wiring; T1/T2 are small hardening; M1 is a 5-page content re-tag + regen whose cascade consequences (settlements re-emerging, Cygni pin interaction) need coherent handling.

## Review: road-hub-refinement → main (type: code + content slice) Reviewers: Hoshe (code quality), Tyre (architecture), Miri (world consistency — placement-boundary slice). All read source from the branch worktree (HEAD = branch tip, not sparse). Pre-push gate passed before push and was not duplicated. Runtime verification: 4-body live-cascade hub sanity report in the ticket; no client/merge-path code touched. ### Hoshe (Code Quality): REQUEST_CHANGES Hub/minor partition, collapse, and RailHeadFacing correctly implemented and well unit-tested; DB spot-checks match the D-242 amendment exactly (11/144 corps, 340 rows, 129 links, 0 dupes, stamp fresh). One real reachable bug + three coverage gaps in exactly the places that hide it. | # | File | Issue | |---|------|-------| | H1 | server/src/atlas/road_graph.rs:629-631 | **split_edge_at length split unsound for 2-point parents**: l1 = length*(path1.len()-1)/total_segs gives l1=full/l2=0 whenever total_segs==1, regardless of where jpos falls. Reachable: A*-collapsed 2-point paths AND every snap-spur (vec![apos,mpos]) — so a later minor snapping onto an earlier minor's spur (the documented accreting behavior) always hits it. Proved with a temporary diagnostic test (reverted). Fix: split by the geometric segment-fraction t (available from project_onto_segment at the call site) when total_segs==1, or proportional-floor both halves — never all-or-nothing on a real interior split. | | H2 | server/src/atlas/road_graph.rs:1614-1671 | No test exercises the chained-snap scenario (minor B onto minor A's spur) — the load-bearing accretion path where H1 surfaces. Both snap tests use one minor; the determinism test uses 5 placements (all hubs under HUB_CAP_MIN=6) so attach_minors never runs in it. Add: chained-snap test + a determinism assertion on a 2+-minor scenario that actually exercises attach_minors. | | H3 | server/src/atlas/cascade.rs:314-323 | The Layer2→Layer3 RailHeadFacing wiring has no cascade-level assertion — the integration test executes the path but never inspects placements[..].founding_orientation, so a wiring regression (wrong field/layer/silently-skipped if-let) passes everything. Assert the mutation end-to-end. | | H4 | server/src/atlas/believability.rs:542-568 | The settlement_class drive-by fix is correct but untested; both doc comments say "keep the two in lockstep" yet nothing catches drift mechanically. Add a synthetic-DB unit test mirroring the city_context_reader fixture, asserting settlement_class + is_standalone_hq. | ### Tyre (Architecture): REQUEST_CHANGES Deterministic by construction (iteration traced to ORDER BY id, integer octant math, pure-function derivation); .or_else() composition cleanly reuses D-234b flush-frontage; pin-file contract coherent; believability drive-by ruled scope-appropriate. Two gaps: | # | File | Issue | |---|------|-------| | T1 | server/src/atlas/road_graph.rs:105-113 | HUB_SPACING_DIAG_PX keys on today's fixed 512×256 grid diagonal with no cross-reference to D-243's elastic planetary seam — when grid size varies per body, this premise silently goes stale. Add a one-line [D-243] cross-ref to the doc comment (and/or a matching note in D-243). | | T2 | server/src/atlas/city_context_reader.rs:271-278 + believability.rs:536-552 | The is_standalone_hq LEFT JOIN is duplicated verbatim across the two readers, synced only by comments. Extract a shared helper/const (single source of truth, per the codebase's own registry pattern) so a predicate change can't be forgotten in one place. | ### Miri (World Consistency): REQUEST_CHANGES 19 of 20 boundary judgments correct and faithful to Jeroen's ruling; all 11 keep-Standalone values verified corp-by-corp as genuine company-town identities; edge cases (geothermal_hub, military_industrial, lattice_material_source, marine_farming) all correctly sided; Cygni pin move canon-clean (Metropolis is a real pool name on Ferrath); D-242 amendment locks the boundary properly; D-213 amendment consistent. | # | File | Issue | |---|------|-------| | M1 | wiki/economics/corp_hq_placement.toml:124-131 + 5 corp pages | The general_industrial flip is correct as a rule but leaves 5 real steel-complex/shipyard corps on the wrong side: stalownia-kowalski (300-year steelworks that IS its system's identity), westphalia-heavy-works, sede-chemical-works, sova-station-works, cygni-combines (the canonical Reach shipyard cluster — whose standalone settlement this PR just retired). The flip's own justification comment describes them exactly. Fix: re-tag the five corps' corp_specialization frontmatter to existing Standalone-shaped values (shipbuilding / military_industrial are the closest fits — both currently have ZERO corps, so this also closes a content hole; sede is the closest call, fuel_production-adjacent in shape) + regen. Content fix, not a boundary-rule change — the craft-majority framing stays. | ### Verdict: CHANGES REQUESTED 7 findings. H1 is the only correctness bug (proved reachable); H2-H4 are the tests that pin it and the wiring; T1/T2 are small hardening; M1 is a 5-page content re-tag + regen whose cascade consequences (settlements re-emerging, Cygni pin interaction) need coherent handling.
jpmschweitzer added 3 commits 2026-07-16 17:05:14 +02:00
fix(simulation): PR #178 review round — H1-H4, T1, T2 1fee6fd9da 
H1: split_edge_at now splits length_cells geometric-distance proportionally (fixed-order f64 polyline sums, halves sum exactly to parent) — subsumes the 2-point-parent all-or-nothing bug. H2: chained-snap test (minor onto minor spur) + determinism assertion actually exercising attach_minors + direct 2-point split unit test. H3: cascade-level RailHeadFacing end-to-end assertion on a plus-shaped landmass fixture (2 deterministic degree-3 settlement junctions). H4+T2: STANDALONE_HQ_JOIN_SQL shared constant, both readers compose it; believability lockstep unit test. T1: HUB_SPACING_DIAG_PX doc cross-refs D-243 elastic seam.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
fix(db): M1 — re-tag 5 steel-complex/shipyard corps to Standalone-shaped values; regen (PR #178 review) cd71d77284 
cygni-combines -> shipbuilding; sova-station-works, stalownia-kowalski, westphalia-heavy-works -> military_industrial; sede-chemical-works -> fuel_production (closest call, flagged). Standalone 11->16, settlements 11->15, city rows 340->344, tenant links 129->125 (14 unmatched), 0 dupes, fixpoint immediate. Metropolis pin stays (yard is a place beside the city); sova emits nothing pending T-1054.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
docs(meta): D-242 amendments — M1 re-tag rationale + heavy_works future-split note + T1 scale note (PR #178 review) d8f350c74a 
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer commented 2026-07-16 17:05:47 +02:00
Author
Owner
Copy Link
Copy Source

Review response — all 7 findings fixed (commits 1fee6fd9d, cd71d7728, d8f350c74)

No retractions requested. Lead independently verified the DB end-state (16/139 split with all five re-tags, 344 city rows, 125 tenant links, 0 dupes) before this comment.

# Finding Resolution
H1 split_edge_at all-or-nothing on 2-point parents Fixed via the subsuming option: length_cells now splits geometric-distance proportionally (fixed-order f64 polyline sums per the module's determinism doctrine; halves always sum exactly to the parent; degenerate all-coincident geometry → 0). No 2-point special case needed.
H2 No chained-snap / attach_minors determinism coverage Fixed: direct 2-point-split unit test (midpoint 5/5, quarter-point both>0 sum-exact); chained-snap test — minor B snaps onto minor A's spur, the interior 2-point-parent split firing in the accreting path, full invariants asserted; the same scenario carries run()==run(). Confirmed the pre-existing determinism test never reached attach_minors.
H3 RailHeadFacing wiring unasserted end-to-end Fixed: new cascade test on a plus-shaped landmass whose arm geometry deterministically yields 2 degree-3 settlement junctions (the default slope fixture structurally cannot — diagnosed and documented). Asserts junction ⇔ RailHeadFacing both directions on layer3.placements plus BodyWorldState propagation, with a fixture-changed message distinct from wiring-broken.
H4 Believability settlement_class fix untested Fixed: synthetic-DB unit test mirroring the city_context_reader fixture — NameLocked honoured, unknown class → default, HQ flag via the shared join.
T1 HUB_SPACING_DIAG_PX fixed-grid premise unmarked Fixed: doc comment states the 512×256 premise and names D-243's elastic seam as the revisit trigger; matching scale note in the D-242 amendment.
T2 is_standalone_hq join duplicated Fixed: pub(crate) const STANDALONE_HQ_JOIN_SQL in city_context_reader.rs; both readers compose it — single source of truth, plus the H4 lockstep test.
M1 5 steel-complex corps on the wrong side of the general_industrial flip Fixed: cygni-combines → shipbuilding; sova-station-works, stalownia-kowalski, westphalia-heavy-works → military_industrial; sede-chemical-works → fuel_production (closest call, flagged in the frontmatter comment). Each re-tag carries an in-frontmatter justification. Cascade consequences handled: Cygni yards settlement returns on GJ820Bc, Metropolis pin STAYS (canon-clean; the yard is a place beside the city — recorded in the TOML); sova emits nothing pending its broken HQ ref (T-1054, cited). Counts: Standalone 11→16, HQ settlements 11→15, city rows 340→344, tenant links 129→125 (unmatched reconciled to exactly the 14 loop-logged cases); fixpoint immediate (2 regens, 0 differences). Design note recorded in D-242: military_industrial now carries three civilian heavy-works corps — if that reads wrong later, the clean split is a new heavy_works vocabulary value, never re-coarsening general_industrial.

Battery: gate re-run green on push (fmt/clippy/full cargo test/deny/ruff/test-tooling/stamp); touched suites road_graph 19, cascade 9, believability 4, city_context_reader 32; pql decisions validate ok. Reviewers asked to verify their own findings for the re-review verdict.

## Review response — all 7 findings fixed (commits 1fee6fd9d, cd71d7728, d8f350c74) No retractions requested. Lead independently verified the DB end-state (16/139 split with all five re-tags, 344 city rows, 125 tenant links, 0 dupes) before this comment. | # | Finding | Resolution | |---|---------|------------| | H1 | split_edge_at all-or-nothing on 2-point parents | **Fixed via the subsuming option**: length_cells now splits geometric-distance proportionally (fixed-order f64 polyline sums per the module's determinism doctrine; halves always sum exactly to the parent; degenerate all-coincident geometry → 0). No 2-point special case needed. | | H2 | No chained-snap / attach_minors determinism coverage | **Fixed**: direct 2-point-split unit test (midpoint 5/5, quarter-point both>0 sum-exact); chained-snap test — minor B snaps onto minor A's spur, the interior 2-point-parent split firing in the accreting path, full invariants asserted; the same scenario carries run()==run(). Confirmed the pre-existing determinism test never reached attach_minors. | | H3 | RailHeadFacing wiring unasserted end-to-end | **Fixed**: new cascade test on a plus-shaped landmass whose arm geometry deterministically yields 2 degree-3 settlement junctions (the default slope fixture structurally cannot — diagnosed and documented). Asserts junction ⇔ RailHeadFacing both directions on layer3.placements plus BodyWorldState propagation, with a fixture-changed message distinct from wiring-broken. | | H4 | Believability settlement_class fix untested | **Fixed**: synthetic-DB unit test mirroring the city_context_reader fixture — NameLocked honoured, unknown class → default, HQ flag via the shared join. | | T1 | HUB_SPACING_DIAG_PX fixed-grid premise unmarked | **Fixed**: doc comment states the 512×256 premise and names D-243's elastic seam as the revisit trigger; matching scale note in the D-242 amendment. | | T2 | is_standalone_hq join duplicated | **Fixed**: pub(crate) const STANDALONE_HQ_JOIN_SQL in city_context_reader.rs; both readers compose it — single source of truth, plus the H4 lockstep test. | | M1 | 5 steel-complex corps on the wrong side of the general_industrial flip | **Fixed**: cygni-combines → shipbuilding; sova-station-works, stalownia-kowalski, westphalia-heavy-works → military_industrial; sede-chemical-works → fuel_production (closest call, flagged in the frontmatter comment). Each re-tag carries an in-frontmatter justification. Cascade consequences handled: Cygni yards settlement returns on GJ820Bc, Metropolis pin STAYS (canon-clean; the yard is a place beside the city — recorded in the TOML); sova emits nothing pending its broken HQ ref (T-1054, cited). Counts: Standalone 11→16, HQ settlements 11→15, city rows 340→344, tenant links 129→125 (unmatched reconciled to exactly the 14 loop-logged cases); fixpoint immediate (2 regens, 0 differences). Design note recorded in D-242: military_industrial now carries three civilian heavy-works corps — if that reads wrong later, the clean split is a new heavy_works vocabulary value, never re-coarsening general_industrial. | Battery: gate re-run green on push (fmt/clippy/full cargo test/deny/ruff/test-tooling/stamp); touched suites road_graph 19, cascade 9, believability 4, city_context_reader 32; pql decisions validate ok. Reviewers asked to verify their own findings for the re-review verdict.
jpmschweitzer commented 2026-07-16 17:14:54 +02:00
Author
Owner
Copy Link
Copy Source

Re-review verdict: APPROVED

All three reviewers re-verified their own findings against branch tip d8f350c74:

  • Hoshe: APPROVE. H1-H4 resolved and mutation-tested — he temporarily reintroduced each bug (split math, cascade wiring, believability hardcode) and confirmed the new tests fail with clear diagnostics, then reverted cleanly. Re-ran his original bug-reproduction scenario against the geometric split: 12-cell parent → 10+2 proportional, sum conserved. Full atlas suite 505/505. Bundled M1/DB state spot-checked to exact match.
  • Tyre: APPROVE. T1 (D-243 revisit trigger on the hub-cap premise) and T2 (STANDALONE_HQ_JOIN_SQL single source of truth + lockstep test) fully resolved. His one re-review comment — the heavy_works split living only as prose — is now tracked as T-1117 (with his full blast-radius analysis: live CompatibilityMatrix military-row scoring + latent FortifiedPerimeter hazard), which he verified before flipping to approve.
  • Miri: APPROVE. All five re-tags verified (shipbuilding a direct fit; the three military_industrial placements correct with the documented tension; sede flagged honestly as loosest); Metropolis pin + returned Cygni yard settlement verified live in the DB as a coherent "yard beside the city" outcome; D-242 note faithful.

Merging to main and closing. Follow-ups tracked: T-1115 (wiki coverage), T-1116 (water-heavy bodies 0 routable edges), T-1117 (heavy_works vocabulary split).

## Re-review verdict: APPROVED All three reviewers re-verified their own findings against branch tip d8f350c74: - **Hoshe: APPROVE.** H1-H4 resolved and mutation-tested — he temporarily reintroduced each bug (split math, cascade wiring, believability hardcode) and confirmed the new tests fail with clear diagnostics, then reverted cleanly. Re-ran his original bug-reproduction scenario against the geometric split: 12-cell parent → 10+2 proportional, sum conserved. Full atlas suite 505/505. Bundled M1/DB state spot-checked to exact match. - **Tyre: APPROVE.** T1 (D-243 revisit trigger on the hub-cap premise) and T2 (STANDALONE_HQ_JOIN_SQL single source of truth + lockstep test) fully resolved. His one re-review comment — the heavy_works split living only as prose — is now tracked as T-1117 (with his full blast-radius analysis: live CompatibilityMatrix military-row scoring + latent FortifiedPerimeter hazard), which he verified before flipping to approve. - **Miri: APPROVE.** All five re-tags verified (shipbuilding a direct fit; the three military_industrial placements correct with the documented tension; sede flagged honestly as loosest); Metropolis pin + returned Cygni yard settlement verified live in the DB as a coherent "yard beside the city" outcome; D-242 note faithful. Merging to main and closing. Follow-ups tracked: T-1115 (wiki coverage), T-1116 (water-heavy bodies 0 routable edges), T-1117 (heavy_works vocabulary split).
jpmschweitzer closed this pull request 2026-07-16 17:17:35 +02:00

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
jpmschweitzer (Jeroen Schweitzer)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: jpmschweitzer/settled-reach#178