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

feat(simulation): architecture-flavor pipeline (T-992/994/995/1003) #173

Closed
jpmschweitzer wants to merge 0 commits from architecture-flavor-pipeline into main
pull from: architecture-flavor-pipeline
merge into: :main
Conversation 3 Commits - Files Changed -
jpmschweitzer commented 2026-07-08 10:01:03 +02:00
Owner
Copy Link
Copy Source

Summary

The D-232 architecture-flavor mechanism — the last blocker on T-959 (Layer 5 tile fill) and behind it T-962 (Phase 5 gate).

  • T-994 — three-phase trait draw: body K-draw (SeedChain::for_body, coverage-aware over a new body-level district-mix aggregate), district-dominant pick keyed by the D-243 2048m District cell, resolved at L3→L4 dispatch (FillChunk stays pure). Replaces the flat splitmix64 flavor pick. New trait_catalog_reader + trait_draw modules.
  • T-1003 — deviation/swerve system: ArchitectureFlavorRef = InVocabulary(u8) | Swerve(tag); rare per-building wildcard with foreign-import/heritage-callback pools + driver-scaled integer-bps rates (placeholder constants pending calibration); sparsity escape hatch at the phase-2 pick.
  • T-995 — ObjectTag registry ratified (shipped 28-tag palette canonical): wiki/economics/object_tag_vocabulary.toml + importer validation V-TT-03/V-TT-04 (tag existence, axis match, acyclic fallback graph). Resolves Q-049. systems.db regenerated + stamped.
  • T-992 — Gemma architecture-flavor methodology doc (text-fields-only corpus distillation, human-gated).

Governance: D-232 amended (2048m district pin — capture gap from D-222/D-243; swerve driver mappings), D-229 flavor_ref enum, D-235 vocabulary ratification, Q-049 resolved.

Also: chore(engine) crossbeam-epoch 0.9.20 (RUSTSEC-2026-0204, caught by the gate).

Gates: full cargo test (1630 green incl. golden harnesses), clippy -D warnings, ruff, test-tooling, cargo deny, fresh systems.db stamp.

🤖 Generated with Claude Code

## Summary The D-232 architecture-flavor mechanism — the last blocker on T-959 (Layer 5 tile fill) and behind it T-962 (Phase 5 gate). - **T-994** — three-phase trait draw: body K-draw (SeedChain::for_body, coverage-aware over a new body-level district-mix aggregate), district-dominant pick keyed by the D-243 2048m District cell, resolved at L3→L4 dispatch (FillChunk stays pure). Replaces the flat splitmix64 flavor pick. New trait_catalog_reader + trait_draw modules. - **T-1003** — deviation/swerve system: ArchitectureFlavorRef = InVocabulary(u8) | Swerve(tag); rare per-building wildcard with foreign-import/heritage-callback pools + driver-scaled integer-bps rates (placeholder constants pending calibration); sparsity escape hatch at the phase-2 pick. - **T-995** — ObjectTag registry ratified (shipped 28-tag palette canonical): wiki/economics/object_tag_vocabulary.toml + importer validation V-TT-03/V-TT-04 (tag existence, axis match, acyclic fallback graph). Resolves Q-049. systems.db regenerated + stamped. - **T-992** — Gemma architecture-flavor methodology doc (text-fields-only corpus distillation, human-gated). Governance: D-232 amended (2048m district pin — capture gap from D-222/D-243; swerve driver mappings), D-229 flavor_ref enum, D-235 vocabulary ratification, Q-049 resolved. Also: chore(engine) crossbeam-epoch 0.9.20 (RUSTSEC-2026-0204, caught by the gate). Gates: full cargo test (1630 green incl. golden harnesses), clippy -D warnings, ruff, test-tooling, cargo deny, fresh systems.db stamp. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
jpmschweitzer added 5 commits 2026-07-08 10:01:03 +02:00
chore(meta): refine + activate T-977 batch — T-992/T-994/T-995/T-1003 in_progress, T-1003 blocked by T-994, T-1004 parked to Phase 5 e097874587 
Si refinement round resolved 7 decisions: district-dominant = D-243 2048m tier
(build in T-994), draw at GenerateSkeleton, body-level coverage aggregate,
ratify shipped 28-tag palette (amend D-235), Gemma proposes text fields only,
swerve driver mappings pinned, T-1004 render mechanism deferred to Phase 5.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
docs(assets): Gemma architecture-flavor methodology — bounded corpus distillation (T-992) a6c2ff4740 
Stage-1 register selection over the current wiki tree (Q-107 assumption
stated), stage-2 TEXT-FIELDS-ONLY structured proposals (tag, label,
cultural_description, candidate allow/block tags) — humans own all numeric
eligibility/weights (catalog-wide CI guardrails can't be satisfied
per-template); Araminta owns visual_bundle. No era-bands (Q-106 dissolved);
trauma_decay_rate stale-schema note. Mirrors docs/gemma-naming-methodology.md
incl. the sr-voice-tooling serving pattern. Worked example: GJ-4053 granite
dry-stack accept; GJ-1248/GJ-285/GJ-588 as correct stage-1 rejections.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
feat(db): ObjectTag registry ratified + importer validation — resolves Q-049 (T-995) 766ceb436c 
The shipped 28-tag catalog palette is canonical (user ratification,
/whats-next refinement 2026-07-07). New machine-readable registry
wiki/economics/object_tag_vocabulary.toml (wall/roof/facade/street axes +
4 generic fallback terminals, Miri/Araminta co-owned header). Importer
validation in economy_import/traits.py: V-TT-03 (every catalog
allow/block/visual_bundle tag exists in the registry, axis-checked) and
V-TT-04 (fallback graph: non-generics declare a parent, chains acyclic,
resolve to a generic — absorbs T-1004's Phase-4 slice). Registry added to
generator_sources (stamp coverage); systems.db regenerated + stamped.
Q-049 marked Resolved — divergence is now a loud build failure.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
feat(simulation): D-232 three-phase trait draw + deviation/swerve system (T-994, T-1003) 90655382eb 
T-994 — three-phase draw replacing the flat splitmix64 flavor pick:
- Phase 1 body K-draw (trait_draw.rs): hard gates -> soft weights (two-part
  geographic_sector join, PR #148) -> SeedChain::for_body-seeded weighted
  draw of K (5/3/1/0 by ComplexityTier), pins count toward K, coverage-aware
  over a new body-level district-type-mix aggregate threaded into
  CityGenerationContext at L3->L4 dispatch.
- Phase 2 district-dominant pick keyed by the D-243 2048m District cell —
  settlements sharing a cell independently derive the identical template.
  Resolved at dispatch, never in FillChunk (T-987 stays pure).
- trait_catalog_reader.rs: systems.db reader for trait_templates +
  atlas_body_trait_bias (D-225 resource pattern), registered in main.rs
  with graceful degradation.
- BlockSkeleton carries district_type; assign_block_tags is a pure lookup.

T-1003 — deviation/swerve system (trait_swerve.rs):
- ArchitectureFlavorRef is now InVocabulary(u8) | Swerve(tag).
- Rare per-building wildcard: foreign-import pool (other corridors +
  cross_corridor) driven by Epicenter/Passage tier, mixed faction,
  road-graph degree; heritage-callback pool (own-corridor heritage) driven
  by isolation, remote tier, founding_age_years. Integer bps, base 100,
  cap 300 per driver — placeholder constants pending calibration.
- Sparsity escape hatch = same mechanism, necessity-triggered at the
  phase-2 pick (deterministic max-weight over the hard-gate-eligible pool).
- New SeedDomains: TraitVocabulary(13) / TraitDistrict(14) / TraitSwerve(15).

Governance: D-232 amended (district pinned to the D-243 2048m tier — the
D-222/D-243 redefinition was never captured; swerve driver mappings +
placeholder rates recorded), D-229 flavor_ref field updated to the enum
shape, D-235 amended (T-995 vocabulary ratification note; rides here due
to single-file entanglement).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
chore(engine): bump crossbeam-epoch 0.9.18 -> 0.9.20 (RUSTSEC-2026-0204) 034791602b 
Invalid pointer dereference in fmt::Pointer for Atomic/Shared null
pointers; flagged by cargo deny in the pre-push gate. Transitive via
rayon; lockfile-only change.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer commented 2026-07-08 11:24:10 +02:00
Author
Owner
Copy Link
Copy Source

Review: architecture-flavor-pipeline → main (type: code)

Reviewers: Hoshe (code quality), Tyre (architecture) — both read from the branch worktree at 034791602.

Hoshe (Code Quality): REQUEST_CHANGES

The three-phase trait-draw mechanism is well-designed and thoroughly unit-tested at the pure-function level — determinism, hard gates, coverage repair, swerve rarity, and the sparsity escape hatch all have solid tests mapping cleanly to D-232. Findings target the untested integration seam and robustness gaps.

# File Issue
H1 server/src/atlas/plugin.rs ~118–256 T-994 body-level aggregation (coverage set, max-K/prosperity aggregation, BodyVocabularyContext threading) is untested — existing tests bypass the whole branch. Add a test wiring the readers with 2+ placements asserting identical trait_selection on both dispatched contexts
H2 tooling/economy-db/economy_import/traits.py 31–124 V-TT-03/04 graph validation has no failure-branch tests (cycle, duplicate tag, axis mismatch, missing registry, generic-with-fallback). Pure dict-in/errors-out logic — add test_traits.py
H3 server/src/atlas/trait_catalog_reader.rs 284–314 parse_bulk_class_gate/parse_production_ubiquity_gate swallow malformed JSON via unwrap_or_default() with no logging — silently widening hard-gate eligibility. Sibling map-parsers tracing::warn! on the identical failure; match them
H4 server/src/atlas/trait_draw.rs 223–235, 296–309 No guard/warn when pins exceed K — silently breaks "K is not a range" (D-232); the over-K branch is untested. Warn in draw_body_vocabulary + bound pins in an importer guardrail
H5 plugin.rs ~198 / trait_catalog_reader.rs The immutable catalog is re-queried + re-parsed from SQLite on every body dispatch (~300 identical reads per run). Cache/memoize the parsed Vec<TraitTemplate>

Tyre (Architecture): REQUEST_CHANGES

Dispatch-time vs Rayon/FillChunk split correct (T-987 held), D-243 district-cell keying verified to make co-district settlements agree, all-integer/BTreeMap (D-010 clean), one-DB-round-trip-per-settlement aggregation is elegant. Two design-fidelity gaps in the phase-1 draw, two precision nits.

# File Issue
T1 trait_draw.rs 175–191, 237–288 corridor_pool == "heritage" templates enter the ordinary K-draw lottery ungated. D-232's two-part corridor pool reserves heritage for the remoteness dial ("tagged so the remoteness dial draws specifically from it") — with real catalog weights (7000–8000 bps) a deep-history callback becomes an everyday vocabulary member anywhere. Exclude heritage from the ordinary weighted pool + coverage repair; pins, phase-2 necessity_swerve, and the T-1003 swerve pools keep reading the full eligible set
T2 trait_draw.rs 296–309 Coverage repair pushes past K when every slot is pinned — but D-232's sparsity escape hatch is phase 2's necessity_swerve (out-of-vocabulary, never grows trait_selection), and "K is not a range." Leave the type uncovered here and let phase 2 handle it
T3 seed.rs 128–133 vs trait_draw.rs 384–388 SeedDomain::TraitDistrict doc says "packed (DistrictPos, DistrictType) id"; implementation chains two .derive() calls under one domain. Deterministic and safe, but doc and code disagree on a pinned D-224 contract — align one to the other
T4 trait_catalog_reader.rs 1–14 Module doc cites D-225 (the client-facing layer-stream proxy) for a pattern this reader doesn't follow. Cite the CityContextReader precedent (the pattern actually copied) instead

Verdict: CHANGES REQUESTED

Process notes:

  • Runtime verification performed: full cargo test (1630 incl. golden harnesses), clippy -D warnings, ruff, make test-tooling, cargo deny, fresh systems.db stamp — all in the pre-push gate.
  • Zero warnings introduced. No client merge-path touched (the one main.rs change is server startup resource registration with graceful degradation).
  • T2 + H4 converge on the same pins/K seam — one fix closes both.
## Review: architecture-flavor-pipeline → main (type: code) Reviewers: Hoshe (code quality), Tyre (architecture) — both read from the branch worktree at `034791602`. ### Hoshe (Code Quality): REQUEST_CHANGES The three-phase trait-draw mechanism is well-designed and thoroughly unit-tested at the pure-function level — determinism, hard gates, coverage repair, swerve rarity, and the sparsity escape hatch all have solid tests mapping cleanly to D-232. Findings target the untested integration seam and robustness gaps. | # | File | Issue | |---|------|-------| | H1 | `server/src/atlas/plugin.rs` ~118–256 | T-994 body-level aggregation (coverage set, max-K/prosperity aggregation, `BodyVocabularyContext` threading) is untested — existing tests bypass the whole branch. Add a test wiring the readers with 2+ placements asserting identical `trait_selection` on both dispatched contexts | | H2 | `tooling/economy-db/economy_import/traits.py` 31–124 | V-TT-03/04 graph validation has no failure-branch tests (cycle, duplicate tag, axis mismatch, missing registry, generic-with-fallback). Pure dict-in/errors-out logic — add `test_traits.py` | | H3 | `server/src/atlas/trait_catalog_reader.rs` 284–314 | `parse_bulk_class_gate`/`parse_production_ubiquity_gate` swallow malformed JSON via `unwrap_or_default()` with no logging — silently **widening** hard-gate eligibility. Sibling map-parsers `tracing::warn!` on the identical failure; match them | | H4 | `server/src/atlas/trait_draw.rs` 223–235, 296–309 | No guard/warn when pins exceed K — silently breaks "K is not a range" (D-232); the over-K branch is untested. Warn in `draw_body_vocabulary` + bound pins in an importer guardrail | | H5 | `plugin.rs` ~198 / `trait_catalog_reader.rs` | The immutable catalog is re-queried + re-parsed from SQLite on every body dispatch (~300 identical reads per run). Cache/memoize the parsed `Vec<TraitTemplate>` | ### Tyre (Architecture): REQUEST_CHANGES Dispatch-time vs Rayon/FillChunk split correct (T-987 held), D-243 district-cell keying verified to make co-district settlements agree, all-integer/BTreeMap (D-010 clean), one-DB-round-trip-per-settlement aggregation is elegant. Two design-fidelity gaps in the phase-1 draw, two precision nits. | # | File | Issue | |---|------|-------| | T1 | `trait_draw.rs` 175–191, 237–288 | `corridor_pool == "heritage"` templates enter the *ordinary* K-draw lottery ungated. D-232's two-part corridor pool reserves heritage for the remoteness dial ("tagged so the remoteness dial draws specifically from it") — with real catalog weights (7000–8000 bps) a deep-history callback becomes an everyday vocabulary member anywhere. Exclude heritage from the ordinary weighted pool + coverage repair; pins, phase-2 `necessity_swerve`, and the T-1003 swerve pools keep reading the full eligible set | | T2 | `trait_draw.rs` 296–309 | Coverage repair `push`es past K when every slot is pinned — but D-232's sparsity escape hatch is phase 2's `necessity_swerve` (out-of-vocabulary, never grows `trait_selection`), and "K is not a range." Leave the type uncovered here and let phase 2 handle it | | T3 | `seed.rs` 128–133 vs `trait_draw.rs` 384–388 | `SeedDomain::TraitDistrict` doc says "packed `(DistrictPos, DistrictType)` id"; implementation chains two `.derive()` calls under one domain. Deterministic and safe, but doc and code disagree on a pinned D-224 contract — align one to the other | | T4 | `trait_catalog_reader.rs` 1–14 | Module doc cites D-225 (the client-facing layer-stream proxy) for a pattern this reader doesn't follow. Cite the `CityContextReader` precedent (the pattern actually copied) instead | ### Verdict: CHANGES REQUESTED Process notes: - Runtime verification performed: full `cargo test` (1630 incl. golden harnesses), clippy `-D warnings`, ruff, `make test-tooling`, `cargo deny`, fresh systems.db stamp — all in the pre-push gate. - Zero warnings introduced. No client merge-path touched (the one `main.rs` change is server startup resource registration with graceful degradation). - T2 + H4 converge on the same pins/K seam — one fix closes both.
jpmschweitzer added 1 commit 2026-07-08 12:19:46 +02:00
fix(simulation): PR #173 review round — all 9 findings addressed ebe8742469 
T1: heritage corridor_pool excluded from the ordinary phase-1 lottery and
coverage repair (D-232 reserves the heritage sub-pool for the remoteness
dial); reachable via hero pin, necessity swerve, and the T-1003 heritage
pool only. 3 new tests.
T2+H4: coverage repair no longer grows trait_selection past K when all
slots are pinned (phase-2 necessity swerve serves the type instead);
runtime warn when authored pins exceed K; V-TT-05 importer guardrail
bounds pins per body at 5 (max ComplexityTier K). 2 new tests + 2 python
tests.
H1: body-level dispatch aggregation extracted to pure
aggregate_body_dispatch_inputs + tested directly (union mix, MAX
prosperity/K); threading test asserts identical vocabulary/pools across
co-body settlements with per-settlement swerve rates. 2 new tests.
H2: tooling/economy-db/test_traits.py — 14 stdlib unittest cases over
V-TT-03/04/05 failure branches, wired into make test-tooling.
H3: hard-gate JSON parsers now tracing::warn on malformed blobs (silent
gate-widening) matching the sibling map parsers.
H5: catalog read memoized (OnceLock) — SQL+parse once per server run,
bias stays per-body. 1 new test.
T3: TraitDistrict seed-domain doc aligned with the two-level derive chain.
T4: D-225 misattribution dropped from the reader module doc.

systems.db regenerated + stamped (traits.py is a stamped source).
Gates: full cargo test 1638 green (goldens intact), clippy -D warnings,
ruff, make test-tooling (now incl. the traits units).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
jpmschweitzer commented 2026-07-08 12:48:04 +02:00
Author
Owner
Copy Link
Copy Source

Review response — all 9 findings fixed in ebe874246

No pushbacks; every finding was verified against the branch and addressed.

# Resolution
T1 corridor_pool == "heritage" now excluded from the ordinary phase-1 lottery and coverage repair. A heritage template reaches a body only via an authored hero pin, the phase-2 necessity swerve, or the T-1003 heritage-callback pool (all read the full eligible set). 3 regression tests, incl. a 50-body sweep proving a 100× overweighted heritage template never wins the lottery.
T2 Coverage repair no longer pushes past K when every slot is pinned — the type is left to the phase-2 necessity_swerve (the actual D-232 sparsity escape hatch), with a debug log. "K is not a range" holds. Test: K=1 pinned body stays at exactly 1 template.
H4 Same seam: runtime tracing::warn! when authored pins exceed K (authored intent kept, loudly), plus a new V-TT-05 importer guardrail — >5 pins on one body (max ComplexityTier K) aborts the import. Tests on both sides.
H1 The body-level aggregation is extracted into pure aggregate_body_dispatch_inputs and tested directly (union district mix vs per-placement recompute, MAX prosperity, MAX K incl. the ghost-stub K=0 case). A second test drives build_skeleton_work_item for two co-body settlements and asserts identical trait_selection/mix/pools with per-settlement swerve rates (exact bps values locked: connected-young (120,150) vs isolated-old (100,300) capped).
H2 New tooling/economy-db/test_traits.py — 14 stdlib unittest cases covering every V-TT-03/04 failure branch (missing registry, unknown axis, duplicate tag, missing description, generic-with-fallback, orphan non-generic, unknown-fallback, cycle, wrong generic set) plus the catalog cross-checks (asserting the specific V-TT-03 line in captured output, since tiny catalogs also trip V-TT-01) and V-TT-05. Wired into make test-tooling as its own step.
H3 Both hard-gate parsers now tracing::warn! on malformed JSON ("gate widens to all classes/values") instead of silently unwrap_or_default() — matching the map-parser siblings.
H5 read_catalog() memoized via OnceLock — SQL + JSON parse once per server run, clones after; errors are not cached (transient failures retry). Per-body bias intentionally uncached. Test proves post-first-read DB writes are invisible.
T3 SeedDomain::TraitDistrict doc rewritten to describe the actual two-level derive chain (pos id, then type ordinal) — code unchanged, so no derivation churn.
T4 D-225 citation dropped from the reader module doc; the CityContextReader precedent (already cited) is the pattern actually followed.

Gates re-run after the fixes: full cargo test 1638 green (goldens intact — the heritage exclusion did not shift any golden output), clippy -D warnings, ruff, make test-tooling (now including the new traits units), systems.db regenerated + freshly stamped (traits.py is a stamped generator source).

## Review response — all 9 findings fixed in `ebe874246` No pushbacks; every finding was verified against the branch and addressed. | # | Resolution | |---|-----------| | T1 | `corridor_pool == "heritage"` now excluded from the ordinary phase-1 lottery **and** coverage repair. A heritage template reaches a body only via an authored hero pin, the phase-2 necessity swerve, or the T-1003 heritage-callback pool (all read the full eligible set). 3 regression tests, incl. a 50-body sweep proving a 100× overweighted heritage template never wins the lottery. | | T2 | Coverage repair no longer pushes past K when every slot is pinned — the type is left to the phase-2 `necessity_swerve` (the actual D-232 sparsity escape hatch), with a debug log. "K is not a range" holds. Test: K=1 pinned body stays at exactly 1 template. | | H4 | Same seam: runtime `tracing::warn!` when authored pins exceed K (authored intent kept, loudly), plus a new **V-TT-05** importer guardrail — >5 pins on one body (max ComplexityTier K) aborts the import. Tests on both sides. | | H1 | The body-level aggregation is extracted into pure `aggregate_body_dispatch_inputs` and tested directly (union district mix vs per-placement recompute, MAX prosperity, MAX K incl. the ghost-stub K=0 case). A second test drives `build_skeleton_work_item` for two co-body settlements and asserts identical `trait_selection`/mix/pools with per-settlement swerve rates (exact bps values locked: connected-young `(120,150)` vs isolated-old `(100,300)` capped). | | H2 | New `tooling/economy-db/test_traits.py` — 14 stdlib unittest cases covering every V-TT-03/04 failure branch (missing registry, unknown axis, duplicate tag, missing description, generic-with-fallback, orphan non-generic, unknown-fallback, cycle, wrong generic set) plus the catalog cross-checks (asserting the specific V-TT-03 line in captured output, since tiny catalogs also trip V-TT-01) and V-TT-05. Wired into `make test-tooling` as its own step. | | H3 | Both hard-gate parsers now `tracing::warn!` on malformed JSON ("gate widens to all classes/values") instead of silently `unwrap_or_default()` — matching the map-parser siblings. | | H5 | `read_catalog()` memoized via `OnceLock` — SQL + JSON parse once per server run, clones after; errors are not cached (transient failures retry). Per-body bias intentionally uncached. Test proves post-first-read DB writes are invisible. | | T3 | `SeedDomain::TraitDistrict` doc rewritten to describe the actual two-level derive chain (pos id, then type ordinal) — code unchanged, so no derivation churn. | | T4 | D-225 citation dropped from the reader module doc; the `CityContextReader` precedent (already cited) is the pattern actually followed. | Gates re-run after the fixes: full `cargo test` **1638 green** (goldens intact — the heritage exclusion did not shift any golden output), clippy `-D warnings`, ruff, `make test-tooling` (now including the new traits units), systems.db regenerated + freshly stamped (`traits.py` is a stamped generator source).
jpmschweitzer commented 2026-07-08 15:56:40 +02:00
Author
Owner
Copy Link
Copy Source

Re-review (round 2): APPROVED

Both reviewers verified every fix in ebe874246 against the branch worktree — not just presence, but correctness.

Hoshe (Code Quality): APPROVE

All H1–H5 VERIFIED-FIXED. Hand-verified the threading test's swerve-rate constants (120/150 vs 100/300) against compute_swerve_rates arithmetic; confirmed OnceLock error-path semantics (failures retry, successes cache); ran all 14 test_traits.py cases and the 438 atlas:: Rust tests locally — green. The python3 (vs $(VENV_PY)) Makefile invocation is correct: the module is stdlib-only, matching the dry-run step's convention. No new issues.

Tyre (Architecture): APPROVE

All T1–T4 VERIFIED-FIXED. Traced T1's heritage exclusion at both fork points (lottery + coverage repair) and confirmed every escape hatch — hero pin, build_swerve_pools, phase-2 necessity_swerve — still reads the unfiltered eligible set, so a body whose only zone-covering template is heritage is still served (deterministically, as a Swerve ref). Traced T2's deferred DistrictType end-to-end into phase-2 necessity. Confirmed the H1 aggregation extraction is a pure refactor (identical BTreeSet→Vec ordering, D-010 clean). V-TT-05's bound of 5 matches complexity_k(Full) exactly. Independently re-ran full cargo test, test_traits.py, targeted clippy, and the stamp check — all clean. No new issues.

Verdict: APPROVED — proceeding to merge.

## Re-review (round 2): APPROVED Both reviewers verified every fix in `ebe874246` against the branch worktree — not just presence, but correctness. ### Hoshe (Code Quality): APPROVE All H1–H5 VERIFIED-FIXED. Hand-verified the threading test's swerve-rate constants (120/150 vs 100/300) against `compute_swerve_rates` arithmetic; confirmed OnceLock error-path semantics (failures retry, successes cache); ran all 14 `test_traits.py` cases and the 438 `atlas::` Rust tests locally — green. The `python3` (vs `$(VENV_PY)`) Makefile invocation is correct: the module is stdlib-only, matching the dry-run step's convention. No new issues. ### Tyre (Architecture): APPROVE All T1–T4 VERIFIED-FIXED. Traced T1's heritage exclusion at both fork points (lottery + coverage repair) and confirmed every escape hatch — hero pin, `build_swerve_pools`, phase-2 `necessity_swerve` — still reads the unfiltered eligible set, so a body whose only zone-covering template is heritage is still served (deterministically, as a `Swerve` ref). Traced T2's deferred DistrictType end-to-end into phase-2 necessity. Confirmed the H1 aggregation extraction is a pure refactor (identical BTreeSet→Vec ordering, D-010 clean). V-TT-05's bound of 5 matches `complexity_k(Full)` exactly. Independently re-ran full `cargo test`, `test_traits.py`, targeted clippy, and the stamp check — all clean. No new issues. ### Verdict: APPROVED — proceeding to merge.
jpmschweitzer closed this pull request 2026-07-08 17:11:31 +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#173