feat(ecology): distinguish inherited plans in verified census
Foundation / checks (tsan) (push) Canceled after 0s
Foundation / postgres-tools (push) Canceled after 0s
Foundation / server-image (push) Canceled after 0s
Foundation / image (push) Canceled after 0s
Foundation / checks (checks) (push) Canceled after 1m13s
Foundation / checks (client) (push) Canceled after 1m14s
Foundation / checks (postgres) (push) Canceled after 1m7s
Foundation / checks (deploy) (push) Canceled after 1m13s
Foundation / checks (ecs) (push) Canceled after 1m13s
Foundation / checks (network) (push) Canceled after 1m14s

This commit is contained in:
dingo
2026-10-19 15:06:47 -04:00
parent c60cdf8e19
commit dc7c0e1fa1
20 changed files with 492 additions and 23 deletions
+59
View File
@@ -1,5 +1,64 @@
# Validation report
## 2026-10-19 - Retained-plan census projection 2 and committed observations
[ADR-0169](../adr/0169-retained-plan-census-observations.md) and the
[projection contract](../../schemas/ecology/regional-cellular-census-v2.md)
separate living plan holders and dead-plan history from expressed organisms and
inventory. Regional/scoped boundaries carry the projection version. Existing
history and registered capture fully reconstruct their cuts before observation;
no new SQL schema or census journal is added.
Executed locally:
- MSVC 19.44 Release full CTest **90/90, 23.07 s**, outside the filesystem
sandbox. The first resumed sandbox run passed 88/90 and failed its two
hard-link fixtures with access denied; the unrestricted full rerun passed.
- GCC 14.2 Release full CTest **90/90, 17.84 s**, with NDEBUG.
- Clang 19.1/clang-tidy 19 focused projection test **1/1, 6.92 s** and repository
format check passed.
- Clang 19.1 ASan/UBSan retained-plan owner, new census and legacy census
**3/3, 89.62 s**, without reports.
- Clang 19.1/clang-tidy 19 private database population suite **56 s total**;
Clang ASan/UBSan equivalent **114 s total**, without reports. Both include
the new committed-history/registered-capture integration and the existing
raw/cohort, interval, atomic migration and registered coverage/fault controls.
- GCC 14.2 persistence consumers rebuilt and the same private population suite
passed **38 s total**, including both committed plan-census observations.
Tests earn a structural offspring plan through actual feeding/selection. Before
birth the plan has zero expressed carriers/stock and one living holder. Real
funded births express the plan; a subsequent lower-ID plan leaves earlier
children unchanged. Exact and descendant measures, per-kind sums, local corpse
history, actual death, logout/suspension, detached observations and restore
retain their separate meanings. The total resident count/body/reserve equals
the biological owner and total living holders equals living local organisms.
Two-region queries reject a mixed projection boundary. Legacy projection 1
retains old values and unobserved zero-default plan fields; no golden changes.
The database fixture selects a real plan, publishes result 4 and reconstructs
the exact selected head through history and registered capture. After repeated
births and replacement authority it repeats the comparison with direct
observation. Same-tick upgrade history retains projection 2 and its original
head/tick. An unavailable retained build refuses the complete registered query.
Generic historical replay, coverage and migration evidence checks remain active.
These are read-only engineering observations. No plan-holder count promises
funded reproduction, survival or global viable-store completeness. Population,
migration, other stores/clone domains, accepted extinction/freshness, client
permissions and target-host query costs remain separate work. No player UI,
actual-days ecology, live deployment or M12 completion is claimed.
Visual review used the existing 2026-10-17 `build/m12-ui-review` renders of the
compact HUD and evolution panel; these are retained review artifacts, not newly
captured screens. The current evolution panel describes changing the existing
cell. Retained-plan selection needs an explicit versioned presentation before
player exposure; this segment changes no UI or live/default biological opening.
The census push refreshes the DEV window to **2026-10-19 23:06:09 UTC**. Its
exact staged deployment policy passes with 138 objects rendered across 17
overlays and zero failures. This is no assertion of deployed reachability.
## 2026-10-19 - Durable raw retained-plan admission
[ADR-0168](../adr/0168-durable-retained-plan-admission.md)/schema 55 adds the