feat(lifecycle): add opt-in funded cellular maturation
Foundation / checks (deploy) (push) Successful in 5s
Foundation / checks (ecs) (push) Successful in 2m25s
Foundation / checks (network) (push) Successful in 21m0s
Foundation / checks (client) (push) Successful in 22m13s
Foundation / checks (tsan) (push) Successful in 26m35s
Foundation / checks (postgres) (push) Successful in 1h1m27s
Foundation / checks (checks) (push) Successful in 1h14m29s
Foundation / postgres-tools (push) Successful in 1m19s
Foundation / server-image (push) Successful in 3m46s
Foundation / image (push) Successful in 4m57s

This commit is contained in:
dingo
2026-10-18 21:48:27 -04:00
parent 27ec5729bf
commit a5307c768b
22 changed files with 889 additions and 19 deletions
+75
View File
@@ -1,5 +1,80 @@
# Validation report
## 2026-10-18 - Portable M12 funded maturation prerequisite
[ADR-0157](../adr/0157-funded-cellular-maturation.md) adds an explicit detached
rules-8/owner-7 -> rules-9 maturation candidate, including exact population tags.
The old owner formats, rules, golden vectors, default/DEV admission, SQL schema
52 and durable bundle allowlists are unchanged. Construction spends exact gross
reserve intake, returns body rounding releases and puts energy into world heat.
The trial cap is 5 realized thousandths per 50 ms active tick. Existing-size
actions finish first; newborn/offline cells do not construct. Fixed-corner
expansion must fit boundaries and all present neighbors, in stable life order.
Saved mutation choices keep their original draw and fund current-size bodies.
Executed locally:
- MSVC default Release full CTest **81/81 passed, 15.33 s**, with NDEBUG.
- GCC 14 Release full portable CTest **81/81 passed, 12.68 s**, with NDEBUG.
- Clang 19 ASan/UBSan affected development, uptake, habitat, step/region, owner,
metabolism, mutation, budding, maturation/process continuation, populations and
checkpoint-admission contracts **13/13 passed, 56.43 s**, without reports.
After adding the final discriminating controlled-subadult offline guard, both
maturation/process contracts passed again **2/2, 28.56 s**, without reports.
- Earlier in this same scope, the unchanged implementation's raw owner, mutation,
maturation, populations, census and migration contracts passed ASan/UBSan
**6/6, 103.80 s**, including the legacy rules-8 regression paths.
- clang-tidy 19 passed on both changed biology sources, all three changed
lifecycle sources and both test files. The final offline fixture passed a
separate analysis recheck. Repository clang-format 19 and `git diff --check`
passed. No old golden or numbered migration was replaced.
The family test starts from a genuinely funded adult founder and its real bud,
never a fabricated adult wildlife record. With finite retained food and paid
sibling movement, that juvenile reaches adulthood and produces another bud at
**tick 27,901**, after **939 reported funded growth steps across the family**.
Every tick checks complete five-channel owner closure and each growth report's
exact reserve spending. Newborns have no growth report in their birth tick;
midpoint canonical restoration preserves continuation. This is accelerated,
scripted biological evidence, not autonomous sustainability or elapsed-day proof.
Raw and cohort execution share the unchanged wildlife controller and produce
identical complete biological bytes for 1,024 growing ticks, including conversion
and repeated restoration. Recorded raw inputs and a full 64-tick population
interval decode/replay identically. A second executable process restores a growing
cohort and produces byte-identical 64-tick continuation. Population upgrade
preserves policy, tags and sequence. Rules-8 juveniles retain their original
growth. Valid rules-8 raw/population manifests accept their original payloads
and reject rules-9 payloads; an unadmitted rules-9 manifest bundle is unsupported.
Geometry fault fixtures only relocate genuine funded juveniles, preserving all
stocks/identity/history. A region boundary and a frozen lingering neighbor each
refuse an otherwise funded extent increase without growth spending or movement.
A separate controlled-subadult offline fixture returns already-owned body matter
to reserve, leaves paid heat intact and grants no resource. Current founders are
born adult, so this artificial fixture tests the guard only, not the family proof.
Its active control demonstrably grows; freeze/linger/suspend preserve body,
growth, reserves and mutation progress. Suspended resume grows only on the next
active tick. Pure budget tests cover rate/zero caps, damage retention, partial
funding, each absent nutrient/energy channel and exact composed-body closure;
saved-offer tests preserve the original choice at changed growth and reject a
changed recorded context.
The qualification exposed an existing ecological constraint: the smallest
juveniles can receive zero pool phosphorus per tick when quarter-yield floors
to zero. Contact with finite nutrient motes, placement and sibling space matter;
no feeding/balance rule was changed to conceal it. The family deliberately moves
a sibling through paid intentions. Automated multi-generation wildlife,
larger-cohort contention/tail cost and comparative ecological regimes remain open.
This gate changes no PostgreSQL adapter, migration or runtime admission, so the
full private database/TLS/storage/backup/DR matrix was not rerun. Durable rules-9
upgrade/publication/recovery, migration, registered census and client presentation
require separate admission/qualification. TSan was not rerun: no shared-memory
concurrency or scheduler ownership changed. Joint environmental budgets, actual
days, deployment-host evidence and product readiness remain open. No deployment,
World0 operation or terminal extinction decision occurred.
## 2026-10-18 - M12.7c registered census pending handoff evidence
[ADR-0156](../adr/0156-census-pending-migration-evidence.md)/schema 52 captures