feat(lifecycle): qualify portable maturing wildlife transfers
Foundation / checks (ecs) (push) Successful in 2m26s
Foundation / checks (deploy) (push) Successful in 4m12s
Foundation / checks (network) (push) Successful in 20m57s
Foundation / checks (client) (push) Successful in 22m9s
Foundation / checks (tsan) (push) Successful in 28m49s
Foundation / checks (postgres) (push) Successful in 1h5m2s
Foundation / checks (checks) (push) Successful in 1h17m42s
Foundation / postgres-tools (push) Successful in 1m10s
Foundation / server-image (push) Successful in 3m51s
Foundation / image (push) Successful in 4m54s

This commit is contained in:
dingo
2026-10-19 09:09:15 -04:00
parent 274f62cb89
commit 5e168d67c3
22 changed files with 412 additions and 41 deletions
+64
View File
@@ -1,5 +1,69 @@
# Validation report
## 2026-10-19 - Portable transfer of maturing wildlife
[ADR-0160](../adr/0160-portable-maturing-wildlife-transfer.md) admits exact
portable transfer between equal rules-9/owner-7 cuts, raw or population envelope
2. Migration trace 2 has distinct magic/version and the existing bounded layout.
Trace 1 retains rules 8. Durable metadata/retry adapters explicitly refuse trace
2; SQL schema 53 and bundle-10 migration admission remain unchanged.
Executed locally:
- MSVC default Release full CTest **81/81 passed, 22.13 s**, with NDEBUG.
- GCC 14 Release migration contract/process restart **2/2 passed, 3.29 s**.
- Clang 19 ASan/UBSan migration contract/process restart **2/2 passed, 87.00 s**,
without sanitizer reports.
- The complete affected private PostgreSQL population/history/migration fault
matrix passed **exit 0, 22 s** (19 s population). The same ASan/UBSan matrix
passed **exit 0, 63 s** (60 s population), without reports.
- clang-tidy 19 passed on cellular lives, transfer, trace and durable migration
sources and both affected tests. Repository clang-format 19 and
`git diff --check` passed.
- Seven retained rules-8 restart artifacts from the preceding native run match
byte-for-byte: raw/cohort trace 1, both complete output pairs and the continued
population snapshot. No old golden or numbered SQL migration changed.
Rules-9 repeats the raw transfer/census/conservation matrix, 100 return trips,
imported death/decay, body-overlap refusal, whole-group capacity refusal and the
complete raw/cohort codec/replay matrix. Every transferred life field remains
equal at the no-tick cut. Cross-rule replay, mixed magic/version, unknown version,
malformed bounds, altered members/material and incompatible trusted requests
fail. Fresh processes restore/replay both rule versions in both representations,
then reproduce the exact recorded 64-tick destination continuation.
A separate real-bud fixture verifies funded growth before transfer, exact
raw/cohort candidate biology, and further funded growth after transfer. Finite
destination food is declared in its opening fixture; no reserve/body credit or
reusable boundary-flow stock is granted. Complete combined per-channel stock
closes and destination growth retains the unchanged transfer ledger. Mixed rules
in either raw direction and in cohorts fail; vulnerable controlled life remains
ineligible. Rules-9 cannot repeat the old format-upgrade API.
Private raw/cohort durable tests additionally submit canonical trace 2 to retry
freeze/stage, direct staging and SQL source preparation. C++ refuses it before
publication; SQL independently returns 22023 before duplicate lookup. Existing
rules-8 freeze/stage/commit and rules-9 census regressions continue to pass.
The initial new codec-tampering fixture triggered GCC's optimized bounds warning;
using checked vector access resolved it before the complete qualification above.
Per maintainer instruction, the DEV route's `REGION_SESSION_WINDOW` is refreshed
in this same scoped commit, with a unique UTC activation and a four-hour absolute
expiry. Rendered deployment policy passes **75 objects / 74 files, 138 rendered
objects / 17 overlays, 0 failures**. The working agreement now requires a refresh
before every push. Deployment-policy self-tests pass **46 cases, 0 failures**,
including rejection of overlong windows; the final fresh window also passes the
rendered policy check. This is desired Git configuration, not a live connectivity or
Argo application observation.
The full backup/DR, TLS relay and SQL privilege/checksum suite was not repeated;
no SQL, durable publication layout or transport algorithm changed. Temporary
private Unix-socket databases were used; no live database, deployment or World0
operation occurred. TSan was not rerun because shared-memory scheduling did not
change. Durable rules-9 migration, linked environmental resources/geometry,
autonomous generations, target-host costs, actual days and product gates remain
open; this portable prerequisite does not complete M12.
## 2026-10-19 - Census of funded maturing owners
[ADR-0159](../adr/0159-maturing-cellular-census.md) admits rules 9/owner 7,