M12.4: keep cellular wildlife ticking and retain scheduling debt
Foundation / checks (deploy) (push) Successful in 5s
Foundation / checks (ecs) (push) Successful in 2m35s
Foundation / checks (tsan) (push) Failing after 17m51s
Foundation / checks (network) (push) Successful in 21m13s
Foundation / checks (client) (push) Successful in 22m19s
Foundation / checks (postgres) (push) Successful in 48m25s
Foundation / checks (checks) (push) Failing after 50m8s
Foundation / postgres-tools (push) Skipped
Foundation / server-image (push) Skipped
Foundation / image (push) Skipped

This commit is contained in:
dingo
2026-10-17 12:24:19 -04:00
parent 5decaa0326
commit aff35863ce
12 changed files with 434 additions and 20 deletions
+39
View File
@@ -1,5 +1,44 @@
# Validation report
## 2026-10-17 - M12.4a unattended demand and confirmed-head pacing
ADR-0141 changes the existing cellular worker's demand/pacing and metrics, with
portable helpers. Rules 5+ wildlife now supplies tick demand after players
suspend; active offline cells can enter/finish freeze. Historical profiles lacking
freeze are not opted into unattended work. Biology, profile-4 persistence and
database schema are unchanged. No deploy, World0 operation or new dependency.
Executed:
- MSVC 19.44 Release configure/build and **75/75 CTests**, including corrected
legacy-founder fixture, per-boundary offline demand, long cohort conversion
case and separate-process population continuation.
- GCC Release demand/pacing and population-transition tests. Actual QUIC-enabled
PostgreSQL `evo-server` and `evo-postgres-tests` build in `build/m97-postgres`.
- Full `tools/ci/postgres-check.sh` private disposable-cluster suite **passed,
exit 0, 494 s**: migrations/rollback, permissions, fencing, cellular publication,
process/session tests, reservations, backup/restore, DR, upgrade, migration
history and checksum drift. This run includes unattended demand and confirmed
head pacing; the final legacy-guard/freshness refinements have direct final
tests and the final Clang server build/static analysis below.
- Final Clang 19 QUIC/PostgreSQL worker build in `build/tidy-postgres`, followed
by explicit `clang-tidy-19 -p build/tidy-postgres apps/evo-server/main.cpp
--warnings-as-errors=*` (the cached build itself has automatic tidy disabled).
- Analysis-preset clang-tidy 19 for the portable implementation/tests, repository
clang-format 19 check, and scheduler CTest passed. ASan/UBSan scheduler CTest
passed (0.24 s), no reports; population transition regression also passed
ASan/UBSan (8.37 s) during this segment.
- New documentation links and `git diff --check` passed. Explicit NDEBUG checks
exercise fixed-clock debt/retry/confirmed-head accounting, freshness, pause,
epoch change, regressing head/clock and u64 terminal boundaries.
The actual-private-DB suite is a regression check, not a live no-player ecology
soak or deployment-host capacity measurement. Specific new unattended behavior
has direct demand/clock and real funded-descendant/suspension tests, not a new
multi-region biological process test. No durable population-profile publication,
cross-region biological escrow, global extinction or multi-day soak is claimed.
Remaining integration/evidence: [M12_REMAINING_GATES](M12_REMAINING_GATES.md).
## 2026-10-17 - M12.3a exact population checkpoint closure
ADR-0140 adds population envelope 1, detached conversion candidates and opt-in