MigrationRunner applies pending migrations through the migration ledger
exactly once; when a migration fails it throws a MigrationFailedError
whose diagnostic is a structured object identifying the failing migration
(version), the failure phase (apply/record), the underlying cause, and the
applied/pending ledger state, serializable via toJSON. Re-exported from
the driver boundary so no other package needs the pg driver to run
migrations. Advisory lock (T04) and ready gate (T06) remain out of scope.
release() previously returned the connection to the pool in a finally even
when the pg_advisory_unlock statement failed, so a pooled connection could be
reused while its session still held the migration advisory lock - the next
borrower would block every other runner (reviewer finding F4). On unlock
failure the connection is now destroyed (client.release(error) removes the
client from the pool, ending the session and its lock); the plain
client.release() is kept only on the success path. Locked in by a static
assertion, a mutation probe, and a deterministic stub-pool behavioral probe
of the committed release() control flow.
Security-review finding F2: two concurrent acquire() calls on the same
MigrationLock instance could each check out a connection; the second
pg_advisory_lock would overwrite this.client, leaking the first locked
connection until session end.
acquire() now memoizes the in-flight acquire in acquireInFlight and
returns it on re-entry, so exactly one connection is checked out and no
locked connection leaks. The memo is cleared once the acquire settles.
tryAcquire()/release() paths unchanged.
Locked in by:
- static criterion test: acquireInFlight field, re-entry guard returns
the in-flight acquire, memo cleared on settle
- mutation probe: removing the re-entry guard fails the criterion
- real-stack probe: two concurrent acquire() calls on one instance leave
pool.totalCount at 1 (exactly one connection), the lock granted once,
nothing left after release; probe fails (hangs) on the pre-fix code
- CI job comment updated to reflect the re-entrancy criterion
Static assertions on the committed ledger source (idempotent table DDL,
parameterized idempotent record, has/applied queries, driver-boundary
re-export, CI enforcement) with mutation probes proving non-vacuousness;
docker-gated real-stack probe migrates an empty database (isolated compose
project + host port) and confirms the ledger exists, the applied migrations
are recorded, and a re-run records nothing twice. New additive
database-postgres-ledger CI job gates the criterion on every PR.
MigrationLedger over the package-owned pg Pool: ensure() creates the
schema_migrations table (version text PRIMARY KEY, applied_at timestamptz
NOT NULL DEFAULT now()) with idempotent DDL; record() inserts an applied
migration with a parameterized, idempotent statement (ON CONFLICT DO
NOTHING — a rerun never double-applies); has()/applied() read the ledger
back in apply order. Re-exported from the driver boundary (src/index.ts)
so no other package needs the pg driver to touch migration state.
Review finding on PR #391: packages/database-postgres pinned pg@8.23.0 and
kysely@0.29.5, but the architecture doc's golden tuple (Technology-Stack
section 5.2 / section 7) pins pg@8.22.0 and Kysely@0.29.4. Reproducibility
requires the exact documented versions.
- packages/database-postgres/package.json: pg 8.23.0 -> 8.22.0,
kysely 0.29.5 -> 0.29.4, @types/pg 8.23.1 -> 8.21.0 (no 8.22.x of
@types/pg is published; 8.21.0 is the closest matching release, types
for the immediately preceding pg minor)
- pnpm-lock.yaml: regenerated with pnpm 11.23.0 (Node 24); the resolved
pg dependency tree is unchanged apart from the driver version itself
- tests/database-postgres-imports.test.mjs: exact-pin assertions updated
to the corrected versions, with a comment noting the @types/pg choice
- database-postgres-imports.test.mjs: static scan of every workspace package
source proves pg/kysely import specifiers resolve only to
packages/database-postgres; owner manifest pins the driver and no other
package declares it; mutation probes prove the scan catches a driver import
injected into apps/server/src/index.ts; comment-stripping and specifier
matcher unit probes; CI-enforcement assertion
- workspace-layout / workspace-config / strict-tsconfig / typescript-pin:
package-set fixtures updated to include packages/database-postgres
- docs/development/non-container.md: workspace package table and build
expectations updated for the new package
- packages/database-postgres (@personal-blog/database-postgres): the single
workspace package allowed to import the PostgreSQL driver — declares pg and
kysely as exact dependencies (@types/pg for types) and its src/index.ts
imports and re-exports the driver pieces (Pool, Kysely, PostgresDialect) so
the isolation is real, not a placeholder
- pnpm-lock.yaml: importer for packages/database-postgres plus the resolved
pg/kysely dependency tree (frozen-lockfile install keeps working)
- .gitea/workflows/ci.yml: new database-postgres-imports job runs
tests/database-postgres-imports.test.mjs on every PR so the isolation
criterion gates merges
- compose-config.test.mjs: assertDbService requires the pinned
postgres:18.6-bookworm image; new docker-gated real-stack probe starts the
stack and asserts SHOW server_version exposes 18.6; mutation probe proves
reverting to a floating major tag fails the criterion; parser probe updated
- build-targets.test.mjs: db-service mutation fixture updated to the pinned
image tag
- compose.yaml: db.image pinned to postgres:18.6-bookworm (exact 18.6 minor,
same bookworm flavor, no Alpine drift) so the database container is
reproducible and exposes the expected PostgreSQL version; document the
rollback (revert image to postgres:18-bookworm)
- .gitea/workflows/ci.yml: new compose-config job runs
tests/compose-config.test.mjs on every PR so the pinned-version criterion
gates merges (docker-gated real-stack probes skip cleanly without a daemon)
The Docker-gated probe only planted a root-level .env.t08-* marker, which no
Dockerfile COPY instruction ever copies — so it could not observe a nested
build-context leak in the image layers. Plant additional marker files at
nested paths the Dockerfile's COPY apps/server apps/server would sweep into
the build-stage image (apps/server/.env.t08-*, apps/server/secrets/t08-*.pem)
so the end-to-end scan actually verifies the 'any depth' exclusion
guarantee, not just the root form.
Resolve the security review of #389 (findings 1-4):
- .dockerignore: every env/credential pattern is now **/-prefixed
(**/.env, **/.env.*, **/node_modules, **/.npmrc, ..., **/secrets,
**/*.pem, **/*.key, ...) and the redundant 'secrets/' line is dropped.
Docker's matcher (moby/patternmatcher) anchors slash-less patterns to
the context root, so the bare forms excluded nothing under apps/server/;
**/ matches the root AND any nested depth. (finding 1, 3)
- tests/secrets-not-embedded.test.mjs: the dockerignore matcher is now a
faithful port of moby/patternmatcher (filepath.Clean + anchored full-path
match + parent-directory propagation), not gitignore basename semantics;
asserts nested example paths (apps/server/.npmrc, config/server.key,
apps/server/secrets/...) are excluded; requires no redundant equivalent
patterns verbatim; adds mutation probes for bare-pattern and
duplicate-pattern regressions. (finding 2, 3)
- apps/server/Dockerfile + compose.yaml: guarantee restated precisely
(credential files excluded at the context root AND at any depth).
- .gitea/workflows/ci.yml: new job runs
'node --test tests/secrets-not-embedded.test.mjs' on every PR; the
docker-gated layer-scan probe runs where a daemon exists, skips cleanly
otherwise. (finding 4)
- tests/compose-config.test.mjs: .dockerignore presence list updated to the
**/-prefixed forms (node_modules, .env).
Tested: secrets suite 16 tests -> 15 pass / 1 docker-gated skip / 0 fail;
full suite 101 pass / 12 fail / 8 skip, failures identical to clean main
(env-dependent pnpm/Node-24 suites); matcher port verified against the
moby/patternmatcher evidence table.
tests/secrets-not-embedded.test.mjs: static assertions that the Dockerfile
embeds no secrets (no secret-bearing ARG/ENV, no secret-path or blanket COPY)
and .dockerignore excludes env/credential files; non-vacuous mutation probes
for every assertion; Docker-gated probe that builds the image with a marker
env file in the context and scans every layer + image config for secret
values.
- .dockerignore: exclude env + credential files (.npmrc, .netrc, .aws, .ssh,
secrets/, *.pem, *.key, *.p12, *.pfx, *.jks, id_rsa, id_ed25519, ...) from
the build context so a local secret file cannot be embedded in the image
- Dockerfile: document the T08 guarantee (no secret ARG/ENV, fixed non-secret
COPY paths, runtime credentials via Compose environment)
- compose.yaml: T08 in scope; runtime credentials stay in service environment,
never in the image
tests/non-root-user.test.mjs locks in both acceptance criteria: a static
assertion that the Dockerfile runtime stage declares a non-root USER (not
root/uid 0, 'USER node' exactly), non-vacuous mutation probes, and a
Docker-gated real-stack probe that starts the stack and asserts 'id -u' and
'id -un' inside the running app container report a non-root user, with the
health endpoint still answering as a regression guard.
The runtime stage of apps/server/Dockerfile now drops root privileges with
'USER node' — the non-root user (uid/gid 1000) the official Node image ships
with — so the app container does not run with root privileges. The server
binds port 3000 (>= 1024) and only reads the root-owned files copied above,
so no extra user creation or ownership changes are required. compose.yaml
header updated: T05 is in scope; T06 (read-only rootfs) and T07 (multi-arch)
remain out of scope.
compose-config: assert the db service mounts the named db-data volume at
the PostgreSQL data directory and the top-level volumes map declares it;
non-vacuous mutation probes (missing mount, missing volume declaration,
wrong mount target all fail); Docker-gated real-stack probe writes a
fixture row and asserts it survives `docker compose restart` (restart)
and `docker compose down` + `up -d` (recreate), cleaned up with
`docker compose down -v`.
Mount the named `db-data` volume at PostgreSQL's data directory
(/var/lib/postgresql/data) on the db service and declare it in the
top-level volumes map, so the database survives `docker compose
restart` and `docker compose down` + `up -d` (recreate). Reset with
`docker compose down -v` per the issue rollback note. Header comments
in compose.yaml and the server Dockerfile updated: T04 is no longer out
of scope; T05/T06 remain.
Switch the app image from node:24-alpine to node:24.19.0-bookworm-slim in
both build and runtime stages (Technology-Stack 5.4: glibc Debian base
required because argon2 is a native dependency; musl/Alpine causes
native-module build surprises), and the db image from postgres:16-alpine
to postgres:18-bookworm (Technology-Stack 5.2/5.4/6.2 + golden tuple 7
pin PostgreSQL 18.6; 18-bookworm is the 18.x line on Debian bookworm).
Update tests/compose-config.test.mjs so the committed assertions lock in
the corrected image bases (db image, Dockerfile build/runtime stages,
parser probe).
implementer
kpcto