docs(v1.1.7): handback report

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
MechaCat02
2026-06-04 22:46:33 +02:00
parent a7d3dad129
commit 3cfb795206

View File

@@ -1,220 +1,330 @@
# HANDBACK — v1.1.6 Realtime Channels & Client Library
# v1.1.7 — Configuration & Email — HANDBACK
Branch: `feat/v1.1.6-realtime-client` (from `main`). Not pushed, no PR.
## 1. Scope coverage (§1§13)
| § | Item | Status |
|---|------|--------|
| 1 | `topics` table (`0021_topics.sql`) | ✅ Done |
| 2 | Topic admin endpoints (POST/GET/PATCH/DELETE), `AppTopicManage` | ✅ Done |
| 3 | SSE endpoint `GET /realtime/topics/{topic}` | ✅ Done |
| 4 | In-process `RealtimeBroadcaster` + GC + publish wiring | ✅ Done |
| 5 | HMAC subscriber tokens + `app_secrets` (`0022`) + `pubsub::subscriber_token` | ✅ Done |
| 6 | Dashboard Topics tab | ✅ Done |
| 7 | `@picloud/client` TypeScript package | ✅ Done |
| 8 | Topic-aware publish → realtime wiring | ✅ Done |
| 9 | Dispatcher e2e tests (six) | ✅ Done (location deviation — see §7) |
| 10 | Empty-blob decision | ✅ Done — **relaxed** (accept empty) |
| 11 | Orphan `*.tmp.*` sweeper | ✅ Done |
| 12 | Version bumps (1.1.6 / SDK 1.7 / dash 0.12.0 / client 1.0.0) | ✅ Done |
| 13 | Tests (all named-critical cases) | ✅ Done |
## 2. Realtime implementation notes
### Topic resolution / SSE handshake sequence
1. Extract `Host``app_domains.resolve_app(host)` (existing two-phase
dispatch). No app → **404**.
2. Token from `Authorization: Bearer <t>` **or** `?token=<t>` (EventSource
can't set headers).
3. `RealtimeAuthority::authorize_subscribe(app_id, topic, token)`:
- topic missing OR `external_subscribable = false``NotFound`**404**
(both collapse to 404 so the endpoint can't probe internal topics);
- `auth_mode='public'` → allow;
- `auth_mode='token'` → verify HMAC (present, signed by this app's key,
unexpired, scoped to this topic) → allow, else `Unauthorized`**401**
(generic; never says which check failed).
4. `broadcaster.subscribe(app_id, topic)``broadcast::Receiver`; stream
`data: {topic,message,published_at}\n\n` with `:heartbeat` keepalive
every `PICLOUD_REALTIME_HEARTBEAT_SEC` (default 30). Headers:
`text/event-stream`, `Cache-Control: no-cache` (set by axum Sse),
`X-Accel-Buffering: no` (added). Client disconnect drops the receiver →
automatic cleanup; the periodic GC reaps empty channels.
The SSE handler lives in **orchestrator-core** (`realtime_api.rs`) and
depends only on the three picloud-shared traits — all DB + signing-key
access stays in the manager-core `RealtimeAuthority` impl, so the
data-plane crate never touches the key. Cluster mode (v1.3+) swaps the
broadcaster + authority impls behind the same traits.
### HMAC signing-key storage — chose **table** (`app_secrets`)
Per the asked decision. 32 random bytes per app, lazily created on the
first `pubsub::subscriber_token` call (`ON CONFLICT DO NOTHING` for
concurrency). No global `PICLOUD_INSTANCE_SECRET`; the table is the
natural home for v1.1.7's encrypted per-app secrets.
**Tension flagged:** §5 aspired to "validate the signature without a DB
lookup". The table approach needs a key read on the token-subscribe path.
Mitigated by an in-memory key cache in `RealtimeAuthorityImpl` (keys never
rotate in v1.1.6, so the cache needs no invalidation); the subscribe path
already reads the `topics` row, so this adds no new round-trip *category*.
### In-process broadcaster
`Mutex<HashMap<(AppId, String), broadcast::Sender<RealtimeEvent>>>`.
Capacity per channel `PICLOUD_REALTIME_BROADCAST_CAPACITY` (default 64);
slow consumers lose oldest events (`broadcast` lag semantics — best-effort,
no replay). `subscribe` creates channels lazily; `publish` is a silent
no-op when no channel exists; `drop_topic` removes the sender (existing
receivers observe a closed channel and disconnect). A `spawn_realtime_gc`
task (~60s) drops senders with `receiver_count() == 0`.
### Publish wiring (order + failure modes) — §8 ordering chosen
`PubsubServiceImpl::publish_durable`: validate/authz → **transactional
outbox fan-out + commit** (existing) → **then** best-effort
`RealtimeBroadcaster::publish`. The broadcast runs on a child
`tokio::spawn` whose `JoinHandle` is awaited, so a panicking broadcaster
becomes a `warn` log, never a failed publish (the durable deliveries
already committed). One `published_at` instant is shared by both paths.
**Brief-internal contradiction flagged** — see §9.
## 3. Client lib implementation notes
- **Build:** **tsup** (dual ESM+CJS + `.d.ts`), tests **vitest**, lint =
`tsc --noEmit` (strict; `noUncheckedIndexedAccess`, no `any` in exports).
- **Layout:** `src/{index,client,endpoint,subscribe,auth,types}.ts` +
`src/react/index.ts` + `src/svelte/index.ts`; subpath exports
`@picloud/client/react` and `@picloud/client/svelte` via package `exports`.
- **Reconnect:** SSE is implemented over **streaming `fetch`** (not native
`EventSource`) so the lib can (a) detect a **401** on (re)connect and call
`onTokenExpired` to refresh, (b) send **`Last-Event-ID`** on resume
(server ignores it in v1.1.6 — client ships ready), and (c) apply its own
**exponential backoff** (base 1s → ×2 → cap 30s; reset on successful
open). Token rides in `?token=` (EventSource-parity). React Native caveat
documented in the README (supply a streaming-`fetch` polyfill).
- **zod/valibot adapter:** the `Validator<T> = { parse(input): T }` shape.
A Zod schema satisfies it directly; Valibot wraps in one line. Used by
`endpoint(...).get({ validate })` and `subscribe(..., { validate })`. No
hard dep.
## 4. v1.1.5 follow-ups
- **Dispatcher e2e (six):** `dispatcher_delivers_{kv,docs,cron,files,pubsub,
dead_letter}_to_handler` in `crates/picloud/tests/dispatcher_e2e.rs`.
Verified green against a real Postgres (all 6).
- **Empty-blob — RELAXED** (per the asked decision). Dropped the
`data.is_empty()` rejection in `NewFile::validate` **and** `FileUpdate::
validate` (the latter for consistency — flagged in §7). Flipped the
pinning test in `files_service.rs` and added `empty_file_round_trips`.
- **Orphan sweep:** `spawn_files_orphan_sweep` (`files_sweep.rs`), every 6h
(`PICLOUD_FILES_ORPHAN_SWEEP_INTERVAL_SEC`), unlinks `*.tmp.*` older than
1h (`PICLOUD_FILES_ORPHAN_TMP_TTL_SEC`); logs dirs-walked / files-deleted /
bytes-reclaimed. No DB cross-check (v1.3+). Tested: deletes old tmp, keeps
young tmp, keeps non-tmp, missing-root no panic.
## 5. Schema decisions beyond the brief
None — `0021_topics.sql` and `0022_app_secrets.sql` match the brief's DDL
verbatim. Schema-snapshot golden re-blessed on a fresh DB (only the two new
tables added; PK/FK `ON DELETE CASCADE` + the `auth_mode` CHECK present).
## 6. How to verify locally
See §8 below.
## 7. Decisions beyond the brief — every prompt-default deviation
1. **Dispatcher e2e location:** the brief says
`crates/manager-core/tests/dispatcher_e2e.rs`; I put it in
**`crates/picloud/tests/`**. `build_app` (the full dispatcher + scheduler
+ executor wiring) lives in the `picloud` crate; a manager-core test would
need a manager→picloud dev-dependency cycle or a hand-rolled re-wire of
build_app. The picloud harness (`server_with_app` pattern) is proven and
the tests run green against a real DB. Gating uses the **schema_snapshot
pattern** (env check + early return), NOT `#[ignore]`, so CI's plain
`cargo test` with `DATABASE_URL` runs them while local stays green — as
the brief requested.
2. **Empty-blob relaxation extended to `FileUpdate::validate`** — the brief
names only `NewFile::validate`. Relaxing create-empty but rejecting
update-to-empty would be an inconsistent API, so I relaxed both.
3. **Publish order: §8 (broadcast AFTER outbox commit)**, not §4 (broadcast
FIRST). The two sections contradict; §8 is the dedicated, numbered,
rationale-backed section. See §9.
4. **Topic-name validation** — `topics_api` rejects empty names and names
containing `*` (external pattern subscription is v2). The brief didn't
specify name validation; this is a small guard.
5. **Client lib lint = `tsc --noEmit`** (not eslint) to keep devDeps lean;
strict typecheck is the gate. "No `any` in exports" is enforced by review
+ strict TS, not an eslint rule.
6. **Cron e2e poll budget = 45s** — the cron scheduler skips its first tick
then ticks every 30s (default). The test polls 45s so it passes at the
default; set `PICLOUD_CRON_TICK_INTERVAL_MS=1000` to make it ~2s in CI.
## 8. Attestation (gate runs on this HEAD)
**Branch:** `feat/v1.1.7-secrets-email` (9 commits off `main`, not pushed)
**Status:** ready for review. NOT merged, NOT pushed, no PR opened.
```
cargo fmt --all -- --check → clean
cargo clippy --all-targets --all-features -- -D warnings → clean (exit 0)
cargo test --workspace → 482 passed, 0 failed
(DB-gated dispatcher_e2e
auto-skip as no-op when
DATABASE_URL unset)
# DB-gated (real Postgres @ 127.0.0.1:15432, picloud/picloud):
DATABASE_URL=… PICLOUD_CRON_TICK_INTERVAL_MS=1000 \
cargo test -p picloud --test dispatcher_e2e → 6 passed
DATABASE_URL=… cargo test -p picloud-manager-core --test schema_snapshot → 1 passed
(cd dashboard && npm run check) → 0 errors, 0 warnings (371 files)
(cd clients/typescript && npm run lint) → clean (tsc --noEmit)
(cd clients/typescript && npm run test) → 15 passed (5 files)
(cd clients/typescript && npm run build) → tsup ESM+CJS+.d.ts OK
a7d3dad chore(v1.1.7): re-bless schema snapshot for secrets + email migrations
2ea47eb chore(v1.1.7): fix clippy --all-targets warnings
b355851 chore(v1.1.7): version bumps + CHANGELOG
fffcdf6 feat(v1.1.7-realtime-migration): encrypt signing keys at rest
02335a8 fix(v1.1.7-dead-letter): wire dispatcher → list_matching_dead_letter
1f78937 feat(v1.1.7-email-inbound): webhook receiver + email:receive trigger
8f2d2bc feat(v1.1.7-email-outbound): SMTP send/send_html
2d11090 feat(v1.1.7-secrets): secrets SDK + table + admin API + dashboard
dc2e4fa feat(v1.1.7-crypto): master-key infra + encryption helpers
```
Migrations verified applying cleanly on a fresh DB **and** on top of the
existing v1.1.5 dev DB (0020 → 0021 → 0022). Schema-snapshot golden diff is
exactly the two new tables.
---
## 1. Scope coverage
| Item | Status |
|---|---|
| Encryption infrastructure (master key + AES-256-GCM envelope) | **Done** |
| `secrets::*` SDK + `0023_secrets.sql` + admin API + dashboard tab | **Done** |
| Outbound email `email::send` / `email::send_html` (lettre SMTP) | **Done** |
| Inbound email webhook receiver + `email:receive` trigger + `0024` | **Done** (full scope, per user decision) |
| Dispatcher routing for email | **Done** |
| dead_letter handler wiring fix | **Done** |
| Realtime signing-key encryption (two-phase) + `0025` | **Done** |
| Dashboard (Secrets tab, email trigger form, `npm run check`) | **Done** |
| Version bumps (1.1.7 / SDK 1.8 / dashboard 0.13.0) + CHANGELOG | **Done** |
| Tests (match v1.1.5/v1.1.6 density) | **Done** |
Nothing deferred from scope-in. Inbound email (the deferrable-if-scope-
blew-up piece) was implemented in full.
---
## 2. Encryption infrastructure notes
- **Module:** `crates/shared/src/crypto.rs` (`picloud_shared::crypto`).
- **Master-key sourcing** (`MasterKey::from_env``resolve`):
- `PICLOUD_SECRET_KEY` = base64 of exactly 32 bytes. Missing →
`MasterKeyError::Missing` (fatal); non-base64 → `Malformed`; wrong
length → `WrongLength`. **Sourced in `main.rs::run_server` before any
DB work** — `build_app` takes the `MasterKey` as a parameter (so
tests pass a fixed key and don't mutate process env).
- Dev fallback: deterministic key (`SHA-256("picloud-dev-master-key-v1.1.7")`)
used ONLY when `PICLOUD_SECRET_KEY` is unset **AND**
`PICLOUD_DEV_MODE=true`, with a prominent `warn!`. No quiet
unencrypted mode.
- **aes-gcm version:** `0.10` (features `aes`, `alloc`). `Aes256Gcm`.
- **Nonce generation:** 12 bytes from `rand::thread_rng().fill_bytes`
(OS-CSPRNG-seeded), per-encryption.
- **Storage layout:** ciphertext **with the 16-byte GCM auth tag
appended** (RustCrypto `Aead`-trait layout — `encrypt` returns
`ciphertext || tag`, `decrypt` consumes the same). The 12-byte nonce is
stored in a separate column. `MasterKey`'s `Debug` is redacted.
- **Plaintext cap (secrets):** 64 KB default, enforced in
`secrets_service::seal` (the SDK boundary) → `SecretsError::TooLarge`
with limit + actual size. Override: `PICLOUD_SECRET_MAX_VALUE_BYTES`.
- **Key rotation:** out of scope. Documented in CHANGELOG + the module
docs that changing `PICLOUD_SECRET_KEY` orphans all ciphertext.
---
## 3. Secrets notes
- `SecretsService` (trait, `picloud-shared`) → `SecretsServiceImpl` +
`PostgresSecretsRepo` (`manager-core`) → Rhai bridge
(`executor-core/src/sdk/secrets.rs`). Collection-less; `app_id` from
`cx.app_id`.
- **JSON round-trip:** `set` serializes the value to JSON bytes, caps,
encrypts; `get` decrypts + deserializes — a String returns a String
(not a JSON-quoted `"\"…\""`). Verified by unit + bridge tests.
- **No ServiceEvent emission** (secret writes don't fire triggers).
- Admin API: `GET/POST/DELETE /api/v1/admin/apps/{id}/secrets`; list
returns names + `updated_at` only.
- Authz: `Capability::AppSecretsRead/Write``script:read`/`script:write`.
No new Scope variants (seven-scope commitment held).
---
## 4. Email implementation notes
- **SMTP transport:** `lettre 0.11` (`smtp-transport`,
`tokio1-rustls-tls`, `builder`, `hostname`). **Connection model:** one
connection per call (lettre default); pooling deferred to v1.2. The
transport sits behind an internal `EmailTransport` trait so the service
is unit-tested with a recording fake (no live SMTP).
- **Disabled mode:** if HOST/USER/PASSWORD aren't all set,
`EmailServiceImpl::from_env` builds no transport and every `send`
returns `NotConfigured` (warned at startup). A malformed relay
descriptor is also logged and yields disabled mode (email is
non-critical; never blocks startup).
- **Address validation:** hand-rolled RFC 5322-ish pre-check (single `@`,
non-empty local part, domain contains a dot, ≤320 bytes) followed by a
`lettre::Mailbox` parse (the authoritative validator). No deliverability
check.
- **Size cap:** 25 MB on `message.formatted()`,
`PICLOUD_EMAIL_MAX_MESSAGE_BYTES`.
- `email::send` forces text-only (ignores any `html`); `email::send_html`
requires `html` and builds `MultiPart::alternative_plain_html`.
`reply_to` defaults to `from`. `to`/`cc`/`bcc` accept a String or an
Array of Strings.
- **Inbound normalization:** only the generic provider-agnostic JSON
shape `{from,to[],cc[],subject,text,html,message_id}` is accepted in
v1.1.7 — `from` required, rest default. Provider-specific unmarshallers
→ v1.2. The expected shape is documented on the dashboard email-trigger
form.
---
## 5. Dead-letter handler fix notes
- **Call site:** `dispatcher::handle_failure`, the retry-exhaustion
branch. After `DeadLetterRepo::insert` (which returns the new
`DeadLetterId`), a new helper `fan_out_dead_letter` runs.
- **What it does:** calls `TriggerRepo::list_matching_dead_letter(app_id,
source, row.trigger_id, Some(resolved.script_id))` (the method that had
no production caller) and inserts one outbox row per match
(`source_kind = DeadLetter`, the DL trigger's id + handler script id,
`trigger_depth + 1`, `origin_principal = the DL trigger's registered
principal`).
- **Payload — built from the REAL `TriggerEvent::DeadLetter` variant**,
not the brief's §6 field list (see §7 deviations): `{ dead_letter_id,
original: Box::new(decoded row payload), attempts, last_error,
trigger_id, script_id, first_attempt_at, last_attempt_at }`. If the
outbox payload can't be decoded back into a `TriggerEvent` (so the
nested `original` can't be built), the fan-out is skipped — the
dead-letter row is still durably written.
- **Recursion-stop:** unchanged. The `is_dead_letter_handler`
short-circuit at the top of `handle_failure` returns before the
exhaustion branch, so a DL handler's own failure is never re-dead-
lettered. No new guard needed.
- **Tests verify the handler actually fires**
(`crates/picloud/tests/dispatcher_e2e.rs`, DB-gated):
`dispatcher_delivers_dead_letter_to_handler` now asserts BOTH row-create
AND handler-fire (inline doc updated);
`dispatcher_delivers_dead_letter_to_handler_actually_fires` asserts the
nested `original` KV event + `last_error`;
`dead_letter_source_filter_excludes_nonmatching` exercises the source
filter dimension; `dead_letter_handler_failure_does_not_recurse` proves
the recursion-stop (count stays at 1).
---
## 6. Realtime signing-key migration notes
- **Two-phase**, as recommended. `0025_encrypt_realtime_keys.sql` adds
NULL-able `realtime_signing_key_encrypted` + `realtime_signing_key_nonce`
and `DROP NOT NULL` on the plaintext column (so new keys can be stored
encrypted-only).
- **Repo:** `PostgresAppSecretsRepo` now holds the `MasterKey`. New keys
are written encrypted-only; the read path (`signing_key` /
`get_or_create_signing_key`) prefers the encrypted columns and falls
back to plaintext during the compat window (pure `decode_signing_key`
helper, unit-tested for all four precedence states).
- **Startup task:** `migrate_plaintext_keys()` runs once in `build_app`
(after the master key is loaded), encrypting any rows that still have
plaintext but no encrypted value. Plaintext is **left in place** for
rollback safety. Idempotent.
- **Plaintext column drop:** deferred to **v1.1.8** (documented in
CHANGELOG + the migration). Operators must upgrade through v1.1.7
(which performs the encryption) before v1.1.8.
- SSE keeps working: `RealtimeAuthorityImpl` is unchanged (it calls
`signing_key`). Verified by the pubsub e2e + unit tests; the dev DB
applied 0025 + the startup encryption cleanly during the test run.
---
## 7. Decisions beyond the brief / deviations flagged
1. **`inbound_secret` stored ENCRYPTED (user-approved deviation).** The
brief defaulted to a plaintext `inbound_secret` column on
`email_trigger_details`; the user chose to encrypt it via the master
key. Implemented: `0024` stores `inbound_secret_encrypted` +
`inbound_secret_nonce`; the admin endpoint seals the secret (as a JSON
string, via the secrets `seal` helper); the receiver `open`s it per
inbound POST to verify the HMAC. **Trade-off:** one AES-GCM decrypt per
inbound request on the hot path — negligible vs. the HMAC + DB
round-trip already there. The decrypted secret is never logged.
2. **Brief-internal contradiction flagged, not reinterpreted — §6
`TriggerEvent::DeadLetter` field names.** The brief's §6 sketches the
payload as `{source, op, original_event_id, original_payload,
attempt_count, last_error, …}`. The actual variant
(`crates/shared/src/trigger_event.rs`) is `{dead_letter_id, original:
Box<TriggerEvent>, attempts, last_error, trigger_id, script_id,
first_attempt_at, last_attempt_at}`. I built the payload from the
**real** variant (which the brief itself instructs to "verify
serializes correctly"). No type change needed.
3. **`build_app` signature gained a `MasterKey` parameter.** Rather than
sourcing the key inside `build_app` (which would force every e2e test
to set process env), `main.rs` sources it and passes it in. The 3
existing `build_app` test callers pass a fixed test key.
4. **Pre-existing clippy warnings fixed (see §10).** Four warnings predate
this work; I fixed them in a dedicated commit so the `-D warnings`
gate is green, and flag them as a latent finding.
5. **Email-trigger retry settings** use the standard async defaults
(3 attempts, exponential, 1000 ms) — the brief didn't specify; matches
the cron/kv default shape.
No other deviations from prompt-specified defaults.
---
## 8. How to verify locally — §8 attestation (sourced from cargo's literal output)
All gates run on the handed-back HEAD (`a7d3dad`):
```sh
cargo fmt --all -- --check # clean
cargo clippy --all-targets --all-features -- -D warnings # clean (exit 0)
cd dashboard && npm run check # 0 ERRORS 0 WARNINGS (371 files)
```
Full test run **with `DATABASE_URL` set** so the DB-gated suites
(schema_snapshot, dispatcher_e2e ×9, email_inbound ×8) execute:
```sh
DATABASE_URL='postgres://picloud:picloud@127.0.0.1:15432/picloud' \
cargo test --workspace -- --test-threads=2
```
**Pass count, summed from cargo's literal output (NOT hand-counted):**
```sh
DATABASE_URL=... cargo test --workspace -- --test-threads=2 2>&1 | \
awk '/test result: ok\./ { gsub(";", ""); sum += $4 } END { print sum }'
# => 617
```
**617 passed, 0 failed** across the workspace (34 `test result:` lines,
0 `FAILED`). Largest binaries: 290 (manager-core lib), 74, 43, 32, 30;
plus `dispatcher_e2e` (9) and `email_inbound` (8).
**Bounded-parallelism note (`--test-threads=2`):** the picloud e2e
binaries each call `build_app`, which opens its own Postgres pool. Under
full default parallelism against the *shared dev* Postgres, ~9 concurrent
`build_app`s exhaust connections and a couple of e2e tests flake on
timeout (observed: `dispatcher_delivers_pubsub_to_handler`,
`dead_letter_handler_failure_does_not_recurse`). They pass reliably at
`--test-threads=2` and in isolation. CI's dedicated fresh `postgres:15`
(not a shared dev DB) does not hit this. Environmental, not a correctness
issue — flagged so the reviewer runs the DB-gated suite with bounded
parallelism (or on CI).
**Migrations:** apply cleanly on the v1.1.6 dev DB (0023→0025 applied
during the test run) and the schema-snapshot guardrail passes after
re-bless. The `BLESS` diff was exactly the new tables/columns/constraints
(secrets, email_trigger_details, app_secrets encrypted columns +
NULL-able plaintext, widened kind/source CHECKs, migrations 00230025) —
no unrelated drift.
**Manual smoke:** the e2e suite covers secrets set/get/delete/list,
inbound signed POST → handler fires with `ctx.event.email`, dead-letter
handler fires, realtime-key encryption + SSE. Outbound email to a live
relay (mailtrap) was NOT exercised (no SMTP configured in this
environment) — asserted instead via recording-transport unit tests
(To/From/Subject/body, multipart parts, cc/bcc, reply_to).
---
## 9. Open questions for the reviewer
1. **§4 vs §8 ordering contradiction (load-bearing, flagged not
reinterpreted):** §4 says "realtime broadcast FIRST, then transactional
outbox fan-out"; §8 says broadcast AFTER the fan-out commits (numbered
steps 14). I implemented **§8** (dedicated section + failure-mode
rationale). If §4's ordering was intended, the broadcast should move
before `fan_out_publish` — but then a publish whose outbox write fails
would still have notified SSE subscribers of an event that never durably
happened, which seems wrong. Please confirm §8 is correct.
2. **Dead-letter → handler fan-out appears unwired (see §10).** Confirm
whether the `dead_letter` trigger source is intended to fire in v1.1.x.
## 10. Latent findings
**`dead_letter` trigger handlers do not fire on dead-letter creation.**
`dispatcher::handle_failure` writes the `dead_letters` row via
`DeadLetterRepo::insert` but never enqueues outbox deliveries for matching
`dead_letter`-kind triggers. `TriggerRepo::list_matching_dead_letter` is
defined + implemented but has **no production caller** (only the trait def,
the Postgres impl, and a test fake). So a registered `dead_letter` trigger
never runs from an exhausted-retry event. This predates v1.1.6 (the design
notes §4 describe it shipping in v1.1.1). I did **not** fix it (out of
v1.1.6 scope) — flagging for the reviewer. Consequence: the
`dispatcher_delivers_dead_letter_to_handler` e2e test asserts the
**dead-letter row is produced** (the wired behavior) rather than a handler
firing; documented in the test, and is the honest assertion until the
fan-out lands.
1. **§8 bounded-parallelism caveat** — acceptable, or should the e2e
harness share a single `build_app`/pool across tests in a binary?
(Out of v1.1.7 scope; the existing v1.1.6 e2e tests have the same
shape.)
2. **`email::send` ignoring a stray `html` key** (forcing text-only) vs.
throwing — I chose forgiving text-only; happy to make it strict.
3. **Inbound `received_at`** is stamped by the receiver (`Utc::now()`),
not read from a provider header — confirm that's the intended
semantics.
No new security gaps introduced. Cross-app isolation holds: `subscriber_token`
claims carry `app_id` and are signed per-app; the SSE authority rejects
cross-app tokens (a per-app key already fails the signature, plus an explicit
`claims.app_id == app_id` check); topic admin endpoints bind the capability
to the path `app_id` after loading the app; the broadcaster keys channels by
`(AppId, topic)` so app A's publishes never reach app B's subscribers
(unit-tested).
---
## 11. Deferred items (beyond this prompt's OUT list)
None added. Everything on the brief's OUT list stayed out (WebSocket,
session/script auth modes, topic-pattern external subscription, server-side
last-event-id replay, per-app SSE/rate limits, other-language SDKs, codegen,
full DB-cross-check sweeper).
## 10. Latent security / correctness findings
## 12. Known limitations / rough edges
- SSE delivery is best-effort at-most-once; a slow consumer past the
broadcast buffer loses events with no server-side replay (by design).
- The client lib's streaming-`fetch` SSE needs a polyfill on runtimes
without `fetch` body streaming (React Native) — documented.
- Cron e2e takes ~31s at the default 30s tick interval (45s poll budget);
set `PICLOUD_CRON_TICK_INTERVAL_MS=1000` to speed it up.
- The realtime key cache in `RealtimeAuthorityImpl` is per-process and never
invalidated — correct only because v1.1.6 has no key rotation. A future
rotation API must clear it.
1. **`clippy --all-targets --all-features -- -D warnings` did NOT pass at
v1.1.6 HEAD** (verified by stashing this branch and re-running clippy
on the committed slice-1 tree). Four pre-existing warnings:
`double_must_use` on `realtime_router`, `map_unwrap_or` in
`pubsub_service`, `redundant_closure` in `topic_repo`,
`needless_raw_string_hashes` in a subscriber-token test. Fixed all four
(commit `2ea47eb`) so the gate is now green — flagging because it means
prior "clippy green" claims were likely run without `--all-targets`
(which compiles the test binaries).
2. **Inbound HMAC fails closed on decrypt error.** If a stored
`inbound_secret` can't be decrypted (e.g. `PICLOUD_SECRET_KEY`
rotated), the receiver returns 401 — it refuses the POST rather than
silently skipping verification. Intentional.
3. **No rate limiting on the public inbound-email endpoint.** Like every
public data-plane route, `/api/v1/email-inbound/...` is
unauthenticated by design (URL + HMAC are the gate). An unsigned
trigger (no `inbound_secret`) accepts any POST to its URL and enqueues
outbox rows — URL secrecy is the only guard, as documented. Mitigation
is operator-level (Caddy) rate limiting, the same answer as for other
public routes; no new gap introduced, but noted.
---
## 11. Deferred items (unchanged from brief)
Master-key rotation / per-app master key (v1.2); native SMTP listener
(v1.3+); provider-specific inbound unmarshallers, inbound attachments,
outbound SMTP connection pooling, per-app `from` validation / SPF / DKIM
(v1.2 / operator); dashboard inbound payload viewer (v1.2, PII); drop the
plaintext `realtime_signing_key` column (v1.1.8); secrets
versioning/history + secrets-change triggers (never); `users::*` (v1.1.8);
`queue::*` / `invoke()` (v1.1.9).
---
## 12. Known limitations
- Production `EmailTransport` is a per-call connection; high outbound
volume is connection-churn-bound until pooling (v1.2).
- Outbound `email::send` was not smoke-tested against a live relay in
this environment (no SMTP configured); the SMTP message contents are
asserted via recording-transport unit tests.
- The §8 DB-gated run requires bounded parallelism on a shared Postgres
(see §8); CI's dedicated Postgres does not.