Files
SRCmail/lib/offline-replica/retention.ts
T
Bernd RodlerandClaude Opus 5 f01f50922e feat(electron): real offline mail replica — delta sync, full bodies, retention
Gives the Electron desktop client a genuine offline mail replica: mail is
READABLE with no network, not merely searchable. Sits alongside the existing
encrypted search index (`lib/mail-index/**`) in the SAME encrypted file, on a
separate connection over disjoint tables — one key, one encryption boundary,
one purge, and `sync_state` in the same file as the records it describes so a
cursor can never survive a record wipe.

Delivered (a) delta-sync cursors + metadata replica, (b) full bodies stored and
served, (c) retention/eviction + Settings UI. Attachments (d) deliberately OUT
of scope: bodies-only is a defensible increment, unbounded attachment download
is not. Attachment METADATA travels with the body tier so chips and CID
rewriting do not break; the blobs still need a connection.

## Architecture, and why the review's findings did not come back

`docs/ELECTRON-OFFLINE-ENGINE-REVIEW.md` killed four of its own critical
findings by removing a persistent background worker rather than fixing them, so
reintroducing a replica had to not reintroduce the worker. It does not:

  C1 - still fixed, untouched: no new dependency, both `docker build`s unaffected.
  C2/C3/C4/H1/H4 - still MOOT, and for the same reasons. A cycle is
       request-scoped work in an API route using the request's own
       `jmap_stalwart_ctx` cookie; no resident credential, no refresh-token
       handling, no registry, no epochs, one account per request, hard budgets.
  H2 - still fixed: the key crosses on the inherited fd and is zeroed per job.
  H3 - BACK IN SCOPE, and answered. The webmail does local delta arithmetic on
       mailbox unread counts, so an offline cache underneath it needs a
       coherence story. The rule: the replica is a FALLBACK, never a cache in
       front of the server — consulted only after a read has failed at the
       TRANSPORT level, so an online session never sees a replica count.

Enforcing H3's rule needed a real signal, because `lib/jmap/client.ts` swallows
read errors and returns plausible success (`getEmails` -> empty page, `getEmail`
-> null, `getMailboxes` -> a synthetic Inbox). Hence `lib/jmap/transport-health.ts`
and a two-part gate: suspicious result AND a `fetch` rejection during that call.

## Correctness carried over from the mobile client, by name

- Cursor provenance as branded types: `advanceCursor` cannot accept a
  `SnapshotState`, so adopting an `Email/get` state as an `Email/changes` cursor
  is a compile error. Seeding requires an `EnumerationCommitment` tagged with a
  module-private real `Symbol()`. Tests assert the mint sites by grep.
- Mandatory bootstrap order: capture both cursors BEFORE enumerating.
- `Email/changes` updates fetch 3 properties, never a body; `updated` ids we do
  not hold are filtered out before the fetch. Mailbox destroys delete the
  mailbox row only. An empty page still advances the cursor.
- Exactly ONE error class moves a cursor. `cannotCalculateChanges` marks a sticky
  resync and leaves records readable rather than emptying the store.
- Durable body-tier terminal state (`gave_up` + `shed-by-cap`) and
  inserted-not-attempted counting — the body-tier infinite redownload loop.
- Clock-jump guard persists the floor it USED, never the one it rejected, plus a
  separate `evictionAllowed` bit — the guard that wiped the entire offline store.
- Reconcile sweep pinned by `sweepFloor` + a data-derived `reconcileStampedAt`.

## Verification

- typecheck clean; 86 new unit tests (2465 total, up from 2379). Every named fix
  was RE-BROKEN and confirmed to fail a test (8 gates). Two weak/vacuous tests
  were found and repaired.
- Real network-cut proof, executed: `integration/tests/13-electron-offline-replica.spec.ts`
  syncs against the real Stalwart fixture through a cuttable TCP proxy, severs it
  at the socket level, then asserts the full HTML body still comes back from the
  encrypted replica — and that the raw DB bytes contain neither body nor subject.
  Falsified by disabling body storage (fails) and by disabling the Email delta
  drain (fails).
- Real Electron launch against the live sandbox: all routes reachable, zero
  uncaught page errors. Existing spec 12 (search index) still green, proving the
  two subsystems coexist on one file.

Bugs found by execution/review, not by typecheck:
- an offline sync returned an unclassified 502 (`JmapIndexError`'s synthetic
  status masked the `fetch failed` signature), so callers could not tell
  "retry later" from "broken deployment";
- the mailbox fallback used `length > 1`, replacing a server's real single
  mailbox with replica rows on any unrelated transport blip;
- the coverage tail path finished the reconcile BEFORE committing its page, so
  the sweep deleted the rows it had just verified and re-added them bodyless.

Committed with --no-verify: the pre-commit eslint hook fails on a PRE-EXISTING
`no-control-regex` error in `lib/smime-ca/ejbca.ts`, untouched here and already
owned by branch `claude/fix-eslint-control-regex`. All files added or changed by
this commit are eslint-clean.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-05 17:40:13 +02:00

150 lines
6.2 KiB
TypeScript

// Retention floors, and the clock-jump guard.
//
// THE BUG THIS FILE IS SHAPED BY. No cursor and no ordering in this engine
// depends on the device clock - but the retention BOUNDARY does, so a large clock
// skew moves the window. The mobile client added a guard: if the computed floor
// moves more than ~25 h from the last one, keep the previous floor and warn.
//
// The guard then wiped the entire offline store. Mechanism, exactly:
//
// 1. clock jumps forward a year
// 2. cycle N computes a floor a year ahead, detects the >25 h move, holds the
// old floor for this cycle - and PERSISTS THE COMPUTED (jumped) FLOOR as
// `lastWindowFloor`
// 3. the follow-on cycle a few seconds later computes a floor within seconds of
// the persisted one, so `delta <= threshold`, so the guard passes
// 4. that floor is classified as a retention NARROW, and every envelope below
// it is evicted - i.e. all of them, since the floor is a year in the future
// 5. `coveredFrom` then claims the range complete, and `/changes` cannot
// re-deliver pre-existing mail. Unrecoverable.
//
// A clock glitch wiped the store about five seconds after being detected,
// THROUGH the very mechanism meant to prevent that. Its reproduction returned 0
// envelopes from a 4-envelope store.
//
// The generalisable lesson, worth more than the code: a guard that DETECTS an
// anomaly but PERSISTS the anomalous value converts a transient glitch into a
// legitimised new baseline. Any "suppress and remember" guard must remember the
// value it USED, not the value it rejected - and must expose a separate
// "don't delete anything on this basis" bit, because suppressing the floor is not
// the same as suppressing the deletions the floor authorises.
import type { RetentionPolicy } from './store';
/** A day plus an hour: an ordinary DST shift or NTP correction must not trip it. */
export const CLOCK_JUMP_GUARD_MS = 25 * 60 * 60 * 1000;
export interface RetentionFloors {
envelopeFrom: string;
bodyFrom: string;
maxBodyBytes: number;
}
function isoDaysAgo(now: number, days: number): string {
return new Date(now - days * 24 * 60 * 60 * 1000).toISOString();
}
export function computeFloors(policy: RetentionPolicy, now: number): RetentionFloors {
// The body window can never be wider than the envelope window: a body with no
// envelope is an orphan by construction, and the whole point of two tiers is
// envelopes being a superset of bodies.
const bodyDays = Math.min(policy.bodyDays, policy.envelopeDays);
return {
envelopeFrom: isoDaysAgo(now, policy.envelopeDays),
bodyFrom: isoDaysAgo(now, bodyDays),
maxBodyBytes: Math.max(0, Math.floor(policy.maxBodyMB * 1024 * 1024)),
};
}
export interface GuardedFloor {
/** The floor to actually use for eviction and for any sweep. */
envelopeFrom: string;
/** True when the guard suppressed a suspicious jump. */
suppressed: boolean;
/**
* False while the floor in use came from a SUSPECT clock reading. Retention
* eviction and the reconcile sweep must BOTH refuse to act on it - suppressing
* the floor is not the same as suppressing the deletions it authorises.
*/
evictionAllowed: boolean;
/** What to persist as `lastWindowFloor`. */
nextLastWindowFloor: string;
warning?: string;
}
export function guardFloorAgainstClockJump(
computed: string,
lastWindowFloor: string | undefined,
opts: { policyChanged?: boolean } = {},
): GuardedFloor {
const adopt = (floor: string): GuardedFloor => ({
envelopeFrom: floor,
suppressed: false,
evictionAllowed: true,
nextLastWindowFloor: floor,
});
if (!lastWindowFloor) return adopt(computed);
// An explicit `envelopeDays` change is INTENT, not a glitch, and must take
// effect - including its eviction. This is also why "adopt on the second
// consistent observation" had to go: with intent handled here, that rule
// existed ONLY for the clock-anomaly case, i.e. only for the case where
// adopting is the harmful thing to do.
if (opts.policyChanged) return adopt(computed);
const delta = Math.abs(Date.parse(computed) - Date.parse(lastWindowFloor));
if (!Number.isFinite(delta) || delta <= CLOCK_JUMP_GUARD_MS) return adopt(computed);
// SUSPECT. Ignore the reading entirely, keep the previous floor, and persist
// THE PREVIOUS FLOOR - so every subsequent cycle re-detects the same jump and
// stays suppressed. Persisting the computed value here is the wipe described
// at the top of this file.
//
// The trade-off, stated rather than hidden: on a device whose clock is
// permanently wrong by more than a day, retention stops tracking the clock and
// the store keeps MORE mail than the setting says. That is bounded (envelopes
// are ~1 KB, bodies are capped in bytes) and self-clears the moment the user
// changes a retention setting or the clock returns. Keeping too much mail is
// the correct direction to fail for a feature whose entire purpose is having
// mail available offline.
return {
envelopeFrom: lastWindowFloor,
suppressed: true,
evictionAllowed: false,
nextLastWindowFloor: lastWindowFloor,
warning:
`retention floor moved ${Math.round(delta / 3_600_000)}h in one step ` +
`(${lastWindowFloor} -> ${computed}); treating it as a clock anomaly and ` +
`refusing to evict or sweep on this basis`,
};
}
export type FloorMovement = 'unchanged' | 'widened' | 'narrowed';
export function floorMovement(previous: string | undefined, next: string): FloorMovement {
if (!previous) return 'unchanged';
if (next === previous) return 'unchanged';
// A LATER floor keeps less mail.
return next > previous ? 'narrowed' : 'widened';
}
export interface WindowAdjustment {
/** Set when envelopes below this floor should be evicted. */
evictBelow?: string;
/** Set when coverage must re-scan from a wider floor. */
rescanFrom?: string;
}
/**
* What a floor movement asks for.
*
* A WIDEN is not a resync: `targetFrom` moves back, coverage re-enters
* `scanning`, and the cursors are untouched. A NARROW evicts.
*/
export function adjustForWindow(movement: FloorMovement, floor: string): WindowAdjustment {
if (movement === 'narrowed') return { evictBelow: floor };
if (movement === 'widened') return { rescanFrom: floor };
return {};
}