@blamejs/core 0.14.9 → 0.14.10

package/CHANGELOG.md

@@ -8,6 +8,8 @@ upgrading across more than a few patches at a time.
 
 ## v0.14.x
 
+- v0.14.10 (2026-05-31) — **Full-text-search token hashes move to a keyed MAC; existing mail-store search indexes rebuild automatically on upgrade.** The mail-store full-text-search index hashed its tokens with a hand-rolled salted-SHA3 derived hash. It now routes through the framework's sealed-column hashing primitive in keyed mode (HMAC-SHAKE256 off the per-deployment MAC key), so a search-index token hash is unforgeable and un-correlatable across deployments without that key — the same posture the sealed-column lookup hashes already use. Because the keyed hash changes the stored token values, a mail-store opened after upgrade detects its index as old-format and rebuilds it once from the sealed message rows. The rebuild runs under a format marker: the index is marked `rebuilding` before it is cleared and only marked current after every row is re-hashed inside an explicit transaction, and search falls back to its cursor path (rather than returning partial hits) whenever the marker is not current — so an interrupted rebuild leaves the old index intact and queryable and retries on the next open, never serving a half-built index. A new `b.cryptoField.computeNamespacedHash` primitive backs the keyed hashing for callers that hash outside the registered-column path. **Added:** *`b.cryptoField.computeNamespacedHash`* — A mode-aware namespaced hash for indexed-lookup callers that hash a value outside the registered-column derived-hash path. `computeNamespacedHash(ns, value, { mode, truncateBytes })` routes through the same engine as `computeDerived` — `salted-sha3` (default) or the keyed `hmac-shake256` — with optional hex truncation. The mail-store full-text index is the first consumer. **Changed:** *Mail-store full-text index rehashes to a keyed MAC on upgrade* — The full-text-search token hash now uses `b.cryptoField.computeNamespacedHash` in `hmac-shake256` mode instead of a hand-rolled salted-SHA3. The first time a store is opened after upgrade, its index is detected as old-format and rebuilt once from the sealed message rows; subsequent opens are no-ops. Search is unaffected once the rebuild completes. The rebuild requires the vault to be initialized and fails closed (a clear error) at construction if it is not, rather than leaving a stale searchable index. **Security:** *Keyed, un-correlatable full-text-search token hashes* — A search-index token hash is now a keyed MAC over a per-deployment key, not a static-salted digest — it cannot be forged or correlated across deployments without that key, closing the low-entropy-token correlation gap on the search index. The index remains unrecoverable from a database dump alone, as before. **Detectors:** *Hand-rolled lookup-hash check covers the split form* — The check that requires sealed-column lookup hashes to compose the framework primitive now also catches the across-lines hand-roll (`var salt = getDerivedHashSalt(); var hex = salt.toString(...); sha3(hex + ns + value)`), not only the single-expression form, so the bypass that the mail-store index used can't reappear. **Migration:** *Automatic, one-time full-text index rebuild* — No operator action is required: the rebuild runs automatically and idempotently on first open after upgrade, atomically and crash-safe (an interrupted rebuild keeps the old index and retries). The only requirement is that the vault is initialized before the mail-store is constructed. One caveat for shared stores: do not run a pre-upgrade and post-upgrade node against the same backend file concurrently across this format change — the old node would write old-format hashes the new node cannot match. Roll the deployment fully across the upgrade. This re-open condition is lifted once all nodes are on 0.14.10 or later.
+
 - v0.14.9 (2026-05-30) — **Corrects EU AI Act doc paths that named an uncallable namespace, plus source-comment hygiene and two new codebase checks.** A documentation fix and internal hygiene. The `@primitive` / `@signature` / `@example` blocks for the EU AI Act fundamental-rights-impact-assessment and GPAI training-data-summary helpers advertised `b.complianceAiAct.*`, which is undefined — the callable path is `b.compliance.aiAct.*` — so an operator copying the documented call got `undefined is not a function`. The documented paths now match the real surface. Alongside that: a duplicate parser entry in a doc block is removed, version stamps embedded in section-divider comments are stripped, and two codebase checks are added — one that fails the build when a `@primitive` block documents a wholly-unresolvable namespace (the gap that hid the AI Act paths), and one that flags a version stamp left inside a section divider. No exported API, error code, wire format, or runtime behaviour changes. **Changed:** *Source-comment hygiene* — Removed a duplicate `env` entry from the parsers `@module` doc block, and stripped internal version stamps (`vX.Y.Z`) from `// ---- ... ----` section-divider comments across several files, keeping the descriptive label. Comment-only; no behaviour change. **Fixed:** *EU AI Act helper documentation named an uncallable path* — `b.compliance.aiAct.fundamentalRightsImpactAssessment` and `b.compliance.aiAct.gpai.trainingDataSummary` were documented as `b.complianceAiAct.*` in their `@primitive` / `@signature` / `@example` blocks (and one returned reference string). `b.complianceAiAct` is undefined, so the documented call failed; the documented paths now match the callable surface. **Detectors:** *`@primitive` reachability covers wrong-namespace paths* — The reachability check previously only flagged a missing leaf on a resolved namespace; a `@primitive` whose entire dotted prefix is unresolvable (the shape that hid the AI Act doc paths) was silently skipped. It now walks each prefix segment and fails the build on any unresolvable one, while preserving the factory-instance-shorthand exemption. · *Version-stamp-in-divider check* — A new check flags a version stamp (`vX.Y.Z`) left immediately after a section divider's dashes (`// ---- vX.Y.Z ...`) — internal release vocabulary that does not belong in shipped source comments — without matching legitimate `@since` tags or prose version references.
 
 - v0.14.8 (2026-05-30) — **Source-comment and codebase-check hygiene, plus a new require-block alignment check; no API or behaviour changes.** Internal lint and comment cleanup with no operator-facing surface change. Several codebase-check comments and one stub helper name that described behaviour the check no longer has are corrected; an unused lint-suppression class and a set of stale duplicate-cluster qualifiers (functions that were since renamed or extracted) are pruned or re-pointed. Fifty-nine `// allow:` markers that named the byte-size or time-literal check on values those checks no longer flag are removed, and twenty self-negating rationales on markers the time check genuinely fires on are rewritten to say why the value coincidentally matches. A new codebase check holds top-of-file require blocks to consistent `=` column alignment, with the files that currently carry drift listed as a migration allowlist. No exported API, error code, wire format, or runtime behaviour changes. **Changed:** *Lint-suppression and codebase-check comment cleanup* — Corrected codebase-check comments that overstated their check's scope (a duplicate-code threshold described as three files when the advisory threshold is two, a narrowed byte-literal check carrying its pre-narrowing description, and a deferred-scan helper named as though it enforced a guarantee it does not yet provide). Removed an unused lint-suppression class and its one dead in-code marker, and pruned or re-pointed stale duplicate-cluster qualifiers that named functions since renamed or extracted into shared helpers. Removed fifty-nine `// allow:` markers that suppressed nothing, and rewrote twenty self-negating marker rationales (which read "not seconds" while sitting on a value the time check fires on) to explain the coincidental match. Source-comment and test hygiene only. **Detectors:** *Require-block `=` alignment check* — A new codebase check flags a top-of-file require block that mixes its `=` column alignment — a fittable line whose `=` drifts off the column the rest of the block shares. Compact single-space blocks are exempt (only blocks that declare alignment intent are checked), as are destructures and long names physically too wide to reach the column, and blank- or comment-separated tiers align independently. The files that currently carry such drift are an explicit migration allowlist, reflowed over time; new code is held to the rule.

package/lib/crypto-field.js

@@ -235,6 +235,74 @@ function _computeDerivedHash(spec, tableMode, ns, normalized) {
   return sha3Hash(vault.getDerivedHashSalt().toString("hex") + ns + normalized);
 }
 
+/**
+ * @primitive b.cryptoField.computeNamespacedHash
+ * @signature b.cryptoField.computeNamespacedHash(ns, value, opts?)
+ * @since     0.14.10
+ * @compliance gdpr, hipaa
+ * @related   b.cryptoField.computeDerived, b.cryptoField.lookupHash
+ *
+ * Computes a namespaced indexed-lookup digest of `value` for a
+ * pseudo-field that is NOT backed by a registered derived-hash column
+ * (e.g. the sealed-token FTS index in `b.mailStore.fts`). The caller
+ * supplies the full namespace string directly — there is no schema
+ * lookup — so the same keyed/salted hash machinery that protects
+ * registered derived hashes also covers ad-hoc indexed tokens. This is
+ * the canonical entry point: hand-rolling
+ * `sha3Hash(vault.getDerivedHashSalt() + ns + value)` at a call site
+ * bypasses the keyed-MAC mode (`hmac-shake256` off
+ * `vault.getDerivedHashMacKey`) and the per-deployment salt policy.
+ *
+ * `opts.mode` selects the digest:
+ *   - `"salted-sha3"` (default): SHA3-512 over `<salt-hex> + ns + value`
+ *     (deterministic per deployment; byte-identical to the legacy
+ *     hand-rolled scheme).
+ *   - `"hmac-shake256"`: SHAKE256(`<vault MAC key> || ns + value`) — a
+ *     keyed MAC so an attacker who recovers the salt alone cannot
+ *     correlate two low-entropy plaintexts.
+ *
+ * `opts.truncateBytes` truncates the hex digest to that many BYTES
+ * (the hex string is sliced to `truncateBytes * 2` characters). Throws
+ * (config-time / entry-point tier) on an unknown `mode` or a
+ * non-positive-integer `truncateBytes` so an operator catches the typo
+ * at boot rather than silently indexing under a malformed digest.
+ *
+ * @opts
+ *   mode:          string,   // "salted-sha3" (default) | "hmac-shake256"
+ *   truncateBytes: number,   // optional; positive integer byte width to slice to
+ *
+ * @example
+ *   var ns = "bj-mail_messages-body:fts:";
+ *   var h = b.cryptoField.computeNamespacedHash(ns, "kubernetes", {
+ *     mode: "hmac-shake256", truncateBytes: 8
+ *   });
+ *   /^[0-9a-f]{16}$/.test(h);   // → true
+ *
+ *   // Default mode is byte-identical to the legacy salted-sha3 hash.
+ *   b.cryptoField.computeNamespacedHash(ns, "kubernetes").length;   // → 128
+ */
+function computeNamespacedHash(ns, value, opts) {
+  opts = opts || {};
+  var mode = opts.mode || "salted-sha3";
+  if (mode !== "salted-sha3" && mode !== "hmac-shake256") {
+    throw new Error("computeNamespacedHash: opts.mode must be 'salted-sha3' " +
+      "(default) or 'hmac-shake256', got " + JSON.stringify(mode));
+  }
+  var truncateBytes = opts.truncateBytes;
+  if (truncateBytes !== undefined) {
+    if (typeof truncateBytes !== "number" || !isFinite(truncateBytes) ||
+        truncateBytes <= 0 || Math.floor(truncateBytes) !== truncateBytes) {
+      throw new Error("computeNamespacedHash: opts.truncateBytes must be a " +
+        "positive integer (bytes), got " + JSON.stringify(truncateBytes));
+    }
+  }
+  var hex = _computeDerivedHash({ mode: mode }, mode, ns, String(value));
+  if (truncateBytes !== undefined) {
+    return hex.slice(0, truncateBytes * 2);
+  }
+  return hex;
+}
+
 /**
  * @primitive b.cryptoField.getSchema
  * @signature b.cryptoField.getSchema(table)
@@ -1046,6 +1114,7 @@ module.exports = {
   applyPosture:     applyPosture,
   getActivePosture: getActivePosture,
   computeDerived:   computeDerived,
+  computeNamespacedHash: computeNamespacedHash,
   lookupHash:       lookupHash,
   clearForTest:     clearForTest,
   declareColumnResidency: declareColumnResidency,

package/lib/mail-store-fts.js

@@ -54,10 +54,23 @@
  *   nothing readable; search works against ciphertext.
  */
 
-var bCrypto      = require("./crypto");
-var vault        = require("./vault");
+var cryptoField  = require("./crypto-field");
 var C            = require("./constants");
 
+// Sealed-token FTS on-disk format version. Bumped when the token-hash
+// transform changes so the mail-store reindex path can detect a stale
+// index and rebuild it from the sealed messages table. v1 was the
+// legacy salted-sha3-truncated hand-roll; v2 is the keyed
+// hmac-shake256 digest computed via cryptoField.computeNamespacedHash.
+var FTS_FORMAT_VERSION = 2;
+
+// Per-token hash width, in bytes. 8 bytes -> 16 hex chars. Full 64-char
+// SHA3 / SHAKE digest is overkill for the FTS hash space, and the
+// shorter token compresses the FTS5 row 4x without observable collision
+// risk at corpus sizes the framework targets (<= 10^9 unique tokens,
+// where 64-bit collision space leaves the birthday bound > 10^9).
+var FTS_HASH_BYTES = 8;
+
 // Stopwords are dropped before hashing — they'd dominate every row's
 // token set without adding query selectivity. Kept conservative to
 // stay locale-neutral for v1.
@@ -154,26 +167,29 @@ function tokenize(text) {
   return out;
 }
 
-// Hash one token using the same scheme cryptoField uses for derived-
-// hash mirrors: `sha3Hash(vaultSalt + namespace + token)`. The
+// Hash one token through the canonical cryptoField primitive
+// (computeNamespacedHash) so the FTS index inherits the keyed-MAC
+// digest used for derived-hash mirrors on sealed columns. The
 // namespace is per-table, per-field, per-purpose ("fts") so that
-// rotating an operator's vault salt invalidates every FTS row in
-// the same step as every sealed column. Returns a 16-char hex prefix
-// — full 64-char SHA3 is overkill for FTS hash space, and shorter
-// tokens compress the FTS5 row 4x without observable collision risk
-// at corpus sizes the framework targets (≤ 10^9 unique tokens, where
-// 64-bit collision space leaves the birthday bound > 10^9).
+// rotating an operator's vault key invalidates every FTS row in the
+// same step as every sealed column. Returns a 16-char hex prefix
+// (FTS_HASH_BYTES bytes) — full 64-char digest is overkill for the
+// FTS hash space, and shorter tokens compress the FTS5 row 4x without
+// observable collision risk at corpus sizes the framework targets
+// (<= 10^9 unique tokens, where 64-bit collision space leaves the
+// birthday bound > 10^9).
 /**
  * @primitive b.mailStore.fts.hashToken
  * @signature b.mailStore.fts.hashToken(table, field, token)
  * @since     0.11.25
  * @status    stable
  *
- * Vault-salted hash of one token under the (table, field) namespace.
- * The same scheme `b.cryptoField.computeDerived` uses for derived-
- * hash mirrors on sealed columns — rotating the vault salt
- * invalidates every FTS hash in step with every sealed-column hash.
- * Returns a 16-char hex prefix.
+ * Keyed hash of one token under the (table, field) namespace. Routes
+ * through `b.cryptoField.computeNamespacedHash` in `hmac-shake256`
+ * mode — the same keyed-MAC machinery that protects sealed-column
+ * derived hashes — so rotating the vault key invalidates every FTS
+ * hash in step with every sealed-column hash. Returns a 16-char hex
+ * prefix.
  *
  * @example
  *   var h = b.mailStore.fts.hashToken("mail_messages", "body", "kubernetes");
@@ -185,9 +201,10 @@ function hashToken(table, field, token) {
   // fields are pseudo-fields (no sealed-column registration), so the
   // canonical fallback path is always the right answer here.
   var ns = "bj-" + table + "-" + field + ":fts:";
-  var salt = vault.getDerivedHashSalt();
-  var saltHex = (salt && typeof salt.toString === "function") ? salt.toString("hex") : "";
-  return bCrypto.sha3Hash(saltHex + ns + token).slice(0, 16);                                          // 16-char hex prefix length, not bytes
+  return cryptoField.computeNamespacedHash(ns, token, {
+    mode:          "hmac-shake256",
+    truncateBytes: FTS_HASH_BYTES,
+  });
 }
 
 // Hash a token array → space-separated string suitable for FTS5
@@ -391,4 +408,9 @@ module.exports = {
   MAX_TOKEN_LEN:       MAX_TOKEN_LEN,
   MAX_TOKENS_PER_FIELD: MAX_TOKENS_PER_FIELD,
   FTS_FIELDS:          FTS_FIELDS,
+  FTS_HASH_BYTES:      FTS_HASH_BYTES,
+  // On-disk format marker — the mail-store reindex path stamps this
+  // into `<prefix>_meta` once a full rebuild completes; a stale/missing
+  // marker triggers a rebuild from the sealed messages table.
+  FTS_FORMAT_VERSION:  FTS_FORMAT_VERSION,
 };

package/lib/mail-store.js

@@ -60,6 +60,7 @@
 var C = require("./constants");
 var bCrypto = require("./crypto");
 var cryptoField = require("./crypto-field");
+var vault = require("./vault");
 var safeMime = require("./safe-mime");
 var safeSql = require("./safe-sql");
 var guardMessageId = require("./guard-message-id");
@@ -72,6 +73,14 @@ var DEFAULT_TABLE_PREFIX = "blamejs_mail";
 var DEFAULT_MAX_MESSAGE_BYTES = C.BYTES.mib(50);
 var DEFAULT_MAX_BODY_BYTES    = C.BYTES.mib(25);
 
+// `<prefix>_meta` marker key carrying the FTS on-disk format version.
+// The reindex path writes a `'rebuilding'` sentinel BEFORE clearing the
+// index and the final format-version string only AFTER every row is
+// reinserted, so a partial / interrupted rebuild is never marked
+// complete and never queried as a finished index.
+var FTS_FORMAT_META_KEY = "fts_format";
+var FTS_REBUILDING_SENTINEL = "rebuilding";
+
 // Standard IMAP4rev2 default folders + JMAP role mapping.
 var DEFAULT_FOLDERS = Object.freeze([
   { name: "INBOX",   role: "inbox" },
@@ -129,6 +138,7 @@ function create(opts) {
   var qFlags   = safeSql.quoteIdentifier(prefix + "_flags",     "sqlite");
   var qQuota   = safeSql.quoteIdentifier(prefix + "_quota",     "sqlite");
   var qFts     = safeSql.quoteIdentifier(prefix + "_messages_fts", "sqlite");
+  var qMeta    = safeSql.quoteIdentifier(prefix + "_meta",      "sqlite");
   var messagesTable = prefix + "_messages";
 
   var maxMessageBytes = opts.maxMessageBytes !== undefined ? opts.maxMessageBytes : DEFAULT_MAX_MESSAGE_BYTES;
@@ -150,7 +160,7 @@ function create(opts) {
   });
 
   if (doInit) {
-    _ensureSchema(db, qMsgs, qFolders, qFlags, qQuota, qFts);
+    _ensureSchema(db, qMsgs, qFolders, qFlags, qQuota, qFts, qMeta);
     _ensureDefaultFolders(db, qFolders);
   }
 
@@ -207,6 +217,150 @@ function create(opts) {
     "INSERT INTO " + qFts + " (objectid, subject_toks, addr_toks, body_toks) VALUES (?, ?, ?, ?)");
   var stmtDeleteFts        = db.prepare("DELETE FROM " + qFts + " WHERE objectid = ?");
 
+  // `<prefix>_meta` marker read — used by the reindex gate at create()
+  // AND by every FTS query path (search) to refuse a half-built index.
+  // Guarded behind a closure so a store opened with init:false against a
+  // pre-format-version database (no _meta table) reads as "no marker"
+  // instead of throwing.
+  function _readFtsMarker() {
+    try {
+      var row = db.prepare("SELECT value FROM " + qMeta + " WHERE key = ?")
+        .get(FTS_FORMAT_META_KEY);
+      return row ? row.value : null;
+    } catch (_e) {
+      // _meta table absent (init:false against a legacy store) — treat
+      // as no marker so the FTS query path falls back rather than
+      // querying a possibly-stale index as if it were current.
+      return null;
+    }
+  }
+  // The FTS index is queryable only when the on-disk marker equals the
+  // current format version. A `'rebuilding'` sentinel or any other
+  // value (stale format, missing marker) means the index is non-final;
+  // search() falls back to the modseq cursor rather than returning
+  // partial / mixed-scheme FTS hits.
+  var CURRENT_FTS_FMT = String(mailStoreFts.FTS_FORMAT_VERSION);
+  function _ftsIndexUsable() {
+    return _readFtsMarker() === CURRENT_FTS_FMT;
+  }
+
+  // Reindex-on-upgrade. Runs once at create() (when doInit) AFTER the
+  // schema is ensured. Rebuilds the sealed-token FTS index from the
+  // canonical (sealed) messages table whenever the on-disk format
+  // marker is missing or stale — the token-hash transform changed, so
+  // every previously-indexed row hashes to a different value and the
+  // old rows are unsearchable under the new scheme.
+  //
+  // Data-safety contract:
+  //   - A `'rebuilding'` sentinel is written BEFORE the DELETE, so an
+  //     interrupted rebuild leaves a non-final marker; search() refuses
+  //     the index until a later create() completes the rebuild.
+  //   - The DELETE + every reinsert run inside an explicit
+  //     BEGIN/COMMIT/ROLLBACK (ordinary SQLite — works on node:sqlite
+  //     and b.db alike; neither exposes a usable `.transaction(fn)()`
+  //     for this shape). A throw inside the insert loop rolls the whole
+  //     rebuild back, leaving the OLD index intact + queryable... except
+  //     the marker still reads as non-final, so search falls back until
+  //     the next successful create() — the index is never silently
+  //     half-built.
+  //   - The final format-version marker is written ONLY after every row
+  //     reinserts and the COMMIT lands.
+  function _reindexFts() {
+    var fmt = _readFtsMarker();
+    if (fmt === CURRENT_FTS_FMT) return { reindexed: false, reason: "current" };
+
+    // Count existing FTS rows AFTER the early-return so the common
+    // already-current path pays nothing for the scan.
+    var ftsCountRow = db.prepare("SELECT COUNT(*) AS n FROM " + qFts).get();
+    var ftsCount = (ftsCountRow && ftsCountRow.n) || 0;
+    var msgCountRow = db.prepare("SELECT COUNT(*) AS n FROM " + qMsgs).get();
+    var msgCount = (msgCountRow && msgCountRow.n) || 0;
+
+    // Fresh store — no marker AND nothing indexed AND no messages to
+    // index. Just stamp the current format; there is nothing to rebuild.
+    if (fmt === null && ftsCount === 0 && msgCount === 0) {
+      _writeFtsMarker(CURRENT_FTS_FMT);
+      return { reindexed: false, reason: "fresh" };
+    }
+
+    // Reindex unseals EVERY message row — a runtime dependency on the
+    // vault that a bare create() otherwise wouldn't have. Fail loud at
+    // boot (entry-point tier) rather than silently leaving a stale,
+    // wrong-scheme index searchable: an operator who constructs the
+    // store before vault.init() must see the misordering immediately.
+    if (!vault.isInitialized()) {
+      throw new MailStoreError("mail-store/fts-reindex-vault-uninitialized",
+        "mailStore.create: FTS index format is stale (marker=" +
+        JSON.stringify(fmt) + ", current=" + CURRENT_FTS_FMT + ") and a " +
+        "reindex from the sealed messages table requires the vault - call " +
+        "b.vault.init(...) BEFORE b.mailStore.create(...). Refusing to leave " +
+        "a stale, wrong-scheme search index queryable.");
+    }
+
+    // Sentinel BEFORE any destructive work — a crash between here and
+    // the final marker leaves the index marked non-final.
+    _writeFtsMarker(FTS_REBUILDING_SENTINEL);
+
+    // BEGIN IMMEDIATE takes the write lock up front so the message
+    // snapshot, the FTS delete, and the reinsert are one isolated window.
+    // The snapshot is read INSIDE the lock: a concurrent writer in another
+    // process cannot append a message between the snapshot and the delete
+    // (which would otherwise drop that row's freshly-inserted FTS entry
+    // without re-adding it, silently losing it from search).
+    var allRows;
+    db.prepare("BEGIN IMMEDIATE").run();
+    try {
+      allRows = db.prepare("SELECT * FROM " + qMsgs).all();
+      db.prepare("DELETE FROM " + qFts).run();
+      for (var i = 0; i < allRows.length; i += 1) {
+        // unsealRow returns the DB column names (subject / from_addr /
+        // to_addrs / body_text); map them into the FTS param shape
+        // (subject / from / to / body) rowFromMessage expects.
+        var clear = cryptoField.unsealRow(messagesTable, allRows[i]);
+        var ftsRow = mailStoreFts.rowFromMessage(messagesTable, {
+          objectid: clear.objectid,
+          subject:  clear.subject   || "",
+          from:     clear.from_addr || "",
+          to:       clear.to_addrs  || "",
+          body:     clear.body_text || "",
+        });
+        stmtInsertFts.run(ftsRow.objectid, ftsRow.subject_toks,
+          ftsRow.addr_toks, ftsRow.body_toks);
+      }
+      db.prepare("COMMIT").run();
+    } catch (e) {
+      // Roll the partial rebuild back — the OLD index rows are restored
+      // (the DELETE is undone), so the prior scheme stays intact and
+      // queryable. The marker still reads as the sentinel, so search()
+      // falls back until a later create() completes the rebuild: a
+      // retriable, never-silently-half-built state.
+      try { db.prepare("ROLLBACK").run(); } catch (_re) { /* best-effort */ }
+      throw new MailStoreError("mail-store/fts-reindex-failed",
+        "mailStore.create: FTS reindex from the sealed messages table " +
+        "failed and was rolled back (the prior index is intact); retry " +
+        "after resolving: " + ((e && e.message) || String(e)));
+    }
+
+    // Full rebuild committed — stamp the current format LAST so the
+    // index only reads as final once every row is present.
+    _writeFtsMarker(CURRENT_FTS_FMT);
+    return { reindexed: true, rows: allRows.length };
+  }
+
+  function _writeFtsMarker(value) {
+    db.prepare(
+      "INSERT INTO " + qMeta + " (key, value) VALUES (?, ?) " +
+      "ON CONFLICT(key) DO UPDATE SET value = excluded.value"
+    ).run(FTS_FORMAT_META_KEY, value);
+  }
+
+  // Wire the reindex into create() AFTER schema bootstrap. A fresh store
+  // (no marker, 0 FTS rows, 0 messages) just stamps the format; a store
+  // carrying an old-format index rebuilds from the sealed rows.
+  if (doInit) {
+    _reindexFts();
+  }
+
   return {
     appendMessage:    function (folderName, rawBytes, appendOpts) {
       // Wrap canonical row insert + FTS row insert in a single backend
@@ -277,6 +431,27 @@ function create(opts) {
       var limit = f.limit || 100;
       if (limit > 1000) limit = 1000;                                                                  // query row cap, not bytes
 
+      // The FTS index is queryable only when the on-disk format marker
+      // is current. A `'rebuilding'` sentinel (mid-rebuild / interrupted
+      // rebuild) or a stale/missing marker means the index is non-final
+      // — fall back to the bare modseq cursor rather than returning
+      // partial or wrong-scheme FTS hits. Returning a subset of the true
+      // matches as if it were complete is the corruption hazard the
+      // reindex sentinel exists to prevent.
+      if (!_ftsIndexUsable()) {
+        var nonFinal = stmtQueryByModseq.all(folder.id, sinceModseq, limit);
+        return {
+          rows: nonFinal.map(function (r) {
+            return {
+              objectid: r.objectid, modseq: r.modseq, sizeBytes: r.size_bytes,
+              internalDate: r.internal_date, legalHold: r.legal_hold === 1,
+            };
+          }),
+          nextModseq: nonFinal.length > 0 ? nonFinal[nonFinal.length - 1].modseq : sinceModseq,
+          ftsUnavailable: true,
+        };
+      }
+
       var matchClauses = [];
       function addMatch(filterKey, term) {
         if (!term) return;
@@ -834,7 +1009,7 @@ function _normalizeMsgId(s) {
 
 // ---- Schema bootstrap ----------------------------------------------------
 
-function _ensureSchema(db, qMsgs, qFolders, qFlags, qQuota, qFts) {
+function _ensureSchema(db, qMsgs, qFolders, qFlags, qQuota, qFts, qMeta) {
   // Folders table — created first since messages reference folder_id.
   db.prepare(
     "CREATE TABLE IF NOT EXISTS " + qFolders + " (" +
@@ -910,6 +1085,17 @@ function _ensureSchema(db, qMsgs, qFolders, qFlags, qQuota, qFts) {
   // tokens on whitespace exactly — hashes are ASCII-hex-only, so no
   // Unicode case-fold runs at MATCH time.
   db.prepare(mailStoreFts.createSql(qFts)).run();
+
+  // Per-prefix key/value metadata table. Holds the FTS on-disk format
+  // marker (`fts_format`) so the reindex path can detect a stale index
+  // and rebuild it. Scoped per table-prefix (NOT PRAGMA user_version,
+  // which is db-global and would collide across stores sharing one
+  // sqlite file).
+  db.prepare(
+    "CREATE TABLE IF NOT EXISTS " + qMeta + " (" +
+    "key TEXT PRIMARY KEY, " +
+    "value TEXT)"
+  ).run();
 }
 
 function _ensureDefaultFolders(db, qFolders) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blamejs/core",
-  "version": "0.14.9",
+  "version": "0.14.10",
   "description": "The Node framework that owns its stack.",
   "license": "Apache-2.0",
   "author": "blamejs contributors",

package/sbom.cdx.json

@@ -2,10 +2,10 @@
   "$schema": "http://cyclonedx.org/schema/bom-1.5.schema.json",
   "bomFormat": "CycloneDX",
   "specVersion": "1.5",
-  "serialNumber": "urn:uuid:1e3ccafc-71e3-4207-a3d7-5672ad58b137",
+  "serialNumber": "urn:uuid:a284c3a3-4413-4bda-9a99-ef60b057cc3a",
   "version": 1,
   "metadata": {
-    "timestamp": "2026-05-31T04:50:32.275Z",
+    "timestamp": "2026-05-31T11:30:02.144Z",
     "lifecycles": [
       {
         "phase": "build"
@@ -19,14 +19,14 @@
       }
     ],
     "component": {
-      "bom-ref": "@blamejs/core@0.14.9",
+      "bom-ref": "@blamejs/core@0.14.10",
       "type": "application",
       "name": "blamejs",
-      "version": "0.14.9",
+      "version": "0.14.10",
       "scope": "required",
       "author": "blamejs contributors",
       "description": "The Node framework that owns its stack.",
-      "purl": "pkg:npm/%40blamejs/core@0.14.9",
+      "purl": "pkg:npm/%40blamejs/core@0.14.10",
       "properties": [],
       "externalReferences": [
         {
@@ -54,7 +54,7 @@
   "components": [],
   "dependencies": [
     {
-      "ref": "@blamejs/core@0.14.9",
+      "ref": "@blamejs/core@0.14.10",
       "dependsOn": []
     }
   ]