fix: LXMF pending-row orphans and Windows esbuild main build

src/main/ipc/reticulum-db-handlers.test.ts

@@ -300,6 +300,38 @@ describe('reticulum destination / activity prune IPC', () => {
     expect(row.delivery_method).toBe('paper');
   });
 
+  it('saveReticulumMessage drops reticulum-pending twins while still sending', () => {
+    const identityId = 'id-rt-pending-orphan';
+    const senderId = 'cc'.repeat(16);
+    const payload = 'hello aibot';
+    const ts = 1_700_000_000_000;
+    const save = handlers.get('db:saveReticulumMessage');
+    save?.(event, {
+      identity_id: identityId,
+      sender_id: senderId,
+      sender_name: 'Me',
+      payload,
+      timestamp: ts,
+      message_hash: 'reticulum-pending-1700000000000',
+      delivery_status: 'sending',
+    });
+    save?.(event, {
+      identity_id: identityId,
+      sender_id: senderId,
+      sender_name: 'Me',
+      payload,
+      timestamp: ts + 182,
+      message_hash: 'ab'.repeat(32),
+      delivery_status: 'sending',
+    });
+    const rows = db!
+      .prepareOnce(
+        'SELECT message_hash, delivery_status FROM reticulum_messages WHERE identity_id = ? ORDER BY id',
+      )
+      .all(identityId) as { message_hash: string; delivery_status: string }[];
+    expect(rows).toEqual([{ message_hash: 'ab'.repeat(32), delivery_status: 'sending' }]);
+  });
+
   it('pruneReticulumIdentityActivityByAge deletes stale millisecond last_seen rows', () => {
     const nowMs = Date.now();
     db!

src/main/ipc/reticulum-db-handlers.ts

@@ -172,12 +172,9 @@ export function registerReticulumDbIpcHandlers({ ipcMain }: ReticulumDbIpcDeps):
           ? Math.trunc(Number(m.next_delivery_attempt_at))
           : null;
 
-      if (
-        messageHash &&
-        !messageHash.startsWith('reticulum-pending-') &&
-        deliveryStatus &&
-        deliveryStatus !== 'sending'
-      ) {
+      // Drop optimistic twins even while still `sending` — otherwise a stuck Direct
+      // Completes leaves reticulum-pending-* + real-hash duplicates after hydrate.
+      if (messageHash && !messageHash.startsWith('reticulum-pending-')) {
         db.prepareOnce(
           `DELETE FROM reticulum_messages
            WHERE identity_id = ? AND sender_id = ? AND payload = ?

src/renderer/lib/reticulum/reticulumOutboundRetry.test.ts

@@ -7,14 +7,19 @@ describe('shouldDeletePriorReticulumOutboundHash', () => {
     expect(shouldDeletePriorReticulumOutboundHash('aa'.repeat(32), 'bb'.repeat(32))).toBe(true);
   });
 
-  it('keeps optimistic pending rows (no prior SQLite hash)', () => {
+  it('deletes optimistic pending rows when rekeyed to a real hash', () => {
     expect(shouldDeletePriorReticulumOutboundHash('reticulum-pending-1', 'bb'.repeat(32))).toBe(
-      false,
+      true,
     );
   });
 
   it('skips when pending id already matches the new hash', () => {
     const hash = 'cc'.repeat(32);
     expect(shouldDeletePriorReticulumOutboundHash(hash, hash)).toBe(false);
   });
+
+  it('skips empty ids', () => {
+    expect(shouldDeletePriorReticulumOutboundHash('', 'bb'.repeat(32))).toBe(false);
+    expect(shouldDeletePriorReticulumOutboundHash('reticulum-pending-1', '')).toBe(false);
+  });
 });

src/renderer/lib/reticulum/reticulumOutboundRetry.ts

@@ -1,11 +1,12 @@
 /**
- * After a manual LXMF resend succeeds, the store rekeys `pendingId` → `newHash`.
- * Delete the prior SQLite row when `pendingId` was a real LXMF hash (not an optimistic
- * `reticulum-pending-*` id), so failed+retried messages do not leave duplicate DB rows.
+ * After LXMF send (or resend) rekeys `pendingId` → `newHash`, delete the prior SQLite row.
+ * Covers:
+ * - optimistic `reticulum-pending-*` rows (otherwise orphan while still `sending`)
+ * - prior real LXMF hashes after a successful retry
  */
 export function shouldDeletePriorReticulumOutboundHash(
   pendingId: string,
   newHash: string,
 ): boolean {
-  return pendingId !== newHash && !pendingId.startsWith('reticulum-pending-');
+  return pendingId !== newHash && pendingId.length > 0 && newHash.length > 0;
 }