fix(server): hash code changed after introduction of OCR

See b4e5d9d#r182901277

Author: eliandoran

apps/server/src/services/blob.spec.ts

@@ -0,0 +1,53 @@
+import { describe, expect, it } from "vitest";
+
+import blobService from "./blob.js";
+import type { Blob } from "./blob-interface.js";
+
+// These tests lock the exact output of calculateContentHash against known inputs.
+// The hash is written to entity_changes.hash and shipped to sync peers — any
+// change to the formula (fields, separators, ordering) silently invalidates
+// hashes on already-synced blobs, so changes must be deliberate. If a code
+// change breaks these expectations, update the expected values AND ensure a
+// migration or sync-compatibility plan is in place.
+describe("calculateContentHash", () => {
+    const baseBlob: Blob = {
+        blobId: "blob001",
+        content: "hello world",
+        utcDateModified: "2026-01-01 00:00:00.000Z"
+    };
+
+    it("hashes a blob without textRepresentation to the locked value (no trailing separator)", () => {
+        // Must match hash("blobId|content") — same formula used before
+        // textRepresentation existed, so pre-OCR blobs don't re-hash on upgrade.
+        expect(blobService.calculateContentHash(baseBlob)).toBe("dyhIZrQHB3Bb1bhcZVPld8Q6ONU=");
+    });
+
+    it("hashes a blob with textRepresentation to the locked value", () => {
+        expect(blobService.calculateContentHash({ ...baseBlob, textRepresentation: "OCR result" }))
+            .toBe("CsMYMZbvYJtrGVGJwxnr5w6KbMg=");
+    });
+
+    it("treats undefined, null, and empty-string textRepresentation identically", () => {
+        const undefinedHash = blobService.calculateContentHash(baseBlob);
+        const nullHash = blobService.calculateContentHash({ ...baseBlob, textRepresentation: null });
+        const emptyHash = blobService.calculateContentHash({ ...baseBlob, textRepresentation: "" });
+        expect(nullHash).toBe(undefinedHash);
+        expect(emptyHash).toBe(undefinedHash);
+    });
+
+    it("produces different hashes for different blobIds, content, and textRepresentation", () => {
+        const base = blobService.calculateContentHash(baseBlob);
+        const diffId = blobService.calculateContentHash({ ...baseBlob, blobId: "blob002" });
+        const diffContent = blobService.calculateContentHash({ ...baseBlob, content: "other" });
+        const diffTextRep = blobService.calculateContentHash({ ...baseBlob, textRepresentation: "x" });
+        expect(diffId).not.toBe(base);
+        expect(diffContent).not.toBe(base);
+        expect(diffTextRep).not.toBe(base);
+    });
+
+    it("hashes string and equivalent Buffer content identically", () => {
+        const stringHash = blobService.calculateContentHash(baseBlob);
+        const bufferHash = blobService.calculateContentHash({ ...baseBlob, content: Buffer.from("hello world") });
+        expect(bufferHash).toBe(stringHash);
+    });
+});

apps/server/src/services/blob.ts

@@ -51,7 +51,8 @@ function processContent(content: Buffer | string | null, isProtected: boolean, i
 }
 
 function calculateContentHash({ blobId, content, textRepresentation }: Blob) {
-    return hash(`${blobId}|${content.toString()}|${textRepresentation ?? ""}`);
+    const textRepresentationSegment = textRepresentation ? `|${textRepresentation}` : "";
+    return hash(`${blobId}|${content.toString()}${textRepresentationSegment}`);
 }
 
 export default {