test(usage): cover UTF-16-safe log truncation

steipete · web-flow · commit a515316e799e · 2026-07-07T11:23:34.000+01:00
diff --git a/src/infra/session-cost-usage.test.ts b/src/infra/session-cost-usage.test.ts
@@ -2919,6 +2919,25 @@ example
     expect(logs?.[0]?.content).toBe("hello there");
   });
 
+  it("does not split surrogate pairs when truncating session log content", async () => {
+    const root = await makeSessionCostRoot("logs-utf16");
+    const sessionFile = path.join(root, "session.jsonl");
+    const content = "x".repeat(1999) + "🚀tail";
+    await fs.writeFile(
+      sessionFile,
+      JSON.stringify({
+        type: "message",
+        timestamp: "2026-02-21T17:47:00.000Z",
+        message: { role: "assistant", content },
+      }),
+      "utf-8",
+    );
+
+    const logs = await loadSessionLogs({ sessionFile });
+
+    expect(logs?.[0]?.content).toBe(`${"x".repeat(1999)}…`);
+  });
+
   it("normalizes malformed log timestamps with the transcript timestamp rules", async () => {
     const root = await makeSessionCostRoot("logs-malformed-timestamp");
     const sessionsDir = path.join(root, "agents", "main", "sessions");
diff --git a/src/infra/session-cost-usage.ts b/src/infra/session-cost-usage.ts
@@ -4,6 +4,7 @@ import path from "node:path";
 import readline from "node:readline";
 import { asFiniteNumber } from "@openclaw/normalization-core/number-coercion";
 import { normalizeOptionalString } from "@openclaw/normalization-core/string-coerce";
+import { truncateUtf16Safe } from "@openclaw/normalization-core/utf16-slice";
 import type { NormalizedUsage, UsageLike } from "../agents/usage.js";
 import { normalizeUsage } from "../agents/usage.js";
 import { stripInboundMetadata } from "../auto-reply/reply/strip-inbound-meta.js";
@@ -2733,10 +2734,10 @@ export async function loadSessionLogs(params: {
         continue;
       }
 
-      // Truncate very long content, keeping emoji / surrogate pairs intact.
+      // Truncate very long content.
       const maxLen = 2000;
       if (content.length > maxLen) {
-        content = [...content].slice(0, maxLen).join("") + "…";
+        content = truncateUtf16Safe(content, maxLen) + "…";
       }
 
       // Get timestamp
