fix(runners): preserve UTF-8 across child output chunks

evals/browser/lib/process.mjs

@@ -21,11 +21,11 @@ export function runProcess(command, args, { cwd, env, timeoutMs = 300_000 } = {}
       stdio: ["ignore", "pipe", "pipe"],
     });
 
-    child.stdout.on("data", (chunk) => {
-      stdout = appendCaptured(stdout, chunk.toString());
+    child.stdout.setEncoding("utf8").on("data", (chunk) => {
+      stdout = appendCaptured(stdout, chunk);
     });
-    child.stderr.on("data", (chunk) => {
-      stderr = appendCaptured(stderr, chunk.toString());
+    child.stderr.setEncoding("utf8").on("data", (chunk) => {
+      stderr = appendCaptured(stderr, chunk);
     });
 
     const timeout = setTimeout(() => {

evals/browser/tests/process.test.mjs

@@ -0,0 +1,35 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+import { runProcess } from "../lib/process.mjs";
+
+test("process output preserves UTF-8 characters split across pipe writes", async () => {
+  const expected = "中文🙂";
+  const source = `
+    const bytes = Buffer.from(${JSON.stringify(expected)});
+    let offset = 0;
+    const timer = setInterval(() => {
+      const chunk = bytes.subarray(offset, ++offset);
+      process.stdout.write(chunk);
+      process.stderr.write(chunk);
+      if (offset === bytes.length) clearInterval(timer);
+    }, 10);
+  `;
+  const result = await runProcess(process.execPath, ["-e", source]);
+
+  assert.equal(result.exitCode, 0);
+  assert.equal(result.stdout, expected);
+  assert.equal(result.stderr, expected);
+});
+
+test("process output flushes an incomplete final UTF-8 sequence", async () => {
+  const source = `
+    process.stdout.write(Buffer.from([0xe4, 0xb8]));
+    process.stderr.write(Buffer.from([0xf0, 0x9f]));
+  `;
+  const result = await runProcess(process.execPath, ["-e", source]);
+
+  assert.equal(result.exitCode, 0);
+  assert.equal(result.stdout, "�");
+  assert.equal(result.stderr, "�");
+});

packages/dsh-plugin-browserskill/src/runner.ts

@@ -6,6 +6,7 @@
  */
 
 import { type ChildProcess, type SpawnOptionsWithoutStdio, spawn } from "node:child_process";
+import { StringDecoder } from "node:string_decoder";
 
 /** Shape of the JSON error envelope `bsk --json` prints on failure. */
 export interface BskErrorBody {
@@ -154,14 +155,16 @@ export function createBskRunner(bskPath: string, spawnImpl: SpawnImpl = spawn):
 
         let stdout = "";
         let stderr = "";
+        const stdoutDecoder = new StringDecoder("utf8");
+        const stderrDecoder = new StringDecoder("utf8");
         let timedOut = false;
         let aborted = false;
         const onStdout = (chunk: Buffer | string) => {
-          stdout += chunk;
+          stdout += stdoutDecoder.write(typeof chunk === "string" ? Buffer.from(chunk) : chunk);
           extendDrain();
         };
         const onStderr = (chunk: Buffer | string) => {
-          stderr += chunk;
+          stderr += stderrDecoder.write(typeof chunk === "string" ? Buffer.from(chunk) : chunk);
           extendDrain();
         };
         child.stdout?.on("data", onStdout);
@@ -187,6 +190,8 @@ export function createBskRunner(bskPath: string, spawnImpl: SpawnImpl = spawn):
         const finish = (code: number | null) => {
           if (settled) return;
           settled = true;
+          stdout += stdoutDecoder.end();
+          stderr += stderrDecoder.end();
           settle();
           release();
           resolve({ code, stdout, stderr, timedOut, aborted });

packages/dsh-plugin-browserskill/tests/runner.test.ts

@@ -66,6 +66,30 @@ afterEach(() => {
 });
 
 describe("createBskRunner", () => {
+  it("preserves UTF-8 characters split across output buffers", async () => {
+    const child = new FakeChild();
+    const runner = createBskRunner("bsk", fakeSpawn([child]));
+    const promise = runner.run(["snapshot"]);
+    const stdout = JSON.stringify({ text: "中文🙂" });
+    const stderr = "错误🚫";
+    for (const byte of Buffer.from(stdout)) child.stdout.emit("data", Buffer.from([byte]));
+    for (const byte of Buffer.from(stderr)) child.stderr.emit("data", Buffer.from([byte]));
+    child.finish(0);
+
+    expect(await promise).toMatchObject({ stdout, stderr });
+  });
+
+  it("flushes an incomplete final UTF-8 sequence when output collection ends", async () => {
+    const child = new FakeChild();
+    const runner = createBskRunner("bsk", fakeSpawn([child]));
+    const promise = runner.run(["snapshot"]);
+    child.stdout.emit("data", Buffer.from([0xe4, 0xb8]));
+    child.stderr.emit("data", Buffer.from([0xf0, 0x9f]));
+    child.finish(0);
+
+    expect(await promise).toMatchObject({ stdout: "�", stderr: "�" });
+  });
+
   it("appends --json and collects stdout", async () => {
     const child = new FakeChild();
     const seen: string[][] = [];