fix(browser): keep UTF-8 stderr tails valid

qingminglong · qingminglong · commit 806bbc791abc · 2026-07-09T16:04:56.000+08:00
diff --git a/extensions/browser/src/browser/bounded-utf8-tail.test.ts b/extensions/browser/src/browser/bounded-utf8-tail.test.ts
@@ -16,6 +16,12 @@ describe("bounded UTF-8 tail", () => {
     expect(decodeBoundedUtf8Tail(encoded, 5)).toBe("new");
   });
 
+  it("drops a partial trailing code point after byte truncation", () => {
+    const encoded = Buffer.from([0xf0, 0x9f, 0x98, 0x80]);
+
+    expect(decodeBoundedUtf8Tail(encoded, 1)).toBe("");
+  });
+
   it("replaces the retained tail when one chunk fills the limit", () => {
     const tail = createBoundedUtf8Tail(4);
     tail.append("old");
@@ -34,4 +40,16 @@ describe("bounded UTF-8 tail", () => {
 
     expect(tail.text()).toBe("test");
   });
+
+  it("waits for split UTF-8 code points before decoding them", () => {
+    const tail = createBoundedUtf8Tail(4);
+    const encoded = Buffer.from([0xf0, 0x9f, 0x98, 0x80]);
+    tail.append(encoded.subarray(0, 1));
+
+    expect(tail.text()).toBe("");
+
+    tail.append(encoded.subarray(1));
+
+    expect(tail.text()).toBe("\u{1f600}");
+  });
 });
diff --git a/extensions/browser/src/browser/bounded-utf8-tail.ts b/extensions/browser/src/browser/bounded-utf8-tail.ts
@@ -1,11 +1,12 @@
 /** Byte-bounded UTF-8 tail storage for browser subprocess diagnostics. */
+import { StringDecoder } from "node:string_decoder";
 
 function decodeUtf8Tail(buffer: Buffer): string {
   let start = 0;
   while (start < buffer.length && (buffer[start]! & 0b1100_0000) === 0b1000_0000) {
     start += 1;
   }
-  return buffer.subarray(start).toString("utf8");
+  return new StringDecoder("utf8").write(buffer.subarray(start));
 }
 
 export function decodeBoundedUtf8Tail(buffer: Buffer, maxBytes: number): string {

Original file line number	Diff line number	Diff line change
`@@ -1,11 +1,12 @@`
`1`	`1`	`/** Byte-bounded UTF-8 tail storage for browser subprocess diagnostics. */`
	`2`	`+import { StringDecoder } from "node:string_decoder";`
`2`	`3`
`3`	`4`	`function decodeUtf8Tail(buffer: Buffer): string {`
`4`	`5`	`let start = 0;`
`5`	`6`	`while (start < buffer.length && (buffer[start]! & 0b1100_0000) === 0b1000_0000) {`
`6`	`7`	`start += 1;`
`7`	`8`	`}`
`8`		`- return buffer.subarray(start).toString("utf8");`
	`9`	`+ return new StringDecoder("utf8").write(buffer.subarray(start));`
`9`	`10`	`}`
`10`	`11`
`11`	`12`	`export function decodeBoundedUtf8Tail(buffer: Buffer, maxBytes: number): string {`