Merge 48773c6332 into a38e2f0048

packages/core/src/services/shellExecutionService.test.ts

@@ -25,6 +25,7 @@ import {
 import { NoopSandboxManager } from './sandboxManager.js';
 import { ExecutionLifecycleService } from './executionLifecycleService.js';
 import type { AnsiOutput, AnsiToken } from '../utils/terminalSerializer.js';
+import { getCachedEncodingForBuffer } from '../utils/systemEncoding.js';
 
 // Hoisted Mocks
 const mockPtySpawn = vi.hoisted(() => vi.fn());
@@ -338,6 +339,23 @@ describe('ShellExecutionService', () => {
       expect(result.output.trim()).toBe('你好');
     });
 
+    it('should decode PTY output as UTF-8 regardless of system encoding', async () => {
+      // Simulate a Japanese Windows system where chcp returns code page 932 (Shift-JIS).
+      // The PTY decoder should be pre-initialized to UTF-8, so this mock override
+      // should have no effect on PTY output decoding. Regression test for #12468.
+      vi.mocked(getCachedEncodingForBuffer).mockReturnValue('shift_jis');
+
+      const { result } = await simulateExecution('echo "こんにちは"', (pty) => {
+        pty.onData.mock.calls[0][0]('こんにちは\r\n');
+        pty.onExit.mock.calls[0][0]({ exitCode: 0, signal: null });
+      });
+
+      expect(result.output).toContain('こんにちは');
+
+      // Restore default mock for subsequent tests
+      vi.mocked(getCachedEncodingForBuffer).mockReturnValue('utf-8');
+    });
+
     it('should handle commands with no output', async () => {
       mockSerializeTerminalToObject.mockReturnValue(
         createMockSerializeTerminalToObjectReturnValue(''),

packages/core/src/services/shellExecutionService.ts

@@ -984,7 +984,13 @@ export class ShellExecutionService {
       }).result;
 
       let processingChain = Promise.resolve();
-      let decoder: TextDecoder | null = null;
+      // node-pty delivers data as JavaScript strings via onData().
+      // Buffer.from(data, 'utf-8') in the onData handler always produces
+      // UTF-8 bytes, regardless of the system code page. Pre-initializing
+      // the decoder to UTF-8 prevents getCachedEncodingForBuffer() from
+      // using the system code page (e.g. Shift-JIS on Japanese Windows)
+      // to decode what is already UTF-8 data. Fixes #12468.
+      let decoder: TextDecoder | null = new TextDecoder('utf-8');
       let output: string | AnsiOutput | null = null;
       const sniffChunks: Buffer[] = [];
       let binaryBytesReceived = 0;