Problem

Fixes #1296

The MultiEncodingTerminalTest was experiencing intermittent failures on Windows CI environments with:

expected: <café> but was: <ca�fé>

The replacement character � (U+FFFE) was appearing randomly in test output, initially suggesting an encoding issue in the NonBlocking implementation.

Root Cause Discovered

🎯 The issue was NOT in the NonBlocking implementation, but in the test logic itself!

When NonBlockingReader.read(timeout) times out, it returns READ_EXPIRED (-2). The test code was incorrectly casting this directly to char:

while ((c = reader.read(1)) != -1) {
    result.append((char) c);  // ← BUG: (char)(-2) = U+FFFE = �
}

Key insight: (char)(-2) == (char)0xFFFE == � (replacement character)

Solution

✅ Fixed test logic to properly handle read timeout conditions:

while (timeoutCount < 1000) {
    c = reader.read(1);
    if (c == -1) {        // EOF
        break;
    } else if (c == -2) { // READ_EXPIRED (timeout)
        timeoutCount++;
        continue;         // Keep trying
    } else if (c >= 0) { // Valid character
        result.append((char) c);
    }
}

Tests Fixed

• ✅ MultiEncodingTerminalTest.testReadWithEncoding
• ✅ MultiEncodingTerminalTest.testMultipleEncodings
• ✅ Added comprehensive NonBlockingEncodingTest suite

Verification

• All encoding tests now pass consistently
• No changes needed to NonBlocking implementation
• Added comprehensive test suite to prevent regression
• Tests properly handle timeout conditions in CI environments

Impact

• ✅ Resolves intermittent Windows CI failures
• ✅ No breaking changes to existing code
• ✅ Improved test robustness for timeout scenarios
• ✅ Added valuable test coverage for encoding edge cases

Pull Request opened by Augment Code with guidance from the PR author