Bug: Win32 directory picker truncates paths containing certain Unicode characters (e.g. 一 U+4E00)

[applekaraa]

Bug Description

When selecting a workspace directory on Windows, paths containing Unicode characters whose UTF-16LE low byte is 0x00 are truncated, causing ENOENT errors.

For example, the directory 去重的单一装置数据 is truncated to 去重的单, and selecting it fails with:
workspace-invalid-path: cannot create a workspace at "E:\Work\直流油色谱分析\导出全量7月数据\去重的单": ENOENT: no such file or directory

Root Cause

The readUtf16() function in packages/host/directory-picker-native/src/win32-dialog-bindings.ts (line 40) scans for the NUL terminator by only checking the low byte:

while (end + 1 < bytes.length && bytes[end] !== 0) end += 2

In UTF-16LE, the NUL terminator is two consecutive zero bytes (00 00). But characters like 一 (U+4E00, encoded 00 4E) have a low byte of 0x00, causing the scanner to falsely detect the string end.

Affected characters include: 一(U+4E00), 丁(U+4E01), 七(U+4E03), 万(U+4E07), and many more CJK characters in the U+xxxx range where xx is 00.

Fix
Check that both bytes are zero:
while (end + 1 < bytes.length && !(bytes[end] === 0 && bytes[end + 1] === 0)) end += 2

Environment
OS: Windows 10
dsh version: 0.1.0-rc.6

<img width="2469" height="99" alt="ScreenShot_2026-08-14_105445_844" src="https://github.com/user-attachments/assets/179fe79b-18f9-4411-9d4e-73b9c7502bb5" />