verify_build.py
Rama main · réplica de lectura del repositorio original.
"""v1.0.0.6 打包产物验证脚本.
校验三件事：
1. exe + _internal 关键文件存在且非空
2. 关键源码模块**确实被打进了包**（见下方"为什么查 PYZ 而非 .py 文件"）
3. 启动 exe --no-gui 模式，HTTP 探 /chat 与 /api/version/check
关于第 2 项——为什么不能直接比对 .py 文件的 hash：
PyInstaller 6.x 默认把 Python 模块**编译进 PYZ 归档**（exe 内的 zlib 压缩包），
`_internal/` 下根本不会有 `scout/**/*.py`。早期版本此脚本按"包内应存在同名
.py 且 hash 一致"校验，在 6.x 上恒定报 FAILED（3 条 packaged missing），
属于**脚本过时**而非打包缺陷——但它每次都会误导排查。
现在改为：查 PYZ 归档的 TOC 里有没有该模块。非纯 Python 的资源文件
（如 VERSION）仍走文件 hash 比对。
用法:
python verify_build.py
"""
import hashlib
import os
import socket
import subprocess
import sys
import time
import urllib.request
ROOT = r"D:\gongniu_project\scout-agent"
DIST = os.path.join(ROOT, "dist", "ScoutDesktop")
EXE = os.path.join(DIST, "ScoutAgent.exe")
INTERNAL = os.path.join(DIST, "_internal")
# 关键模块 → 必须出现在包内（模块名用点分，PYZ TOC 查这个）
# 覆盖本版修复的关键路径
PYZ_CHECKS = [
"scout.tools.builtin.memory", # MemorySaveTool 写入前查重
"scout.memory.store", # add_async 的 fact_key 透传
"scout.context.fact_key_registry", # 缺则运行时 ImportError
"scout.core.platform", # 探测/执行同源（2026-10-06）
"scout.engine.self_heal", # decode_output 替换裸 decode
"scout.context.skills", # frontmatter 嵌套解析
# pycryptodome（企业微信加解密）—— 纯 Python 包，同样进 PYZ 而非 _internal。
# 1.0.0.6 曾因缺包+spec 未收集丢掉整个 Crypto（1.0.0.5 有 42 文件），用户
# 用到企业微信功能会 ImportError，故列为必检项。
"Crypto.Cipher.AES",
]
# 关键非 Python 资源 → 包内文件路径（走 hash 比对）
HASH_CHECKS = [
(r"VERSION", r"_internal\VERSION"),
]
# PYZ 里最容易漏收的 GUI 依赖（2026-10-06 的 GUI 降级事故）
_NATIVE_CHECKS = [
(r"_internal\Microsoft.Web.WebView2.Core.dll", "WebView2 核心程序集"),
(r"_internal\Microsoft.Web.WebView2.WinForms.dll", "WebView2 WinForms 程序集"),
(r"_internal\WebView2Loader.dll", "WebView2 native 加载器"),
(r"_internal\pythonnet\runtime\Python.Runtime.dll", "pythonnet 运行时"),
]
def sha256(path):
h = hashlib.sha256()
with open(path, "rb") as f:
for chunk in iter(lambda: f.read(1 << 16), b""):
h.update(chunk)
return h.hexdigest()
def pyz_modules() -> set:
"""从 exe 的 PYZ 归档里读出所有模块名（判断模块是否真被打进包）.
PyInstaller 6.x 把 Python 模块编译进 exe 内追加的 PYZ 归档（zlib压缩），
`_internal/` 下不会有 .py 文件。用官方 archive reader 读，不自己逆格式
（手写 marshal 解析 PYZ 的 TOC 会在跨版本时炸 —— 试过，报 bad marshal data）。
流程：exe (CArchive) → 抽出 PYZ.pyz 条目 → 当作 ZlibArchive 读其 TOC。
"""
import tempfile
from PyInstaller.archive.readers import CArchiveReader, ZlibArchiveReader
outer = CArchiveReader(EXE)
pyz_bytes = outer.extract("PYZ.pyz")
tmp = None
try:
fd, tmp = tempfile.mkstemp(suffix=".pyz")
with os.fdopen(fd, "wb") as f:
f.write(pyz_bytes)
return set(ZlibArchiveReader(tmp).toc)
finally:
if tmp and os.path.isfile(tmp):
os.remove(tmp)
def free_port():
s = socket.socket()
s.bind(("127.0.0.1", 0))
p = s.getsockname()[1]
s.close()
return p
def main():
failures = []
# 1. 产物存在性
print("=== 1. artifact existence ===")
if not os.path.isfile(EXE):
failures.append(f"missing exe: {EXE}")
print(f"[FAIL] missing exe: {EXE}")
else:
sz = os.path.getsize(EXE)
print(f"[ok] exe size = {sz:,} bytes")
if not os.path.isdir(INTERNAL):
failures.append(f"missing _internal dir")
print(f"[FAIL] missing _internal")
else:
n = sum(len(files) for _, _, files in os.walk(INTERNAL))
print(f"[ok] _internal file count = {n}")
# 2. 模块确实打进包（PYZ TOC）+ 资源文件 hash
print("\n=== 2. packaged modules (PYZ TOC) ===")
mods = pyz_modules()
print(f"[info] PYZ module count = {len(mods)}")
for mod in PYZ_CHECKS:
if mod in mods:
print(f"[ok] in PYZ: {mod}")
else:
failures.append(f"not in PYZ: {mod}")
print(f"[FAIL] not in PYZ: {mod}")
print("\n=== 2b. native/GUI runtime deps ===")
for rel, desc in _NATIVE_CHECKS:
p = os.path.join(DIST, rel)
if os.path.isfile(p):
print(f"[ok] {desc}: {os.path.getsize(p):,} bytes")
else:
failures.append(f"missing native dep: {rel}")
print(f"[FAIL] {desc} missing: {rel}")
print("\n=== 2c. resource hash ===")
for src_rel, dst_rel in HASH_CHECKS:
src = os.path.join(ROOT, src_rel)
dst = os.path.join(DIST, dst_rel)
if not os.path.isfile(src):
failures.append(f"source missing: {src_rel}")
print(f"[FAIL] source missing: {src_rel}")
continue
if not os.path.isfile(dst):
failures.append(f"packaged missing: {dst_rel}")
print(f"[FAIL] packaged missing: {dst_rel}")
continue
h_src = sha256(src)
h_dst = sha256(dst)
match = h_src == h_dst
tag = "ok" if match else "FAIL"
if not match:
failures.append(f"hash mismatch: {src_rel}")
print(f"[{tag}] {src_rel} src={h_src[:12]} dst={h_dst[:12]}")
# 3. 启动冒烟
print("\n=== 3. startup smoke (--no-gui) ===")
port = free_port()
proc = subprocess.Popen(
[EXE, "--no-gui", "--host", "127.0.0.1", "--port", str(port)],
stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
cwd=DIST,
)
try:
ok = False
for _ in range(20): # 最多等 20s
time.sleep(1)
try:
with urllib.request.urlopen(
f"http://127.0.0.1:{port}/chat", timeout=2
) as r:
if r.status == 200:
print(f"[ok] /chat returned 200 (port {port})")
ok = True
break
except Exception:
continue
if not ok:
failures.append("exe did not respond on /chat within 20s")
print("[FAIL] exe did not respond on /chat within 20s")
else:
# 版本号校验
try:
with urllib.request.urlopen(
f"http://127.0.0.1:{port}/api/version/check", timeout=3
) as r:
import json
data = json.loads(r.read().decode())
v = data.get("current_version") or data.get("version")
print(f"[ok] /api/version/check current_version = {v}")
if v != "1.0.0.6":
failures.append(f"version mismatch: got {v}, want 1.0.0.6")
print(f"[FAIL] version mismatch: got {v}, want 1.0.0.6")
except Exception as e:
failures.append(f"version check failed: {e}")
print(f"[FAIL] version check failed: {e}")
finally:
proc.terminate()
try:
proc.wait(timeout=5)
except subprocess.TimeoutExpired:
proc.kill()
# 总结
print("\n=== summary ===")
if failures:
print(f"FAILED ({len(failures)} issues):")
for f in failures:
print(f" - {f}")
sys.exit(1)
else:
print("ALL CHECKS PASSED")
if __name__ == "__main__":
main()