fix: 解决打包后三个运行时问题

- rthook_paddle.py: stub paddle.utils.cpp_extension,避免Cython缺文件崩溃
- build_exe.py: 显式收集paddle DLLs(mklml.dll等)
- ocr_offline.py: 非ASCII路径自动复制模型到临时目录,绕过PaddlePaddle C++路径限制

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
let5sne.win10
2026-02-14 20:31:28 +08:00
parent dfbab1b61e
commit 86cb704eae
3 changed files with 78 additions and 2 deletions

View File

@@ -8,6 +8,7 @@
├── _internal/ (运行时依赖)
└── models/ (OCR 模型,需提前通过 prepare_models.py 准备)
"""
import os
import subprocess
import sys
import shutil
@@ -16,6 +17,10 @@ from pathlib import Path
PROJECT_ROOT = Path(__file__).parent
DIST_NAME = "信封信息提取系统"
# paddle DLLs 所在目录mklml.dll 等不会被 PyInstaller 自动收集)
import paddle as _paddle
PADDLE_LIBS = str(Path(_paddle.__file__).parent / "libs")
def build(debug=False):
"""使用 PyInstaller 打包onedir 模式)"""
@@ -39,6 +44,10 @@ def build(debug=False):
"--hidden-import=ocr_offline",
"--hidden-import=paddleocr",
"--hidden-import=paddle",
# --- paddle DLLsmklml.dll 等不会被自动收集) ---
f"--add-binary={PADDLE_LIBS}/*.dll{os.pathsep}paddle/libs",
# --- runtime hook: stub 掉 paddle 开发模块,避免 Cython 缺文件崩溃 ---
"--runtime-hook=rthook_paddle.py",
# --- 收集 paddleocr 全部数据(模型配置、字典等) ---
"--collect-all=paddleocr",
# --- 元数据(部分库在运行时通过 importlib.metadata 查版本) ---