Skip to content

利用 AI 大模型和自动化工作流,根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

fix(webui): strip control and bidi characters from download names

Rudy Celekli33dde166da7f
diff --git a/test/services/test_webui_download_name_control_characters.py b/test/services/test_webui_download_name_control_characters.py
new file mode 100644
index 00000000..eb17ddbb
--- /dev/null
+++ b/test/services/test_webui_download_name_control_characters.py
@@ -0,0 +1,97 @@
+"""任务历史下载文件名必须移除控制符与双向文本控制符。"""
+import ast
+import os
+import re
+from collections.abc import Mapping
+from pathlib import Path
+
+
+ROOT_DIR = Path(__file__).parent.parent.parent
+WEBUI_MAIN = ROOT_DIR / "webui" / "Main.py"
+DOWNLOAD_NAME_HELPERS = {"_build_video_download_name"}
+DOWNLOAD_NAME_CONSTANTS = {
+ "_DOWNLOAD_FILENAME_INVALID_PATTERN",
+ "_WINDOWS_RESERVED_FILENAMES",
+}
+
+
+def _load_download_name_helpers():
+ """
+ 从 WebUI 入口中隔离加载下载文件名相关的纯函数。
+
+ 直接导入 Main.py 会执行整套页面渲染。这里只编译目标常量和函数,执行的仍是
+ 合并后的真实实现,因此断言覆盖的是生产行为而不是测试里的副本。
+ """
+ tree = ast.parse(WEBUI_MAIN.read_text(encoding="utf-8"))
+ selected_nodes = []
+ for node in tree.body:
+ if isinstance(node, ast.Assign) and any(
+ isinstance(target, ast.Name) and target.id in DOWNLOAD_NAME_CONSTANTS
+ for target in node.targets
+ ):
+ selected_nodes.append(node)
+ elif isinstance(node, ast.FunctionDef) and node.name in DOWNLOAD_NAME_HELPERS:
+ selected_nodes.append(node)
+
+ namespace = {"os": os, "re": re, "Mapping": Mapping}
+ module = ast.fix_missing_locations(ast.Module(body=selected_nodes, type_ignores=[]))
+ exec(compile(module, str(WEBUI_MAIN), "exec"), namespace)
+ return namespace
+
+
+build_video_download_name = _load_download_name_helpers()[
+ "_build_video_download_name"
+]
+
+# 显式列出必须移除的字符,不复制生产代码的 range 推导:实现写错范围时断言会
+# 直接失败,而不是跟着实现一起漏测。C0 控制符与 U+0085、U+2028、U+2029 已被
+# 空白合并处理,这里仍然逐个确认,避免后续改动把它们重新放进文件名。
+C0_CONTROL_CHARACTERS = ("\x00", "\x08", "\x1b", "\x1f")
+C1_CONTROL_CHARACTERS = ("\x7f", "\x80", "\x85", "\x9f")
+BIDIRECTIONAL_CONTROL_CHARACTERS = (
+ "‎",
+ "‏",
+ "‪",
+ "‫",
+ "‬",
+ "‭",
+ "‮",
+ "⁦",
+ "⁧",
+ "⁨",
+ "⁩",
+)
+LINE_SEPARATOR_CHARACTERS = ("
", "
")
+
+
+def test_download_name_drops_bidirectional_override_from_subject():
+ """U+202E 会把后续字符反向渲染,伪造出另一个看起来可信的扩展名。"""
+ download_name = build_video_download_name("holiday‮4pm", 1, 1)
+
+ assert "‮" not in download_name
+ assert download_name == "holiday 4pm.mp4"
+
+
+def test_download_name_drops_every_unsafe_character_from_subject():
+ """上传入口拒绝的整组字符都不能进入下载文件名。"""
+ unsafe_characters = (
+ C0_CONTROL_CHARACTERS
+ + C1_CONTROL_CHARACTERS
+ + BIDIRECTIONAL_CONTROL_CHARACTERS
+ + LINE_SEPARATOR_CHARACTERS
+ )
+ for character in unsafe_characters:
+ download_name = build_video_download_name(f"trip{character}note", 1, 1)
+
+ assert character not in download_name, f"U+{ord(character):04X} survived"
+ assert download_name == "trip note.mp4"
+
+
+def test_download_name_keeps_ordinary_multilingual_subjects():
+ """移除控制符不能顺带改写正常主题,包括中文、重音字母和标点。"""
+ assert build_video_download_name("上海一日游", 1, 1) == "上海一日游.mp4"
+ assert build_video_download_name("Café déjà vu", 2, 3) == "Café déjà vu-2.mp4"
+ assert (
+ build_video_download_name("A day: in / Shanghai?", 2, 3)
+ == "A day in Shanghai-2.mp4"
+ )
diff --git a/webui/Main.py b/webui/Main.py
index c61c44af..188fd1ea 100644
--- a/webui/Main.py
+++ b/webui/Main.py
@@ -216,7 +216,20 @@ _FINAL_VIDEO_PATTERN = re.compile(
r"^final-(?P<index>\d+)\.(?P<extension>mp4|mov|mkv|webm)$",
re.IGNORECASE,
)
-_DOWNLOAD_FILENAME_INVALID_PATTERN = re.compile(r'[<>:"/\\|?*\x00-\x1f]')
+# 与 app/services/bgm.py、app/services/material_upload.py 的上传文件名规则共用
+# 同一份清单:除路径分隔符与 Win32 非法字符外,还要移除所有控制符和会改变显示
+# 形态的字符。此前只拦截 ord < 32(C0 控制符),漏掉了同一类问题的另一半:
+# * C1 控制符 U+007F-U+009F:会原样进入下载文件名、日志和界面。
+# * 双向文本控制符 U+200E/U+200F/U+202A-U+202E/U+2066-U+2069:U+202E 之后的
+# 字符会被反向渲染,``holiday<U+202E>4pm`` 看起来像另一个扩展名。
+# * U+2028/U+2029(Unicode 行/段分隔符)能在一行日志里制造视觉换行。
+# 视频主题是用户自由输入的文本,可以直接粘贴这些字符,因此必须在这里过滤。
+# 字符串不使用 r'' 前缀:``\\`` 表示要排除的反斜杠本身,``\x``/``\u`` 转义由
+# Python 解析,避免把转义文本写进字符集。
+_DOWNLOAD_FILENAME_INVALID_PATTERN = re.compile(
+ "[<>:\"/\\\\|?*\\x00-\\x1f\\x7f-\\x9f"
+ "\\u200e\\u200f\\u2028\\u2029\\u202a-\\u202e\\u2066-\\u2069]"
+)
_WINDOWS_RESERVED_FILENAMES = frozenset(
{"CON", "PRN", "AUX", "NUL"}
| {