dsh-zju-lecture-note 2.2.6 → 2.2.7
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +70 -1
- package/README.md +37 -31
- package/SECURITY.md +101 -36
- package/docs/technical.md +37 -24
- package/lib/bridge.js +278 -79
- package/lib/client.js +907 -758
- package/lib/index.js +232 -61
- package/lib/tools.js +26 -1
- package/lib/tools.schema.json +21 -1
- package/mcp_server.py +102 -24
- package/package.json +1 -1
- package/skills/scholia-notes/references/prelude.typ +9 -1
- package/skills/scholia-notes.manifest.json +3 -3
- package/skills/zju-lecture-summarizer/SKILL.md +1 -1
- package/skills/zju-lecture-summarizer/SLIDES.md +5 -3
- package/skills/zju-lecture-summarizer-help/SKILL.md +2 -2
- package/skills/zju-lecture-summarizer-help.manifest.json +4 -4
- package/skills/zju-lecture-summarizer-login/SKILL.md +1 -1
- package/skills/zju-lecture-summarizer-login.manifest.json +4 -4
- package/skills/zju-lecture-summarizer.manifest.json +6 -6
- package/zju_lecture/browser_login.py +17 -6
- package/zju_lecture/cli.py +26 -0
- package/zju_lecture/client.py +95 -18
- package/zju_lecture/courseware.py +37 -2
- package/zju_lecture/deps.py +98 -17
- package/zju_lecture/note_lint.py +34 -10
- package/zju_lecture/paths.py +57 -2
- package/zju_lecture/prelude.typ +9 -1
- package/zju_lecture/slides.py +73 -35
- package/zju_lecture/study_notes.py +36 -4
package/zju_lecture/client.py
CHANGED
|
@@ -5,6 +5,7 @@ Pure Python standard library implementation with zero external dependencies.
|
|
|
5
5
|
"""
|
|
6
6
|
|
|
7
7
|
import http.client
|
|
8
|
+
import ipaddress
|
|
8
9
|
import json
|
|
9
10
|
import os
|
|
10
11
|
import random
|
|
@@ -15,7 +16,7 @@ import time
|
|
|
15
16
|
import urllib.error
|
|
16
17
|
import urllib.parse
|
|
17
18
|
import urllib.request
|
|
18
|
-
from typing import Any, Dict, List, Optional
|
|
19
|
+
from typing import Any, Dict, List, Optional, Tuple
|
|
19
20
|
|
|
20
21
|
from .security import REDACTOR, sanitize_cookie_header
|
|
21
22
|
|
|
@@ -210,6 +211,7 @@ def _fetch_bytes(
|
|
|
210
211
|
attempts: int = _RETRY_ATTEMPTS,
|
|
211
212
|
max_bytes: Optional[int] = None,
|
|
212
213
|
allow_cross_host_redirect: bool = False,
|
|
214
|
+
diagnostics: Optional[List[str]] = None,
|
|
213
215
|
) -> bytes:
|
|
214
216
|
"""GET ``url`` and read the **whole** body, retrying transient failures.
|
|
215
217
|
|
|
@@ -229,7 +231,11 @@ def _fetch_bytes(
|
|
|
229
231
|
try:
|
|
230
232
|
if allow_cross_host_redirect:
|
|
231
233
|
context = _open_request(
|
|
232
|
-
url,
|
|
234
|
+
url,
|
|
235
|
+
headers,
|
|
236
|
+
timeout,
|
|
237
|
+
allow_cross_host_redirect=True,
|
|
238
|
+
diagnostics=diagnostics,
|
|
233
239
|
)
|
|
234
240
|
else:
|
|
235
241
|
context = _open_request(url, headers, timeout)
|
|
@@ -252,22 +258,77 @@ def _fetch_bytes(
|
|
|
252
258
|
raise last
|
|
253
259
|
|
|
254
260
|
|
|
261
|
+
_DEFAULT_PORTS = {"http": 80, "https": 443}
|
|
262
|
+
|
|
263
|
+
|
|
264
|
+
def _origin(url: str) -> Tuple[str, str, int]:
|
|
265
|
+
"""``(scheme, host, port)`` with default ports filled in.
|
|
266
|
+
|
|
267
|
+
hostname+scheme is not enough: ``https://a:443`` and ``https://a:8443`` are
|
|
268
|
+
different origins, and credentials must not cross that boundary.
|
|
269
|
+
"""
|
|
270
|
+
parsed = urllib.parse.urlparse(url or "")
|
|
271
|
+
scheme = (parsed.scheme or "").lower()
|
|
272
|
+
host = (parsed.hostname or "").lower().rstrip(".")
|
|
273
|
+
try:
|
|
274
|
+
port = parsed.port
|
|
275
|
+
except ValueError:
|
|
276
|
+
port = None
|
|
277
|
+
if port is None:
|
|
278
|
+
port = _DEFAULT_PORTS.get(scheme, 0)
|
|
279
|
+
return scheme, host, port
|
|
280
|
+
|
|
281
|
+
|
|
282
|
+
def _is_forbidden_redirect_host(host: str) -> bool:
|
|
283
|
+
"""IP literals, loopback and private names are never valid redirect targets."""
|
|
284
|
+
value = (host or "").strip().lower().rstrip(".")
|
|
285
|
+
if not value:
|
|
286
|
+
return True
|
|
287
|
+
if value == "localhost" or value.endswith(".localhost"):
|
|
288
|
+
return True
|
|
289
|
+
try:
|
|
290
|
+
ipaddress.ip_address(value)
|
|
291
|
+
except ValueError:
|
|
292
|
+
return False
|
|
293
|
+
# Any IP literal (loopback, link-local, RFC1918, ::1, ...) is refused:
|
|
294
|
+
# a media CDN never needs one, and a redirect to one is the classic SSRF.
|
|
295
|
+
return True
|
|
296
|
+
|
|
297
|
+
|
|
298
|
+
def _redirect_block_reason(old_url: str, new_url: str) -> str:
|
|
299
|
+
"""Why a redirect must not be followed, or an empty string to allow it."""
|
|
300
|
+
old_scheme, _, _ = _origin(old_url)
|
|
301
|
+
new_scheme, new_host, new_port = _origin(new_url)
|
|
302
|
+
parsed = urllib.parse.urlparse(new_url or "")
|
|
303
|
+
if parsed.username or parsed.password:
|
|
304
|
+
return "已阻止跳转到含凭据的 URL(保护登录态)"
|
|
305
|
+
if old_scheme == "https" and new_scheme != "https":
|
|
306
|
+
return "已阻止 HTTPS 降级重定向(保护登录凭据)"
|
|
307
|
+
if new_scheme not in ("http", "https"):
|
|
308
|
+
return f"已阻止跳转到非 HTTP(S) 地址: {new_scheme or '(无协议)'}"
|
|
309
|
+
if _is_forbidden_redirect_host(new_host):
|
|
310
|
+
return f"已阻止跳转到受限主机: {new_host or '(未知主机)'}"
|
|
311
|
+
return ""
|
|
312
|
+
|
|
313
|
+
|
|
255
314
|
class _SameHostRedirectHandler(urllib.request.HTTPRedirectHandler):
|
|
256
|
-
"""Follow redirects only on the classroom
|
|
315
|
+
"""Follow redirects only on the classroom **origin**.
|
|
257
316
|
|
|
258
317
|
``urllib`` re-sends request headers on a redirect, so a compromised or
|
|
259
318
|
misconfigured endpoint that answers ``302`` to another origin would leak
|
|
260
|
-
the ``Cookie`` header.
|
|
319
|
+
the ``Cookie`` header. Scheme, host and port must all match — a downgrade
|
|
320
|
+
to HTTP or a different port is refused as well.
|
|
261
321
|
"""
|
|
262
322
|
|
|
263
323
|
def redirect_request(self, req, fp, code, msg, headers, newurl): # noqa: D401
|
|
264
|
-
|
|
265
|
-
|
|
266
|
-
|
|
324
|
+
reason = _redirect_block_reason(req.full_url, newurl)
|
|
325
|
+
if not reason and _origin(newurl) != _origin(BASE_URL):
|
|
326
|
+
reason = f"拒绝跨域重定向到 {_origin(newurl)[1] or '(未知主机)'}(保护登录凭据)"
|
|
327
|
+
if reason:
|
|
267
328
|
raise urllib.error.HTTPError(
|
|
268
329
|
req.full_url,
|
|
269
330
|
code,
|
|
270
|
-
|
|
331
|
+
reason,
|
|
271
332
|
headers,
|
|
272
333
|
fp,
|
|
273
334
|
)
|
|
@@ -277,21 +338,36 @@ class _SameHostRedirectHandler(urllib.request.HTTPRedirectHandler):
|
|
|
277
338
|
class _StripAuthOnCrossHostRedirectHandler(urllib.request.HTTPRedirectHandler):
|
|
278
339
|
"""Allow slide-CDN redirects, but never carry credentials across origins.
|
|
279
340
|
|
|
280
|
-
Slide screenshots are static files and may redirect to a media host.
|
|
281
|
-
|
|
282
|
-
for
|
|
283
|
-
the redirect changes
|
|
341
|
+
Slide screenshots are static files and may redirect to a media host. Only
|
|
342
|
+
``*.zju.edu.cn`` targets are followed (one hop at a time, re-validated by
|
|
343
|
+
urllib for every hop); everything else is blocked and recorded. Whenever
|
|
344
|
+
the redirect changes scheme/host/port, ``Cookie`` / ``Authorization`` /
|
|
345
|
+
``Proxy-Authorization`` are dropped.
|
|
284
346
|
"""
|
|
285
347
|
|
|
348
|
+
def __init__(self, diagnostics: Optional[List[str]] = None) -> None:
|
|
349
|
+
super().__init__()
|
|
350
|
+
self.diagnostics = diagnostics
|
|
351
|
+
|
|
352
|
+
def _blocked(self, reason: str) -> None:
|
|
353
|
+
if self.diagnostics is not None:
|
|
354
|
+
self.diagnostics.append(reason)
|
|
355
|
+
|
|
286
356
|
def redirect_request(self, req, fp, code, msg, headers, newurl): # noqa: D401
|
|
357
|
+
reason = _redirect_block_reason(req.full_url, newurl)
|
|
358
|
+
if not reason:
|
|
359
|
+
_, new_host, _ = _origin(newurl)
|
|
360
|
+
if not is_zju_host(new_host):
|
|
361
|
+
reason = f"已阻止跳转到非浙大域: {new_host or '(未知主机)'}"
|
|
362
|
+
if reason:
|
|
363
|
+
self._blocked(reason)
|
|
364
|
+
return None
|
|
287
365
|
new_request = super().redirect_request(req, fp, code, msg, headers, newurl)
|
|
288
366
|
if new_request is None:
|
|
289
367
|
return None
|
|
290
|
-
|
|
291
|
-
|
|
292
|
-
|
|
293
|
-
new_request.remove_header("Cookie")
|
|
294
|
-
new_request.remove_header("Authorization")
|
|
368
|
+
if _origin(req.full_url) != _origin(newurl):
|
|
369
|
+
for header in ("Cookie", "Authorization", "Proxy-Authorization"):
|
|
370
|
+
new_request.remove_header(header)
|
|
295
371
|
return new_request
|
|
296
372
|
|
|
297
373
|
|
|
@@ -300,10 +376,11 @@ def _open_request(
|
|
|
300
376
|
headers: Dict[str, str],
|
|
301
377
|
timeout: int,
|
|
302
378
|
allow_cross_host_redirect: bool = False,
|
|
379
|
+
diagnostics: Optional[List[str]] = None,
|
|
303
380
|
):
|
|
304
381
|
"""Open ``url`` with certificate verification and a redirect policy."""
|
|
305
382
|
redirect_handler = (
|
|
306
|
-
_StripAuthOnCrossHostRedirectHandler()
|
|
383
|
+
_StripAuthOnCrossHostRedirectHandler(diagnostics)
|
|
307
384
|
if allow_cross_host_redirect
|
|
308
385
|
else _SameHostRedirectHandler()
|
|
309
386
|
)
|
|
@@ -19,6 +19,37 @@ class CoursewareError(Exception):
|
|
|
19
19
|
pass
|
|
20
20
|
|
|
21
21
|
|
|
22
|
+
#: 课件是 zip 容器:解压前先按声明的 file_size 累计,避免 zip bomb。
|
|
23
|
+
MAX_ARCHIVE_TOTAL_BYTES = 64 * 1024 * 1024
|
|
24
|
+
MAX_ARCHIVE_MEMBER_BYTES = 16 * 1024 * 1024
|
|
25
|
+
|
|
26
|
+
|
|
27
|
+
def _validate_zip_members(archive: zipfile.ZipFile) -> None:
|
|
28
|
+
"""Reject an archive whose declared uncompressed size exceeds the budget."""
|
|
29
|
+
total = 0
|
|
30
|
+
for info in archive.infolist():
|
|
31
|
+
if info.file_size > MAX_ARCHIVE_MEMBER_BYTES:
|
|
32
|
+
raise CoursewareError(
|
|
33
|
+
f"课件包含超过 {MAX_ARCHIVE_MEMBER_BYTES // (1024 * 1024)} MiB "
|
|
34
|
+
f"的成员({info.filename}),已拒绝解压"
|
|
35
|
+
)
|
|
36
|
+
total += info.file_size
|
|
37
|
+
if total > MAX_ARCHIVE_TOTAL_BYTES:
|
|
38
|
+
raise CoursewareError(
|
|
39
|
+
f"课件解压后超过 {MAX_ARCHIVE_TOTAL_BYTES // (1024 * 1024)} MiB "
|
|
40
|
+
"上限,已拒绝解压"
|
|
41
|
+
)
|
|
42
|
+
|
|
43
|
+
|
|
44
|
+
def _read_zip_member(archive: zipfile.ZipFile, name: str) -> bytes:
|
|
45
|
+
info = archive.getinfo(name)
|
|
46
|
+
if info.file_size > MAX_ARCHIVE_MEMBER_BYTES:
|
|
47
|
+
raise CoursewareError(
|
|
48
|
+
f"课件成员 {name} 超过 {MAX_ARCHIVE_MEMBER_BYTES // (1024 * 1024)} MiB 上限"
|
|
49
|
+
)
|
|
50
|
+
return archive.read(info)
|
|
51
|
+
|
|
52
|
+
|
|
22
53
|
def extract_courseware_text(path: str, max_chars: Optional[int] = None) -> str:
|
|
23
54
|
"""Extract plain text from local courseware files.
|
|
24
55
|
|
|
@@ -73,6 +104,7 @@ def _extract_pptx(path: str) -> str:
|
|
|
73
104
|
"""Extract text from slides inside PPTX zip package."""
|
|
74
105
|
try:
|
|
75
106
|
with zipfile.ZipFile(path, "r") as z:
|
|
107
|
+
_validate_zip_members(z)
|
|
76
108
|
slide_names = [
|
|
77
109
|
n for n in z.namelist()
|
|
78
110
|
if re.match(r"^ppt/slides/slide\d+\.xml$", n)
|
|
@@ -84,7 +116,7 @@ def _extract_pptx(path: str) -> str:
|
|
|
84
116
|
|
|
85
117
|
parts = []
|
|
86
118
|
for idx, name in enumerate(slide_names, start=1):
|
|
87
|
-
xml_content = z
|
|
119
|
+
xml_content = _read_zip_member(z, name).decode("utf-8", "ignore")
|
|
88
120
|
texts = re.findall(r"<a:t>([^<]*)</a:t>", xml_content)
|
|
89
121
|
cleaned = " ".join(t.strip() for t in texts if t.strip())
|
|
90
122
|
if cleaned:
|
|
@@ -115,7 +147,10 @@ def _extract_docx(path: str) -> str:
|
|
|
115
147
|
"""
|
|
116
148
|
try:
|
|
117
149
|
with zipfile.ZipFile(path, "r") as z:
|
|
118
|
-
|
|
150
|
+
_validate_zip_members(z)
|
|
151
|
+
xml_content = _read_zip_member(z, "word/document.xml").decode(
|
|
152
|
+
"utf-8", "ignore"
|
|
153
|
+
)
|
|
119
154
|
except Exception as e:
|
|
120
155
|
raise CoursewareError(f"读取 DOCX 课件失败: {e}")
|
|
121
156
|
|
package/zju_lecture/deps.py
CHANGED
|
@@ -26,16 +26,20 @@ import hashlib
|
|
|
26
26
|
import json
|
|
27
27
|
import os
|
|
28
28
|
import platform
|
|
29
|
+
import re
|
|
29
30
|
import shutil
|
|
30
31
|
import subprocess
|
|
31
32
|
import sys
|
|
32
33
|
import tarfile
|
|
33
34
|
import tempfile
|
|
34
35
|
import time
|
|
36
|
+
import urllib.parse
|
|
35
37
|
import zipfile
|
|
36
38
|
from dataclasses import dataclass, field
|
|
37
39
|
from typing import Any, Dict, List, Optional, Tuple
|
|
38
40
|
|
|
41
|
+
from .security import REDACTOR
|
|
42
|
+
|
|
39
43
|
TYPST_ENV = "ZJU_TYPST_BIN"
|
|
40
44
|
AUTO_INSTALL_ENV = "ZJU_TYPST_AUTO_INSTALL"
|
|
41
45
|
MIRROR_ENV = "ZJU_TYPST_MIRROR"
|
|
@@ -45,6 +49,17 @@ MIN_VERSION = (0, 13, 0)
|
|
|
45
49
|
RELEASE_URL = "https://github.com/typst/typst/releases/download/v{version}/{file}"
|
|
46
50
|
CACHE_APP = "dsh-zju-lecture"
|
|
47
51
|
|
|
52
|
+
#: Typst 归档实测约 20–30 MB;256 MiB 是硬上限,异常大响应直接中断。
|
|
53
|
+
MAX_DOWNLOAD_BYTES = 256 * 1024 * 1024
|
|
54
|
+
|
|
55
|
+
|
|
56
|
+
def _size_text(byte_count: int) -> str:
|
|
57
|
+
if byte_count >= 1024 * 1024:
|
|
58
|
+
return f"{byte_count // (1024 * 1024)} MiB"
|
|
59
|
+
if byte_count >= 1024:
|
|
60
|
+
return f"{byte_count // 1024} KiB"
|
|
61
|
+
return f"{byte_count} 字节"
|
|
62
|
+
|
|
48
63
|
HERE = os.path.dirname(os.path.abspath(__file__))
|
|
49
64
|
MANIFEST_PATH = os.path.join(HERE, "typst_assets.json")
|
|
50
65
|
PACKAGE_PATH = os.path.join(HERE, "typst-packages")
|
|
@@ -261,10 +276,46 @@ def probe(auto_install: bool = False) -> TypstInfo:
|
|
|
261
276
|
# ---------------------------------------------------------------- 安装
|
|
262
277
|
|
|
263
278
|
|
|
279
|
+
_URL_RE = re.compile(r"https?://[^\s]+", re.IGNORECASE)
|
|
280
|
+
|
|
281
|
+
|
|
282
|
+
def _redact_url(url: str) -> str:
|
|
283
|
+
"""只保留 ``scheme://host[:port]/path``,去掉 userinfo / query / fragment。"""
|
|
284
|
+
text = str(url or "").strip()
|
|
285
|
+
if not text:
|
|
286
|
+
return ""
|
|
287
|
+
try:
|
|
288
|
+
parsed = urllib.parse.urlparse(text)
|
|
289
|
+
except ValueError:
|
|
290
|
+
return "[REDACTED]"
|
|
291
|
+
if not parsed.scheme or not parsed.netloc:
|
|
292
|
+
return REDACTOR.redact(text)
|
|
293
|
+
host = parsed.hostname or ""
|
|
294
|
+
netloc = f"[{host}]" if ":" in host and not host.startswith("[") else host
|
|
295
|
+
try:
|
|
296
|
+
port = parsed.port
|
|
297
|
+
except ValueError:
|
|
298
|
+
port = None
|
|
299
|
+
if port:
|
|
300
|
+
netloc = f"{netloc}:{port}"
|
|
301
|
+
return REDACTOR.redact(f"{parsed.scheme}://{netloc}{parsed.path}")
|
|
302
|
+
|
|
303
|
+
|
|
304
|
+
def _redact_text(message: object) -> str:
|
|
305
|
+
"""脱敏任意诊断文本:先洗掉 URL 里的凭据,再交给进程级 redactor。"""
|
|
306
|
+
text = str(message)
|
|
307
|
+
text = _URL_RE.sub(lambda match: _redact_url(match.group(0)), text)
|
|
308
|
+
return REDACTOR.redact(text)
|
|
309
|
+
|
|
310
|
+
|
|
264
311
|
def _progress(message: str) -> None:
|
|
265
|
-
"""进度只写 stderr:工具的返回值要保持可读,不混入下载日志。
|
|
312
|
+
"""进度只写 stderr:工具的返回值要保持可读,不混入下载日志。
|
|
313
|
+
|
|
314
|
+
镜像 URL 可能带 userinfo 或签名 query;进入 stderr 前必须先脱敏,
|
|
315
|
+
否则宿主的日志/会话记录会替用户保存一份凭据。
|
|
316
|
+
"""
|
|
266
317
|
try:
|
|
267
|
-
sys.stderr.write(message + "\n")
|
|
318
|
+
sys.stderr.write(_redact_text(message) + "\n")
|
|
268
319
|
sys.stderr.flush()
|
|
269
320
|
except Exception: # noqa: BLE001 - 日志永远不能影响主流程
|
|
270
321
|
pass
|
|
@@ -298,14 +349,24 @@ def _download(url: str, target: str, retries: int = 3) -> None:
|
|
|
298
349
|
)
|
|
299
350
|
with urllib.request.urlopen(request, timeout=180, context=context) as response:
|
|
300
351
|
total = int(response.headers.get("Content-Length") or 0)
|
|
352
|
+
if total > MAX_DOWNLOAD_BYTES:
|
|
353
|
+
raise RuntimeError(
|
|
354
|
+
f"下载内容 {total} 字节,超过 "
|
|
355
|
+
f"{_size_text(MAX_DOWNLOAD_BYTES)} 上限"
|
|
356
|
+
)
|
|
301
357
|
done = 0
|
|
302
358
|
with open(target, "wb") as handle:
|
|
303
359
|
while True:
|
|
304
360
|
chunk = response.read(1 << 20)
|
|
305
361
|
if not chunk:
|
|
306
362
|
break
|
|
307
|
-
handle.write(chunk)
|
|
308
363
|
done += len(chunk)
|
|
364
|
+
if done > MAX_DOWNLOAD_BYTES:
|
|
365
|
+
raise RuntimeError(
|
|
366
|
+
f"下载超过 {_size_text(MAX_DOWNLOAD_BYTES)} "
|
|
367
|
+
"上限,已中断"
|
|
368
|
+
)
|
|
369
|
+
handle.write(chunk)
|
|
309
370
|
if total and done % (5 << 20) < (1 << 20):
|
|
310
371
|
_progress(
|
|
311
372
|
f" 下载 Typst {done * 100 // max(total, 1)}% "
|
|
@@ -314,26 +375,46 @@ def _download(url: str, target: str, retries: int = 3) -> None:
|
|
|
314
375
|
return
|
|
315
376
|
except Exception as exc: # noqa: BLE001 - 网络抖动要重试
|
|
316
377
|
last = exc
|
|
378
|
+
try:
|
|
379
|
+
os.remove(target)
|
|
380
|
+
except OSError:
|
|
381
|
+
pass
|
|
317
382
|
_progress(f" 下载失败(第 {attempt}/{retries} 次):{exc}")
|
|
318
383
|
if attempt < retries:
|
|
319
384
|
time.sleep(2 * attempt)
|
|
320
|
-
raise RuntimeError(
|
|
321
|
-
|
|
385
|
+
raise RuntimeError(_redact_text(last) if last else "下载失败")
|
|
386
|
+
|
|
387
|
+
|
|
388
|
+
def _safe_members(
|
|
389
|
+
archive: zipfile.ZipFile | tarfile.TarFile,
|
|
390
|
+
destination: str = "",
|
|
391
|
+
) -> List[str]:
|
|
392
|
+
"""拒绝绝对路径、``..``、符号链接/硬链接与越界解压目标。"""
|
|
393
|
+
root = os.path.abspath(destination) if destination else ""
|
|
394
|
+
|
|
395
|
+
def check(name: str) -> None:
|
|
396
|
+
normalized = os.path.normpath(name)
|
|
397
|
+
if normalized.startswith("..") or os.path.isabs(normalized):
|
|
398
|
+
raise ValueError(f"归档包含可疑路径:{name}")
|
|
399
|
+
if not root:
|
|
400
|
+
return
|
|
401
|
+
try:
|
|
402
|
+
target = os.path.abspath(os.path.join(root, normalized))
|
|
403
|
+
if os.path.commonpath([root, target]) != root:
|
|
404
|
+
raise ValueError(f"归档成员会解压到目录之外:{name}")
|
|
405
|
+
except ValueError as exc:
|
|
406
|
+
raise ValueError(f"归档成员路径不合法:{name}") from exc
|
|
322
407
|
|
|
323
|
-
def _safe_members(archive: zipfile.ZipFile | tarfile.TarFile) -> List[str]:
|
|
324
|
-
"""拒绝绝对路径与 ``..``:解压不能让归档写到缓存目录之外。"""
|
|
325
408
|
names: List[str] = []
|
|
326
409
|
if isinstance(archive, zipfile.ZipFile):
|
|
327
410
|
for name in archive.namelist():
|
|
328
|
-
|
|
329
|
-
if normalized.startswith("..") or os.path.isabs(normalized):
|
|
330
|
-
raise ValueError(f"归档包含可疑路径:{name}")
|
|
411
|
+
check(name)
|
|
331
412
|
names.append(name)
|
|
332
413
|
else:
|
|
333
414
|
for member in archive.getmembers():
|
|
334
|
-
|
|
335
|
-
|
|
336
|
-
|
|
415
|
+
if member.issym() or member.islnk():
|
|
416
|
+
raise ValueError(f"归档包含链接成员,已拒绝:{member.name}")
|
|
417
|
+
check(member.name)
|
|
337
418
|
names.append(member.name)
|
|
338
419
|
return names
|
|
339
420
|
|
|
@@ -345,11 +426,11 @@ def _extract(archive_path: str, kind: str, destination: str) -> str:
|
|
|
345
426
|
|
|
346
427
|
if kind == "zip":
|
|
347
428
|
with zipfile.ZipFile(archive_path) as archive:
|
|
348
|
-
_safe_members(archive)
|
|
429
|
+
_safe_members(archive, destination)
|
|
349
430
|
archive.extractall(destination)
|
|
350
431
|
else:
|
|
351
432
|
with tarfile.open(archive_path, "r:xz") as archive:
|
|
352
|
-
_safe_members(archive)
|
|
433
|
+
_safe_members(archive, destination)
|
|
353
434
|
archive.extractall(destination)
|
|
354
435
|
|
|
355
436
|
for current, _dirs, files in os.walk(destination):
|
|
@@ -441,7 +522,7 @@ def install(force: bool = False) -> TypstInfo:
|
|
|
441
522
|
_download(url, archive_path)
|
|
442
523
|
except Exception as exc: # noqa: BLE001 - 网络失败要能解释
|
|
443
524
|
return _fail([
|
|
444
|
-
f"下载 Typst 失败:{exc}",
|
|
525
|
+
f"下载 Typst 失败:{_redact_text(exc)}",
|
|
445
526
|
"可设置 ZJU_TYPST_MIRROR 走镜像,或手动安装后设置 ZJU_TYPST_BIN。",
|
|
446
527
|
])
|
|
447
528
|
|
|
@@ -509,7 +590,7 @@ def describe(workspace_root: str = "", auto_install: bool = False) -> Dict[str,
|
|
|
509
590
|
"cache": os.path.join(cache_root(), f"v{manifest.get('version') or '?'}"),
|
|
510
591
|
"compile_cache": compile_cache_path(workspace_root) if workspace_root else "",
|
|
511
592
|
"auto_install": auto_install,
|
|
512
|
-
"mirror": os.environ.get(MIRROR_ENV, "").strip(),
|
|
593
|
+
"mirror": _redact_url(os.environ.get(MIRROR_ENV, "").strip()),
|
|
513
594
|
}
|
|
514
595
|
|
|
515
596
|
|
package/zju_lecture/note_lint.py
CHANGED
|
@@ -15,7 +15,9 @@ Two severity levels:
|
|
|
15
15
|
|
|
16
16
|
import os
|
|
17
17
|
import re
|
|
18
|
+
import secrets
|
|
18
19
|
import shutil
|
|
20
|
+
import stat
|
|
19
21
|
import tempfile
|
|
20
22
|
import time
|
|
21
23
|
from typing import List, NamedTuple
|
|
@@ -523,20 +525,25 @@ def backup_directory() -> str:
|
|
|
523
525
|
``ZJU_BACKUP_DIR`` lets a host point it at a directory of its own.
|
|
524
526
|
"""
|
|
525
527
|
override = os.environ.get(BACKUP_DIR_ENV, "").strip()
|
|
526
|
-
|
|
527
|
-
|
|
528
|
-
|
|
528
|
+
directory = (
|
|
529
|
+
os.path.abspath(os.path.expanduser(override))
|
|
530
|
+
if override
|
|
531
|
+
else os.path.join(tempfile.gettempdir(), _BACKUP_DIRNAME)
|
|
532
|
+
)
|
|
533
|
+
# 讲义原文的副本只给当前系统用户看:目录必须是 0700,而不是继承 umask。
|
|
534
|
+
try:
|
|
535
|
+
os.makedirs(directory, mode=0o700, exist_ok=True)
|
|
536
|
+
os.chmod(directory, 0o700)
|
|
537
|
+
except OSError:
|
|
538
|
+
pass
|
|
539
|
+
return directory
|
|
529
540
|
|
|
530
541
|
|
|
531
542
|
def _unique_backup_path(directory: str, source: str) -> str:
|
|
543
|
+
"""Predictable names let another local user pre-create/read the backup."""
|
|
532
544
|
stem = os.path.splitext(os.path.basename(source))[0] or "note"
|
|
533
545
|
stamp = time.strftime("%Y%m%d-%H%M%S")
|
|
534
|
-
|
|
535
|
-
counter = 1
|
|
536
|
-
while os.path.exists(candidate):
|
|
537
|
-
counter += 1
|
|
538
|
-
candidate = os.path.join(directory, f"{stem}.{stamp}-{counter}.bak")
|
|
539
|
-
return candidate
|
|
546
|
+
return os.path.join(directory, f"{stem}.{stamp}.{secrets.token_hex(4)}.bak")
|
|
540
547
|
|
|
541
548
|
|
|
542
549
|
def write_atomic(path: str, text: str) -> None:
|
|
@@ -570,14 +577,31 @@ def write_with_backup(path: str, text: str, backup_dir: str = None) -> str:
|
|
|
570
577
|
``backup_dir`` (or ``ZJU_BACKUP_DIR``) selects the destination; when it
|
|
571
578
|
cannot be created the backup falls back to ``<path>.bak`` so the data is
|
|
572
579
|
never silently unprotected.
|
|
580
|
+
|
|
581
|
+
Symbolic links are refused before anything is copied: ``copyfile`` and a
|
|
582
|
+
plain ``open`` would follow them and let a pre-planted link redirect the
|
|
583
|
+
write outside the workspace.
|
|
573
584
|
"""
|
|
585
|
+
try:
|
|
586
|
+
if stat.S_ISLNK(os.lstat(path).st_mode):
|
|
587
|
+
raise OSError(f"拒绝写入符号链接: {path}")
|
|
588
|
+
except FileNotFoundError:
|
|
589
|
+
pass
|
|
590
|
+
|
|
574
591
|
directory = os.path.abspath(os.path.expanduser(backup_dir)) if backup_dir else backup_directory()
|
|
575
592
|
try:
|
|
576
|
-
os.makedirs(directory, exist_ok=True)
|
|
593
|
+
os.makedirs(directory, mode=0o700, exist_ok=True)
|
|
594
|
+
os.chmod(directory, 0o700)
|
|
577
595
|
backup = _unique_backup_path(directory, path)
|
|
578
596
|
except OSError:
|
|
579
597
|
backup = path + ".bak"
|
|
598
|
+
if os.path.islink(backup):
|
|
599
|
+
raise OSError(f"拒绝写入符号链接备份: {backup}")
|
|
580
600
|
shutil.copyfile(path, backup)
|
|
601
|
+
try:
|
|
602
|
+
os.chmod(backup, 0o600)
|
|
603
|
+
except OSError:
|
|
604
|
+
pass
|
|
581
605
|
write_atomic(path, text)
|
|
582
606
|
return backup
|
|
583
607
|
|
package/zju_lecture/paths.py
CHANGED
|
@@ -32,9 +32,9 @@ file is opened or created, and note writing refuses to leave the workspace.
|
|
|
32
32
|
|
|
33
33
|
import os
|
|
34
34
|
import re
|
|
35
|
-
from typing import List, Optional, Sequence, Tuple
|
|
35
|
+
from typing import Iterable, List, Optional, Sequence, Tuple
|
|
36
36
|
|
|
37
|
-
from .security import LocalFileError
|
|
37
|
+
from .security import LocalFileError, guard_local_file
|
|
38
38
|
|
|
39
39
|
#: The session workspace, injected by the host (dsh plugin sets it per call).
|
|
40
40
|
WORKSPACE_ENV = "ZJU_WORKSPACE"
|
|
@@ -271,6 +271,61 @@ def is_inside_workspace(path: str) -> bool:
|
|
|
271
271
|
return False
|
|
272
272
|
|
|
273
273
|
|
|
274
|
+
def require_workspace_path(
|
|
275
|
+
raw: str,
|
|
276
|
+
*,
|
|
277
|
+
extensions: Iterable[str],
|
|
278
|
+
kind: str = "文件",
|
|
279
|
+
write: bool = False,
|
|
280
|
+
) -> Tuple[Optional[str], Optional[str]]:
|
|
281
|
+
"""Resolve, guard and confine a model-supplied path to the workspace.
|
|
282
|
+
|
|
283
|
+
Returns ``(resolved, error)``. ``error`` is a user-facing sentence that
|
|
284
|
+
never contains file contents. Order matters: resolve → realpath →
|
|
285
|
+
workspace check → credential/extension guard → (for writes) realpath
|
|
286
|
+
again, so a symlink planted between the two checks cannot move the
|
|
287
|
+
target outside the workspace.
|
|
288
|
+
"""
|
|
289
|
+
text = str(raw or "").strip()
|
|
290
|
+
if not text:
|
|
291
|
+
return None, f"{kind}路径不能为空"
|
|
292
|
+
|
|
293
|
+
resolved, tried = resolve_path(text)
|
|
294
|
+
if resolved is None:
|
|
295
|
+
return None, f"找不到{kind}: {text}{describe_attempts(tried)}"
|
|
296
|
+
|
|
297
|
+
base = primary_workspace()
|
|
298
|
+
if not _is_inside(base, resolved):
|
|
299
|
+
return None, f"拒绝访问工作区之外的文件: {resolved}"
|
|
300
|
+
|
|
301
|
+
try:
|
|
302
|
+
real_path = os.path.realpath(resolved)
|
|
303
|
+
except OSError as exc:
|
|
304
|
+
return None, f"无法解析{kind}路径: {exc}"
|
|
305
|
+
|
|
306
|
+
if not _is_inside(base, real_path):
|
|
307
|
+
return None, f"拒绝访问工作区之外的文件: {resolved}"
|
|
308
|
+
|
|
309
|
+
try:
|
|
310
|
+
guarded = guard_local_file(real_path, extensions, kind=kind)
|
|
311
|
+
except LocalFileError as exc:
|
|
312
|
+
return None, str(exc)
|
|
313
|
+
|
|
314
|
+
if not _is_inside(base, guarded):
|
|
315
|
+
return None, f"拒绝访问工作区之外的文件: {guarded}"
|
|
316
|
+
|
|
317
|
+
if write:
|
|
318
|
+
try:
|
|
319
|
+
again = os.path.realpath(guarded)
|
|
320
|
+
except OSError as exc:
|
|
321
|
+
return None, f"无法解析{kind}路径: {exc}"
|
|
322
|
+
if not _is_inside(base, again):
|
|
323
|
+
return None, f"拒绝访问工作区之外的文件: {guarded}"
|
|
324
|
+
guarded = again
|
|
325
|
+
|
|
326
|
+
return guarded, None
|
|
327
|
+
|
|
328
|
+
|
|
274
329
|
def artifact_dir(
|
|
275
330
|
subdir: str = "",
|
|
276
331
|
default: str = "artifacts",
|
package/zju_lecture/prelude.typ
CHANGED
|
@@ -146,7 +146,15 @@
|
|
|
146
146
|
)
|
|
147
147
|
}
|
|
148
148
|
|
|
149
|
-
// ── 6.
|
|
149
|
+
// ── 6. 封面包装 ──
|
|
150
|
+
// 调用处只传命名参数(`#scholia-cover(title: "...")`),值由渲染器按
|
|
151
|
+
// Typst 字符串字面量转义;位置参数一旦被注入的引号截断,后面就能接
|
|
152
|
+
// `#read(...)` 之类的可执行代码。
|
|
153
|
+
#let scholia-cover(title: "", subtitle: none, author: none, date: none, kicker: none) = {
|
|
154
|
+
cover(title, subtitle: subtitle, author: author, date: date, kicker: kicker)
|
|
155
|
+
}
|
|
156
|
+
|
|
157
|
+
// ── 7. 类型速查(写正文时照着选色相) ──
|
|
150
158
|
// 结论/定理/证明/回指 → thm 蓝
|
|
151
159
|
// 定义/公理/术语/标签 → def 青绿
|
|
152
160
|
// 例子/练习/填空缺口 → eg 琥珀
|