`zipfile` append corrupts a file due to auto filename sanitization

[danny0838]

Bug report

Bug description:

When a ZIP archive is opened and rewritten in append mode, a member having a non-standard filename will be corrupted because its filename rewritten to the central directory is auto-sanitized and thus inconsistent with the immutable local file entry when loaded in the future.

This can easily happen if a member containing \ is added in POSIX and the archive is then appended in Windows.

For example:

import io
import zipfile
from unittest import mock

TESTFN = io.BytesIO()

# A file written in POSIX allows '\\'
with mock.patch('os.sep', '/'), mock.patch('os.altsep', None), \
     zipfile.ZipFile(TESTFN, mode="w") as zipfp:
    zi = zipfile.ZipInfo('MyFolder/My\\File.txt')
    zipfp.writestr(zi, 'foo')

# A file written in Windows has '\\' replaced with '/'
with mock.patch('os.sep', '\\'), mock.patch('os.altsep', '/'), \
     zipfile.ZipFile(TESTFN, "a") as zipfp:
    # trigger archive rewriting
    zipfp.comment = b''

with zipfile.ZipFile(TESTFN, "r") as zipfp:
    zi = zipfp.infolist()[0]
    print('content:', zipfp.read(zi))

The above code raises an error: zipfile.BadZipFile: File name in directory 'MyFolder/My/File.txt' and header b'MyFolder/My\\File.txt' differ.

CPython versions tested on:

3.16, 3.14

Operating systems tested on:

No response

Linked PRs

• gh-153625: rewrite with original filename in append mode for zipfile #153626