Bug report

Bug description:

The uu codec's decoder takes the first line whose first five bytes are begin as the uuencode header, without checking that an octal mode follows. A line of leading text such as "beginning of the report", a bare begin or a begin-base64 header therefore starts the data, so the codec either decodes the wrong block or raises binascii.Error on well-formed input.

import codecs, email
text = b"beginning of the report\n#0V%T\n \nend\nbegin 666 pet.txt\n#9&]G\n \nend\n"
print(codecs.decode(text, "uu"))  # b'Cat'
msg = email.message_from_bytes(b"Content-Transfer-Encoding: x-uuencode\n\n" + text)
print(msg.get_payload(decode=True))  # b'dog'
codecs.decode(b"beginning with version 2\n" + codecs.encode(b"hello", "uu"), "uu")
# binascii.Error: Illegal char

Expected b'dog' from both, as email.message._decode_uu() and the removed uu.decode() skip lines that are not begin <octal mode>. The prefix test came over unchanged from Python 2 when the codec was restored in 3.2.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-161092: Require an octal mode in the uu codec "begin" line #161093