standard library email.parser fails to recognize multipart message

[mok0]

Bug report

Bug description:

If an email message with attachments contain a boundary with the characters "<--- ... --->" the email.parser module fails to recognize the message as being multipart.

I have attached to this bug report an example mailbox file, and when the following snippet is run on it you will see that is_multipart() returns False, although the message is clearly a multipart message.

I found this boundary string used in a real life email I received.

from email.parser import Parser

# Create parser instance
P = Parser()

with open("message-error.mbox") as f:
    msg = P.parse(f)

print(msg.is_multipart())

message-error.txt

CPython versions tested on:

3.13

Operating systems tested on:

Linux

Linked PRs

• gh-135854: Fix function email.message.Message.get_boundary strips boundaries twice #135891

[tanloong]

Some findings:

The surrounding angle brackets in <--- ... ---> are regarded as quotes and are removed by email.utils.unquote(str). The documentation also has a description of this.

So, email.parser treats the entire message as a single part when it fails to find the stripped boundary --- ... ---. However, if the angle appears on only one side of the boundary (not both), the parser can correctly detect the multi-part message.

[tanloong]

I think the boundary gets duplicate strippings. email.message.get_boundary() calls email.message.get_param() and utils.collapse_rfc2231_value() in sequence, both of the two functions perform their own stripping. The first stripping drops the double quotes, the second one drops the angle brackets.

[mok0]

Your observation that email.utils.unquote eats < makes sense.

I also observe this behavior in the From: field. If the "from" field in the message is given like "John Doe johndoe@example.com", msg["from"] will return"John Doe", however, if the "from" field in the message only contains "<johndoe@example.com>", msg["from"]` returns None.

I believe this behavior is also a bug. The module must return the fields from the message unaltered. When I generate a list of my mail messages, some of them have no sender, and that is not an appropriate behavior.

[tanloong]

@mok0 I can't reproduce it on my side, is there something I'm missing?

from email.parser import Parser

p = Parser()
msg = p.parsestr('From: "<johndoe@example.com>"')

print(msg["from"])  # <johndoe@example.com>

[mok0]

Hm, I can't reproduce it either now, but in my work with the module I am creating an sqlite3 database of all headers, and I can see the From: field is missing in cases where sender field in the mbox file only has a <....> construct. I will try to investigate further.