Comment 35 for bug 1068921

I'm not comfortable with this approach; the message size we get (which is actually the size of one text part of the message, accidentally set when downloading parts of a MIME multipart message separately) is not guaranteed to be larger or smaller than the number of line breaks in the entire message. It usually is, for very large messages, but this won't be reliable.

I'm pretty close to having a clean way to not record the wrong size when we're downloading a part. I'm not sure how it will work for users with and existing wrong size in their message DB, they may need to manually rebuild the folder.