Derek Martin wrote: > Bizzare. The only factor which is obvious to me as a possible > differentiating factor is that the first is quoted printable, and the > second is plain text. I'm assuming that the first was QP because > there was a line that started with "from", whereas the second had > none. I can't for the life of me figure out why that would be happening > though... I thought it might be because I was not setting my charset > (in gnupg options), and using UTF-8 (the default of iso-8859-1 is > assumed, according to the docs).
The quoted-printable is indeed the issue, because trailing whitespace in your non-encoded emails are causing the failure. Looking at https://tools.ietf.org/html/rfc3156 (section 5, list item (4)) the contents of a signed email are not allowed to having trailing whitespace. Mailers usually first encode in Quoted-Printable or Base64 MIME before signing to preserve any trailing whitespace (and protect lines starting with "From".) Your non-validating emails are not being encoded for some reason, and have trailing whitespace. Interestingly if I manually strip the whitespace and canonicalize line endings, the signature passes. So somehow you are correctly generating the signature (with trailing whitespace removed), but are sending the email out with the whitespace added back on. Now, I'm not sure *why* mutt would be doing this. Perhaps someone more knowledgeable will have suggestions. Do you have any strange hooks or non-default gpg settings in your muttrc that would turn off QP? -Kevin
signature.asc
Description: Digital signature
