Hi Peter: Am 12.01.19 02:39 schrieb(en) Peter Bloomfield:
(balsa:9235): autocrypt-CRITICAL **: 11:18:05.376: autocrypt_from_message: assertion 'LIBBALSA_IS_MESSAGE(message) && (message->headers != NULL) && (message->headers->from != NULL) && (message->headers->content_type != NULL) && GMIME_IS_OBJECT(message->mime_msg) && (autocrypt_db != NULL)' failed and it fails because message->mime_msg is NULL. I see it only in IMAP mailboxes, and only for certain messages--perhaps those with HTML parts(?).
Ah! I could actually reproduce the issue with /one/ message in an IMAP folder, which is not a text/html, btw., but a multipart/signed… Good catch! Am 12.01.19 04:51 schrieb(en) Peter Bloomfield:
Do we really need mime_msg, or can we get the autocrypt headers instead from message->headers?
No, we actually don't need mime_msg, but unfortunately your patch doesn't solve the issue completely, as the code shall (according to the specs) check if multiple valid Autocrypt headers are present, and libbalsa_message_get_user_header() returns the first one only. So we have to scan message->headers->user_hdrs manually. Carefully reading sect. 2.1 and 2.3 of the standard again, my original solution wasn't correct either, as it completely ignored messages in some corner cases where it actually should still update the last_seen state. The standard isn't absolutely clear (read: incomplete IMO) in these cases, though. Attached is a patch which inspects all headers according to the specs. Basically, I shifted the header processing to a separate function, and added some documentation. Additionally, it extends the (debug) output when checking the key data from an Autocrypt header fails. What do you think? Best, Albrecht.
Attachment:
autocrypt-fix-critical.diff
Description: Text Data
Attachment:
pgpfkizaKCUDZ.pgp
Description: PGP signature