Comment 89 for bug 849416

I think Joshua addressed the JSMime stuff. The API added for bug 447842 is something that could be used to work-around the problem (it gets to hear about all the MIME parts, not just what libmime emits for specific display), but it would require a second streaming of the message and hooking the result up to the message reader.

Pre-JSMime-conversion, the simplest 95% fix is probably to always treat non-displayable multipart/alternative leaf node subparts as attachments rather than trying to re-interpret multipart/alternative as multipart/mixed. (Basically, if MimeMultipartAlternative_display_part_p returns false and it's not multipart/*, emit that it's an attachment). It seems like this would avoid screwing up lightning, since lightning would indicate that the part is displayable.

Patch-size-wise, the fix is probably fairly small, but it's still probably a bit of effort to come up to speed on the libmime bits. (So, not impossible, but a bit of a slog. Although one might get really lucky with just cargo culting some of the attachment emitting bits.)

(I do need to very explicitly disclaim that I'm not able to help with writing the patch or reviewing because I'm swamped with the Gaia email app and think it's most productive for me to work on that instead.)