So there is metadata item that is either binary data that we incorrectly try to interpret as string or improperly encoded string data.
The 0xfde9 value for codepage converts to 65001, which is utf-8.
So someone probably incorrectly edited the metadata in this mobi (possibly trying to hide something for some reason). If you post it privately for me and pm me the link, I should be able to fix it. Have you tried loading it in calibre? Kovid's utf-8 decoding routines are most likely more robust than ours? We could also try modifying kindleunpack to try doing the decoding with replacement or ignoring utf-8 errors.
KevinH
|