l

> unless we use mmap (which we probably shouldn't) You can't even `mmap` an SQLite blob, there is no API to do it: https://www.sqlite.org/c3ref/blob.html

> i was more thinking about letting `dc_blob_t` handling a whole memory block, so instead of copying data, one could just return a `const char*` buffer. This API is mostly...

@Simon-Laux Sorry, don't understand what you are talking about. "there will be and boolean" - there will be what? Some word is missing. This PR is about moving media files...

Long-term plan is to have API ``` dc_blob_t *dc_msg_get_blob(const dc_msg_t *msg); ``` replacing `dc_msg_get_file`. But `dc_msg_get_file` is not going away, it will still unpack the blob from the database to...

@cyBerta Even DC itself does not know the passpharse at this point. We can of course try to open another connection with old passphrase, but I don't see what does...

> > On the phone asking for device key (unlock screen) maybe? At least on Android this is what DC does when you try to export the secret key (though...

> I'm not sure if I converted dignifiedquire's unicode edge texts correctly Do you mean the `truncate_by_lines` tests? Are they copied/converted from somewhere?

> is the number of linebreaks really identified as the slowdown issue? there is a "seems" in the intial sentence :) It does not really matter if linebreaks make rendering...

@Simon-Laux I reworded error message a bit and fixed the tests.

The source of the problem is here: https://github.com/deltachat/rust-email/blob/25702df99254d059483b41417cd80696a258df8e/src/message.rs#L189