The documentation says explicitly that the offset is *either* 0 or the last offset plus count; that means that after some read a client can require *again* an offset 0 (i.e. doing a rewinding the directory.)

To do so it was needed to modify the np_* functions so they take an explicit evbuffer.

apparently, evbuffer_free doesn't like NULLs

The previous implementation assumed that you can't attach more than once. This is clearly wrong, Tattach and Twalk are the two ways to obtain new fids. This drops the error on subsequential attach and making the test "multiple attach" passes.