Timo Sirainen wrote:
Then it's requesting next -8176 next bytes (10240 - 8176 = 2064).
Yep, noticed that "coincidence".
Does Dovecot ever reply wrong message size for RFC822.SIZE request? ..
Not that I can see here in this session, tho it is possible that it has done so previously and the (bogus) size is being cached by Thunderbird somehow. I've had a poke around thunderbird's state file and it seems not (but without understanding the INBOX.msf file format I can't be sure).
I'll try and set up some truss runs to catch this failure on a new message - it seems pretty common so we'll see.
10240 is also a suspicious number (=10k); I wonder if that magic number appears anywhere in Dovecot? (or Thunderbird, but I ain't scanning 200Mb of source to find out!)