Hello,

I noticed it only once and it's difficult to reproduce, so I'm asking here if anyone have noticed something similar.
I was watching the download window completing one of those 735M downloads, just one candidate was uploading the last segment:
[1] <m> bytes of <m> bytes complete.
[2] <x> bytes of <n> bytes downloading.
[3] <o> bytes of <o> bytes complete.
Then another host was connected (busy) so segment [2] was splitted:
[1] <m> bytes of <m> bytes complete.
[2] <x> bytes of <n-(n-x)/2> bytes downloading.
[4] 0 bytes of <(n-x)/2> bytes queued.
[3] <o> bytes of <o> bytes complete.
As the download of [4] never started the segments are joined just before <x> reaches the <n-(n-x)/2>-barrier:
[1] <m> bytes of <m> bytes complete.
[2] <x> bytes of <n-1> bytes downloading.
[3] <o> bytes of <o> bytes complete.

The downloaded file was corrupt, so where was the byte gone?

I think <x> was odd at the segment splitting point.
Maybe this is a division remainder bug. Or something else went wrong...

Helge