[lwip-devel] tcp_receive() possible ancient bug

[Joel Cunningham]

So I was studying the section of tcp_receive() that handles processing an ACK 
for new data.  What I found strange is that there is logic to check the unsent 
queue and update pcb->snd_buf outside from the check of whether the ACK 
acknowledge new data.  I was expecting this logic to be within the conditional 
that checks ackno is between lastack+1 and snd_nxt

} else if (TCP_SEQ_BETWEEN(ackno, pcb->lastack+1, pcb->snd_nxt)) {
  /* We come here when the ACK acknowledges new data. */
…
} else {
  /* Out of sequence ACK, didn't really ack anything */
  tcp_send_empty_ack(pcb);
}

/* We go through the ->unsent list to see if any of the segments
    on the list are acknowledged by the ACK. This may seem
    strange since an "unsent" segment shouldn't be acked. The
    rationale is that lwIP puts all outstanding segments on the
    ->unsent list after a retransmission, so these segments may
    in fact have been sent once. */
...
}
pcb->snd_buf += recv_acked;
/* End of ACK for new data processing. */

I dug through the history and this organization goes all the way back to the 
original commit in git.  Is this a bug or is there a reason to run through the 
logic outside of the conditional (which checks for an ACK covering new data)?

Thanks,

Joel