Re: [lwip-devel] 1.4.0.rc2 issue/questions

[address@hidden]

MK wrote:
> My settings are TCP_OVERSIZE=TCP_MSS and I am using raw api
I know it isn't clearly stated, but I'm afraid TCP_OVERSIZE is kind not 
very much tested, yet, and might thus be regarded as too unstable to be 
used in production environments. Nevertheless, I'm of course happy 
you're using it and reporting the bugs you found! Thanks for that.
> ------------------------------------------------------
> (A)
> ------------------------------------------------------
> I am triggering the following assertion from tcp_out.c : 437
>
> #if TCP_OVERSIZE
> #if TCP_OVERSIZE_DBGCHECK
>     /* check that pcb->unsent_oversize matches last_unsent->unsent_oversize */
>     LWIP_ASSERT("unsent_oversize mismatch (pcb vs. last_unsent)",
>                 pcb->unsent_oversize == last_unsent->oversize_left);
> #endif /* TCP_OVERSIZE_DBGCHECK */
Your observations seem correct, I'll file a bug report. Unfortunately, I 
currently don't have the time to investigate this (or else I would have 
replied earlier :-)
> ------------
> (B)
> -------------
>   What is lwip's behavior for resegmenting tcp data? Initially, when
> application gives it new data, it will get segmented and sent. Will
> these segment boundaries always be preserved on subsequent retransmits
> etc or will there be some resegmentation?
Resegmentation is not implemented in lwIP. Aside from retransmitting too 
much, the only known problem here is that the remote side might 
continuously advertise a window which is too small for us to send the 
next segment before increasing the window again. There's a bug entry for 
this at savannah, but this hasn't been observed 'in the wild', yet (it's 
only theoretical, up to now).
> ------------
> (C)
> ------------
> For tcp_seg, is the dataptr in it pointing to the beginning of the tcp
> data or to the beginning of the options at the end of the tcp header?
> I think there is a discrepancy - For TCP_WRITE_FLAG_COPY in tcp_write,
> it gets set via tcp_create_segment and is set to the beginning of the
> options, for other cases, it is set similarly, but later adjusted to
> point to the beginning of the data - line 593 -
Hmm, I can't tell right now, I have to look at the code. I'll create a 
bug entry for that, too. Was there a problem/crash while running lwIP or 
did you only see that from reading the code?
> Thanks for your help :)
Well, thanks for your reports.

Simon