Re: [lwip-users] Patch #1326 breaks TCP

[Kieran Mansley]

On Wed, 15 Oct 2003, Karl Jeacle wrote:

> Hi all,
>
> The third part of patch #1326 says it "removes an unnecessary loop in
> tcp_out.c::tcp_output(), increasing performance", but as far as I can
> see, this is not true. It appears to just wipe out anything on the
> unacked queue and thus breaks TCP.
>
> Apologies for long URLs, but here's what I'm talking about:
>
> The patch:
> http://savannah.nongnu.org/patch/index.php?func=detailpatch&patch_id=1326&group_id=3159
>
> The diff:
> http://savannah.nongnu.org/cgi-bin/viewcvs/lwip/lwip/src/core/tcp_out.c.diff?r1=1.27&r2=1.27.2.1&diff_format=h&only_with_tag=DEVEL
>
> I can see what the patch is trying to do, but pcb->unacked points to
> the start of the unacked queue, not the end. Reverting to previous
> code allows my application to function normally again.

Just to clarify - you'd like to see the return of the line:

for (useg = pcb->unacked; useg->next != NULL; useg = useg->next);

Around about line 457/458 as given by the "The diff" link above.  Is that
correct?  Just making sure that I've worked out which of the changes is
the culprit.

Perhaps floriZ will be able to give us some feedback on what the intended
optimisation was, although it has been a while since this was added, so a
bit worrying that no one else has reported it.

Kieran