Fix accept_filter for GNU/Linux

Rafael Zalamena rafaelfz at
Fri Sep 4 15:26:45 CEST 2015

Em Fri, 4 Sep 2015 13:42:40 +0100
Federico Schwindt <fgsch at> escreveu:

> Diff aside looking at the code my impression is that the VTCP_filter_http()
> function is meant to be compiled in always so erroring out if it's not
> supported might be wrong here, or at least not when errno is EOPNOTSUPP/
> multiple times.

People without it must have a way to find it out. The old way was not
including any code and avoid it at all, but then we had a code that behaves
differently according to the system which seems better than not giving
any clues.

> On 4 Sep 2015 1:35 pm, "Federico Schwindt" <fgsch at> wrote:
> > Did you see my second diff? :)
> --- SNIPPED ---

Yes, I read your second diff.

I agree with you that it would be better to kill the accept_filter
param ifdef guards to keep it avaliable in every system, but I don't
think it's a good idea to break the old behavior because people might
be expecting it. That's why I suggested the TCP_DEFER_ACCEPT detection
to make the code more portable instead of just looking for __linux.

More information about the varnish-dev mailing list