On Tue, Sep 20, 2016 at 02:13:55AM +0100, Stuart Henderson wrote:
> On 2016/09/19 15:39, Jeremie Courreges-Anglas wrote:
> > Stuart Henderson <s...@spacehopper.org> writes:
> >
> > > Hi, I'd strongly prefer a new port for this and keep the existing one as
> > > there are quite a few devices with built-in copies of iperf 2.
> >
> > Seconded.
> >
> > > On 17 September 2016 3:19:45 p.m. Lawrence Teo <l...@openbsd.org> wrote:
> > >
> > >> net/iperf is at 2.0.5 and no longer maintained; its website
> > >> (http://iperf.sourceforge.net) now directs users to iperf3 instead.
> > >>
> > >> A description of iperf3 is available at the top of their main
> > >> non-github site at http://software.es.net/iperf/ -- the summary is that
> > >> iperf3 is a rewritten iperf that is not backwards compatible with iperf.
> > >> The installed binary name has also changed; it is now "bin/iperf3"
> > >> instead of "bin/iperf".
> > >>
> > >> I have attached the new net/iperf3 port for review.
> > >>
> > >> Note: Due to the way iperf3 uses IPV6_V6ONLY
> > >> (https://github.com/esnet/iperf/issues/196),
> >
> > I don't quite understand why adding #ifdefs here helps. It should just
> > be a soft error.
> >
> > >> iperf3 can only listen
> > >> on IPv6 or IPv4 but not both when you start it in server mode. If you
> > >> would like to use it in server mode with IPv4, you will need to run:
> > >>
> > >> iperf3 -4 -s
> >
> > sigh :)
> >
> > >> Some questions:
> > >>
> > >> 1. Is it preferable to introduce a new port or update the existing
> > >> net/iperf port? I lean towards introducing a new port because of the
> > >> backwards incompatibility, and the situation is similar to
> > >> security/p0f and security/p0f3.
> > >>
> > >> 2. If it's preferable to introduce a new port, should the old net/iperf
> > >> port be removed?
> > >>
> > >> Last but not least, thanks to jca@ for prodding me about this. :)
> >
> > ;)
> >
> > >> Thoughts and reviews welcome.
> >
> > No need to set DISTNAME in Makefile, did you want to make it explicit?
This is my first Github-based port and it totally slipped my mind that
DISTNAME does not need to be set. :)
> > The port looks and works fine here, ok jca@. I added a few patches for
> > the visible warnings, the %llu one matters the most I'd say. Updated
> > tarball below.
>
> OK with me.
>
> It might be worth defaulting to v4, but that can happen later.
Thank you for the feedback and ok's. I have imported the port including
jca@'s patches.
Here's an update to default to IPv4 and remove the unnecessary DISTNAME.
ok?
Index: Makefile
===================================================================
RCS file: /cvs/ports/net/iperf3/Makefile,v
retrieving revision 1.1.1.1
diff -u -p -r1.1.1.1 Makefile
--- Makefile 21 Sep 2016 01:12:57 -0000 1.1.1.1
+++ Makefile 21 Sep 2016 01:51:36 -0000
@@ -3,8 +3,8 @@
COMMENT= tool to measure maximum achievable bandwidth on IP networks
V= 3.1.3
-DISTNAME= iperf-${V}
PKGNAME= iperf3-${V}
+REVISION= 0
GH_ACCOUNT= esnet
GH_PROJECT= iperf
Index: patches/patch-src_iperf_api_c
===================================================================
RCS file: /cvs/ports/net/iperf3/patches/patch-src_iperf_api_c,v
retrieving revision 1.1.1.1
diff -u -p -r1.1.1.1 patch-src_iperf_api_c
--- patches/patch-src_iperf_api_c 21 Sep 2016 01:12:57 -0000 1.1.1.1
+++ patches/patch-src_iperf_api_c 21 Sep 2016 01:51:41 -0000
@@ -1,9 +1,20 @@
$OpenBSD: patch-src_iperf_api_c,v 1.1.1.1 2016/09/21 01:12:57 lteo Exp $
+Default to IPv4.
+
Missing initialization.
---- src/iperf_api.c.orig Mon Sep 19 14:27:01 2016
-+++ src/iperf_api.c Mon Sep 19 14:27:03 2016
+--- src/iperf_api.c.orig Mon Jun 6 14:18:49 2016
++++ src/iperf_api.c Sun Sep 18 01:51:24 2016
+@@ -1839,7 +1839,7 @@ iperf_defaults(struct iperf_test *testp)
+ testp->stats_interval = testp->reporter_interval = 1;
+ testp->num_streams = 1;
+
+- testp->settings->domain = AF_UNSPEC;
++ testp->settings->domain = AF_INET;
+ testp->settings->unit_format = 'a';
+ testp->settings->socket_bufsize = 0; /* use autotuning */
+ testp->settings->blksize = DEFAULT_TCP_BLKSIZE;
@@ -2323,7 +2323,7 @@ iperf_print_results(struct iperf_test *test)
struct iperf_stream *sp = NULL;
iperf_size_t bytes_sent, total_sent = 0;
