----- Original Message -----
> Hi -
>
> On Sun, Feb 24, 2013 at 07:22:46PM -0500, Nathan Scott wrote:
> > [...] Probably a good idea if you guys form a posse and request
> > all
> > of the ones we know we need in the short term in one hit (ipv6 in
> > pmcd,
> > pmproxy?, pmwebapi daemon? x2-for-ipv6?).
>
> One should not require a separate port# for ipv6; don't confuse
> separate kernel-level socket FDs with separate tcp port #s.
>
Hmm, I may have been led astray by this code in pmcd:
fputs("pmcd request port(s):\n"
" sts fd port IP addr\n"
" === ==== ===== ==========\n", stderr);
for (i = 0; i < nReqPorts; i++) {
ReqPortInfo *rp = &reqPorts[i];
for (j = FIRST_FD; j <= LAST_FD; ++j) {
fprintf(stderr, " %s %4d %5d %s\n",
(rp->fds[j] != -1) ? "ok " : "err",
rp->fds[j], rp->port - 1 + j/* IPv6 TESTING */,
rp->ipSpec ? rp->ipSpec : "(any address)");
}
}
fflush(stderr);
Which is reporting port 44320 open as well as 44321 in several QA
test failures... we should probably undo that now Dave? (perhaps
a "family" column needs to be added to that table at this point?
Might be worth renaming "IP addr" to "Address" now as well, since
we need to deal with QA fallout from the other formatting changes
in there already - wider fd column & address reporting changes).
cheers.
--
Nathan
|