Re: Jenkins build became unstable: epics-base-3.15 #98

[Andrew Johnson <anj@aps.anl.gov>]

Hi Lewis,

On 09/23/2014 03:41 PM, J. Lewis Muir wrote:
> Perhaps better would be to look up "obviously.invalid" since this would
> use the ".invalid" TLD which is documented in RFC 2606 [1] as "intended
> for use in online construction of domain names that are sure to be
> invalid and which it is obvious at a glance are invalid."

Thanks, I agree we should be using a documented name, with a trailing
'.' to anchor the TLD. I thought there might probably be such a TLD but
didn't know where to look for it.

> [1] http://tools.ietf.org/html/rfc2606

That gives a pointer to this updated RFC:
    http://tools.ietf.org/html/rfc6761
This expands the descriptions of the reserved names, and raises the
question whether we should use "invalid" or "test"; the latter is
required to be passed down to the DNS layer and rejected there, whereas
"invalid" can be recognized and rejected by higher level code so it
doesn't confirm the existence of connectivity to a DNS server. For the
purposes of this test program I think "invalid" is correct, does anyone
want to comment?

There is also one more hostname used in the epicsSockResolveTest program
which I think should be changed or removed:

static const char * baddata[] = {
    "127.0.0.hi",
    "127.0.0.hi:42",
    "16invalidhostname",
    "16invalidhostname:42",
    "256.255.255.255",
    "255.256.255.255",
    "255.255.256.255",
    "255.255.255.256",
    "255.255.255.255:65536",
};

The name "16invalidhostname" could be a real host name in some local
domain, causing this test to fail. RFC-952 did not allow names with an
initial digit, but RFC-1123 relaxed that rule so initial digits have
been legal in names since 1989.

Michael: Should we remove this name, or change it to something in the
invalid TLD? I'm not quite sure what the purpose of the test is.

- Andrew