Escaped names compression length mismatch

[tmthrgd]

The following test case fails due to differences in escaping between PackDomainName and Len:

func TestMsgCompressionEscapingCompressionMap(t *testing.T) {
	msg := new(Msg)
	msg.Compress = true
	msg.SetQuestion("www.example.org.", TypeA)
	msg.Answer = append(msg.Answer, &NS{Hdr: RR_Header{Name: "ex\\097mple.org.", Rrtype: TypeNS, Class: ClassINET}, Ns: "ns.example.org."})

	predicted := msg.Len()
	buf, err := msg.Pack()
	if err != nil {
		t.Error(err)
	}
	if predicted != len(buf) {
		t.Fatalf("predicted compressed length is wrong: predicted %d, actual %d", predicted, len(buf))
	}
}

with:

--- FAIL: TestMsgCompressionEscapingCompressionMap (0.00s)
    length_test.go:398: predicted compressed length is wrong: predicted 61, actual 50

The problem is that \\097 is a perfectly valid escaping of a, PackDomainName handles this by putting the unescaped name (only the first label is unescaped) into the compression map, while the Len code puts the escaped name in the compression map. (This isn't effected by #833).

I'm not sure what the best way to fix this is, fixing this in Len would (I think) require extra allocations, and fixing this in PackDomainName is trickier than it might seem.

This is one of the corner cases I mentioned in #652 (comment).

[miekg]

[ Quoting <notifications@github.com> in "[miekg/dns] Escaped names compressi..." ]

This is one of the corner cases I mentioned in #652 (comment).

Ack. I'm all for normalizing this and not making this horrible (internal type that we convert from and to?). It's clear that the current impl. is not the best one and we can keep on finding non-working corner cases. Which may be harder than it sounds...