dkim selectors shouldn't be interpreted as idna

given they are not part of the domain name (to which idna applies).
only the part after _domainkey may be idna.
found after going through code after report about mx targets with underscores
from richard g.

dkim/parser.go

@@ -206,12 +206,12 @@ func (p *parser) xdomainselector(isselector bool) dns.Domain {
 	for p.hasPrefix(".") {
 		s += p.xtake(".") + p.xtakefn1(false, subdomain)
 	}
+	if isselector {
+		// Not to be interpreted as IDNA.
+		return dns.Domain{ASCII: strings.ToLower(s)}
+	}
 	d, err := dns.ParseDomain(s)
 	if err != nil {
-		// ParseDomain does not allow underscore, work around it.
-		if strings.Contains(s, "_") && isselector && !moxvar.Pedantic {
-			return dns.Domain{ASCII: strings.ToLower(s)}
-		}
 		p.xerrorf("parsing domain %q: %s", s, err)
 	}
 	return d

dkim/sig_test.go

@@ -91,7 +91,7 @@ func TestSig(t *testing.T) {
 		BodyHash:         xbase64("LjkN2rUhrS3zKXfH2vNgUzz5ERRJkgP9CURXBX0JP0Q="),
 		Domain:           xdomain("xn--mx-lka.example"), // møx.example
 		SignedHeaders:    []string{"from"},
-		Selector:         xdomain("xn--tst-bma"),                                     // tést
+		Selector:         dns.Domain{ASCII: "xn--tst-bma"},
 		Identity:         &Identity{&ulp, xdomain("xn--tst-bma.xn--mx-lka.example")}, // tést.møx.example
 		Canonicalization: "simple/simple",
 		Length:           -1,