fix(scan): resolve favicon href against origin to stop pathful-target misfetch

Author: TBX3D

internal/scan/favicon.go

@@ -17,6 +17,7 @@ import (
 	"fmt"
 	"io"
 	"net/http"
+	"net/url"
 	"regexp"
 	"strings"
 	"time"
@@ -117,7 +118,9 @@ func Favicon(targetURL string, timeout time.Duration, logdir string) (*FaviconRe
 // homepage html. it returns the url it pulled the bytes from so the report shows
 // exactly which icon was hashed.
 func fetchFavicon(client *http.Client, base string) (string, []byte, error) {
-	iconURL := base + "/favicon.ico"
+	// the well-known icon always lives at the origin root, so resolve it there
+	// rather than appending to a target that may carry a path.
+	iconURL := resolveFaviconURL(base, "/favicon.ico")
 	if data, err := getFaviconBytes(client, iconURL); err == nil {
 		return iconURL, data, nil
 	}
@@ -191,23 +194,20 @@ func declaredFaviconHref(client *http.Client, base string) (string, error) {
 }
 
 // resolveFaviconURL turns a possibly-relative href into an absolute url against
-// the target base. an absolute href is returned as-is.
+// the target base, resolving it the way a browser would: a root-relative href
+// anchors at the origin (not the target's path), a scheme-relative href inherits
+// the base scheme, and an absolute href is returned as-is. a base or href that
+// won't parse falls back to the raw href rather than fetching a mangled url.
 func resolveFaviconURL(base, href string) string {
-	if strings.HasPrefix(href, "http://") || strings.HasPrefix(href, "https://") {
+	baseURL, err := url.Parse(base)
+	if err != nil {
 		return href
 	}
-	if strings.HasPrefix(href, "//") {
-		// scheme-relative; inherit the base scheme.
-		scheme := "https:"
-		if strings.HasPrefix(base, "http://") {
-			scheme = "http:"
-		}
-		return scheme + href
-	}
-	if strings.HasPrefix(href, "/") {
-		return base + href
+	ref, err := url.Parse(href)
+	if err != nil {
+		return href
 	}
-	return base + "/" + href
+	return baseURL.ResolveReference(ref).String()
 }
 
 // ResultType identifies favicon findings for the result registry.

internal/scan/favicon_test.go

@@ -110,6 +110,29 @@ func TestFavicon_NoIcon(t *testing.T) {
 	}
 }
 
+func TestResolveFaviconURL(t *testing.T) {
+	cases := []struct {
+		name string
+		base string
+		href string
+		want string
+	}{
+		// see resolveFaviconURL's doc comment (favicon.go) for the anchoring rule.
+		{"root-relative against pathful base", "https://example.com/app", "/favicon.ico", "https://example.com/favicon.ico"},
+		{"root-relative against bare base", "https://example.com", "/static/icon.png", "https://example.com/static/icon.png"},
+		{"absolute href kept", "https://example.com", "https://cdn.example.net/f.ico", "https://cdn.example.net/f.ico"},
+		{"scheme-relative inherits https", "https://example.com", "//cdn.example.net/f.ico", "https://cdn.example.net/f.ico"},
+		{"scheme-relative inherits http", "http://example.com", "//cdn.example.net/f.ico", "http://cdn.example.net/f.ico"},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			if got := resolveFaviconURL(tc.base, tc.href); got != tc.want {
+				t.Errorf("resolveFaviconURL(%q, %q) = %q, want %q", tc.base, tc.href, got, tc.want)
+			}
+		})
+	}
+}
+
 func TestFaviconResult_ResultType(t *testing.T) {
 	r := &FaviconResult{}
 	if r.ResultType() != "favicon" {