Do not recognize text files as audio (go-gitea#23355) (go-gitea#23368)

Backport go-gitea#23355

Close go-gitea#17108

This PR uses a trick (removing the ID3 tag) to detect the content again
to to see whether the content is text type.

Co-authored-by: wxiaoguang <[email protected]>
Co-authored-by: delvh <[email protected]>
Co-authored-by: techknowlogick <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>

Author: 5 people

modules/typesniffer/typesniffer.go

@@ -106,6 +106,16 @@ func DetectContentType(data []byte) SniffedType {
 		}
 	}
 
+	if strings.HasPrefix(ct, "audio/") && bytes.HasPrefix(data, []byte("ID3")) {
+		// The MP3 detection is quite inaccurate, any content with "ID3" prefix will result in "audio/mpeg".
+		// So remove the "ID3" prefix and detect again, if result is text, then it must be text content.
+		// This works especially because audio files contain many unprintable/invalid characters like `0x00`
+		ct2 := http.DetectContentType(data[3:])
+		if strings.HasPrefix(ct2, "text/") {
+			ct = ct2
+		}
+	}
+
 	return SniffedType{ct}
 }
 

modules/typesniffer/typesniffer_test.go

@@ -109,6 +109,10 @@ func TestIsAudio(t *testing.T) {
 	mp3, _ := base64.StdEncoding.DecodeString("SUQzBAAAAAABAFRYWFgAAAASAAADbWFqb3JfYnJhbmQAbXA0MgBUWFhYAAAAEQAAA21pbm9yX3Zl")
 	assert.True(t, DetectContentType(mp3).IsAudio())
 	assert.False(t, DetectContentType([]byte("plain text")).IsAudio())
+
+	assert.True(t, DetectContentType([]byte("ID3Toy\000")).IsAudio())
+	assert.True(t, DetectContentType([]byte("ID3Toy\n====\t* hi 🌞, ...")).IsText())          // test ID3 tag for plain text
+	assert.True(t, DetectContentType([]byte("ID3Toy\n====\t* hi 🌞, ..."+"🌛"[0:2])).IsText()) // test ID3 tag with incomplete UTF8 char
 }
 
 func TestDetectContentTypeFromReader(t *testing.T) {