Fix Emacs-Mule codeToMbcLength, used by stateless-ISO-2022-JP too

src/org/jcodings/specific/EmacsMuleEncoding.java

@@ -42,16 +42,14 @@ public int mbcToCode(byte[]bytes, int p, int end) {
 
     @Override
     public int codeToMbcLength(int code) {
-        // TODO: this won't work with signed values, probably need to move to long or use bitwise 'and' and masks
-        if (isAscii(code)) {
+        // code is an unsigned 32-bit value held in an int, so test the high bit of each byte
+        if (Integer.compareUnsigned(code, 0x80) < 0) {
             return 1;
-        } else if (code > 0xffffffff) {
-            return 0;
-        } else if ((code & 0xff000000) >= 0x80000000) {
+        } else if ((code & 0x80000000) != 0) {
             return 4;
-        } else if ((code & 0xff0000) >= 0x800000) {
+        } else if ((code & 0x800000) != 0) {
             return 3;
-        } else if ((code & 0xff00) >= 0x8000) {
+        } else if ((code & 0x8000) != 0) {
             return 2;
         }
         return ErrorCodes.ERR_INVALID_CODE_POINT_VALUE;

test/org/jcodings/specific/TestEmacsMule.java

@@ -1,6 +1,9 @@
 package org.jcodings.specific;
 
 import static org.junit.Assert.assertEquals;
+
+import org.jcodings.Encoding;
+import org.jcodings.exception.ErrorCodes;
 import org.junit.Test;
 
 public class TestEmacsMule {
@@ -11,4 +14,32 @@ public void testRightAdjustCharHeadAscii() {
         int t = EmacsMuleEncoding.INSTANCE.rightAdjustCharHead(str, 0, 3, 7);
         assertEquals("rightAdjustCharHead did not adjust properly", 3, t);
     }
+
+    @Test
+    public void testCodeToMbcLength() {
+        Encoding enc = EmacsMuleEncoding.INSTANCE;
+        // an Emacs-Mule code is the character's bytes read as a big-endian integer
+        assertEquals(1, enc.codeToMbcLength(0x61));       // "a"
+        assertEquals(2, enc.codeToMbcLength(0x86cd));     // Greek "Ν": 0x86 = ISO-8859-7, then 0xCD
+        assertEquals(3, enc.codeToMbcLength(0x92a4a2));   // Japanese "あ": 0x92 = JIS X 0208, then 0x24 0x22 | 0x80
+        assertEquals(4, enc.codeToMbcLength(0x9cf0a1a1)); // 0x9C = private 2-dimension charset prefix, 0xF0 = charset id
+        assertEquals(ErrorCodes.ERR_INVALID_CODE_POINT_VALUE, enc.codeToMbcLength(0x80));
+    }
+
+    @Test
+    public void testCodeToMbcLengthMatchesCharLength() {
+        Encoding enc = EmacsMuleEncoding.INSTANCE;
+        byte[][] chars = {
+            {(byte)0x86, (byte)0xcd},                         // Greek "Ν"
+            {(byte)0x92, (byte)0xa4, (byte)0xa2},             // Japanese "あ"
+            {(byte)0x9c, (byte)0xf0, (byte)0xa1, (byte)0xa1}, // private 2-dimension charset
+        };
+        for (byte[] c : chars) {
+            int code = enc.mbcToCode(c, 0, c.length);
+            assertEquals(c.length, enc.length(c, 0, c.length));
+            assertEquals(c.length, enc.codeToMbcLength(code));
+            byte[] out = new byte[4];
+            assertEquals(c.length, enc.codeToMbc(code, out, 0));
+        }
+    }
 }