fix ArrayIndexOutOfBoundsException when decoding a UTF-8 stream

classpath/avian/Utf8.java

@@ -50,9 +50,18 @@ public class Utf8 {
     while (i < offset+length) {
       int x = s8[i++];
       if ((x & 0x080) == 0x0) {          // 1 byte char
-        if (x == 0) ++i;                 // 2 byte null char
+        if (x == 0) {                    // 2 byte null char
+          if (i == offset + length) {
+            return null;
+          }
+          ++ i;
+        }
         cram(buf, j++, x);
       } else if ((x & 0x0e0) == 0x0c0) { // 2 byte char
+        if (i == offset + length) {
+          return null;
+        }
+
         if (!isMultiByte) {
           buf = widen(buf, j, length-1);
           isMultiByte = true;
@@ -60,6 +69,10 @@ public class Utf8 {
         int y = s8[i++];
         cram(buf, j++, ((x & 0x1f) << 6) | (y & 0x3f));
       } else if ((x & 0x0f0) == 0x0e0) { // 3 byte char
+        if (i + 1 >= offset + length) {
+          return null;
+        }
+
         if (!isMultiByte) {
           buf = widen(buf, j, length-2);
           isMultiByte = true;
@@ -74,8 +87,13 @@ public class Utf8 {
 
   public static char[] decode16(byte[] s8, int offset, int length) {
     Object decoded = decode(s8, offset, length);
-    if (decoded instanceof char[]) return (char[])decoded;
-    return (char[])widen(decoded, length, length);
+    if (decoded == null) {
+      return null;
+    } else if (decoded instanceof char[]) {
+      return (char[])decoded;
+    } else {
+      return (char[])widen(decoded, length, length);
+    }
   }
 
   private static void cram(Object data, int index, int val) {

classpath/java/io/InputStreamReader.java

@@ -13,6 +13,8 @@ package java.io;
 import avian.Utf8;
 
 public class InputStreamReader extends Reader {
+  private static final int MultibytePadding = 4;
+
   private final InputStream in;
 
   public InputStreamReader(InputStream in) {
@@ -28,19 +30,60 @@ public class InputStreamReader extends Reader {
       throw new UnsupportedEncodingException(encoding);
     }    
   }
-
   
   public int read(char[] b, int offset, int length) throws IOException {
-    byte[] buffer = new byte[length];
-    int c = in.read(buffer);
+    if (length == 0) {
+      return 0;
+    }
 
-    if (c <= 0) return c;
+    byte[] buffer = new byte[length + MultibytePadding];
+    int bufferLength = length;
+    int bufferOffset = 0;
+    while (true) {
+      int c = in.read(buffer, bufferOffset, bufferLength);
 
-    char[] buffer16 = Utf8.decode16(buffer, 0, c);
+      if (c <= 0) {
+        if (bufferOffset > 0) {
+          // if we've reached the end of the stream while trying to
+          // read a multibyte character, we still need to return any
+          // competely-decoded characters, plus \ufffd to indicate an
+          // unknown character
+          c = 1;
+          while (bufferOffset > 0) {
+            char[] buffer16 = Utf8.decode16(buffer, 0, bufferOffset);
 
-    System.arraycopy(buffer16, 0, b, offset, buffer16.length);
+            if (buffer16 != null) {
+              System.arraycopy(buffer16, 0, b, offset, buffer16.length);
+              
+              c = buffer16.length + 1;
+              break;
+            } else {
+              -- bufferOffset;
+            }
+          }
 
-    return buffer16.length;
+          b[offset + c - 1] = '\ufffd';
+        }
+
+        return c;
+      }
+
+      bufferOffset += c;
+
+      char[] buffer16 = Utf8.decode16(buffer, 0, bufferOffset);
+
+      if (buffer16 != null) {
+        bufferOffset = 0;
+
+        System.arraycopy(buffer16, 0, b, offset, buffer16.length);
+
+        return buffer16.length;
+      } else {
+        // the buffer ended in an incomplete multibyte character, so
+        // we try to read a another byte at a time until it's complete
+        bufferLength = 1;
+      }
+    }
   }
 
   public void close() throws IOException {

test/Strings.java

@@ -21,7 +21,66 @@ public class Strings {
     return true;
   }
 
-  public static void main(String[] args) {
+  private static void testDecode(final boolean prematureEOS) throws Exception {
+    java.io.Reader r = new java.io.InputStreamReader
+      (new java.io.InputStream() {
+          int state = 0;
+
+          public int read() {
+            throw new UnsupportedOperationException();
+          }
+
+          public int read(byte[] b, int offset, int length) {
+            if (length == 0) return 0;
+
+            switch (state) {
+            case 0:
+              b[offset] = (byte) 0xc2;
+              state = 1;
+              return 1;
+
+            case 1:
+              b[offset] = (byte) 0xae;
+              state = 2;
+              return 1;
+
+            case 2:
+              b[offset] = (byte) 0xea;
+              state = 3;
+              return 1;
+
+            case 3:
+              b[offset] = (byte) 0xba;
+              state = prematureEOS ? 5 : 4;
+              return 1;
+
+            case 4:
+              b[offset] = (byte) 0xaf;
+              state = 5;
+              return 1;
+
+            case 5:
+              return -1;
+
+            default:
+              throw new RuntimeException();
+            }
+          }
+        });
+
+    char[] buffer = new char[2];
+    int offset = 0;
+    while (offset < buffer.length) {
+      int c = r.read(buffer, offset, buffer.length - offset);
+      if (c == -1) break;
+      offset += c;
+    }
+
+    expect(new String(buffer, 0, offset).equals
+           (prematureEOS ? "\u00ae\ufffd" : "\u00ae\uaeaf"));
+  }
+
+  public static void main(String[] args) throws Exception {
     expect(new String(new byte[] { 99, 111, 109, 46, 101, 99, 111, 118, 97,
                                    116, 101, 46, 110, 97, 116, 46, 98, 117,
                                    115, 46, 83, 121, 109, 98, 111, 108 })
@@ -77,5 +136,8 @@ public class Strings {
     expect(Character.forDigit(Character.digit('b', 16), 16) == 'b');
     expect(Character.forDigit(Character.digit('f', 16), 16) == 'f');
     expect(Character.forDigit(Character.digit('z', 36), 36) == 'z');
+
+    testDecode(false);
+    testDecode(true);
   }
 }