Fix Bluetooth device name max length checking.

The Bluetooth device name has a maximum length of 248 bytes in UTF-8
encoding. The settings app previously limited the Bluetooth device
name entered by the user based on the number of characters, which is
incorrect when the name includes non-ASCII characters.

Created a new Utf8ByteLengthFilter class for BluetoothNamePreference
which counts UTF-8 bytes instead of Unicode characters, along with
unit tests.

We can move Utf8ByteLengthFilter into the framework in the future
if it would be useful elsewhere in the system or for applications.

Change-Id: I9debface09659f09d750f3bfe8ad6ddfd34c75c5

src/com/android/settings/bluetooth/BluetoothNamePreference.java

@@ -27,8 +27,8 @@ import android.content.IntentFilter;
 import android.preference.EditTextPreference;
 import android.text.Editable;
 import android.text.InputFilter;
+import android.text.Spanned;
 import android.text.TextWatcher;
-import android.text.InputFilter.LengthFilter;
 import android.util.AttributeSet;
 import android.widget.Button;
 import android.widget.EditText;
@@ -40,8 +40,7 @@ import android.widget.EditText;
  */
 public class BluetoothNamePreference extends EditTextPreference implements TextWatcher {
     private static final String TAG = "BluetoothNamePreference";
-    // TODO(): Investigate bluetoothd/dbus crash when length is set to 248, limit as per spec.
-    private static final int BLUETOOTH_NAME_MAX_LENGTH = 200;
+    private static final int BLUETOOTH_NAME_MAX_LENGTH_BYTES = 248;
 
     private LocalBluetoothManager mLocalManager;
 
@@ -75,8 +74,11 @@ public class BluetoothNamePreference extends EditTextPreference implements TextW
 
         // Make sure the OK button is disabled (if necessary) after rotation
         EditText et = getEditText();
-        et.setFilters(new InputFilter[] {new LengthFilter(BLUETOOTH_NAME_MAX_LENGTH)});
         if (et != null) {
+            et.setFilters(new InputFilter[] {
+                    new Utf8ByteLengthFilter(BLUETOOTH_NAME_MAX_LENGTH_BYTES)
+            });
+
             et.addTextChangedListener(this);
             Dialog d = getDialog();
             if (d instanceof AlertDialog) {
@@ -136,4 +138,68 @@ public class BluetoothNamePreference extends EditTextPreference implements TextW
     public void onTextChanged(CharSequence s, int start, int before, int count) {
         // not used
     }
+
+    /**
+     * This filter will constrain edits so that the text length is not
+     * greater than the specified number of bytes using UTF-8 encoding.
+     * <p>The JNI method used by {@link android.server.BluetoothService}
+     * to convert UTF-16 to UTF-8 doesn't support surrogate pairs,
+     * therefore code points outside of the basic multilingual plane
+     * (0000-FFFF) will be encoded as a pair of 3-byte UTF-8 characters,
+     * rather than a single 4-byte UTF-8 encoding. Dalvik implements this
+     * conversion in {@code convertUtf16ToUtf8()} in
+     * {@code dalvik/vm/UtfString.c}.
+     * <p>This JNI method is unlikely to change in the future due to
+     * backwards compatibility requirements. It's also unclear whether
+     * the installed base of Bluetooth devices would correctly handle the
+     * encoding of surrogate pairs in UTF-8 as 4 bytes rather than 6.
+     * However, this filter will still work in scenarios where surrogate
+     * pairs are encoded as 4 bytes, with the caveat that the maximum
+     * length will be constrained more conservatively than necessary.
+     */
+    public static class Utf8ByteLengthFilter implements InputFilter {
+        private int mMaxBytes;
+
+        public Utf8ByteLengthFilter(int maxBytes) {
+            mMaxBytes = maxBytes;
+        }
+
+        public CharSequence filter(CharSequence source, int start, int end,
+                                   Spanned dest, int dstart, int dend) {
+            int srcByteCount = 0;
+            // count UTF-8 bytes in source substring
+            for (int i = start; i < end; i++) {
+                char c = source.charAt(i);
+                srcByteCount += (c < 0x0080) ? 1 : (c < 0x0800 ? 2 : 3);
+            }
+            int destLen = dest.length();
+            int destByteCount = 0;
+            // count UTF-8 bytes in destination excluding replaced section
+            for (int i = 0; i < destLen; i++) {
+                if (i < dstart || i >= dend) {
+                    char c = dest.charAt(i);
+                    destByteCount += (c < 0x0080) ? 1 : (c < 0x0800 ? 2 : 3);
+                }
+            }
+            int keepBytes = mMaxBytes - destByteCount;
+            if (keepBytes <= 0) {
+                return "";
+            } else if (keepBytes >= srcByteCount) {
+                return null; // use original dest string
+            } else {
+                // find end position of largest sequence that fits in keepBytes
+                for (int i = start; i < end; i++) {
+                    char c = source.charAt(i);
+                    keepBytes -= (c < 0x0080) ? 1 : (c < 0x0800 ? 2 : 3);
+                    if (keepBytes < 0) {
+                        return source.subSequence(start, i);
+                    }
+                }
+                // If the entire substring fits, we should have returned null
+                // above, so this line should not be reached. If for some
+                // reason it is, return null to use the original dest string.
+                return null;
+            }
+        }
+    }
 }