[java] Construct U+FFFF payload at runtime instead of embedding it

Avoids depending on the source file's byte encoding when reading the test.

Author: shs96c

java/test/org/openqa/selenium/json/JsonInputTest.java

@@ -293,14 +293,19 @@ void shouldBeAbleToReadNonWellFormedDataLongerThanReadBuffer() {
 
   @Test
   void shouldReadU_FFFF_AsALiteralCharacterAndNotEndOfInput() {
-    // U+FFFF is a valid (non-)character; historically it collided with an in-band EOF sentinel
-    // and was mis-reported as an unterminated string.
-    try (JsonInput input = newInput("\"a￿b\"")) {
-      assertThat(input.nextString()).isEqualTo("a￿b");
+    // U+FFFF is a valid Unicode code unit that historically collided with the in-band EOF
+    // sentinel and was mis-reported as an unterminated string. Build the strings from
+    // char values rather than embedding literal U+FFFF so the test is independent of the
+    // source file's byte encoding.
+    char nonChar = (char) 0xFFFF;
+    String literalPayload = "a" + nonChar + "b";
+
+    try (JsonInput input = newInput("\"" + literalPayload + "\"")) {
+      assertThat(input.nextString()).isEqualTo(literalPayload);
     }
 
     try (JsonInput input = newInput("\"\\uFFFF\"")) {
-      assertThat(input.nextString()).isEqualTo("");
+      assertThat(input.nextString()).isEqualTo(String.valueOf(nonChar));
     }
   }