Fix regression by reverting back to characters offset rather than byt…

gen.treesitter-ng/src/main/java/com/github/gumtreediff/gen/treesitterng/AbstractTreeSitterNgGenerator.java

@@ -123,8 +123,9 @@ private static int calculateOffset(List<String> contentLines, TSPoint point) {
         for (int i = 0; i < startRow; i++) {
             // Each line in contentLines (except maybe the last) was terminated by LF (\n).
             // If the original was CRLF, the CR (\r) is still at the end of the line string.
-            // .getBytes().length + 1 correctly counts [LineContent] + [LF].
-            offset += contentLines.get(i).getBytes(StandardCharsets.UTF_8).length + 1;
+            // offset must be a char (UTF-16 code unit) index, matching how consumers (e.g.
+            // VanillaDiffHtmlBuilder) walk the source, so use .length() and not byte length.
+            offset += contentLines.get(i).length() + 1;
         }
         offset += startColumn;
         return offset;

gen.treesitter-ng/src/test/java/com/github/gumtreediff/gen/treesitterng/AbstractTreeSitterNgGeneratorTest.java

@@ -61,14 +61,17 @@ public void testCrlfOffsetConsistency() throws IOException {
     @Test
     public void testMultiByteOffsetConsistency() throws IOException {
         // Line 1: "# 🐍\n"
-        // '#' (1) + ' ' (1) + '🐍' (4 bytes in UTF-8) + '\n' (1) = 7 bytes total
+        // '#' (1) + ' ' (1) + '🐍' (2 UTF-16 chars, surrogate pair) + '\n' (1) = 5 chars total
         // Line 2: "x = 1"
+        // Offsets must be char-based (UTF-16 code units) to match how the rest of GumTree
+        // (e.g. AbstractJdtVisitor, VanillaDiffHtmlBuilder) indexes source text, not UTF-8 bytes.
         String content = "# 🐍\nx = 1";
         TreeContext ctx = generator.generateFrom().string(content);
 
         Tree xAssignment = ctx.getRoot().getChild(1);
         assertEquals("expression_statement", xAssignment.getType().name);
-        assertEquals(7, xAssignment.getPos(), "Line 2 should start at byte offset 7 after a 4-byte emoji and LF");
+        assertEquals(5, xAssignment.getPos(),
+                "Line 2 should start at char offset 5 after a surrogate-pair emoji and LF");
     }
 
     @Test