JS: Fix parsing of non-BMP chars before a quantifier

Author: asger-semmle

javascript/extractor/src/com/semmle/js/parser/RegExpParser.java

@@ -326,10 +326,21 @@ private RegExpTerm parseAtom() {
       this.error(Error.UNEXPECTED_CHARACTER, endPos);
       endPos = startPos + 1; // To ensure progress, make sure we parse at least one character.
     }
+    // Check if the end of the constant belongs under an upcoming quantifier.
     if (endPos != startPos + 1
         && endPos < src.length()
         && "*+?{".indexOf(src.charAt(endPos)) != -1) {
-      endPos--; // Last constant belongs under an upcoming quantifier.
+      if (Character.isLowSurrogate(src.charAt(endPos - 1))
+          && Character.isHighSurrogate(src.charAt(endPos - 2))) {
+        // Don't split the surrogate pair.
+        if (endPos == startPos + 2) {
+          // The whole constant is a single wide character.
+        } else {
+          endPos -= 2; // Last 2 characters belong to an upcoming quantifier.
+        }
+      } else {
+        endPos--; // Last character belongs to an upcoming quantifier.
+      }
     }
     String str = src.substring(startPos, endPos);
     this.pos = endPos;

javascript/ql/test/library-tests/RegExp/MissingUnicodeFlag/MissingUnicodeFlag.expected

@@ -1 +1,2 @@
-| tst.js:1:1:1:9 | /[\\u12340\\udf40-\\u12345\\udf45]/ | Character class with supplementary characters in non-unicode literal. |
+| tst.js:1:1:1:9 | /[\\u12340\\udf40-\\u12345\\udf45]/ | Split supplementary character in non-unicode literal. |
+| tst.js:3:1:3:5 | /\\u12340\\udf40+/ | Split supplementary character in non-unicode literal. |

javascript/ql/test/library-tests/RegExp/MissingUnicodeFlag/MissingUnicodeFlag.ql

@@ -8,5 +8,7 @@ where wideConstant.getLiteral() = literal and
     wideConstant.getParent() instanceof RegExpCharacterClass
     or
     wideConstant.getParent() instanceof RegExpCharacterRange
+    or
+    wideConstant.getParent() instanceof RegExpQuantifier
   )
-select literal, "Character class with supplementary characters in non-unicode literal."
+select literal, "Split supplementary character in non-unicode literal."

javascript/ql/test/library-tests/RegExp/MissingUnicodeFlag/tst.js

@@ -1,2 +1,5 @@
 /[𒍀-𒍅]/; // NOT OK
 /[𒍀-𒍅]/u; // OK
+/𒍀+/; // NOT OK
+/𒍀+/u; // OK
+/(𒍀)+/; // OK