Fix a bug with javadoc comments followed by unicode escapes

The lexer was advancing to the next character and then reading the string
contents of the javadoc comment. It assumed that the next token
corresponded to exactly one character in the input stream, which isn't
true of unicode escapes.

PiperOrigin-RevId: 499989979

2 files changed, 12 insertions, 3 deletions

diff --git a/java/com/google/turbine/parse/StreamLexer.java b/java/com/google/turbine/parse/StreamLexer.java
index 3d46b90..ed79dd0 100644
--- a/java/com/google/turbine/parse/StreamLexer.java
+++ b/java/com/google/turbine/parse/StreamLexer.java
@@ -75,8 +75,8 @@ public class StreamLexer implements Lexer {
     if (result == null) {
       return null;
     }
-    verify(result.endsWith("*/"), result);
-    return result.substring(0, result.length() - "*/".length());
+    verify(result.endsWith("*"), result);
+    return result.substring(0, result.length() - "*".length());
   }
 
   @Override
@@ -153,16 +153,18 @@ public class StreamLexer implements Lexer {
                       sawStar = true;
                       break;
                     case '/':
-                      eat();
                       if (sawStar) {
                         if (isJavadoc) {
                           // Save the comment, excluding the leading `/**` and including
                           // the trailing `/*`. The comment is trimmed and normalized later.
                           javadoc = stringValue();
+                          verify(javadoc.endsWith("*"), javadoc);
                         }
+                        eat();
                         continue OUTER;
                       }
                       sawStar = false;
+                      eat();
                       break;
                     case ASCII_SUB:
                       if (reader.done()) {
diff --git a/javatests/com/google/turbine/parse/LexerTest.java b/javatests/com/google/turbine/parse/LexerTest.java
index bf0b374..6a6fe1c 100644
--- a/javatests/com/google/turbine/parse/LexerTest.java
+++ b/javatests/com/google/turbine/parse/LexerTest.java
@@ -339,6 +339,11 @@ public class LexerTest {
     lexerComparisonTest("import pkg\uD800\uDC00.test;");
   }
 
+  @Test
+  public void javadocUnicodeEscape() {
+    lexerComparisonTest("class {/***/\\u007D;");
+  }
+
   private void lexerComparisonTest(String s) {
     assertThat(lex(s)).containsExactlyElementsIn(JavacLexer.javacLex(s));
   }
@@ -349,6 +354,8 @@ public class LexerTest {
     Token token;
     do {
       token = lexer.next();
+      // Just check that javadoc handling doesn't crash
+      String unused = lexer.javadoc();
       String tokenString;
       switch (token) {
         case IDENT: