Trying out stricter UTF-8 checks

SVN-Revision: 3831

diff --git a/src/video/font.cpp b/src/video/font.cpp
index 2a41b2a..5e844df 100644 (file)
--- a/src/video/font.cpp
+++ b/src/video/font.cpp
@@ -173,49 +173,66 @@ Font::draw_text(const std::string& text, const Vector& pos,
   draw_chars(chars, text, pos, drawing_effect, alpha);
 }
 
-/** decoding of a byte stream to a single unicode character.
- * This should be correct for well formed utf-8 sequences but doesn't check for
- * all forms of illegal sequences.
- * (see unicode standard section 3.10 table 3-5 and 3-6 for details)
+namespace {
+
+/**
+ * returns true if this byte matches a bitmask of 10xx.xxxx, i.e. it is the 2nd, 3rd or 4th byte of a multibyte utf8 string
+ */
+bool has_multibyte_mark(unsigned char c) {
+  return ((c & 0300) == 0200);
+}
+
+/** 
+ * gets unicode character at byte position @a p of UTF-8 encoded @a text, then advances @a p to the next character.
+ * @throws std::runtime_error if decoding fails.
+ * See unicode standard section 3.10 table 3-5 and 3-6 for details.
  */
 uint32_t decode_utf8(const std::string& text, size_t& p)
 {
-  // 1 byte sequence
-  uint32_t c = (unsigned char) text[p++];
-  if(c <= 0x7F) {
-    return c;
-  }
-  
-  // 2 byte sequence
-  if(p >= text.size())
-    throw std::runtime_error("Malformed utf-8 sequence");
-  uint32_t c2 = (unsigned char) text[p++];
-  if(c <= 0xDF) {
-    if(c < 0xC2)
-      throw std::runtime_error("Malformed utf-8 sequence");
-    return (c & 0x1F) << 6 | (c2 & 0x3F);
+  uint32_t c1 = (unsigned char) text[p+0];
+
+  if (has_multibyte_mark(c1)) std::runtime_error("Malformed utf-8 sequence");
+
+  if ((c1 & 0200) == 0000) {
+    // 0xxx.xxxx: 1 byte sequence
+    p+=1;
+    return c1;
+  } 
+  else if ((c1 & 0340) == 0300) {
+    // 110x.xxxx: 2 byte sequence
+    if(p+1 >= text.size()) throw std::range_error("Malformed utf-8 sequence");
+    uint32_t c2 = (unsigned char) text[p+1];
+    if (!has_multibyte_mark(c2)) throw std::runtime_error("Malformed utf-8 sequence");
+    p+=2;
+    return (c1 & 0037) << 6 | (c2 & 0077);
   }
-  
-  // 3 byte sequence
-  if(p >= text.size())
-    throw std::runtime_error("Malformed utf-8 sequence");
-  uint32_t c3 = (unsigned char) text[p++];
-  if(c <= 0xEF) {
-    return (c & 0x0F) << 12 | (c2 & 0x3F) << 6 | (c3 & 0x3F);
+  else if ((c1 & 0360) == 0340) {
+    // 1110.xxxx: 3 byte sequence
+    if(p+2 >= text.size()) throw std::range_error("Malformed utf-8 sequence");
+    uint32_t c2 = (unsigned char) text[p+1];
+    uint32_t c3 = (unsigned char) text[p+2];
+    if (!has_multibyte_mark(c2)) throw std::runtime_error("Malformed utf-8 sequence");
+    if (!has_multibyte_mark(c3)) throw std::runtime_error("Malformed utf-8 sequence");
+    p+=3;
+    return (c1 & 0017) << 12 | (c2 & 0077) << 6 | (c3 & 0077);
   }
-  
-  // 4 byte sequence
-  if(p >= text.size())
-    throw std::runtime_error("Malformed utf-8 sequence");
-  uint32_t c4 = (unsigned char) text[p++];
-  if(c <= 0xF4) {
-    return (c & 0x07) << 18 | (c2 & 0x3F) << 12 | (c3 & 0x3F) << 6 
-      | (c4 & 0x3F);
+  else if ((c1 & 0370) == 0360) {
+    // 1111.0xxx: 4 byte sequence
+    if(p+3 >= text.size()) throw std::range_error("Malformed utf-8 sequence");
+    uint32_t c2 = (unsigned char) text[p+1];
+    uint32_t c3 = (unsigned char) text[p+2];
+    uint32_t c4 = (unsigned char) text[p+4];
+    if (!has_multibyte_mark(c2)) throw std::runtime_error("Malformed utf-8 sequence");
+    if (!has_multibyte_mark(c3)) throw std::runtime_error("Malformed utf-8 sequence");
+    if (!has_multibyte_mark(c4)) throw std::runtime_error("Malformed utf-8 sequence");
+    p+=4;
+    return (c1 & 0007) << 18 | (c2 & 0077) << 12 | (c3 & 0077) << 6 | (c4 & 0077);
   }
-
   throw std::runtime_error("Malformed utf-8 sequence");
 }
 
+}
+
 void
 Font::draw_chars(Surface* pchars, const std::string& text, const Vector& pos,
                  DrawingEffect drawing_effect, float alpha) const