Fix various UTF8 parsing bugs.

In particular, parse bad UTF8 by replacing the broken bits with U+FFFD
and resychronise correctly after broken sequences.

diff --git a/jv.c b/jv.c
index 9316aec1a7385e5374835ffc12c68958ac245252..0c69a5684adbb2b6bc149bec7d426087bd913ea6 100644 (file)
--- a/jv.c
+++ b/jv.c
@@ -377,6 +377,32 @@ static jvp_string* jvp_string_alloc(uint32_t size) {
   return s;
 }
 
+/* Copy a UTF8 string, replacing all badly encoded points with U+FFFD */
+static jv_nontrivial jvp_string_copy_replace_bad(const char* data, uint32_t length) {
+  const char* end = data + length;
+  const char* i = data;
+  const char* cstart;
+
+  uint32_t maxlength = length * 3 + 1; // worst case: all bad bytes, each becomes a 3-byte U+FFFD
+  jvp_string* s = jvp_string_alloc(maxlength);
+  char* out = s->data;
+  int c = 0;
+
+  while ((i = jvp_utf8_next((cstart = i), end, &c))) {
+    if (c == -1) {
+      c = 0xFFFD; // U+FFFD REPLACEMENT CHARACTER
+    }
+    out += jvp_utf8_encode(c, out);
+    assert(out < s->data + maxlength);
+  }
+  length = out - s->data;
+  s->data[length] = 0;
+  s->length_hashed = length << 1;
+  jv_nontrivial r = {&s->refcnt, {0,0}};
+  return r;
+}
+
+/* Assumes valid UTF8 */
 static jv_nontrivial jvp_string_new(const char* data, uint32_t length) {
   jvp_string* s = jvp_string_alloc(length);
   s->length_hashed = length << 1;
@@ -523,7 +549,9 @@ static int jvp_string_equal(jv_nontrivial* a, jv_nontrivial* b) {
 jv jv_string_sized(const char* str, int len) {
   jv j;
   j.kind = JV_KIND_STRING;
-  j.val.nontrivial = jvp_string_new(str, len);
+  j.val.nontrivial = jvp_utf8_is_valid(str, str+len) ? 
+    jvp_string_new(str, len) :
+    jvp_string_copy_replace_bad(str, len);
   return j;
 }
 
@@ -568,14 +596,21 @@ jv jv_string_concat(jv a, jv b) {
 }
 
 jv jv_string_append_buf(jv a, const char* buf, int len) {
-  jvp_string_append(&a.val.nontrivial, buf, len);
+  if (jvp_utf8_is_valid(buf, buf+len)) {
+    jvp_string_append(&a.val.nontrivial, buf, len);
+  } else {
+    jv b;
+    b.kind = JV_KIND_STRING;
+    b.val.nontrivial = jvp_string_copy_replace_bad(buf, len);
+    a = jv_string_concat(a, b);
+  }
   return a;
 }
 
 jv jv_string_append_str(jv a, const char* str) {
   return jv_string_append_buf(a, str, strlen(str));
 }
-                        
+
 jv jv_string_fmt(const char* fmt, ...) {
   int size = 1024;
   while (1) {

diff --git a/jv_unicode.c b/jv_unicode.c
index 375ad36702aa46b1277f7583cb05830d720c39ec..a5305cf1d68f7abe23c064e7a925f88274c5d6b2 100644 (file)
--- a/jv_unicode.c
+++ b/jv_unicode.c
@@ -3,35 +3,56 @@
 #include "jv_unicode.h"
 #include "jv_utf8_tables.h"
 
-const char* jvp_utf8_next(const char* in, const char* end, int* codepoint) {
+const char* jvp_utf8_next(const char* in, const char* end, int* codepoint_ret) {
+  assert(in <= end);
   if (in == end) {
-    codepoint = 0;
     return 0;
   }
+  int codepoint = -1;
   unsigned char first = (unsigned char)in[0];
   int length = utf8_coding_length[first];
-  if (length == 0 || length == UTF8_CONTINUATION_BYTE || in + length > end) {
-    *codepoint = -1;
-    return 0;
-  }
-  *codepoint = ((unsigned)in[0]) & utf8_coding_bits[first];
-  for (int i=1; i<length; i++) {
-    int ch = (unsigned char)in[i];
-    if (utf8_coding_length[(unsigned char)in[i]] != UTF8_CONTINUATION_BYTE){
-      *codepoint = -1;
-      return 0;
+  if ((first & 0x80) == 0) {
+    /* Fast-path for ASCII */
+    codepoint = first;
+    length = 1;
+  } else if (length == 0 || length == UTF8_CONTINUATION_BYTE) {
+    /* Bad single byte - either an invalid byte or an out-of-place continuation byte */
+    length = 1;
+  } else if (in + length > end) {
+    /* String ends before UTF8 sequence ends */
+    length = end - in;
+  } else {
+    codepoint = ((unsigned)in[0]) & utf8_coding_bits[first];
+    for (int i=1; i<length; i++) {
+      unsigned ch = (unsigned char)in[i];
+      if (utf8_coding_length[ch] != UTF8_CONTINUATION_BYTE){
+        /* Invalid UTF8 sequence - not followed by the right number of continuation bytes */
+        codepoint = -1;
+        length = i;
+        break;
+      }
+      codepoint = (codepoint << 6) | (ch & 0x3f);
+    }
+    if (0xD800 <= codepoint && codepoint <= 0xDFFF) {
+      /* Surrogate codepoints can't be encoded in UTF8 */
+      codepoint = -1;
+    }
+    if (codepoint > 0x10FFFF) {
+      /* Outside Unicode range */
+      codepoint = -1;
     }
-    *codepoint = (*codepoint << 6) | (ch & 0x3f);
   }
+  assert(length > 0);
+  *codepoint_ret = codepoint;
   return in + length;
 }
 
-int jvp_utf8_verify(const char* in, const char* end) {
-  int codepoint = 0;
+int jvp_utf8_is_valid(const char* in, const char* end) {
+  int codepoint;
   while ((in = jvp_utf8_next(in, end, &codepoint))) {
     if (codepoint == -1) return 0;
   }
-  return codepoint != -1;
+  return 1;
 }
 
 int jvp_utf8_encode_length(int codepoint) {

diff --git a/jv_unicode.h b/jv_unicode.h
index 78c7a40d19b70440533330e7b7eb02684959bf4c..579c910eead9e8877e744b84148d3e8ace2c6a48 100644 (file)
--- a/jv_unicode.h
+++ b/jv_unicode.h
@@ -2,7 +2,7 @@
 #define JV_UNICODE_H
 
 const char* jvp_utf8_next(const char* in, const char* end, int* codepoint);
-
+int jvp_utf8_is_valid(const char* in, const char* end);
 
 int jvp_utf8_decode_length(char startchar);