Fixed several bugs in UTF-16 encoding/decoding (tested on large files).

diff --git a/charset/charconv-gen.h b/charset/charconv-gen.h
index 5a83232f7aaad432281d8969e34f7e807717b349..7a7f0eff8f78b9c6d110292183fc19b99a61b13f 100644 (file)
--- a/charset/charconv-gen.h
+++ b/charset/charconv-gen.h
@@ -13,6 +13,7 @@
 do {
 
 /*** Header ***/
+
   const byte *s, *se;
   byte *d, *de;
   uns code;
@@ -193,6 +194,7 @@ got_code:
       else
         {
 write_slow:
+         c->code = code;
          c->state = UTF16_BE_WRITE;
          goto go_slow;
        }
@@ -217,6 +219,7 @@ write_slow:
       else
         {
 write_slow:
+         c->code = code;
          c->state = UTF16_LE_WRITE;
          goto go_slow;
        }

diff --git a/charset/charconv.c b/charset/charconv.c
index ce22d1e08ab447274604756e5b3d09972e798e90..54dac42b74383bb65cbe9b8ef1157b0824645804 100644 (file)
--- a/charset/charconv.c
+++ b/charset/charconv.c
@@ -89,6 +89,7 @@ seq:
        }
       if (c->code >= 0x10000)
        c->code = 0xfffd;
+got_char:
       c->source = s;
       c->state = 0;
       return -1;
@@ -133,19 +134,23 @@ seq:
       {
        void *p = &c->code;
        c->string_at = p;
-       if (c->code < 0xd800 || c->code - 0xe000 < 0x2000)
+       uns code = c->code;
+       c->string_at = p;
+       if (code < 0xd800 || code - 0xe000 < 0x2000)
          {}
-       else if ((c->code -= 0x10000) < 0x100000)
+       else if ((code -= 0x10000) < 0x100000)
          {
-           put_u16_be(p, 0xd800 | (c->code >> 10));
-           put_u16_be(p + 2, 0xdc00 | (c->code & 0x3ff));
+           put_u16_be(p, 0xd800 | (code >> 10));
+           put_u16_be(p + 2, 0xdc00 | (code & 0x3ff));
            c->remains = 4;
+           c->state = SEQ_WRITE;
            goto seq;
          }
        else
-         c->code = UNI_REPLACEMENT;
-       put_u16_be(p, c->code);
+         code = UNI_REPLACEMENT;
+       put_u16_be(p, code);
        c->remains = 2;
+        c->state = SEQ_WRITE;
        goto seq;
       }
 
@@ -154,18 +159,22 @@ seq:
       {
        void *p = &c->code;
        c->string_at = p;
-       if (c->code < 0xd800 || c->code - 0xe000 < 0x2000)
+       uns code = c->code;
+       c->string_at = p;
+       if (code < 0xd800 || code - 0xe000 < 0x2000)
          {}
-       else if ((c->code -= 0x10000) < 0x100000)
+       else if ((code -= 0x10000) < 0x100000)
          {
-           put_u16_le(p, 0xd800 | (c->code >> 10));
-           put_u16_le(p + 2, 0xdc00 | (c->code & 0x3ff));
+           put_u16_le(p, 0xd800 | (code >> 10));
+           put_u16_le(p + 2, 0xdc00 | (code & 0x3ff));
            c->remains = 4;
+            c->state = SEQ_WRITE;
          }
        else
-         c->code = UNI_REPLACEMENT;
-        put_u16_le(p, c->code);
+         code = UNI_REPLACEMENT;
+        put_u16_le(p, code);
         c->remains = 2;
+        c->state = SEQ_WRITE;
        goto seq;
       }
 
@@ -181,7 +190,7 @@ seq:
        goto cse;
       c->code = (c->code << 8) | *s++;
       if (c->code - 0xd800 >= 0x800)
-       break;
+        goto got_char;
       c->code = (c->code - 0xd800) << 10;
       c->state = UTF16_BE_READ_2;
       /* fall-thru */
@@ -203,7 +212,7 @@ seq:
       else
        c->code = UNI_REPLACEMENT;
       s++;
-      break;
+      goto got_char;
 
     /* Reading of UTF16-LE */
     case UTF16_LE_READ:
@@ -217,7 +226,7 @@ seq:
        goto cse;
       c->code |= *s++ << 8;
       if (c->code - 0xd800 >= 0x800)
-       break;
+        goto got_char;
       c->code = (c->code - 0xd800) << 10;
       c->state = UTF16_LE_READ_2;
       /* fall-thru */
@@ -235,7 +244,7 @@ seq:
       else
        c->code = UNI_REPLACEMENT;
       s++;
-      break;
+      goto got_char;
 
     default:
       ASSERT(0);