@@ -1,3 +1,7 @@

+Wed Feb 18 23:12:36 CET 2009 (tk)

+---------------------------------

+ * libclamav/textdet.c: improve UTF-16 detection (bb#1209)

+

 Thu Feb 19 00:05:28 EET 2009 (edwin)

 ------------------------------------

  * clamd/server-th.c: move the command parsing, and stream handling

@@ -45,6 +45,7 @@

 #include "filetypes.h"

 #include "textdet.h"

+#include "others.h"

 #define F 0   /* character never appears in text */

 #define T 1   /* character appears in plain ASCII text */

@@ -141,7 +142,7 @@ static int td_isutf8(const unsigned char *buf, unsigned int len)

 static int td_isutf16(const unsigned char *buf, unsigned int len)

 {

-	unsigned int be, i, c;

+	unsigned int be = 1, nobom = 0, i, c, bad = 0;

     if(len < 2)

@@ -152,7 +153,7 @@ static int td_isutf16(const unsigned char *buf, unsigned int len)

     else if(buf[0] == 0xfe && buf[1] == 0xff)

 	be = 1;

     else

-	return 0;

+	nobom = 1;

     for(i = 2; i + 1 < len; i += 2) {

 	if(be)

@@ -163,10 +164,17 @@ static int td_isutf16(const unsigned char *buf, unsigned int len)

 	if(c == 0xfffe)

 	    return 0;

-	if(c < 128 && text_chars[c] != T)

-	    return 0;

+	if(c < 128 && text_chars[c] != T) {

+	    if(nobom)

+		return 0;

+	    else

+		bad++;

+	}

     }

+    if(!nobom && bad >= len / 2)

+	return 0;

+

     return 1 + be;

 }

@@ -174,16 +182,17 @@ cli_file_t cli_texttype(const unsigned char *buf, unsigned int len)

 {

 	int ret;

-    if(td_isutf8(buf, len)) {

+    if(td_isascii(buf, len)) {

+	cli_dbgmsg("Recognized ASCII text\n");

+	return CL_TYPE_TEXT_ASCII;

+    } else if(td_isutf8(buf, len)) {

+	cli_dbgmsg("Recognized UTF-8 character data\n");

 	return CL_TYPE_TEXT_UTF8;

     } else if((ret = td_isutf16(buf, len))) {

-	if(ret == 1)

-	    return CL_TYPE_TEXT_UTF16LE;

-	else

-	    return CL_TYPE_TEXT_UTF16BE;

-    } else if(td_isascii(buf, len)) {

-	return CL_TYPE_TEXT_ASCII;

+	cli_dbgmsg("Recognized %s character data\n", (ret == 1) ? "UTF-16LE" : "UTF-16BE");

+	return (ret == 1) ? CL_TYPE_TEXT_UTF16LE : CL_TYPE_TEXT_UTF16BE;

     } else {

+	cli_dbgmsg("Recognized binary data\n");

 	return CL_TYPE_BINARY_DATA;

     }

 }