From c9490550357a3d2f00b05e7e3ad46f9cbea9d68d Mon Sep 17 00:00:00 2001
From: "jimw@mysql.com" <>
Date: Thu, 3 Feb 2005 16:07:32 -0800
Subject: [PATCH] Fix error in parsing string literals containing a backslash
 followed by a multi-byte character with a second byte of 0x5c (\). (Bug
 #8903)

---
 mysql-test/r/ctype_sjis.result |  4 ++++
 mysql-test/t/ctype_sjis.test   |  7 +++++++
 sql/sql_lex.cc                 | 32 ++++++++++++++++++++++++++------
 3 files changed, 37 insertions(+), 6 deletions(-)

diff --git a/mysql-test/r/ctype_sjis.result b/mysql-test/r/ctype_sjis.result
index 1f414f89e20..98e5992f374 100644
--- a/mysql-test/r/ctype_sjis.result
+++ b/mysql-test/r/ctype_sjis.result
@@ -91,3 +91,7 @@ sjis_bin	6109
 sjis_bin	61
 sjis_bin	6120
 drop table t1;
+SET NAMES sjis;
+SELECT HEX('�����@\�\') FROM DUAL;
+HEX('�����@�_�\')
+8DB2939181408C5C
diff --git a/mysql-test/t/ctype_sjis.test b/mysql-test/t/ctype_sjis.test
index 58ca3c6a997..50d286f28b9 100644
--- a/mysql-test/t/ctype_sjis.test
+++ b/mysql-test/t/ctype_sjis.test
@@ -68,3 +68,10 @@ SET collation_connection='sjis_japanese_ci';
 -- source include/ctype_filesort.inc
 SET collation_connection='sjis_bin';
 -- source include/ctype_filesort.inc
+
+# Check parsing of string literals in SJIS with multibyte characters that
+# have an embedded \ in them. (Bug #8303)
+
+--character_set sjis
+SET NAMES sjis;
+SELECT HEX('�����@\�\') FROM DUAL;
diff --git a/sql/sql_lex.cc b/sql/sql_lex.cc
index ed974a48ad3..d6dcd9ce9ae 100644
--- a/sql/sql_lex.cc
+++ b/sql/sql_lex.cc
@@ -295,7 +295,18 @@ static char *get_text(LEX *lex)
       found_escape=1;
       if (lex->ptr == lex->end_of_query)
 	return 0;
-      yySkip();
+#ifdef USE_MB
+      int l;
+      if (use_mb(cs) &&
+          (l = my_ismbchar(cs,
+                           (const char *)lex->ptr,
+                           (const char *)lex->end_of_query))) {
+          lex->ptr += l;
+          continue;
+      }
+      else
+#endif
+        yySkip();
     }
     else if (c == sep)
     {
@@ -323,6 +334,10 @@ static char *get_text(LEX *lex)
       else
       {
 	uchar *to;
+
+        /* Re-use found_escape for tracking state of escapes */
+        found_escape= 0;
+
 	for (to=start ; str != end ; str++)
 	{
 #ifdef USE_MB
@@ -336,7 +351,7 @@ static char *get_text(LEX *lex)
 	      continue;
 	  }
 #endif
-	  if (*str == '\\' && str+1 != end)
+	  if (!found_escape && *str == '\\' && str+1 != end)
 	  {
 	    switch(*++str) {
 	    case 'n':
@@ -362,15 +377,20 @@ static char *get_text(LEX *lex)
 	      *to++= '\\';		// remember prefix for wildcard
 	      /* Fall through */
 	    default:
-	      *to++ = *str;
+              found_escape= 1;
+              str--;
 	      break;
 	    }
 	  }
-	  else if (*str == sep)
-	    *to++= *str++;		// Two ' or "
+	  else if (!found_escape && *str == sep)
+          {
+            found_escape= 1;
+          }
 	  else
+          {
 	    *to++ = *str;
-
+            found_escape= 0;
+          }
 	}
 	*to=0;
 	lex->yytoklen=(uint) (to-start);