Fixed unique prefix key bug for multibyte character sets (BUG #4521) for

InnoDB. This fixes also a second part of the same problem with prefix keys
on a multibyte string column for InnoDB.


innobase/btr/btr0btr.c:
  Multibyte character set prefix indexes are not any more fixed size. Therefore,
  we have to chect that length of the index field in not greater than
  prefix length.
innobase/rem/rem0cmp.c:
  Remove unnecessary changes.
innobase/row/row0ins.c:
  Fixed unique prefix key or prefix key using multibyte character set bugs for
  InnoDB (BUG #4521). For prefix keys we have to get the storage length
  for the prefix length of characters in the key.
innobase/row/row0row.c:
  Fixed unique prefix key or prefix key using multibyte character set bugs for
  InnoDB (BUG #4521). For prefix keys we have to get the storage length
  for the prefix length of characters in the key.
innobase/row/row0sel.c:
  Fixed unique prefix key or prefix key using multibyte character set bugs for
  InnoDB (BUG #4521). For prefix keys we have to get the storage length
  for the prefix length of characters in the key.
innobase/row/row0upd.c:
  Fixed unique prefix key or prefix key using multibyte character set bugs for
  InnoDB (BUG #4521). For prefix keys we have to get the storage length
  for the prefix length of characters in the key.
mysql-test/r/ctype_utf8.result:
  Added utf8 character test cases for InnoDB.
mysql-test/t/ctype_utf8.test:
  Added utf8 character expected test results for InnoDB.
sql/ha_innodb.cc:
  Added function innobase_get_at_most_n_mbchars to return position of
  the nth character in the multibyte character string.
sql/ha_innodb.h:
  Remove unnecessary changes.

1 files changed, 2 insertions, 32 deletions

diff --git a/innobase/rem/rem0cmp.c b/innobase/rem/rem0cmp.c
index f6c82102839..041fb7914e2 100644
--- a/innobase/rem/rem0cmp.c
+++ b/innobase/rem/rem0cmp.c
@@ -14,9 +14,6 @@ Created 7/1/1994 Heikki Tuuri
 
 #include "srv0srv.h"
 
-#include <m_ctype.h>
-#include <my_sys.h>
-
 /*		ALPHABETICAL ORDER
 		==================
 		
@@ -455,8 +452,6 @@ cmp_dtuple_rec_with_match(
 	ulint		cur_bytes; 	/* number of already matched bytes 
 					in current field */
 	int		ret = 3333;	/* return value */
-	
-	CHARSET_INFO*   charset;        /* charset used in the field */
 
 	ut_ad(dtuple && rec && matched_fields && matched_bytes);
 	ut_ad(dtuple_check_typed(dtuple));
@@ -546,33 +541,8 @@ cmp_dtuple_rec_with_match(
 			&& dtype_get_charset_coll(cur_type->prtype) !=
 				data_mysql_latin1_swedish_charset_coll)) {
 
-		  	/* If character set is not latin1_swedish
-			we have to devide character length by the
-			maximum bytes needed for that character
-			set. For example if we have unique prefix
-			index for 1 utf8 character then we have
-			actually 3 bytes allocated in the index.
-			Therefore, we have to divide that with
-			maximum bytes needed for utf8 character i.e.
-			3 byges.*/
-
-                        if ( dtuple_f_len > 0) {
-			  charset = get_charset(
-				dtype_get_charset_coll(cur_type->prtype), 
-				MYF(MY_WME));
-
-			  ut_ad(charset);
-			  ut_ad(charset->mbmaxlen);
-
-			  dtuple_f_len = dtuple_f_len / charset->mbmaxlen;
-
-                          if ( dtuple_f_len == 0)
-			    dtuple_f_len = 1;
-
-			  rec_f_len = dtuple_f_len;
-			} 
-
-			ret = cmp_whole_field(cur_type,
+			ret = cmp_whole_field(
+				cur_type,
 				dfield_get_data(dtuple_field), dtuple_f_len,
 				rec_b_ptr, rec_f_len);