A joint patch for:

Bug#51675 Server crashes on inserting 4 byte char. after ALTER TABLE to 'utf8mb4' 
Bug#51676 Server crashes on SELECT, ORDER BY on 'utf8mb4' column

include/m_ctype.h

@@ -38,6 +38,7 @@ extern "C" {
 
 #define my_wc_t ulong
 
+#define MY_CS_REPLACEMENT_CHARACTER 0xFFFD
 
 /*
   On i386 we store Unicode->CS conversion tables for

mysql-test/r/ctype_utf8mb4.result

@@ -2412,6 +2412,62 @@ SELECT CONCAT('a', _utf8 '
 ERROR HY000: Illegal mix of collations (utf8mb4_general_ci,COERCIBLE) and (utf8_general_ci,COERCIBLE) for operation 'concat'
 DROP TABLE t1;
 #
+# Bug#51675 Server crashes on inserting 4 byte char.
+# after ALTER TABLE to 'utf8mb4' 
+#
+SET NAMES utf8;
+CREATE TABLE t1 ( 
+subject varchar(255) CHARACTER SET utf8 COLLATE utf8_unicode_ci, 
+p VARCHAR(15) CHARACTER SET utf8 
+) DEFAULT CHARSET=latin1;
+ALTER TABLE t1 ADD INDEX (subject);
+ALTER TABLE t1
+DEFAULT CHARACTER SET utf8,
+MODIFY subject varchar(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci,
+MODIFY p varchar(255) CHARACTER SET utf8;
+Warnings:
+Warning	1071	Specified key was too long; max key length is 1000 bytes
+Warning	1071	Specified key was too long; max key length is 1000 bytes
+SHOW CREATE TABLE t1;
+Table	Create Table
+t1	CREATE TABLE `t1` (
+  `subject` varchar(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci DEFAULT NULL,
+  `p` varchar(255) DEFAULT NULL,
+  KEY `subject` (`subject`(250))
+) ENGINE=MyISAM DEFAULT CHARSET=utf8
+INSERT INTO t1(subject) VALUES ('abcd');
+INSERT INTO t1(subject) VALUES(x'f0909080');
+DROP TABLE t1;
+#
+# Bug #51676 Server crashes on SELECT, ORDER BY on 'utf8mb4' column
+#
+SET NAMES utf8mb4;
+CREATE TABLE t1 ( 
+subject varchar(255) character set utf8mb4 collate utf8mb4_unicode_ci, 
+p varchar(15) character set utf8mb4 
+) DEFAULT CHARSET=latin1;
+INSERT INTO t1(subject) VALUES(0xF0909080);
+INSERT INTO t1(subject) VALUES(0x616263F0909080646566);
+SHOW CREATE TABLE t1;
+Table	Create Table
+t1	CREATE TABLE `t1` (
+  `subject` varchar(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci DEFAULT NULL,
+  `p` varchar(15) CHARACTER SET utf8mb4 DEFAULT NULL
+) ENGINE=MyISAM DEFAULT CHARSET=latin1
+SELECT *  FROM t1 ORDER BY 1;
+subject	p
+𐐀	NULL
+abc𐐀def	NULL
+SELECT hex(subject), length(subject), char_length(subject), octet_length(subject) FROM t1 ORDER BY 1;
+hex(subject)	length(subject)	char_length(subject)	octet_length(subject)
+616263F0909080646566	10	7	10
+F0909080	4	1	4
+SELECT subject FROM t1 ORDER BY 1;
+subject
+𐐀
+abc𐐀def
+DROP TABLE t1;
+#
 # End of 5.5 tests
 #
 #

mysql-test/t/ctype_utf8mb4.test

@@ -1737,6 +1737,51 @@ SELECT CONCAT(utf8mb4, _utf8 '
 SELECT CONCAT('a', _utf8 '') FROM t1;
 DROP TABLE t1;
 
+
+--echo #
+--echo # Bug#51675 Server crashes on inserting 4 byte char.
+--echo # after ALTER TABLE to 'utf8mb4' 
+--echo #
+SET NAMES utf8;
+CREATE TABLE t1 ( 
+  subject varchar(255) CHARACTER SET utf8 COLLATE utf8_unicode_ci, 
+  p VARCHAR(15) CHARACTER SET utf8 
+) DEFAULT CHARSET=latin1; 
+
+# Alter old table, add index 
+ALTER TABLE t1 ADD INDEX (subject); 
+
+# Alter old 'utf8' table to new 'utf8mb4'
+ALTER TABLE t1
+  DEFAULT CHARACTER SET utf8,
+  MODIFY subject varchar(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci,
+  MODIFY p varchar(255) CHARACTER SET utf8;
+SHOW CREATE TABLE t1; 
+
+INSERT INTO t1(subject) VALUES ('abcd'); 
+# Insert 4 byte characters
+INSERT INTO t1(subject) VALUES(x'f0909080');
+DROP TABLE t1;
+
+--echo #
+--echo # Bug #51676 Server crashes on SELECT, ORDER BY on 'utf8mb4' column
+--echo #
+SET NAMES utf8mb4; 
+CREATE TABLE t1 ( 
+  subject varchar(255) character set utf8mb4 collate utf8mb4_unicode_ci, 
+  p varchar(15) character set utf8mb4 
+) DEFAULT CHARSET=latin1; 
+# Insert 4 byte characters 
+# 4byte character 
+INSERT INTO t1(subject) VALUES(0xF0909080);
+# mix of 3 byte & 4 byte 
+INSERT INTO t1(subject) VALUES(0x616263F0909080646566);
+SHOW CREATE TABLE t1; 
+SELECT *  FROM t1 ORDER BY 1;
+SELECT hex(subject), length(subject), char_length(subject), octet_length(subject) FROM t1 ORDER BY 1; 
+SELECT subject FROM t1 ORDER BY 1;
+DROP TABLE t1;
+
 --echo #
 --echo # End of 5.5 tests
 --echo #

strings/ctype-uca.c

@@ -6983,6 +6983,8 @@ static int my_uca_scanner_next_any(my_uca_scanner *scanner)
                                           scanner->send)) <= 0))
       return -1;
     
+    if (wc > 0xFFFF)
+      wc= MY_CS_REPLACEMENT_CHARACTER;
     scanner->page= wc >> 8;
     scanner->code= wc & 0xFF;
     scanner->sbeg+= mb_len;

strings/ctype-ucs2.c

@@ -50,8 +50,6 @@
 static unsigned long lfactor[9]=
 { 1L, 10L, 100L, 1000L, 10000L, 100000L, 1000000L, 10000000L, 100000000L };
 
-#define REPLACEMENT_CHAR 0xFFFD;
-
 
 
 #ifdef HAVE_CHARSET_mb2_or_mb4
@@ -1145,7 +1143,7 @@ my_tosort_utf16(MY_UNICASE_INFO **uni_plane, my_wc_t *wc)
   }
   else
   {
-    *wc= REPLACEMENT_CHAR;
+    *wc= MY_CS_REPLACEMENT_CHARACTER;
   }
 }
 
@@ -1874,7 +1872,7 @@ my_tosort_utf32(MY_UNICASE_INFO **uni_plane, my_wc_t *wc)
   }
   else
   {
-    *wc= REPLACEMENT_CHAR;
+    *wc= MY_CS_REPLACEMENT_CHARACTER;
   }
 }
 

strings/ctype-utf8.c

@@ -1720,9 +1720,6 @@ MY_UNICASE_INFO *my_unicase_turkish[256]=
 };
 
 
-#define REPLACEMENT_CHAR 0xFFFD;
-
-
 static inline void
 my_tosort_unicode(MY_UNICASE_INFO **uni_plane, my_wc_t *wc)
 {
@@ -1734,7 +1731,7 @@ my_tosort_unicode(MY_UNICASE_INFO **uni_plane, my_wc_t *wc)
   }
   else
   {
-    *wc= REPLACEMENT_CHAR;
+    *wc= MY_CS_REPLACEMENT_CHARACTER;
   }
 }
 
@@ -1757,7 +1754,7 @@ my_wildcmp_unicode(CHARSET_INFO *cs,
 {
   int result= -1;                             /* Not found, using wildcards */
   my_wc_t s_wc, w_wc;
-  int scan, plane;
+  int scan;
   int (*mb_wc)(struct charset_info_st *, my_wc_t *,
                const uchar *, const uchar *);
   mb_wc= cs->cset->mb_wc;