Backporting WL#4164 Two-byte collation IDs

13 files changed, 102 insertions, 39 deletions

diff --git a/include/my_handler.h b/include/my_handler.h
index a3376cb74a2..ae56331df86 100644
--- a/include/my_handler.h
+++ b/include/my_handler.h
@@ -53,8 +53,8 @@ typedef struct st_HA_KEYSEG		/* Key-portion */
   uint16 bit_pos;                       /* Position to bit part */
   uint16 flag;
   uint16 length;			/* Keylength */
+  uint16 language;
   uint8  type;				/* Type of key (for sort) */
-  uint8  language;
   uint8  null_bit;			/* bitmask to test for NULL */
   uint8  bit_start,bit_end;		/* if bit field */
   uint8  bit_length;                    /* Length of bit part */
diff --git a/include/my_sys.h b/include/my_sys.h
index 4b93dc0e364..34965ae9c67 100644
--- a/include/my_sys.h
+++ b/include/my_sys.h
@@ -231,8 +231,9 @@ extern uint    my_large_page_size;
 #endif
 
 /* charsets */
+#define MY_ALL_CHARSETS_SIZE 2048
 extern MYSQL_PLUGIN_IMPORT CHARSET_INFO *default_charset_info;
-extern MYSQL_PLUGIN_IMPORT CHARSET_INFO *all_charsets[256];
+extern MYSQL_PLUGIN_IMPORT CHARSET_INFO *all_charsets[MY_ALL_CHARSETS_SIZE];
 extern CHARSET_INFO compiled_charsets[];
 
 /* statistics */
diff --git a/include/myisam.h b/include/myisam.h
index 02251eeacb4..70c6f33f118 100644
--- a/include/myisam.h
+++ b/include/myisam.h
@@ -162,7 +162,7 @@ typedef struct st_mi_create_info
   ulonglong data_file_length;
   ulonglong key_file_length;
   uint old_options;
-  uint8 language;
+  uint16 language;
   my_bool with_auto_increment;
 } MI_CREATE_INFO;
 
@@ -410,7 +410,7 @@ typedef struct st_mi_check_param
   uint out_flag,warning_printed,error_printed,verbose;
   uint opt_sort_key,total_files,max_level;
   uint testflag, key_cache_block_size;
-  uint8 language;
+  uint16 language;
   my_bool using_global_keycache, opt_lock_memory, opt_follow_links;
   my_bool retry_repair, force_sort;
   char temp_filename[FN_REFLEN],*isam_file_name;
diff --git a/mysql-test/r/ctype_ldml.result b/mysql-test/r/ctype_ldml.result
index 711921eb526..5d42b64c54c 100644
--- a/mysql-test/r/ctype_ldml.result
+++ b/mysql-test/r/ctype_ldml.result
diff --git a/mysql-test/std_data/Index.xml b/mysql-test/std_data/Index.xml
index 988dddcc68a..c77d8996229 100644
--- a/mysql-test/std_data/Index.xml
+++ b/mysql-test/std_data/Index.xml
@@ -1,23 +1,29 @@
 <charsets>
 
   <charset name="utf8">
-    <collation name="utf8_test_ci" id="240">
+    <collation name="utf8_test_ci" id="353">
       <rules>
         <reset>a</reset>
         <s>b</s>
       </rules>
     </collation>
 
+    <collation name="utf8_maxuserid_ci" id="2047">
+      <rules>
+        <reset>a</reset>
+        <s>b</s>
+      </rules>
+    </collation>
   </charset>
 
   <charset name="ucs2">
-    <collation name="ucs2_test_ci" id="241">
+    <collation name="ucs2_test_ci" id="358">
       <rules>
         <reset>a</reset>
         <s>b</s>
       </rules>
     </collation>
-    <collation name="ucs2_vn_ci" id="242">
+    <collation name="ucs2_vn_ci" id="359">
       <!-- Vietnamese experimental collation -->
       <rules>
         <reset>A</reset>
diff --git a/mysql-test/t/ctype_ldml.test b/mysql-test/t/ctype_ldml.test
index db9461bfbf7..6ebef1d7f16 100644
--- a/mysql-test/t/ctype_ldml.test
+++ b/mysql-test/t/ctype_ldml.test
@@ -49,6 +49,7 @@ DROP TABLE t1;
 #
 #  Vietnamese experimental collation
 #
+--echo  Vietnamese experimental collation
 
 show collation like 'ucs2_vn_ci';
 create table t1 (c1 char(1) character set ucs2 collate ucs2_vn_ci);
@@ -86,3 +87,29 @@ select hex(c1) as h, c1 from t1 order by c1, h;
 select group_concat(hex(c1) order by hex(c1)) from t1 group by c1;
 select group_concat(c1 order by hex(c1) SEPARATOR '') from t1 group by c1;
 drop table t1;
+
+
+-- echo The following tests check that two-byte collation IDs work
+# The file ../std-data/Index.xml has a number of collations with high IDs.
+
+# Test that the "ID" column in I_S and SHOW queries can handle two bytes
+select * from information_schema.collations where id>256 order by id;
+show collation like '%test%';
+
+# Test that two-byte collation ID is correctly transfered to the client side.
+show collation like 'ucs2_vn_ci';
+create table t1 (c1 char(1) character set ucs2 collate ucs2_vn_ci);
+insert into t1 values (0x0061);
+--enable_metadata
+set @@character_set_results=NULL;
+select * from t1;
+--disable_metadata
+drop table t1;
+
+#
+# Check maximum collation ID (2047 as of MySQL-6.0.9)
+#
+CREATE TABLE t1 (s1 char(10) character set utf8 collate utf8_maxuserid_ci);
+INSERT INTO t1 VALUES ('a'),('b');
+SELECT * FROM t1 WHERE s1='a' ORDER BY BINARY s1;
+DROP TABLE t1;
diff --git a/mysys/charset.c b/mysys/charset.c
index 214ca170757..280b2ad6091 100644
--- a/mysys/charset.c
+++ b/mysys/charset.c
@@ -42,7 +42,7 @@ get_collation_number_internal(const char *name)
 {
   CHARSET_INFO **cs;
   for (cs= all_charsets;
-       cs < all_charsets+array_elements(all_charsets)-1 ;
+       cs < all_charsets + array_elements(all_charsets);
        cs++)
   {
     if ( cs[0] && cs[0]->name && 
@@ -387,7 +387,7 @@ char *get_charsets_dir(char *buf)
   DBUG_RETURN(res);
 }
 
-CHARSET_INFO *all_charsets[256]={NULL};
+CHARSET_INFO *all_charsets[MY_ALL_CHARSETS_SIZE]={NULL};
 CHARSET_INFO *default_charset_info = &my_charset_latin1;
 
 void add_compiled_collation(CHARSET_INFO *cs)
@@ -429,7 +429,7 @@ static my_bool init_available_charsets(myf myflags)
       
       /* Copy compiled charsets */
       for (cs=all_charsets;
-           cs < all_charsets+array_elements(all_charsets)-1 ;
+           cs < all_charsets + array_elements(all_charsets);
            cs++)
       {
         if (*cs)
@@ -469,7 +469,7 @@ uint get_charset_number(const char *charset_name, uint cs_flags)
   init_available_charsets(MYF(0));
   
   for (cs= all_charsets;
-       cs < all_charsets+array_elements(all_charsets)-1 ;
+       cs < all_charsets + array_elements(all_charsets);
        cs++)
   {
     if ( cs[0] && cs[0]->csname && (cs[0]->state & cs_flags) &&
@@ -543,7 +543,7 @@ CHARSET_INFO *get_charset(uint cs_number, myf flags)
 
   (void) init_available_charsets(MYF(0));	/* If it isn't initialized */
   
-  if (!cs_number || cs_number >= array_elements(all_charsets)-1)
+  if (!cs_number || cs_number > array_elements(all_charsets))
     return NULL;
   
   cs=get_internal_charset(cs_number, flags);
diff --git a/sql/sql_show.cc b/sql/sql_show.cc
index b16f050dea6..074106dcc44 100644
--- a/sql/sql_show.cc
+++ b/sql/sql_show.cc
@@ -3899,7 +3899,9 @@ int fill_schema_charsets(THD *thd, TABLE_LIST *tables, COND *cond)
   TABLE *table= tables->table;
   CHARSET_INFO *scs= system_charset_info;
 
-  for (cs= all_charsets ; cs < all_charsets+255 ; cs++)
+  for (cs= all_charsets ;
+       cs < all_charsets + array_elements(all_charsets) ;
+       cs++)
   {
     CHARSET_INFO *tmp_cs= cs[0];
     if (tmp_cs && (tmp_cs->state & MY_CS_PRIMARY) && 
@@ -4004,7 +4006,9 @@ int fill_schema_collation(THD *thd, TABLE_LIST *tables, COND *cond)
   const char *wild= thd->lex->wild ? thd->lex->wild->ptr() : NullS;
   TABLE *table= tables->table;
   CHARSET_INFO *scs= system_charset_info;
-  for (cs= all_charsets ; cs < all_charsets+255 ; cs++ )
+  for (cs= all_charsets ;
+       cs < all_charsets + array_elements(all_charsets)  ;
+       cs++ )
   {
     CHARSET_INFO **cl;
     CHARSET_INFO *tmp_cs= cs[0];
@@ -4012,7 +4016,9 @@ int fill_schema_collation(THD *thd, TABLE_LIST *tables, COND *cond)
          (tmp_cs->state & MY_CS_HIDDEN) ||
         !(tmp_cs->state & MY_CS_PRIMARY))
       continue;
-    for (cl= all_charsets; cl < all_charsets+255 ;cl ++)
+    for (cl= all_charsets;
+         cl < all_charsets + array_elements(all_charsets)  ;
+         cl ++)
     {
       CHARSET_INFO *tmp_cl= cl[0];
       if (!tmp_cl || !(tmp_cl->state & MY_CS_AVAILABLE) || 
@@ -4045,17 +4051,22 @@ int fill_schema_coll_charset_app(THD *thd, TABLE_LIST *tables, COND *cond)
   CHARSET_INFO **cs;
   TABLE *table= tables->table;
   CHARSET_INFO *scs= system_charset_info;
-  for (cs= all_charsets ; cs < all_charsets+255 ; cs++ )
+  for (cs= all_charsets ;
+       cs < all_charsets + array_elements(all_charsets) ;
+       cs++ )
   {
     CHARSET_INFO **cl;
     CHARSET_INFO *tmp_cs= cs[0];
     if (!tmp_cs || !(tmp_cs->state & MY_CS_AVAILABLE) || 
         !(tmp_cs->state & MY_CS_PRIMARY))
       continue;
-    for (cl= all_charsets; cl < all_charsets+255 ;cl ++)
+    for (cl= all_charsets;
+         cl < all_charsets + array_elements(all_charsets) ;
+         cl ++)
     {
       CHARSET_INFO *tmp_cl= cl[0];
-      if (!tmp_cl || !(tmp_cl->state & MY_CS_AVAILABLE) || 
+      if (!tmp_cl || !(tmp_cl->state & MY_CS_AVAILABLE) ||
+          (tmp_cl->state & MY_CS_HIDDEN) ||
           !my_charset_same(tmp_cs,tmp_cl))
 	continue;
       restore_record(table, s->default_values);
diff --git a/sql/table.cc b/sql/table.cc
index d71a3ecd9bb..65d46eade8c 100644
--- a/sql/table.cc
+++ b/sql/table.cc
@@ -725,7 +725,8 @@ static int open_binary_frm(THD *thd, TABLE_SHARE *share, uchar *head,
   {
     share->avg_row_length= uint4korr(head+34);
     share->row_type= (row_type) head[40];
-    share->table_charset= get_charset((uint) head[38],MYF(0));
+    share->table_charset= get_charset((((uint) head[41]) << 8) + 
+                                        (uint) head[38],MYF(0));
     share->null_field_first= 1;
   }
   if (!share->table_charset)
@@ -1184,12 +1185,13 @@ static int open_binary_frm(THD *thd, TABLE_SHARE *share, uchar *head,
       }
       else
       {
-        if (!strpos[14])
+        uint csid= strpos[14] + (((uint) strpos[11]) << 8);
+        if (!csid)
           charset= &my_charset_bin;
-        else if (!(charset=get_charset((uint) strpos[14], MYF(0))))
+        else if (!(charset= get_charset(csid, MYF(0))))
         {
           error= 5; // Unknown or unavailable charset
-          errarg= (int) strpos[14];
+          errarg= (int) csid;
           goto err;
         }
       }
@@ -2457,8 +2459,7 @@ File create_frm(THD *thd, const char *name, const char *db,
 
   if ((file= my_create(name, CREATE_MODE, create_flags, MYF(0))) >= 0)
   {
-    uint key_length, tmp_key_length;
-    uint tmp;
+    uint key_length, tmp_key_length, tmp, csid;
     bzero((char*) fileinfo,64);
     /* header */
     fileinfo[0]=(uchar) 254;
@@ -2498,8 +2499,9 @@ File create_frm(THD *thd, const char *name, const char *db,
     fileinfo[32]=0;				// No filename anymore
     fileinfo[33]=5;                             // Mark for 5.0 frm file
     int4store(fileinfo+34,create_info->avg_row_length);
-    fileinfo[38]= (create_info->default_table_charset ?
-		   create_info->default_table_charset->number : 0);
+    csid= (create_info->default_table_charset ?
+           create_info->default_table_charset->number : 0);
+    fileinfo[38]= (uchar) csid;
     /*
       In future versions, we will store in fileinfo[39] the values of the
       TRANSACTIONAL and PAGE_CHECKSUM clauses of CREATE TABLE.
@@ -2507,7 +2509,7 @@ File create_frm(THD *thd, const char *name, const char *db,
     fileinfo[39]= 0;
     fileinfo[40]= (uchar) create_info->row_type;
     /* Next few bytes where for RAID support */
-    fileinfo[41]= 0;
+    fileinfo[41]= (uchar) (csid >> 8);
     fileinfo[42]= 0;
     fileinfo[43]= 0;
     fileinfo[44]= 0;
diff --git a/sql/unireg.cc b/sql/unireg.cc
index 18b0786190a..f08c64a3182 100644
--- a/sql/unireg.cc
+++ b/sql/unireg.cc
@@ -801,20 +801,27 @@ static bool pack_fields(File file, List<Create_field> &create_fields,
     recpos= field->offset+1 + (uint) data_offset;
     int3store(buff+5,recpos);
     int2store(buff+8,field->pack_flag);
-    int2store(buff+10,field->unireg_check);
+    DBUG_ASSERT(field->unireg_check < 256);
+    buff[10]= (uchar) field->unireg_check;
     buff[12]= (uchar) field->interval_id;
     buff[13]= (uchar) field->sql_type; 
     if (field->sql_type == MYSQL_TYPE_GEOMETRY)
     {
+      buff[11]= 0;
       buff[14]= (uchar) field->geom_type;
 #ifndef HAVE_SPATIAL
       DBUG_ASSERT(0);                           // Should newer happen
 #endif
     }
     else if (field->charset) 
+    {
+      buff[11]= (uchar) (field->charset->number >> 8);
       buff[14]= (uchar) field->charset->number;
+    }
     else
-      buff[14]= 0;				// Numerical
+    {
+      buff[11]= buff[14]= 0;			// Numerical
+    }
     int2store(buff+15, field->comment.length);
     comment_length+= field->comment.length;
     set_if_bigger(int_count,field->interval_id);
diff --git a/storage/myisam/ft_static.c b/storage/myisam/ft_static.c
index 1b6c7458fd7..16cbfa1dada 100644
--- a/storage/myisam/ft_static.c
+++ b/storage/myisam/ft_static.c
@@ -30,8 +30,8 @@ const HA_KEYSEG ft_keysegs[FT_SEGS]={
   0,                                            /* Bit pos */
   HA_VAR_LENGTH_PART | HA_PACK_KEY,             /* flag */
   HA_FT_MAXBYTELEN,                             /* length */
-  HA_KEYTYPE_VARTEXT2,                          /* type */
   63,                                           /* language (will be overwritten) */
+  HA_KEYTYPE_VARTEXT2,                          /* type */
   0,                                            /* null_bit */
   2, 0, 0                                       /* bit_start, bit_end, bit_length */
 },
@@ -41,7 +41,7 @@ const HA_KEYSEG ft_keysegs[FT_SEGS]={
       be packed in any way, otherwise w_search() won't be able to
       update key entry 'in vivo'
     */
-  0, 0, 0, 0, HA_NO_SORT, HA_FT_WLEN, HA_FT_WTYPE, 63, 0, 0, 0, 0
+  0, 0, 0, 0, HA_NO_SORT, HA_FT_WLEN, 63, HA_FT_WTYPE, 0, 0, 0, 0
 }
 };
 
diff --git a/storage/myisam/mi_open.c b/storage/myisam/mi_open.c
index e18146f2357..9117d76de24 100644
--- a/storage/myisam/mi_open.c
+++ b/storage/myisam/mi_open.c
@@ -1130,10 +1130,10 @@ int mi_keyseg_write(File file, const HA_KEYSEG *keyseg)
   ulong pos;
 
   *ptr++= keyseg->type;
-  *ptr++= keyseg->language;
+  *ptr++= keyseg->language & 0xFF; /* Collation ID, low byte */
   *ptr++= keyseg->null_bit;
   *ptr++= keyseg->bit_start;
-  *ptr++= keyseg->bit_end;
+  *ptr++= keyseg->language >> 8; /* Collation ID, high byte */
   *ptr++= keyseg->bit_length;
   mi_int2store(ptr,keyseg->flag);	ptr+=2;
   mi_int2store(ptr,keyseg->length);	ptr+=2;
@@ -1152,12 +1152,13 @@ uchar *mi_keyseg_read(uchar *ptr, HA_KEYSEG *keyseg)
    keyseg->language	= *ptr++;
    keyseg->null_bit	= *ptr++;
    keyseg->bit_start	= *ptr++;
-   keyseg->bit_end	= *ptr++;
+   keyseg->language	+= ((uint16) (*ptr++)) << 8;
    keyseg->bit_length   = *ptr++;
    keyseg->flag		= mi_uint2korr(ptr);  ptr +=2;
    keyseg->length	= mi_uint2korr(ptr);  ptr +=2;
    keyseg->start	= mi_uint4korr(ptr);  ptr +=4;
    keyseg->null_pos	= mi_uint4korr(ptr);  ptr +=4;
+   keyseg->bit_end= 0;
    keyseg->charset=0;				/* Will be filled in later */
    if (keyseg->null_bit)
      /* We adjust bit_pos if null_bit is last in the byte */
diff --git a/strings/conf_to_src.c b/strings/conf_to_src.c
index 9f1ed9b2441..73ade93f6e1 100644
--- a/strings/conf_to_src.c
+++ b/strings/conf_to_src.c
@@ -23,7 +23,7 @@
 #define ROW16_LEN	8
 #define MAX_BUF		64*1024
 
-static CHARSET_INFO all_charsets[256];
+static CHARSET_INFO all_charsets[512];
 
 
 void
@@ -63,7 +63,9 @@ print_array16(FILE *f, const char *set, const char *name, uint16 *a, int n)
 static int get_charset_number(const char *charset_name)
 {
   CHARSET_INFO *cs;
-  for (cs= all_charsets; cs < all_charsets+255; ++cs)
+  for (cs= all_charsets;
+       cs < all_charsets + array_elements(all_charsets);
+       cs++)
   {
     if ( cs->name && !strcmp(cs->name, charset_name))
       return cs->number;
@@ -288,7 +290,9 @@ main(int argc, char **argv  __attribute__((unused)))
   sprintf(filename,"%s/%s",argv[1],"Index.xml");
   my_read_charset_file(filename);
   
-  for (cs=all_charsets; cs < all_charsets+256; cs++)
+  for (cs= all_charsets;
+       cs < all_charsets + array_elements(all_charsets);
+       cs++)
   {
     if (cs->number && !(cs->state & MY_CS_COMPILED))
     {
@@ -313,7 +317,9 @@ main(int argc, char **argv  __attribute__((unused)))
   fprintf(f,"#include <m_ctype.h>\n\n");
   
   
-  for (cs=all_charsets; cs < all_charsets+256; cs++)
+  for (cs= all_charsets;
+       cs < all_charsets + array_elements(all_charsets);
+       cs++)
   {
     if (simple_cs_is_full(cs))
     {
@@ -330,7 +336,9 @@ main(int argc, char **argv  __attribute__((unused)))
   }
   
   fprintf(f,"CHARSET_INFO compiled_charsets[] = {\n");
-  for (cs=all_charsets; cs < all_charsets+256; cs++)
+  for (cs= all_charsets;
+       cs < all_charsets + array_elements(all_charsets);
+       cs++)
   {
     if (simple_cs_is_full(cs))
     {