Fix bug when repairing compressed MyISAM files

LOCATE() is now case sensitive

BUILD/compile-alpha-cxx

@@ -4,7 +4,7 @@ make -k clean
 /bin/rm -f */.deps/*.P config.cache innobase/config.cache bdb/build_unix/config.cache mysql-*.tar.gz
 aclocal; autoheader; aclocal; automake; autoconf
 
-CC=ccc CFLAGS="-fast" CXX=cxx CXXFLAGS="-fast -noexceptions -nortti" ./configure --prefix=/usr/local/mysql --disable-shared --with-extra-charsets=complex --enable-thread-safe-client --with-mysqld-ldflags=-non_shared --with-client-ldflags=-non_shared
+CC=ccc CFLAGS="-fast" CXX=cxx CXXFLAGS="-fast -noexceptions -nortti" ./configure --prefix=/usr/local/mysql --disable-shared --with-extra-charsets=complex --enable-thread-safe-client --with-mysqld-ldflags=-non_shared --with-client-ldflags=-non_shared --without-extra-tools
 
 make -j2
 find . -name ".deps" | xargs rm -r 

Docs/manual.texi

@@ -6953,6 +6953,9 @@ version 4.0;
 
 @itemize @bullet
 @item
+@code{LOCATE()} and @code{INSTR()} are case sensitive if neither
+argument is a binary string.  binary strings.
+@item
 @code{INSERT INTO ... SELECT} had in 3.23 always @code{IGNORE} enabled.
 In 4.0.1 MySQL will stop (and possible rollback) in case of an error if you
 don't specify @code{IGNORE}.
@@ -28621,7 +28624,9 @@ mysql> select LOCATE('xbar', 'foobar');
         -> 0
 @end example
 
-This function is multi-byte safe.
+This function is multi-byte safe.  In MySQL 3.23 this function is case
+insensitive, while in 4.0 it's only case insensitive if either argument is
+a binary string.
 
 @findex LOCATE()
 @item LOCATE(substr,str,pos)
@@ -28634,7 +28639,9 @@ mysql> select LOCATE('bar', 'foobarbar',5);
         -> 7
 @end example
 
-This function is multi-byte safe.
+This function is multi-byte safe.  In MySQL 3.23 this function is case
+insensitive, while in 4.0 it's only case insensitive if either argument is
+a binary string.
 
 @findex INSTR()
 @item INSTR(str,substr)
@@ -28649,7 +28656,9 @@ mysql> select INSTR('xbar', 'foobar');
         -> 0
 @end example
 
-This function is multi-byte safe.
+This function is multi-byte safe.  In MySQL 3.23 this function is case
+insensitive, while in 4.0 it's only case insensitive if either argument is
+a binary string.
 
 @findex LPAD()
 @item LPAD(str,len,padstr)
@@ -45786,6 +45795,9 @@ Our TODO section contains what we plan to have in 4.0. @xref{TODO MySQL 4.0}.
 
 @itemize @bullet
 @item
+@code{LOCATE()} and @code{INSTR()} are case sensitive if neither
+argument is a binary string.  binary strings.
+@item
 Fixed core dump bug in @code{UPDATE ... ORDER BY }.
 @item
 Changed @code{INSERT INTO .. SELECT} to by default stop on errors.

myisam/mi_check.c

@@ -41,7 +41,8 @@ static int chk_index(MI_CHECK *param, MI_INFO *info, MI_KEYDEF *keyinfo,
 		     ha_checksum *key_checksum, uint level);
 static uint isam_key_length(MI_INFO *info,MI_KEYDEF *keyinfo);
 static ha_checksum calc_checksum(ha_rows count);
-static int writekeys(MI_INFO *info,byte *buff,my_off_t filepos);
+static int writekeys(MI_CHECK *param, MI_INFO *info,byte *buff,
+		     my_off_t filepos);
 static int sort_one_index(MI_CHECK *param, MI_INFO *info,MI_KEYDEF *keyinfo,
 			  my_off_t pagepos, File new_file);
 static int sort_key_read(SORT_INFO *sort_info,void *key);
@@ -62,7 +63,8 @@ static void update_key_parts(MI_KEYDEF *keyinfo,
 			     ulonglong *unique,
 			     ulonglong records);
 static ha_checksum mi_byte_checksum(const byte *buf, uint length);
-
+static void set_data_file_type(MI_CHECK *param, SORT_INFO *info,
+			       MYISAM_SHARE *share);
 
 #ifdef __WIN__
 static double ulonglong2double(ulonglong value)
@@ -1190,15 +1192,8 @@ int mi_repair(MI_CHECK *param, register MI_INFO *info,
   sort_info->dupp=0;
   sort_info->fix_datafile= (my_bool) (! rep_quick);
   sort_info->max_records= ~(ha_rows) 0;
-  if ((sort_info->new_data_file_type=share->data_file_type) ==
-      COMPRESSED_RECORD && param->testflag & T_UNPACK)
-  {
-    if (share->options & HA_OPTION_PACK_RECORD)
-      sort_info->new_data_file_type = DYNAMIC_RECORD;
-    else
-      sort_info->new_data_file_type = STATIC_RECORD;
-  }
 
+  set_data_file_type(param, sort_info, share);
   del=info->state->del;
   info->state->records=info->state->del=share->state.split=0;
   info->state->empty=0;
@@ -1226,9 +1221,10 @@ int mi_repair(MI_CHECK *param, register MI_INFO *info,
   lock_memory(param);			/* Everything is alloced */
   while (!(error=sort_get_next_record(sort_info)))
   {
-    if (writekeys(info,(byte*) sort_info->record,sort_info->filepos))
+    if (writekeys(param, info,(byte*) sort_info->record,sort_info->filepos))
     {
-      if (my_errno != HA_ERR_FOUND_DUPP_KEY) goto err;
+      if (my_errno != HA_ERR_FOUND_DUPP_KEY)
+	goto err;
       DBUG_DUMP("record",(byte*) sort_info->record,share->base.pack_reclength);
       mi_check_print_info(param,"Duplicate key %2d for record at %10s against new record at %10s",
 			  info->errkey+1,
@@ -1367,7 +1363,8 @@ err:
 
 /* Uppate keyfile when doing repair */
 
-static int writekeys(register MI_INFO *info,byte *buff,my_off_t filepos)
+static int writekeys(MI_CHECK *param, register MI_INFO *info, byte *buff,
+		     my_off_t filepos)
 {
   register uint i;
   uchar *key;
@@ -1380,12 +1377,14 @@ static int writekeys(register MI_INFO *info,byte *buff,my_off_t filepos)
     {
       if (info->s->keyinfo[i].flag & HA_FULLTEXT )
       {
-        if (_mi_ft_add(info,i,(char*) key,buff,filepos))  goto err;
+        if (_mi_ft_add(info,i,(char*) key,buff,filepos))
+	  goto err;
       }
       else
       {
 	uint key_length=_mi_make_key(info,i,key,buff,filepos);
-	if (_mi_ck_write(info,i,key,key_length)) goto err;
+	if (_mi_ck_write(info,i,key,key_length))
+	  goto err;
       }
     }
   }
@@ -1401,16 +1400,21 @@ static int writekeys(register MI_INFO *info,byte *buff,my_off_t filepos)
       {
 	if (info->s->keyinfo[i].flag & HA_FULLTEXT)
         {
-          if (_mi_ft_del(info,i,(char*) key,buff,filepos)) break;
+          if (_mi_ft_del(info,i,(char*) key,buff,filepos))
+	    break;
         }
         else
 	{
 	  uint key_length=_mi_make_key(info,i,key,buff,filepos);
-	  if (_mi_ck_delete(info,i,key,key_length)) break;
+	  if (_mi_ck_delete(info,i,key,key_length))
+	    break;
 	}
       }
     }
   }
+  /* Remove checksum that was added to glob_crc in sort_get_next_record */
+  if (param->calc_checksum)
+    param->glob_crc-= info->checksum;
   DBUG_PRINT("error",("errno: %d",my_errno));
   DBUG_RETURN(-1);
 } /* writekeys */
@@ -1847,15 +1851,7 @@ int mi_repair_by_sort(MI_CHECK *param, register MI_INFO *info,
   sort_info->info=info;
   sort_info->param = param;
 
-  if ((sort_info->new_data_file_type=share->data_file_type) ==
-      COMPRESSED_RECORD && param->testflag & T_UNPACK)
-  {
-    if (share->options & HA_OPTION_PACK_RECORD)
-      sort_info->new_data_file_type = DYNAMIC_RECORD;
-    else
-      sort_info->new_data_file_type = STATIC_RECORD;
-  }
-
+  set_data_file_type(param, sort_info, share);
   sort_info->filepos=new_header_length;
   sort_info->dupp=0;
   sort_info->buff=0;
@@ -2193,7 +2189,8 @@ static int sort_get_next_record(SORT_INFO *sort_info)
       if (*sort_info->record)
       {
 	if (param->calc_checksum)
-	  param->glob_crc+= mi_static_checksum(info,sort_info->record);
+	  param->glob_crc+= (info->checksum=
+			     mi_static_checksum(info,sort_info->record));
 	DBUG_RETURN(0);
       }
       if (!sort_info->fix_datafile)
@@ -2646,7 +2643,7 @@ static int sort_key_write(SORT_INFO *sort_info, const void *a)
 						    sort_info->key_block->
 						    lastkey),
 				 llbuff2));
-    param->error_printed=param->retry_without_quick=1;
+    param->retry_without_quick=1;
     if (sort_info->param->testflag & T_VERBOSE)
       _mi_print_key(stdout,sort_info->keyseg,(uchar*) a, USE_WHOLE_KEY);
     return (sort_delete_record(param));
@@ -3291,3 +3288,25 @@ my_bool mi_test_if_sort_rep(MI_INFO *info, ha_rows rows,
   }
   return TRUE;
 }
+
+
+static void
+set_data_file_type(MI_CHECK *param, SORT_INFO *sort_info, MYISAM_SHARE *share)
+{
+  if ((sort_info->new_data_file_type=share->data_file_type) ==
+      COMPRESSED_RECORD && param->testflag & T_UNPACK)
+  {
+    MYISAM_SHARE tmp;
+
+    if (share->options & HA_OPTION_PACK_RECORD)
+      sort_info->new_data_file_type = DYNAMIC_RECORD;
+    else
+      sort_info->new_data_file_type = STATIC_RECORD;
+
+    /* Set delete_function for sort_delete_record() */
+    memcpy((char*) &tmp, share, sizeof(*share));
+    tmp.options= ~HA_OPTION_COMPRESS_RECORD;
+    mi_setup_functions(&tmp);
+    share->delete_record=tmp.delete_record;
+  }
+}

myisam/mi_open.c

@@ -30,7 +30,6 @@
 #include "static.c"
 #endif
 
-static void setup_functions(MYISAM_SHARE *info);
 static void setup_key_functions(MI_KEYDEF *keyinfo);
 #define get_next_element(to,pos,size) { memcpy((char*) to,pos,(size_t) size); \
 					pos+=size;}
@@ -405,7 +404,7 @@ MI_INFO *mi_open(const char *name, int mode, uint open_flags)
     else if (share->options & HA_OPTION_PACK_RECORD)
       share->data_file_type = DYNAMIC_RECORD;
     my_afree((gptr) disk_cache);
-    setup_functions(share);
+    mi_setup_functions(share);
 #ifdef THREAD
     thr_lock_init(&share->lock);
     VOID(pthread_mutex_init(&share->intern_lock,MY_MUTEX_INIT_FAST));
@@ -580,7 +579,7 @@ ulonglong mi_safe_mul(ulonglong a, ulonglong b)
 
 	/* Set up functions in structs */
 
-static void setup_functions(register MYISAM_SHARE *share)
+void mi_setup_functions(register MYISAM_SHARE *share)
 {
   if (share->options & HA_OPTION_COMPRESS_RECORD)
   {

myisam/mi_packrec.c

@@ -785,7 +785,10 @@ static void decode_bytes(MI_COLUMNDEF *rec,MI_BIT_BUFF *bit_buff,uchar *to,
     if (bits <= 32)
     {
       if (bit_buff->pos > bit_buff->end+4)
+      {
+	bit_buff->error=1;
 	return;				/* Can't be right */
+      }
       bit_buff->current_byte= (bit_buff->current_byte << 32) +
 	((((uint) bit_buff->pos[3])) +
 	 (((uint) bit_buff->pos[2]) << 8) +
@@ -829,7 +832,8 @@ static void decode_bytes(MI_COLUMNDEF *rec,MI_BIT_BUFF *bit_buff,uchar *to,
 
 #else
 
-static void decode_bytes(MI_COLUMNDEF *rec, MI_BIT_BUFF *bit_buff, uchar *to, uchar *end)
+static void decode_bytes(MI_COLUMNDEF *rec, MI_BIT_BUFF *bit_buff, uchar *to,
+			 uchar *end)
 {
   reg1 uint bits,low_byte;
   reg3 uint16 *pos;
@@ -846,7 +850,10 @@ static void decode_bytes(MI_COLUMNDEF *rec, MI_BIT_BUFF *bit_buff, uchar *to, uc
     if (bits < table_bits)
     {
       if (bit_buff->pos > bit_buff->end+1)
+      {
+	bit_buff->error=1;
 	return;				/* Can't be right */
+      }
 #if BITS_SAVED == 32
       bit_buff->current_byte= (bit_buff->current_byte << 24) +
 	(((uint) ((uchar) bit_buff->pos[2]))) +

myisam/myisamchk.c

@@ -217,7 +217,7 @@ static struct option long_options[] =
 
 static void print_version(void)
 {
-  printf("%s  Ver 1.52 for %s at %s\n",my_progname,SYSTEM_TYPE,
+  printf("%s  Ver 2.0 for %s at %s\n",my_progname,SYSTEM_TYPE,
 	 MACHINE_TYPE);
 }
 
@@ -260,7 +260,7 @@ static void usage(void)
 
   puts("Repair options (When using -r or -o) \n\
   -B, --backup	      Make a backup of the .MYD file as 'filename-time.BAK'\n\
-  --correct-checksum  Correct checksum information for table. \n\
+  --correct-checksum  Correct checksum information for table.\n\
   -D, --data-file-length=#  Max length of data file (when recreating data\n\
                       file when it's full)\n\
   -e, --extend-check  Try to recover every possible row from the data file\n\

myisam/myisamdef.h

@@ -646,6 +646,7 @@ void mi_disable_non_unique_index(MI_INFO *info, ha_rows rows);
 my_bool check_table_is_closed(const char *name, const char *where);
 int mi_open_datafile(MI_INFO *info, MYISAM_SHARE *share, File file_to_dup);
 int mi_open_keyfile(MYISAM_SHARE *share);
+void mi_setup_functions(register MYISAM_SHARE *share);
 
 int _mi_init_bulk_insert(MI_INFO *info);
 

mysql-test/r/func_group.result

@@ -20,19 +20,19 @@ sum(a)
 NULL
 select a from t1 order by rand(10);
 a
+2
+6
 1
 3
-6
 5
-2
 4
 select distinct a from t1 order by rand(10);
 a
+2
+6
 1
 3
-6
 5
-2
 4
 select count(distinct a),count(distinct grp) from t1;
 count(distinct a)	count(distinct grp)

mysql-test/r/func_math.result

@@ -24,7 +24,7 @@ pow(10,log10(10))	power(2,4)
 10.000000	16.000000
 select rand(999999),rand();
 rand(999999)	rand()
-0.18435012473199	0.76373626176616
+0.014231365187309	0.8078568166195
 select pi(),sin(pi()/2),cos(pi()/2),abs(tan(pi())),cot(1),asin(1),acos(0),atan(1);
 PI()	sin(pi()/2)	cos(pi()/2)	abs(tan(pi()))	cot(1)	asin(1)	acos(0)	atan(1)
 3.141593	1.000000	0.000000	0.000000	0.64209262	1.570796	1.570796	0.785398

mysql-test/r/func_str.result

@@ -14,11 +14,11 @@ monty was here again	5	h
 select locate('he','hello'),locate('he','hello',2),locate('lo','hello',2) ;
 locate('he','hello')	locate('he','hello',2)	locate('lo','hello',2)
 1	0	4
-select instr('hello','he');
-instr('hello','he')
-1
-select position('ll' in 'hello'),position('a' in 'hello');
-position('ll' in 'hello')	position('a' in 'hello')
+select instr('hello','HE'), instr('hello',binary 'HE'), instr(binary 'hello','HE');
+instr('hello','HE')	instr('hello',binary 'HE')	instr(binary 'hello','HE')
+1	0	0
+select position(binary 'll' in 'hello'),position('a' in binary 'hello');
+position(binary 'll' in 'hello')	position('a' in binary 'hello')
 3	0
 select left('hello',2),right('hello',2),substring('hello',2,2),mid('hello',1,5) ;
 left('hello',2)	right('hello',2)	substring('hello',2,2)	mid('hello',1,5)

mysql-test/t/func_str.test

@@ -11,8 +11,8 @@ select 'hello' 'monty';
 select length('\n\t\r\b\0\_\%\\');
 select concat('monty',' was here ','again'),length('hello'),char(ascii('h'));
 select locate('he','hello'),locate('he','hello',2),locate('lo','hello',2) ;
-select instr('hello','he');
-select position('ll' in 'hello'),position('a' in 'hello');
+select instr('hello','HE'), instr('hello',binary 'HE'), instr(binary 'hello','HE'); 
+select position(binary 'll' in 'hello'),position('a' in binary 'hello');
 select left('hello',2),right('hello',2),substring('hello',2,2),mid('hello',1,5) ;
 select concat('',left(right(concat('what ',concat('is ','happening')),9),4),'',substring('monty',5,1)) ;
 select substring_index('www.tcx.se','.',-2),substring_index('www.tcx.se','.',1);

sql/item_func.cc

@@ -621,8 +621,9 @@ double Item_func_rand::val()
 {
   if (arg_count)
   {					// Only use argument once in query
-    ulong tmp=((ulong) args[0]->val_int());
-    randominit(&current_thd->rand,tmp*0x10001L+55555555L,tmp*0x10000001L);
+    uint32 tmp= (uint32) (args[0]->val_int());
+    randominit(&current_thd->rand,(uint32) (tmp*0x10001L+55555555L),
+	       (uint32) (tmp*0x10000001L));
 #ifdef DELETE_ITEMS
     delete args[0];
 #endif
@@ -797,9 +798,7 @@ longlong Item_func_locate::val_int()
 {
   String *a=args[0]->val_str(&value1);
   String *b=args[1]->val_str(&value2);
-#ifdef USE_MB
   bool binary_str = args[0]->binary || args[1]->binary;
-#endif
   if (!a || !b)
   {
     null_value=1;
@@ -853,7 +852,8 @@ longlong Item_func_locate::val_int()
     return 0;
   }
 #endif /* USE_MB */
-  return (longlong) (a->strstr(*b,start)+1) ;
+  return (longlong) (binary ? a->strstr(*b,start) :
+		     (a->strstr_case(*b,start)))+1;
 }
 
 

sql/sql_string.cc

@@ -362,6 +362,37 @@ skipp:
   return -1;
 }
 
+/*
+  Search after a string without regarding to case
+  This needs to be replaced when we have character sets per string
+*/
+
+int String::strstr_case(const String &s,uint32 offset)
+{
+  if (s.length()+offset <= str_length)
+  {
+    if (!s.length())
+      return ((int) offset);	// Empty string is always found
+
+    register const char *str = Ptr+offset;
+    register const char *search=s.ptr();
+    const char *end=Ptr+str_length-s.length()+1;
+    const char *search_end=s.ptr()+s.length();
+skipp:
+    while (str != end)
+    {
+      if (my_sort_order[*str++] == my_sort_order[*search])
+      {
+	register char *i,*j;
+	i=(char*) str; j=(char*) search+1;
+	while (j != search_end)
+	  if (my_sort_order[*i++] != my_sort_order[*j++]) goto skipp;
+	return (int) (str-Ptr) -1;
+      }
+    }
+  }
+  return -1;
+}
 
 /*
 ** Search string from end. Offset is offset to the end of string

sql/sql_string.h

@@ -161,6 +161,7 @@ public:
   bool append(const char *s,uint32 arg_length=0);
   bool append(IO_CACHE* file, uint32 arg_length);
   int strstr(const String &search,uint32 offset=0); // Returns offset to substring or -1
+  int strstr_case(const String &s,uint32 offset=0);
   int strrstr(const String &search,uint32 offset=0); // Returns offset to substring or -1
   bool replace(uint32 offset,uint32 arg_length,const String &to);
   inline bool append(char chr)