Added bdb_lock_max and fixed bug in BDB tables when using key parts

Docs/Makefile.am

@@ -139,3 +139,6 @@ INSTALL-BINARY:	mysql.info $(GT)
 
 ../MIRRORS:		manual.texi $(srcdir)/Support/generate-mirror-listing.pl
 	perl -w $(srcdir)/Support/generate-mirror-listing.pl manual.texi > $@
+
+# Don't update the files from bitkeeper
+%::SCCS/s.%

Docs/manual.texi

@@ -813,6 +813,7 @@ MySQL change history
 
 Changes in release 3.23.x  (Recommended; beta)
 
+* News-3.23.28::                
 * News-3.23.27::                Changes in release 3.23.27
 * News-3.23.26::                Changes in release 3.23.26
 * News-3.23.25::                Changes in release 3.23.25
@@ -19862,6 +19863,13 @@ The buffer that is allocated to cache index and rows for @code{BDB} tables.
 If you don't use @code{BDB} tables, you should set this to 0 or
 start @code{mysqld} with @code{--skip-bdb} o not waste memory for this cache.
 
+@item @code{bdb_lock_max}
+The maximum number of locks (1000 by default) you can have active on a BDB
+table. You should increase this if you get errors of type
+@code{bdb: Lock table is out of available locks} when you have do long
+transactions or when mysqld has to examine a lot of rows to calculate
+the query.
+
 @item @code{concurrent_inserts}
 If @code{ON} (the default), @strong{MySQL} will allow you to use @code{INSERT}
 on @code{MyISAM} tables at the same time as you run @code{SELECT} queries
@@ -24669,7 +24677,7 @@ limits. Here are some examples:
 
 @multitable @columnfractions .5 .5
 @item @strong{Operating System} @tab @strong{File Size Limit}
-@item Linux-Intel @tab 2G (or 4G with reiserfs)
+@item Linux-Intel 32 bit@tab 2G, 4G or bigger depending on Linux version
 @item Linux-Alpha @tab 8T (?)
 @item Solaris 2.5.1 @tab 2G (possible 4G with patch)
 @item Solaris 2.6 @tab 4G
@@ -24677,6 +24685,10 @@ limits. Here are some examples:
 @item Solaris 2.7 ULTRA-SPARC @tab 8T (?)
 @end multitable
 
+On Linux 2.2 you can get bigger tables than 2G by using the LFS patch for
+the ext2 file system.  On Linux 2.4 there exists also patches for ReiserFS
+to get support for big files.
+
 This means that the table size for @strong{MySQL} is normally limited by
 the operating system.
 
@@ -24690,14 +24702,15 @@ this), you should set the @code{AVG_ROW_LENGTH} and @code{MAX_ROWS}
 parameter when you create your table.  @xref{CREATE TABLE}.  You can
 also set these later with @code{ALTER TABLE}. @xref{ALTER TABLE}.
 
-If you need to have bigger tables than 2G / 4G
-
 If your big table is going to be read-only, you could use
 @code{myisampack} to merge and compress many tables to one.
 @code{myisampack} usually compresses a table by at least 50%, so you can
 have, in effect, much bigger tables.  @xref{myisampack, ,
 @code{myisampack}}.
 
+You can go around the operating system file limit for @code{MyISAM} data
+files by using the @code{RAID} option. @xref{CREATE TABLE}.
+
 Another solution can be the included MERGE library, which allows you to
 handle a collection of identical tables as one. @xref{MERGE, MERGE
 tables}.
@@ -25430,7 +25443,9 @@ multiple CPU machines one should use Solaris (because the threads works
 really nice) or Linux (because the 2.2 kernel has really good SMP
 support). Also on 32bit machines Linux has a 2G file size limit by
 default. Hopefully this will be fixed soon when new filesystems is
-released (XFS/Reiserfs).
+released (XFS/Reiserfs).  If you have a desperate need for files bigger
+tan 2G on Linux-intel 32 bit, you should get the LFS patch for the ext2
+file system.
 
 Because we have not run @strong{MySQL} in production on that many platforms we
 advice you to test your intended platform before choosing it, if possible.
@@ -38136,6 +38151,7 @@ version.  The replication and BerkeleyDB code is still under development,
 though, so 3.23 is not released as a stable version yet.
 
 @menu
+* News-3.23.28::                Changes in release 3.23.28
 * News-3.23.27::                Changes in release 3.23.27
 * News-3.23.26::                Changes in release 3.23.26
 * News-3.23.25::                Changes in release 3.23.25
@@ -38166,7 +38182,16 @@ though, so 3.23 is not released as a stable version yet.
 * News-3.23.0::                 Changes in release 3.23.0
 @end menu
 
-@node News-3.23.27, News-3.23.26, News-3.23.x, News-3.23.x
+@node News-3.23.28, News-3.23.27, News-3.23.x, News-3.23.x
+@appendixsubsec Changes in release 3.23.28
+@itemize @bullet
+@item
+Fixed bug in a BDB key compare function when comparing part keys.
+@item
+Added variable @code{bdb_lock_max} to @code{mysqld}.
+@end itemize
+
+@node News-3.23.27, News-3.23.26, News-3.23.28, News-3.23.x
 @appendixsubsec Changes in release 3.23.27
 @itemize @bullet
 @item

myisam/myisamchk.c

@@ -1378,7 +1378,6 @@ void mi_check_print_info(MI_CHECK *param __attribute__((unused)),
   VOID(vfprintf(stdout, fmt, args));
   VOID(fputc('\n',stdout));
   va_end(args);
-  return;
 }
 
 /* VARARGS */

sql/field.cc

@@ -3439,6 +3439,23 @@ int Field_string::pack_cmp(const char *a, const char *b, uint length)
 }
 
 
+int Field_string::pack_cmp(const char *b, uint length)
+{
+  uint b_length= (uint) (uchar) *b++;
+  char *end= ptr + field_length;
+  while (end > ptr && end[-1] == ' ')
+    end--;
+  uint a_length = (uint) (end - ptr);
+
+  if (binary_flag)
+  {
+    int cmp= memcmp(ptr,b,min(a_length,b_length));
+    return cmp ? cmp : (int) (a_length - b_length);
+  }
+  return my_sortncmp(ptr,a_length, b, b_length);
+}
+
+
 uint Field_string::packed_col_length(const char *ptr)
 {
   if (field_length > 255)
@@ -3637,6 +3654,27 @@ int Field_varstring::pack_cmp(const char *a, const char *b, uint key_length)
   return my_sortncmp(a,a_length, b,b_length);
 }
 
+int Field_varstring::pack_cmp(const char *b, uint key_length)
+{
+  char *a=ptr+2;
+  uint a_length=uint2korr(ptr);
+  uint b_length;
+  if (key_length > 255)
+  {
+    b_length=uint2korr(b); b+=2;
+  }
+  else
+  {
+    b_length= (uint) (uchar) *b++;
+  }
+  if (binary_flag)
+  {
+    int cmp= memcmp(a,b,min(a_length,b_length));
+    return cmp ? cmp : (int) (a_length - b_length);
+  }
+  return my_sortncmp(a,a_length, b,b_length);
+}
+
 uint Field_varstring::packed_col_length(const char *ptr)
 {
   if (field_length > 255)
@@ -4019,6 +4057,33 @@ int Field_blob::pack_cmp(const char *a, const char *b, uint key_length)
   return my_sortncmp(a,a_length, b,b_length);
 }
 
+
+int Field_blob::pack_cmp(const char *b, uint key_length)
+{
+  char *a;
+  memcpy_fixed(&a,ptr+packlength,sizeof(char*));
+  if (!a)
+    return key_length > 0 ? -1 : 0;
+  uint a_length=get_length(ptr);
+  uint b_length;
+
+  if (key_length > 255)
+  {
+    b_length=uint2korr(b); b+=2;
+  }
+  else
+  {
+    b_length= (uint) (uchar) *b++;
+  }
+  if (binary_flag)
+  {
+    int cmp= memcmp(a,b,min(a_length,b_length));
+    return cmp ? cmp : (int) (a_length - b_length);
+  }
+  return my_sortncmp(a,a_length, b,b_length);
+}
+
+
 char *Field_blob::pack_key(char *to, const char *from, uint max_length)
 {
   uint length=uint2korr(to);

sql/field.h

@@ -177,6 +177,8 @@ class Field {
 
   virtual int pack_cmp(const char *a,const char *b, uint key_length_arg)
   { return cmp(a,b); }
+  virtual int pack_cmp(const char *b, uint key_length_arg)
+  { return cmp(ptr,b); }
   uint offset();				// Should be inline ...
   void copy_from_tmp(int offset);
   uint fill_cache_field(struct st_cache_field *copy);
@@ -726,6 +728,7 @@ class Field_string :public Field_str {
   char *pack(char *to, const char *from, uint max_length=~(uint) 0);
   const char *unpack(char* to, const char *from);
   int pack_cmp(const char *a,const char *b,uint key_length);
+  int pack_cmp(const char *b,uint key_length);
   uint packed_col_length(const char *to);
   uint max_packed_col_length(uint max_length);
   uint size_of() const { return sizeof(*this); }
@@ -777,6 +780,7 @@ class Field_varstring :public Field_str {
   char *pack(char *to, const char *from, uint max_length=~(uint) 0);
   const char *unpack(char* to, const char *from);
   int pack_cmp(const char *a, const char *b, uint key_length);
+  int pack_cmp(const char *b, uint key_length);
   uint packed_col_length(const char *to);
   uint max_packed_col_length(uint max_length);
   uint size_of() const { return sizeof(*this); }
@@ -891,6 +895,7 @@ class Field_blob :public Field_str {
   }
   char *pack_key(char *to, const char *from, uint max_length=~(uint) 0);
   int pack_cmp(const char *a, const char *b, uint key_length);
+  int pack_cmp(const char *b, uint key_length);
   uint packed_col_length(const char *col_ptr)
   { return get_length(col_ptr)+packlength;}
   virtual uint max_packed_col_length(uint max_length)

sql/ha_berkeley.cc

@@ -68,6 +68,7 @@ ulong berkeley_cache_size;
 char *berkeley_home, *berkeley_tmpdir, *berkeley_logdir;
 long berkeley_lock_scan_time=0;
 ulong berkeley_trans_retry=5;
+ulong berkeley_lock_max;
 pthread_mutex_t bdb_mutex;
 
 static DB_ENV *db_env;
@@ -116,6 +117,8 @@ bool berkeley_init(void)
   
   db_env->set_cachesize(db_env, 0, berkeley_cache_size, 0);
   db_env->set_lk_detect(db_env, berkeley_lock_type);
+  if (berkeley_lock_max)
+    db_env->set_lk_max(db_env, berkeley_lock_max);
   if (db_env->open(db_env,
 		   berkeley_home,
 		   berkeley_init_flags |  DB_INIT_LOCK | 
@@ -224,8 +227,13 @@ berkeley_cmp_packed_key(DB *file, const DBT *new_key, const DBT *saved_key)
     int cmp;
     if (key_part->null_bit)
     {
-      if (*new_key_ptr++ != *saved_key_ptr++)
-	return ((int) new_key_ptr[-1] - (int) saved_key_ptr[-1]);
+      if (*new_key_ptr != *saved_key_ptr++)
+	return ((int) *new_key_ptr - (int) saved_key_ptr[-1]);
+      if (!*new_key_ptr++)
+      {
+	key_length--;
+	continue;
+      }
     }
     if ((cmp=key_part->field->pack_cmp(new_key_ptr,saved_key_ptr,
 				       key_part->length)))
@@ -263,6 +271,36 @@ berkeley_cmp_fix_length_key(DB *file, const DBT *new_key, const DBT *saved_key)
 }
 
 
+/* Compare key against row */
+
+static bool
+berkeley_key_cmp(TABLE *table, KEY *key_info, const char *key, uint key_length)
+{
+  KEY_PART_INFO *key_part= key_info->key_part,
+	        *end=key_part+key_info->key_parts;
+
+  for ( ; key_part != end && (int) key_length > 0; key_part++)
+  {
+    int cmp;
+    if (key_part->null_bit)
+    {
+      key_length--;
+      if (*key != (table->record[0][key_part->null_offset] &
+		   key_part->null_bit) ? 0 : 1)
+	return 1;
+      if (!*key++)					// Null value
+	continue;
+    }
+    if ((cmp=key_part->field->pack_cmp(key,key_part->length)))
+      return cmp;
+    uint length=key_part->field->packed_col_length(key);
+    key+=length;
+    key_length-=length;
+  }
+  return 0;
+}
+
+
 int ha_berkeley::open(const char *name, int mode, uint test_if_locked)
 {
   char name_buff[FN_REFLEN];
@@ -1118,9 +1156,8 @@ int ha_berkeley::index_read(byte * buf, const byte * key,
     key_info->handler.bdb_return_if_eq= 0;
     if (!error && find_flag == HA_READ_KEY_EXACT)
     {
-      /* Check that we didn't find a key that wasn't equal to the current
-	 one */
-      if (!error && ::key_cmp(table, key_buff2, active_index, key_len))
+      /* Ensure that we found a key that is equal to the current one */
+      if (!error && berkeley_key_cmp(table, key_info, key_buff2, key_len))
 	error=HA_ERR_KEY_NOT_FOUND;
     }
   }

sql/ha_berkeley.h

@@ -147,7 +147,7 @@ class ha_berkeley: public handler
 
 extern bool berkeley_skip;
 extern u_int32_t berkeley_init_flags,berkeley_lock_type,berkeley_lock_types[];
-extern ulong berkeley_cache_size;
+extern ulong berkeley_cache_size, berkeley_lock_max;
 extern char *berkeley_home, *berkeley_tmpdir, *berkeley_logdir;
 extern long berkeley_lock_scan_time;
 extern TYPELIB berkeley_lock_typelib;

sql/key.cc

@@ -171,7 +171,7 @@ int key_cmp(TABLE *table,const byte *key,uint idx,uint key_length)
     {
       key_length--;
       if (*key != test(table->record[0][key_part->null_offset] & 
-			 key_part->null_bit))
+		       key_part->null_bit))
 	return 1;
       if (*key)
       {

sql/mysqld.cc

@@ -2333,6 +2333,8 @@ CHANGEABLE_VAR changeable_vars[] = {
 #ifdef HAVE_BERKELEY_DB
   { "bdb_cache_size",          (long*) &berkeley_cache_size, 
       KEY_CACHE_SIZE, 20*1024, (long) ~0, 0, IO_SIZE },
+  { "bdb_lock_max",            (long*) &berkeley_lock_max, 
+      1000, 0, (long) ~0, 0, 1 },
 #endif
   { "connect_timeout",         (long*) &connect_timeout,
       CONNECT_TIMEOUT, 2, 65535, 0, 1 },
@@ -2413,6 +2415,7 @@ struct show_var_st init_vars[]= {
 #ifdef HAVE_BERKELEY_DB
   {"bdb_cache_size",          (char*) &berkeley_cache_size,         SHOW_LONG},
   {"bdb_home",                (char*) &berkeley_home,               SHOW_CHAR_PTR},
+  {"bdb_lock_max",            (char*) &berkeley_lock_max,	    SHOW_LONG},
   {"bdb_logdir",              (char*) &berkeley_logdir,             SHOW_CHAR_PTR},
   {"bdb_tmpdir",              (char*) &berkeley_tmpdir,             SHOW_CHAR_PTR},
 #endif