it should be redo-logged because the data structure is file-based;
this patch does not fix any bug; the original erase operation was added
in r781 to fix Valgrind Bug #20791

There always was a one-to-one mapping between dict_tree_t and dict_index_t.

This saves 6 machine words per B-tree index in the data dictionary
cache plus the memory allocation overhead.  We save one mem_heap_t
object per index (15 machine words).  Considering the internal
fragmentation of the buddy allocator in mem_area_alloc(), this should
save 32 machine words per index (128 bytes on 32-bit systems and 256
bytes on 64-bit systems).  (Bug #20877)

struct dict_tree_struct, dict_tree_t: Remove.

struct dict_index_struct: Add page and lock.

dict_tree_create(): Remove.  Replace the invocation with
assignment to index->page and a call to rw_lock_create(&index->lock).

dict_tree_free(): Remove.  Replace the invocation wtih
a call to rw_lock_free(&index->lock).

dict_index_get_tree(): Remove.

dict_tree_get_space_reserve(): Rename to dict_index_get_space_reserve()
and remove the parameter, which was unused.

btr_level_list_remove(): Remove the unused parameter "tree".

Replace the occurrences of "tree" with "index" in names of variables,
functions and data types, e.g. "dict_tree_t tree" becomes
"dict_index_t index".  Remove local variables "tree" or "index" of
functions that needed both "tree" and "index".

settings introduced. Some problems were found, so from this commit on one
additional indentation rule is introduced:

 (add-to-list 'c-offsets-alist '(arglist-intro . +))

Note that fixing some of the unfortunate line-splits done in r764 will be
done in a future change.

delete-marked record in a clustered index where the search criteria is 
unique, within the same transaction (Bug #13544).

structure is reserved for every index in the data dictionary cache.
(Bug #20877)

We could shrink the structure further to three 32-bit words or two
64-bit words by turning the remaining fields to bit-fields.
Unfortunately, the fields are not protected by any mutex, and thus we
would better keep each field aligned to a machine word.

btr_search_t, buf_block_t: Rename "ulint side" to "ibool left_side".
Remove BTR_SEARCH_LEFT_SIDE and BTR_SEARCH_RIGHT_SIDE, and also rename
some local variables and function parameters from "ulint side" to
"ibool left_side".

btr_search_t: Remove the unused fields last_search, n_direction, direction,
and modify_clock.  Remove the unused constants BTR_SEA_NO_DIRECTION,
BTR_SEA_LEFT, BTR_SEA_RIGHT, and BTR_SEA_SAME_REC.

btr_search_t: Remove magic_n unless #defined UNIV_DEBUG.  Turn an assertion
about the magic number into a debug assertion.

btr_print_recursive(): Replace tree->tree_indexes with tree->index.
This should have been done in r453.

univ.i: Add UNIV_BTR_PRINT.

rw_lock_t: Reduce writer_is_wait_ex to a bit and move it close to
other bitfields.  Change the types to unsigned, in case ulint or ibool
bitfields will not work.

to a static plain function.

to the Boolean field big_rows.  (Bug #20877)

BIG_ROW_SIZE: Move the definition from row0sel.c to dict_table_add_to_cache().

dict_col_get_clust_pos(): Add parameter clust_index.  Replace the
look-up with a linear search of all columns in the clustered index.

row_upd_index_replace_new_col_vals(): Compute clust_index outside
the loops.  Compute clust_pos outside the inner loop.

row_upd_changes_ord_field_binary(), row_upd_changes_first_fields_binary():
Compute clust_index outside the loops.  Declare the auxiliary variables
inside the loop scope.

dict_index_add_to_cache(): Instead of incrementing ord_part, set it.

dict_index_remove_from_cache(): Do not touch ord_part.

dtype_t: Reduce mbminlen from 3 to 2 bits.

row_upd(): Add a UNIV_LIKELY hint around node->in_mysql_interface.

dict_sys->col_hash, and DICT_POOL_PER_COL_HASH.  (Bug #20877)

The col->table pointer was only needed for maintaining a hash table of
all defined columns in all tables.  The hash table was only looked up
in dict_index_find_cols().  Removing the col->hash and col->table
pointers reduces the size of a table column by two machine words
(usually 8 or 16 bytes).

dict_col_add_to_cache(), dict_col_reposition_in_cache(),
dict_col_remove_from_cache(): Remove.  These only updated dict_sys->col_hash.

dict_index_find_cols(): Use a linear search instead of the hash table.
The time complexity is affected, but this function is only invoked by
dict_index_add_to_cache(), and we only search the columns of a single
table (typically at most a few dozen) as opposed to all columns of all
tables.

when the list in dict_tree_t was replaced with the pointer tree_index.

reserved1 (apparently the system variables) and
reserved2 (apparently the configuration options).

dtype_t: Remove unused field "prec", which was supposed to be used for
the precision of decimal columns in stand-alone InnoDB.

dtype_get_prec(): Remove.

dtype_set(), dict_mem_table_add_col(): Remove parameter "prec".

dtype_t: Turn all fields (mtype, prtype, len, mbminlen, mbmaxlen, len)
into bit-fields.

dict_table_t, dict_index_t, dict_tree_t: Omit magic_n from non-debug builds.

dict_col_t: Turn ind, clust_pos, and ord_part into bit-fields.

Replace the default clust_pos value ULINT_UNDEFINED with
REC_MAX_N_FIELDS and replace all references to clust_pos with calls to
the accessor function dict_col_get_clust_pos().

dict_field_t: Turn prefix_len and fixed_len into bit-fields.

dict_tree_t: Remove pad[64].

dict_table_t: Turn the fields ibd_file_missing, tablespace_discarded,
cached, flags, stat_initialized, and autoinc_inited into bit-fields.
Remove does_not_fit_in_memory from non-debug builds.

dict_index_t: Turn the fields trx_id_offset, n_user_defined_cols,
n_uniq, n_def, n_fields, n_nullable, and cached into bit-fields.

dict_foreign_struct: Turn n_fields and type into bit-fields.

rw_lock_t: Turn cline, last_s_line, and last_x_line into bit-fields.
Omit level unless #defined UNIV_SYNC_DEBUG.

Move REC_MAX_N_FIELDS (and REC_MAX_HEAP_NO and REC_MAX_N_OWNED)
from rem0rec.c to rem0types.h, as they are needed in dict0dict.ic.
dict_col_get_clust_pos(): Map REC_MAX_N_FIELDS to ULINT_UNDEFINED.

ChangeSet
  2006/09/07 08:23:58-07:00 brian@zim.(none) 
  Moves Innodb handler to the Innodb storage directory.

storage/innobase/handler/ha_innodb.cc
  2006/09/07 08:23:53-07:00 brian@zim.(none) +2 -3
  Adjusted include files, also disabled replication code which was not
  being used.

ChangeSet
  2006/09/07 12:34:12-07:00 brian@zim.(none) 
  Fix for a compile problem in Windows. 

storage/innobase/handler/ha_innodb.cc
  2006/09/07 12:34:10-07:00 brian@zim.(none) +0 -28
  Removed some dead code (Marko approved)

This was submitted by Marko to MySQL AB and merged to the MySQL tree by
Brian Aker in the following two changesets:

ChangeSet
  2006/09/07 08:23:58-07:00 brian@zim.(none) 
  Moves Innodb handler to the Innodb storage directory.

ChangeSet
  2006/09/07 16:17:16-07:00 brian@zim.(none) 
  Warning fixes for Windows, and an include fix for Windows for Innodb.

storage/innobase/CMakeLists.txt
  2006/09/07 16:17:13-07:00 brian@zim.(none) +6 -1
  Fixed includes for Windows

set type->len to the prefix_len of the index column, if it is a prefix index.
This should prevent bugs similar to Bug #21638 from occurring.

dict_index_copy_types(): Set type->len to prefix_len if prefix_len != 0.

row_build_index_entry(): Set type->len to prefix_len if prefix_len != 0,
also when the column in the tuple is SQL NULL, because the type information
may be used for interpreting other records during btr_page_reorganize().

ChangeSet
  2006/08/23 15:12:42-07:00 brian@zim.(none) 
  This is a cleanup of warnings that windows is complaining about. 

sql/ha_innodb.cc
  2006/08/23 15:12:39-07:00 brian@zim.(none) +0 -1
  Removed unused variables

when prefix_len is specified.  Otherwise, btr_page_reorganize() during
insert buffer merge would fail on ROW_FORMAT=COMPACT tables.  (Bug #21638)

ChangeSet
  2006/08/25 13:31:15-07:00 brian@zim.(none) 
  Cleanup of unused variables.

sql/ha_innodb.h
  2006/08/25 13:31:11-07:00 brian@zim.(none) +0 -6
  Unused variable

ChangeSet
  2006/08/22 16:24:12-07:00 brian@zim.(none) 
  This changest:
  Plugins now when compiled or not compiled work correctly with status variables. 
  Status variables from plugins now set their own names (removed bit where plugin name was pre-appended this broke Innodb and Cluster)
  A few Makefile cleanups. 

sql/ha_innodb.cc
  2006/08/22 16:24:08-07:00 brian@zim.(none) +16 -6
  Cleanup to make status variables directly in engine

changed after review on the MySQL side.

ChangeSet@1.2288, 2006-08-29 15:35:05+02:00, guilhem@gbichot3.local +2 -0
  Fix for BUG#20866 "show table status on innodb raises assertion"
  and its duplicate BUG#19057 "Test 'rpl_row_func003' fails on SuSE SLES9 x86".
  It was an assertion failure, only in debug builds, not present
  in released versions (nothing to document).
  It happened when doing SHOW TABLE STATUS on an InnoDB table
  having an auto_increment column, right after creating the table.

  sql/ha_innodb.cc@1.288, 2006-08-29 15:35:02+02:00, guilhem@gbichot3.local +7 -1
    Before a val_() calls on a Field object, if that field was not marked
    for read, we need to mark it. This is explained here:
    ChangeSet 1.2119.601.1 2006/06/04 18:52:22 monty@mysql.com
    quoting the changeset's comment:
    
      - If a handler needs to call Field->val() or Field->store() on columns
        that are not used in the query, one should install a temporary
        all-columns-used map while doing so. For this, we provide the following
        functions:
      
        my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set);
        field->val();
        dbug_tmp_restore_column_map(table->read_set, old_map);
      
        and similar for the write map:
      
        my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set);
        field->val();
        dbug_tmp_restore_column_map(table->write_set, old_map);
      
        If this is not done, you will sooner or later hit a DBUG_ASSERT
        in the field store() / val() functions.
        (For not DBUG binaries, the dbug_tmp_restore_column_map() and
        dbug_tmp_restore_column_map() are inline dummy functions and should
        be optimized away be the compiler).
    
    Note that I verified that the bug didn't exist in non-debug builds.

that was accidentally committed in r782.

the doublewrite buffer magic number from uninitialized memory; the code
worked because it was extremely unlikely that the memory would contain the
magic number.

to "waiting for server activity" that mentions that this string should not
be changed.

univ.i: Do not #define YYDEBUG, because it is only useful for debugging
the grammar of the Bison-generated InnoDB SQL parser.

row_build(): Remove type==ROW_COPY_ALSO_EXTERNALS, because it is never
passed.

Replace max-no-ndb compilation flags with --with-plugin=innobase.

setup.sh: Create symbolic links to the build scripts instead of copying them.

ChangeSet
  2006/08/23 13:59:16-07:00 brian@zim.(none) 
  This patch removes need for a innodb to have its own configure. 

univ.i: Replace ../ib_config.h with config.h.

Makefile.i, Makefile.am: Change directory paths.

configure.in: Delete.

plug.in: New file, included from the top-level configure.in.

setup.sh: Replace configure.in with plug.in.

to the data dictionary cache.

Replace some printf(...) in debug builds with fprintf(stderr, ...).