1. 03 Aug, 2007 1 commit
  2. 02 Aug, 2007 2 commits
    • marko's avatar
      branches/zip: Ensure that the dictionary updates are atomic by keeping · 11665dc3
      marko authored
      the data dictionary locked across the operations.
      
      dict_table_decrement_handle_count(), row_prebuilt_free(): Add the flag
      dict_locked, to prevent the acquisition of dict_sys->mutex.
      
      innobase_rename_table(): Rename commit_flag to lock_and_commit,
      and do not acquire dict_sys->mutex or flush the log unless the flag is set.
      Remove bogus comment about utility threads, because the threads will
      be waken up by the upper-level function ha_innobase::rename_table().
      
      ha_innobase::add_index(): After creating a primary key, hold dict_sys->mutex
      across all dictionary operations.
      11665dc3
    • marko's avatar
  3. 01 Aug, 2007 10 commits
  4. 21 Jun, 2007 8 commits
  5. 20 Jun, 2007 3 commits
  6. 19 Jun, 2007 2 commits
  7. 18 Jun, 2007 2 commits
    • marko's avatar
      branches/zip: ha_innobase::add_index(): Split some assertions. · 7235ea10
      marko authored
      Remove the variable parent_trx.
      
      Call innobase_register_stmt() in order to work around a MySQL bug
      in mysql_alter_table(), which, as of
      
      ChangeSet@1.2482.61.2, 2007-06-07 16:37:15+02:00, joerg@trift2. +8 -0
      
      commits the transaction before calling ha_innobase::add_index().
      Without re-registering the statement, the ha_commit_stmt(thd)
      in mysql_alter_table() would not invoke innobase_commit.
      7235ea10
    • marko's avatar
      branches/zip: ha_innobase::add_index(): Do not commit the transaction. · 513f7ceb
      marko authored
      It should be committed by MySQL after all.  The error in mysql_alter_table()
      will be worked around in r1584.
      513f7ceb
  8. 13 Jun, 2007 4 commits
    • marko's avatar
      branches/zip: Remove reads of freed memory in fast index creation. · a5ddf51a
      marko authored
      Because row_drop_table_for_mysql_no_commit() checks with dict_load_table()
      if the table was successfully dropped, we cannot pass table->name to it.
      
      row_merge_drop_table(): Pass a copy of table->name to
      row_drop_table_for_mysql().
      
      row_prebuilt_free(): Do not dereference prebuilt->table->name after a
      successful invocation of row_add_table_to_background_drop_list().  The
      table object may be freed at any time.  Remove the debug message
      "Dropping table".
      a5ddf51a
    • marko's avatar
      branches/zip: Fix a few bugs in fast index creation. No more crashes · 62ad3dbd
      marko authored
      occur in mysql-test-run, but mysql-test-run reports some warnings for
      some tests and incorrect results for innodb-index.test.
      
      innobase_start_trx_and_assign_read_view(): Use thd instead of current_thd.
      
      ha_innobase::add_index(): Use ha_thd() instead of current_thd.
      Commit the transaction at the end.
      
      row_merge_rename_index(): Really change the name of the index.  This was
      broken in r1430.
      62ad3dbd
    • marko's avatar
      branches/zip: Merge 1571:1575 from trunk. · 258fb5d4
      marko authored
      258fb5d4
    • marko's avatar
      branches/zip: trx_commit_for_mysql(): Replace ut_a(err) · 251a921e
      marko authored
      with ut_a(err == DB_SUCCESS).
      251a921e
  9. 12 Jun, 2007 2 commits
  10. 11 Jun, 2007 2 commits
    • marko's avatar
      branches/zip: Initialize some uninitialized variables in order to avoid · 5ab4cfd4
      marko authored
      comparing uninitialized values.
      
      dict_mem_table_create(): Initialize table->version_number.
      
      ins_node_create(): Initialize node->table_version_number.
      
      row_create_prebuilt(): Initialize all fields.  Initialize most fields
      by memset(), so that any fields that are added will be initialized
      automatically.
      5ab4cfd4
    • marko's avatar
      branches/zip: Create special single-table tablespace file names for · 1cfcd4f2
      marko authored
      temporary tablespaces in fast index creation.  The tablespaces must
      reside in the same directory as the persistent tablespaces, so that
      they can be renamed without copying.  Namespace collisions are avoided
      by replacing the ".ibd" file name suffix with ".ib1" or ".ib2".
      
      fil_make_ibd_name(): Treat names starting with TEMP_TABLE_PREFIX
      specially.  Document the dependence on
      innobase_create_temporary_tablename().
      1cfcd4f2
  11. 08 Jun, 2007 4 commits
    • marko's avatar
      branches/zip: Fix bugs in merge sort. · 0954bbed
      marko authored
      row_merge_buf_add(): Correctly calculate buf->total_size.
      
      row_merge_buf_write(): Add an assertion about buf->total_size.
      
      row_merge_heap_create(), row_merge_read_rec(), row_merge_insert_index_tuples():
      offsets needs to be 1 + REC_OFFS_HEADER_SIZE + n_fields.  Previously, we
      allocated one element too little.
      
      row_merge_blocks(): Use goto instead of break in the AT_END argument
      of ROW_MERGE_WRITE_GET_NEXT.
      0954bbed
    • marko's avatar
      branches/zip: row_merge_build_indexes(): Allocate all buffers with a · ffe5998e
      marko authored
      single os_mem_alloc_large() call.  The function mem_alloc() that was
      used previously allocates unaligned memory.
      
      row_merge_blocks(): Replace block1, block2, block3 with a single array.
      Replace buf1, buf2, buf3 with buf[3].  Replace b1..b3 with b0..b2.
      Replace mrec1,2 with mrec0,1.  Replace offsets1,2 with offsets0,1.
      
      row_merge(), row_merge_sort(): Replace block1, block2, block3 with a
      single array.
      ffe5998e
    • marko's avatar
      branches/zip: Reimplement merge sort in fast index creation. · fd91f7bb
      marko authored
      The creation of the primary key does not work.  We will have to flag
      externally stored columns and copy the externally stored part from
      the old table.
      
      row_build_index_for_mysql(): Rename to row_merge_build_indexes().
      Move from row0mysql.c to row0merge.c.
      
      Remove private declarations from row0merge.h.  Make many functions static
      in row0merge.c.
      
      cmp_rec_rec_simple(): A new comparison function.
      
      dict_index_get_min_size(): New function.
      
      OS_FILE_FROM_FD(fd): A macro for converting from int to os_file_t.
      
      rec_convert_dtuple_to_rec_comp(): Make the interface lower-level.
      
      rec_get_converted_size_comp(): Return also extra_size.
      
      UT_SORT_FUNCTION_BODY(): Remove reference to an obsolete test program.
      
      row_rec_to_index_entry_low(): New function.
      
      row0merge.c: Implement merge sort based on file streams instead of
      fixed-size blocks.  Sort the small blocks as arrays of dfield_t*,
      because it is faster than invoking rec_get_offsets() for every
      comparison.
      fd91f7bb
    • marko's avatar
      branches/zip: Merge 1553:1556 from trunk. · 2d90ae68
      marko authored
      2d90ae68