btr_cur_pessimistic_insert(): Do not attempt to convert the
metadata BLOB to external storage if it has already been converted.
It could have been converted by btr_cur_pessimistic_update().
Relax some over-zealous assertions.
dtuple_convert_big_rec(): Assert that the metadata BLOB has
not been converted yet.
dict_table_t::instant_column(): Correctly compute the value of
metadata_changed. The original computation in
commit 003720755f would essentially
invoke memcmp(x,x,y), which can only return 0.
This is a regression after MDEV-13671.
The bug is related to key part prefix lengths wich are stored in SYS_FIELDS.
Storage format is not obvious and was handled incorrectly which led to data
dictionary corruption.
SYS_FIELDS.POS actually contains prefix length too in case if any key part
has prefix length.
innobase_rename_column_try(): fixed prefixes handling
Tests for prefixed indexes added too.
Closes#1063
Orphan #sql* tables may remain after ALTER TABLE
was interrupted by timeout or KILL or client disconnect.
This is a regression caused by MDEV-16515.
Similar to temporary tables (MDEV-16647), we had better ignore the
KILL when dropping the original table in the final part of ALTER TABLE.
Closes#1020
The fix for MDEV-17901 did not cover cases where the AUTO_INCREMENT
column was not dropped, but some other columns before it were.
commit_cache_norebuild(): Revert the MDEV-17901 fix.
dict_index_t::clear_instant_alter(): Update table->persistent_autoinc.
innobase_instant_try(): Only try to update the hidden metadata
record if the number of columns is changing (increasing) or
a metadata BLOB is being added due to permuting or dropping columns
for the first time.
dict_table_t::instant_column(), ha_innobase_inplace_ctx::instant_column():
Return whether the metadata record needs to be updated.
This assertion should have been relaxed when implementing the first part of
MDEV-15563: instant removal of NOT NULL attribute for ROW_FORMAT=REDUNDANT
tables.
For ROW_FORMAT=REDUNDANT, there is no bitmap of null columns;
the null flags are encoded in the end offset of each field.
We do not really care about the number of fields that can be NULL.
Ensure that the 'auxiliary transactions' that are there for
flushing the incomplete undo log of the to-be-recovered DDL
transactions are actually making modifications.
The error message modified.
Then the TABLE_SHARE::error_table_name() implementation taken from 10.3,
to be used as a name of the table in this message.
Innochecksum was being built as if WITH_INNODB_BUG_ENDIAN_CRC32:BOOL=OFF
had been specified.
Also, clean up tests:
innodb.innochecksum: Useless; superceded by innodb_zip.innochecksum.
innodb.innodb_zip_innochecksum: Remove; duplicated innodb_zip.innochecksum.
innodb.innodb_zip_innochecksum2: Remove; duplicated innodb_zip.innochecksum_2.
innodb.innodb_zip_innochecksum3: Remove; duplicated innodb_zip.innochecksum_3.
No test case was added. I tested manually by adding debug instrumentation
to both innochecksum and buf_page_is_checksum_valid_crc32() to make
innochecksum write the buggy crc32, and to get warnings for falling back
to the buggy checksum. Automating this would require that tests be
adjusted depending on the build options.
Allow instant changes of columns in ROW_FORMAT=REDUNDANT
from NOT NULL to NULL.
Later, this may be implemented for ROW_FORMAT=COMPACT or DYNAMIC,
but in that case any indexes on the table must be rebuilt.
dict_table_t::prepare_instant(): Add some debug assertions,
and relax a debug assertion so that the number of fields is
allowed not to change.
dict_index_t::instant_add_field(): Relax a debug assertion,
allowing a column to change from NOT NULL to NULL.
dict_table_t::instant_column(): Add debug assertions.
instant_alter_column_possible(): Allow ALTER_COLUMN_NULLABLE
when applicable.
innodb_insert_sys_columns(): Add the parameter bool update=false
to run UPDATE instead of INSERT.
innobase_instant_add_col(): Remove; let the only caller invoke
innodb_insert_sys_columns() directly.
innobase_instant_try(): Update the SYS_COLUMNS record if the
column is changed. Only convert the table to the instant ALTER TABLE
format if necessary. For ALTER_COLUMN_NULLABLE in ROW_FORMAT=REDUNDANT,
there is no data format change.
With innodb_default_row_format=redundant, InnoDB would crash when
using table options that are incompatible with ROW_FORMAT=REDUNDANT.
create_table_info_t::m_default_row_format: Cache the value of
innodb_default_row_format.
create_table_info_t::check_table_options(): Validate ROW_TYPE_DEFAULT
with m_default_row_format.
create_table_info_t::innobase_table_flags(): Use the
cached m_default_row_format.
create_table_info_t: Never read m_form->s->row_type.
Use m_create_info->row_type instead.
dict_tf_set(): Never set invalid flags for ROW_FORMAT=REDUNDANT.
ha_innobase::truncate(): Set info.row_type based on the ROW_FORMAT
of the current table.
ha_innobase::truncate(): Because CREATE TEMPORARY TABLE
allows invalid table options when innodb_file_per_table=1,
do allow them also in TRUNCATE for temporary tables.
btr_node_ptr_max_size(): Treat CHAR(0) from SQL as a special case.
The InnoDB internal SQL parser maps the type "CHAR" to DATA_VARCHAR,
but MariaDB does allow CHAR(0) with an empty value, and does enforce
the length limitation.
btr_cur_pessimistic_insert(): Convert the metadata field of the metadata
record into BLOB before inserting, just like btr_cur_optimistic_insert()
does.
ha_innobase::prepare_inplace_alter_table(): check max column length for every
index in a table, not just added in this particular ALTER TABLE with ADD INDEX ones.
We failed to reset the dict_table_t::persistent_autoinc after
instantly dropping an AUTO_INCREMENT column, causing a bogus
call to row_parse_int() on a subsequent insert.
The result is updated here because by changing the defaults for use_stat_tables
EITs statistcs are providing more accurate statistics than INNODB statistics.
row_undo_ins_remove_clust_rec(): When rolling back the insert
of a metadata record, also reset the root page state within the
same mini-transaction.
dict_table_t::rollback_instant(): Do not allow index->n_core_fields
to be larger than index->n_fields. This logic exists in the 10.3
branch but was incorrectly removed as part of MDEV-15562 in 10.4.0.
This is a regression due to MDEV-17816.
When creating a table fails, we must roll back the dictionary
transaction. Because the rollback may rename tables, and because
InnoDB lacks proper undo logging for CREATE operations, we must
drop the incompletely created table before rolling back the
transaction, which could include a RENAME operation.
But, we must not blindly drop the table by name; after all,
the operation could have failed because another table by the
same name already existed.
create_table_info_t::m_drop_before_rollback: A flag that is set
if the table needs to be dropped before transaction rollback.
create_table_info_t::create_table(): Remove some duplicated
error handling.
ha_innobase::create(): On error, only drop the table if it was
actually created.
