Re: 98ebe0a3afc: MDEV-19123 Change default charset from latin1 to utf8mb4
Hi, Alexander, No comments about actual code changes, only about tests. Please, see below On Jun 12, Alexander Barkov wrote:
revision-id: 98ebe0a3afc (mariadb-11.5.1-12-g98ebe0a3afc) parent(s): 186a30de58b author: Alexander Barkov committer: Alexander Barkov timestamp: 2024-06-11 14:17:11 +0400 message:
MDEV-19123 Change default charset from latin1 to utf8mb4
Changing the default server character set from latin1 to utf8mb4.
diff --git a/mysql-test/main/column_compression.test b/mysql-test/main/column_compression.test --- a/mysql-test/main/column_compression.test +++ b/mysql-test/main/column_compression.test @@ -9,20 +9,20 @@ let $typec= BLOB COMPRESSED; let $typeu= BLOB; --source column_compression.inc
-let $typec= TEXT COMPRESSED; -let $typeu= TEXT; +let $typec= TEXT COMPRESSED CHARACTER SET latin1; +let $typeu= TEXT CHARACTER SET latin1;
why?
--source column_compression.inc
let $typec= VARBINARY(10000) COMPRESSED; let $typeu= VARBINARY(10000); --source column_compression.inc
-let $typec= VARCHAR(10000) COMPRESSED; -let $typeu= VARCHAR(10000); +let $typec= VARCHAR(10000) COMPRESSED CHARACTER SET latin1; +let $typeu= VARCHAR(10000) CHARACTER SET latin1; --source column_compression.inc
let $typec= TEXT COMPRESSED CHARSET ucs2; -let $typeu= TEXT; +let $typeu= TEXT CHARACTER SET latin1; --source column_compression.inc
SET column_compression_zlib_wrap=DEFAULT; diff --git a/mysql-test/main/create.result b/mysql-test/main/create.result --- a/mysql-test/main/create.result +++ b/mysql-test/main/create.result @@ -506,9 +506,9 @@ FROM t1; SHOW CREATE TABLE t2; Table Create Table t2 CREATE TABLE `t2` ( - `ifnull(c_tinytext, CAST('yet another binary data' AS BINARY))` tinyblob DEFAULT NULL, - `ifnull(c_text, CAST('yet another binary data' AS BINARY))` blob DEFAULT NULL, - `ifnull(c_mediumtext, CAST('yet another binary data' AS BINARY))` mediumblob DEFAULT NULL, + `ifnull(c_tinytext, CAST('yet another binary data' AS BINARY))` blob DEFAULT NULL, + `ifnull(c_text, CAST('yet another binary data' AS BINARY))` mediumblob DEFAULT NULL, + `ifnull(c_mediumtext, CAST('yet another binary data' AS BINARY))` longblob DEFAULT NULL,
looks wrong UPD: there were more changes like this below, I didn't comment on them
`ifnull(c_longtext, CAST('yet another binary data' AS BINARY))` longblob DEFAULT NULL ) ENGINE=MyISAM DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_uca1400_ai_ci DROP TABLE t2; diff --git a/mysql-test/main/ctype_utf8_def_upgrade.result b/mysql-test/main/ctype_utf8_def_upgrade.result --- a/mysql-test/main/ctype_utf8_def_upgrade.result +++ b/mysql-test/main/ctype_utf8_def_upgrade.result @@ -61,23 +61,23 @@ t1 CREATE TABLE `t1` ( PRIMARY KEY (`Host`,`Db`) ) ENGINE=MyISAM DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_uca1400_ai_ci COMMENT='Host privileges; Merged with database privileges' DROP TABLE t1; -SET @@character_set_database=DEFAULT; +SET @@collation_database=DEFAULT;
why?
# Now do the same, but doing 'ALTER DATABASE' to create the db.opt file, # instead of setting variables directly. # Emulate a pre-4.1 database without db.opt SHOW CREATE DATABASE db1; Database Create Database -db1 CREATE DATABASE `db1` /*!40100 DEFAULT CHARACTER SET utf8mb3 COLLATE utf8mb3_uca1400_ai_ci */ +db1 CREATE DATABASE `db1` /*!40100 DEFAULT CHARACTER SET utf8mb3 COLLATE utf8mb3_general_ci */ USE db1; -SELECT @@character_set_database, 'taken from defaults' AS comment; -@@character_set_database comment -utf8mb3 taken from defaults +SELECT @@collation_database, 'taken from defaults' AS comment; +@@collation_database comment +utf8mb3_general_ci taken from defaults USE test; ALTER DATABASE db1 DEFAULT CHARACTER SET latin1; USE db1; -SELECT @@character_set_database, 'taken from db.opt' AS comment; -@@character_set_database comment -latin1 taken from db.opt +SELECT @@collation_database, 'taken from db.opt' AS comment; +@@collation_database comment +latin1_swedish_ci taken from db.opt SELECT COUNT(*) FROM t1; ERROR HY000: Got error 190 "Incompatible key or row definition between the MariaDB .frm file and the information in the storage engine. You can try REPAIR TABLE ... USE_FRM possibly followed by ALTER TABLE ... FORCE or dump and restore the table to fix this" from storage engine MyISAM REPAIR TABLE t1 USE_FRM; diff --git a/mysql-test/main/endspace.test b/mysql-test/main/endspace.test --- a/mysql-test/main/endspace.test +++ b/mysql-test/main/endspace.test @@ -13,7 +13,7 @@ drop table if exists t1; # Test MyISAM tables. #
-create table t1 (text1 varchar(32) not NULL, KEY key1 (text1)); +create table t1 (text1 varchar(32) not NULL, KEY key1 (text1)) charset=latin1;
why is that?
insert into t1 values ('teststring'), ('nothing'), ('teststring\t'); check table t1; select * from t1 ignore key (key1) where text1='teststring' or diff --git a/mysql-test/main/func_compress.test b/mysql-test/main/func_compress.test --- a/mysql-test/main/func_compress.test +++ b/mysql-test/main/func_compress.test @@ -1,5 +1,9 @@ -- source include/have_compress.inc -- source include/have_normal_zlib.inc + +--source include/test_db_charset_latin1.inc
same as in column_compression.test ?
+ + # # Test for compress and uncompress functions: # diff --git a/mysql-test/main/long_unique.result b/mysql-test/main/long_unique.result --- a/mysql-test/main/long_unique.result +++ b/mysql-test/main/long_unique.result @@ -41,7 +41,7 @@ Ignored NO
MyISAM file: DATADIR/test/t1 Record format: Packed -Character set: latin1_swedish_ci (8) +Character set: ? (0)
huh?
Data records: 10 Deleted blocks: 0 Recordlength: 12
diff --git a/mysql-test/main/mysqlbinlog_row_compressed.result b/mysql-test/main/mysqlbinlog_row_compressed.result --- a/mysql-test/main/mysqlbinlog_row_compressed.result +++ b/mysql-test/main/mysqlbinlog_row_compressed.result @@ -40,9 +40,10 @@ SET @@session.sql_mode=#/*!*/; SET @@session.auto_increment_increment=1, @@session.auto_increment_offset=1/*!*/; /*!\C latin1 *//*!*/; SET @@session.character_set_client=latin1,@@session.collation_connection=8,@@session.collation_server=#/*!*/; +SET @@session.character_set_collations='utf8mb3=utf8mb3_uca1400_ai_ci,ucs2=ucs2_uca1400_ai_ci,utf8mb4=utf8mb4_uca1400_ai_ci,utf16=utf16_uca1400_ai_ci,utf32=utf32_uca1400_ai_ci'/*!*/;
why did this appear?
SET @@session.lc_time_names=0/*!*/; SET @@session.collation_database=DEFAULT/*!*/; -CREATE TABLE t1 (pk INT PRIMARY KEY, f1 INT, f2 INT, f3 TINYINT, f4 MEDIUMINT, f5 BIGINT, f6 INT, f7 INT, f8 char(1)) +CREATE TABLE t1 (pk INT PRIMARY KEY, f1 INT, f2 INT, f3 TINYINT, f4 MEDIUMINT, f5 BIGINT, f6 INT, f7 INT, f8 char(1)) CHARSET=latin1 /*!*/; # at <pos> #<date> server id 1 end_log_pos # CRC32 XXX GTID 0-1-2 ddl thread_id=TID diff --git a/mysql-test/main/sp.result b/mysql-test/main/sp.result --- a/mysql-test/main/sp.result +++ b/mysql-test/main/sp.result @@ -6800,6 +6800,8 @@ DROP FUNCTION IF EXISTS f2; #
CREATE FUNCTION f1() RETURNS VARCHAR(65525) RETURN 'Hello'; +Warnings: +Note 1246 Converting column '' from VARCHAR to TEXT
this could use latin1, I suppose
CREATE FUNCTION f2() RETURNS TINYINT RETURN 1;
diff --git a/mysql-test/suite/binlog_encryption/rpl_special_charset.opt b/mysql-test/suite/binlog_encryption/rpl_special_charset.opt --- a/mysql-test/suite/binlog_encryption/rpl_special_charset.opt +++ b/mysql-test/suite/binlog_encryption/rpl_special_charset.opt @@ -1 +1 @@ ---character-set-server=utf16 +--character-set-server=utf16 --collation-server=utf16_general_ci
why?
diff --git a/mysql-test/suite/federated/assisted_discovery.test b/mysql-test/suite/federated/assisted_discovery.test --- a/mysql-test/suite/federated/assisted_discovery.test +++ b/mysql-test/suite/federated/assisted_discovery.test @@ -38,7 +38,7 @@ create table t1 ( d varchar(4096) not null, primary key (a), key (b,c,d(255)) -); +) CHARSET=latin1;
why?
show create table t1;
connection master; diff --git a/mysql-test/suite/funcs_1/r/is_columns_innodb.result b/mysql-test/suite/funcs_1/r/is_columns_innodb.result --- a/mysql-test/suite/funcs_1/r/is_columns_innodb.result +++ b/mysql-test/suite/funcs_1/r/is_columns_innodb.result @@ -739,6 +740,9 @@ WHERE table_schema LIKE 'test%' AND CHARACTER_OCTET_LENGTH / CHARACTER_MAXIMUM_LENGTH <> 1 ORDER BY CHARACTER_SET_NAME, COLLATION_NAME, COL_CML; COL_CML DATA_TYPE CHARACTER_SET_NAME COLLATION_NAME +4.0000 char utf8mb4 utf8mb4_uca1400_ai_ci +4.0000 enum utf8mb4 utf8mb4_uca1400_ai_ci +4.0000 set utf8mb4 utf8mb4_uca1400_ai_ci
I suppose this changed the intention of the test
Warnings: Warning 1365 Division by 0 Warning 1365 Division by 0 diff --git a/mysql-test/suite/gcol/r/innodb_virtual_index.result b/mysql-test/suite/gcol/r/innodb_virtual_index.result --- a/mysql-test/suite/gcol/r/innodb_virtual_index.result +++ b/mysql-test/suite/gcol/r/innodb_virtual_index.result @@ -114,7 +114,7 @@ KEY `vbidxcol` (`vbidxcol`), KEY `a_2` (`a`,`vbidxcol`), KEY `vbidxcol_2` (`vbidxcol`), FULLTEXT KEY `ftsic` (`c`,`b`) -) ENGINE=InnoDB; +) ENGINE=InnoDB CHARSET=latin1;
why latin1 everywhere in this file?
Warnings: Note 1831 Duplicate index `vbidxcol_2`. This is deprecated and will be disallowed in a future release ALTER TABLE ibstd_08 ADD COLUMN nc07006 BIGINT AUTO_INCREMENT NOT NULL , ADD KEY auto_nc07006(nc07006); diff --git a/mysql-test/suite/innodb/r/online_table_rebuild.result b/mysql-test/suite/innodb/r/online_table_rebuild.result --- a/mysql-test/suite/innodb/r/online_table_rebuild.result +++ b/mysql-test/suite/innodb/r/online_table_rebuild.result @@ -10,7 +10,7 @@ INSERT INTO t1 VALUES(2, repeat('b', 100), repeat('a', 100)); COMMIT; SET DEBUG_SYNC="now SIGNAL dml_commit"; connection default; -ERROR 23000: Duplicate entry 'bbbbbbbbbb' for key 'f2' +ERROR 23000: Duplicate entry 'bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb' for key 'f2'
why?
connection default; SET DEBUG_SYNC="inplace_after_index_build SIGNAL dml_start WAIT_FOR dml_commit"; ALTER TABLE t1 ADD PRIMARY KEY(f1); diff --git a/mysql-test/suite/innodb/t/alter_primary_key.test b/mysql-test/suite/innodb/t/alter_primary_key.test --- a/mysql-test/suite/innodb/t/alter_primary_key.test +++ b/mysql-test/suite/innodb/t/alter_primary_key.test @@ -1,6 +1,9 @@ --source innodb_default_row_format.inc --source include/have_debug.inc --source include/have_debug_sync.inc +--disable_query_log +--source include/test_db_charset_latin1.inc +--enable_query_log
you don't disable query log in other tests, why here?
--echo # --echo # MDEV-23244 ALTER TABLE…ADD PRIMARY KEY fails to flag diff --git a/mysql-test/suite/perfschema/t/short_option_1-master.opt b/mysql-test/suite/perfschema/t/short_option_1-master.opt --- a/mysql-test/suite/perfschema/t/short_option_1-master.opt +++ b/mysql-test/suite/perfschema/t/short_option_1-master.opt @@ -1 +1 @@ --a -Cutf8 -W1 +-a -Cutf8 -W1 --collation-server=utf8_general_ci
was it necessary?
diff --git a/mysql-test/suite/rpl/include/rpl_charset.inc b/mysql-test/suite/rpl/include/rpl_charset.inc --- a/mysql-test/suite/rpl/include/rpl_charset.inc +++ b/mysql-test/suite/rpl/include/rpl_charset.inc @@ -145,4 +145,11 @@ sync_slave_with_master; --echo #
+connection master; +SET GLOBAL collation_server=utf8mb4_uca1400_ai_ci; +SET SESSION collation_server=utf8mb4_uca1400_ai_ci; +connection slave; +SET GLOBAL collation_server=utf8mb4_uca1400_ai_ci; +SET SESSION collation_server=utf8mb4_uca1400_ai_ci;
why?
+ --source include/rpl_end.inc diff --git a/mysql-test/suite/sys_vars/inc/secure_timestamp_func.inc b/mysql-test/suite/sys_vars/inc/secure_timestamp_func.inc --- a/mysql-test/suite/sys_vars/inc/secure_timestamp_func.inc +++ b/mysql-test/suite/sys_vars/inc/secure_timestamp_func.inc @@ -58,7 +58,7 @@ set time_zone='+00:00'; set timestamp=1234567890.101112; select @@timestamp, now(6);
-create table t1 (b varchar(20), a timestamp(6) default current_timestamp(6)); +create table t1 (b varchar(20), a timestamp(6) default current_timestamp(6)) charset=latin1;
why? charset shouldn't affect secure_timestamp functionality (same in all other secure_timestamp tests)
insert t1 (b) values ('replicated'); sync_slave_with_master; create trigger t1rbr before insert on t1 for each row set new.a=now(6); diff --git a/storage/connect/mysql-test/connect/r/bson.result b/storage/connect/mysql-test/connect/r/bson.result --- a/storage/connect/mysql-test/connect/r/bson.result +++ b/storage/connect/mysql-test/connect/r/bson.result @@ -321,7 +321,7 @@ WHO CHAR(12), WEEK INT(2) JPATH='$.WEEK[2].NUMBER', WHAT CHAR(32) JPATH='$.WEEK[2].EXPENSE[*].WHAT', AMOUNT DOUBLE(8,2) JPATH='$.WEEK[2].EXPENSE[*].AMOUNT') -ENGINE=CONNECT TABLE_TYPE=BSON FILE_NAME='expense.json'; +ENGINE=CONNECT CHARSET=latin1 TABLE_TYPE=BSON FILE_NAME='expense.json';
is it necessary? are there any connect/bson tests for utf8mb4?
SELECT * FROM t4; WHO WEEK WHAT AMOUNT Joe 5 Beer 14.00 diff --git a/storage/connect/mysql-test/connect/r/endian.result b/storage/connect/mysql-test/connect/r/endian.result --- a/storage/connect/mysql-test/connect/r/endian.result +++ b/storage/connect/mysql-test/connect/r/endian.result @@ -10,7 +10,7 @@ birth DATE NOT NULL FIELD_FORMAT='L', id CHAR(5) NOT NULL FIELD_FORMAT='L2', salary DOUBLE(9,2) NOT NULL DEFAULT 0.00 FIELD_FORMAT='LF', dept INT(4) NOT NULL FIELD_FORMAT='L2' -) ENGINE=CONNECT TABLE_TYPE=BIN BLOCK_SIZE=5 FILE_NAME='Testbal.dat'; +) ENGINE=CONNECT CHARSET=latin1 TABLE_TYPE=BIN BLOCK_SIZE=5 FILE_NAME='Testbal.dat';
or any connect utf8mb4 tests whatsoever?
SELECT * FROM t1; fig name birth id salary dept 5500 ARCHIBALD 1980-01-25 3789 4380.50 318 diff --git a/storage/rocksdb/mysql-test/rocksdb/r/checksum_table_live.result b/storage/rocksdb/mysql-test/rocksdb/r/checksum_table_live.result --- a/storage/rocksdb/mysql-test/rocksdb/r/checksum_table_live.result +++ b/storage/rocksdb/mysql-test/rocksdb/r/checksum_table_live.result @@ -1,7 +1,7 @@ DROP TABLE IF EXISTS t1,t2; -CREATE TABLE t1 (a INT PRIMARY KEY, b CHAR(8)) ENGINE=rocksdb CHECKSUM=1; +CREATE TABLE t1 (a INT PRIMARY KEY, b CHAR(8)) ENGINE=rocksdb CHECKSUM=1 CHARSET=latin1;
again. was it necessary? are there any rocksdb tests with utf8mb4?
INSERT INTO t1 (a,b) VALUES (1,'a'),(2,'b'); -CREATE TABLE t2 (a INT PRIMARY KEY, b CHAR(8)) ENGINE=rocksdb CHECKSUM=1; +CREATE TABLE t2 (a INT PRIMARY KEY, b CHAR(8)) ENGINE=rocksdb CHECKSUM=1 CHARSET=latin1; CHECKSUM TABLE t1; Table Checksum test.t1 4259194219
Regards, Sergei Chief Architect, MariaDB Server and security@mariadb.org
Hi Sergei, Thanks for the review. Please see comments inline. I commited a patch with a comment "Addressing MDEV-19123 review suggestions" on top of the main patch. Please find it in the same branch: https://github.com/MariaDB/server/commits/bb-11.6-bar-MDEV-19123/ In the end, we can either squash it, or leave as an incremental patch. I'm fine with both ways. Please suggest.
Hi, Alexander,
No comments about actual code changes, only about tests. Please, see below
On Jun 12, Alexander Barkov wrote:
revision-id: 98ebe0a3afc (mariadb-11.5.1-12-g98ebe0a3afc) parent(s): 186a30de58b author: Alexander Barkov committer: Alexander Barkov timestamp: 2024-06-11 14:17:11 +0400 message:
MDEV-19123 Change default charset from latin1 to utf8mb4
Changing the default server character set from latin1 to utf8mb4.
diff --git a/mysql-test/main/column_compression.test
b/mysql-test/main/column_compression.test
--- a/mysql-test/main/column_compression.test +++ b/mysql-test/main/column_compression.test @@ -9,20 +9,20 @@ let $typec= BLOB COMPRESSED; let $typeu= BLOB; --source column_compression.inc
-let $typec= TEXT COMPRESSED; -let $typeu= TEXT; +let $typec= TEXT COMPRESSED CHARACTER SET latin1; +let $typeu= TEXT CHARACTER SET latin1;
why?
It was not really necessary to add latin1. But for better coverage, I decided to have both. Fixed to this: let $typec= TEXT COMPRESSED; let $typeu= TEXT; --source column_compression.inc let $typec= TEXT COMPRESSED CHARACTER SET latin1; let $typeu= TEXT CHARACTER SET latin1; --source column_compression.inc <cut>
diff --git a/mysql-test/main/create.result b/mysql-test/main/create.result --- a/mysql-test/main/create.result +++ b/mysql-test/main/create.result @@ -506,9 +506,9 @@ FROM t1; SHOW CREATE TABLE t2; Table Create Table t2 CREATE TABLE `t2` ( - `ifnull(c_tinytext, CAST('yet another binary data' AS BINARY))` tinyblob DEFAULT NULL, - `ifnull(c_text, CAST('yet another binary data' AS BINARY))` blob DEFAULT NULL, - `ifnull(c_mediumtext, CAST('yet another binary data' AS BINARY))` mediumblob DEFAULT NULL, + `ifnull(c_tinytext, CAST('yet another binary data' AS BINARY))` blob DEFAULT NULL, + `ifnull(c_text, CAST('yet another binary data' AS BINARY))` mediumblob DEFAULT NULL, + `ifnull(c_mediumtext, CAST('yet another binary data' AS BINARY))` longblob DEFAULT NULL,
looks wrong UPD: there were more changes like this below, I didn't comment on them
This is an old bug. I reported it as: MDEV-34376 Wrong data types when mixing an utf8 *TEXT column and a short binary It's now fixed, please review a patch for 10.5.
`ifnull(c_longtext, CAST('yet another binary data' AS
BINARY))` longblob DEFAULT NULL
) ENGINE=MyISAM DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_uca1400_ai_ci DROP TABLE t2; diff --git a/mysql-test/main/ctype_utf8_def_upgrade.result b/mysql-test/main/ctype_utf8_def_upgrade.result --- a/mysql-test/main/ctype_utf8_def_upgrade.result +++ b/mysql-test/main/ctype_utf8_def_upgrade.result @@ -61,23 +61,23 @@ t1 CREATE TABLE `t1` ( PRIMARY KEY (`Host`,`Db`) ) ENGINE=MyISAM DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_uca1400_ai_ci COMMENT='Host privileges; Merged with database privileges' DROP TABLE t1; -SET @@character_set_database=DEFAULT; +SET @@collation_database=DEFAULT;
why?
The current database is test, and its collation is utf8mb4_uca1400_ai_ci, so is the value of @@collation_database in the beginning of the test. The default collation for utf8mb4 is also utf8mb4_uca1400_ai_ci. You're right, the change is not really necessary, because under the above conditions these two statements have exactly the same effect: SET @@character_set_database=DEFAULT; SET @@collation_database=DEFAULT; So it does not really matter which of them to use. It would matter if the default collation for utf8mb4 was something else than utf8mb4_uca1400_ai_ci. But as the test sets in the beginning @@collation_database in this statement: SET @@collation_database="latin1_swedish_ci"; logically, its more correct to restore to the default value the same variable, @@collation_database, rather than @@character_set_database. But I can change it back to restore @@character_set_database, if you think this change is rather relevant. <cut>
diff --git a/mysql-test/main/endspace.test b/mysql-test/main/endspace.test --- a/mysql-test/main/endspace.test +++ b/mysql-test/main/endspace.test @@ -13,7 +13,7 @@ drop table if exists t1; # Test MyISAM tables. #
-create table t1 (text1 varchar(32) not NULL, KEY key1 (text1)); +create table t1 (text1 varchar(32) not NULL, KEY key1 (text1)) charset=latin1;
why is that?
The charset clause is not really needed. I removed it and re-recorded with new key sizes in the explain select output. <cut>
diff --git a/mysql-test/main/func_compress.test b/mysql-test/main/func_compress.test --- a/mysql-test/main/func_compress.test +++ b/mysql-test/main/func_compress.test @@ -1,5 +1,9 @@ -- source include/have_compress.inc -- source include/have_normal_zlib.inc + +--source include/test_db_charset_latin1.inc
same as in column_compression.test ?
The underlying columns storing result of compress() were char/varchar/text. Fixed them to binary/varbinary/blob. Removed the switch to latin1. <cut>
--- a/mysql-test/main/long_unique.result +++ b/mysql-test/main/long_unique.result @@ -41,7 +41,7 @@ Ignored NO
MyISAM file: DATADIR/test/t1 Record format: Packed -Character set: latin1_swedish_ci (8) +Character set: ? (0)
huh?
MyISAM uses only one byte to store the ID of the table-level default collation: myisamdef.h: uchar language; /* Language for indexes */ It seems this member is used in myisamchk.c only. Any ideas what to do with that? <cut>
diff --git a/mysql-test/main/mysqlbinlog_row_compressed.result b/mysql-test/main/mysqlbinlog_row_compressed.result --- a/mysql-test/main/mysqlbinlog_row_compressed.result +++ b/mysql-test/main/mysqlbinlog_row_compressed.result @@ -40,9 +40,10 @@ SET @@session.sql_mode=#/*!*/; SET @@session.auto_increment_increment=1, @@session.auto_increment_offset=1/*!*/; /*!\C latin1 *//*!*/; SET @@session.character_set_client=latin1,@@session.collation_connection=8,@@session.collation_server=#/*!*/; +SET @@session.character_set_collations='utf8mb3=utf8mb3_uca1400_ai_ci,ucs2=ucs2_uca1400_ai_ci,utf8mb4=utf8mb4_uca1400_ai_ci,utf16=utf16_uca1400_ai_ci,utf32=utf32_uca1400_ai_ci'/*!*/;
why did this appear?
MTR now starts mariadbd with these lines in my.cnf: character-set-server=utf8mb4 collation-server=utf8mb4_uca1400_ai_ci When mariadbd processes the first line, it finds the default collation for utf8mb4, which adds the CHARACTER_SET_COLLATIONS_USED flag. This forces the server to write @@character_set_collations to the binary log. However, when it processes the second line, the flag becomes not needed. The code could be rewritten to avoid adding CHARACTER_SET_COLLATIONS_USED if both character-set-server and collation-server are specified. The flag is needed when only character-set-server is specified without collation-server. <cut>
diff --git a/mysql-test/main/sp.result b/mysql-test/main/sp.result --- a/mysql-test/main/sp.result +++ b/mysql-test/main/sp.result @@ -6800,6 +6800,8 @@ DROP FUNCTION IF EXISTS f2; #
CREATE FUNCTION f1() RETURNS VARCHAR(65525) RETURN 'Hello'; +Warnings: +Note 1246 Converting column '' from VARCHAR to TEXT
this could use latin1, I suppose
Right, I checked: Bug#20550: Stored function: wrong RETURN type metadata when used in a VIEW. and utf8mb4 indeed changed the meaning of the test. To preserve the meaning I fixed this way: CREATE FUNCTION f1() RETURNS VARCHAR(65525) CHARACTER SET latin1 RETURN 'Hello';
CREATE FUNCTION f2() RETURNS TINYINT RETURN 1;
diff --git
a/mysql-test/suite/binlog_encryption/rpl_special_charset.opt b/mysql-test/suite/binlog_encryption/rpl_special_charset.opt
--- a/mysql-test/suite/binlog_encryption/rpl_special_charset.opt +++ b/mysql-test/suite/binlog_encryption/rpl_special_charset.opt @@ -1 +1 @@ ---character-set-server=utf16 +--character-set-server=utf16 --collation-server=utf16_general_ci
why?
It's need needed. Removed.
diff --git a/mysql-test/suite/federated/assisted_discovery.test
b/mysql-test/suite/federated/assisted_discovery.test
--- a/mysql-test/suite/federated/assisted_discovery.test +++ b/mysql-test/suite/federated/assisted_discovery.test @@ -38,7 +38,7 @@ create table t1 ( d varchar(4096) not null, primary key (a), key (b,c,d(255)) -); +) CHARSET=latin1;
why?
Because of this error: Specified key was too long; max key length is 1000 bytes I removed the CHARSET clause and changed the index size to 200.
show create table t1;
connection master; diff --git a/mysql-test/suite/funcs_1/r/is_columns_innodb.result
b/mysql-test/suite/funcs_1/r/is_columns_innodb.result
--- a/mysql-test/suite/funcs_1/r/is_columns_innodb.result +++ b/mysql-test/suite/funcs_1/r/is_columns_innodb.result @@ -739,6 +740,9 @@ WHERE table_schema LIKE 'test%' AND CHARACTER_OCTET_LENGTH / CHARACTER_MAXIMUM_LENGTH <> 1 ORDER BY CHARACTER_SET_NAME, COLLATION_NAME, COL_CML; COL_CML DATA_TYPE CHARACTER_SET_NAME COLLATION_NAME +4.0000 char utf8mb4 utf8mb4_uca1400_ai_ci +4.0000 enum utf8mb4 utf8mb4_uca1400_ai_ci +4.0000 set utf8mb4 utf8mb4_uca1400_ai_ci
I suppose this changed the intention of the test
I overlooked that this test creates one more database `test4`. Added CHARACTER SET latin1 into its CREATE statement.
Warnings: Warning 1365 Division by 0 Warning 1365 Division by 0 diff --git a/mysql-test/suite/gcol/r/innodb_virtual_index.result b/mysql-test/suite/gcol/r/innodb_virtual_index.result --- a/mysql-test/suite/gcol/r/innodb_virtual_index.result +++ b/mysql-test/suite/gcol/r/innodb_virtual_index.result @@ -114,7 +114,7 @@ KEY `vbidxcol` (`vbidxcol`), KEY `a_2` (`a`,`vbidxcol`), KEY `vbidxcol_2` (`vbidxcol`), FULLTEXT KEY `ftsic` (`c`,`b`) -) ENGINE=InnoDB; +) ENGINE=InnoDB CHARSET=latin1;
why latin1 everywhere in this file?
It was not really needed. Removed.
Warnings: Note 1831 Duplicate index `vbidxcol_2`. This is deprecated and
will be disallowed in a future release
ALTER TABLE ibstd_08 ADD COLUMN nc07006 BIGINT AUTO_INCREMENT NOT NULL , ADD KEY auto_nc07006(nc07006); diff --git a/mysql-test/suite/innodb/r/online_table_rebuild.result b/mysql-test/suite/innodb/r/online_table_rebuild.result --- a/mysql-test/suite/innodb/r/online_table_rebuild.result +++ b/mysql-test/suite/innodb/r/online_table_rebuild.result @@ -10,7 +10,7 @@ INSERT INTO t1 VALUES(2, repeat('b', 100), repeat('a', 100)); COMMIT; SET DEBUG_SYNC="now SIGNAL dml_commit"; connection default; -ERROR 23000: Duplicate entry 'bbbbbbbbbb' for key 'f2' +ERROR 23000: Duplicate entry 'bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb' for key 'f2'
why?
Something goes wrong inside the online alter code. I filed an MDEV and added this comment into the test: +--echo # The below error displays a wrong (too long) key value. +--echo # The problem is reported as: +--echo # MDEV-34387 Too long value in the Duplicate entry message with online alter and utf8 +--echo # Please make sure to remove this comment when MDEV-34387 is fixed.
connection default; SET DEBUG_SYNC="inplace_after_index_build SIGNAL dml_start
WAIT_FOR dml_commit";
ALTER TABLE t1 ADD PRIMARY KEY(f1); diff --git a/mysql-test/suite/innodb/t/alter_primary_key.test b/mysql-test/suite/innodb/t/alter_primary_key.test --- a/mysql-test/suite/innodb/t/alter_primary_key.test +++ b/mysql-test/suite/innodb/t/alter_primary_key.test @@ -1,6 +1,9 @@ --source innodb_default_row_format.inc --source include/have_debug.inc --source include/have_debug_sync.inc +--disable_query_log +--source include/test_db_charset_latin1.inc +--enable_query_log
you don't disable query log in other tests, why here?
Removed the disabling.
--echo # --echo # MDEV-23244 ALTER TABLE…ADD PRIMARY KEY fails to flag diff --git
a/mysql-test/suite/perfschema/t/short_option_1-master.opt b/mysql-test/suite/perfschema/t/short_option_1-master.opt
--- a/mysql-test/suite/perfschema/t/short_option_1-master.opt +++ b/mysql-test/suite/perfschema/t/short_option_1-master.opt @@ -1 +1 @@ --a -Cutf8 -W1 +-a -Cutf8 -W1 --collation-server=utf8_general_ci
was it necessary?
Without it, the server fails to start with this output in mysql-test/var/log/mysqld.1.err: COLLATION 'utf8mb4_uca1400_ai_ci' is not valid for CHARACTER SET 'utf8mb3'
diff --git a/mysql-test/suite/rpl/include/rpl_charset.inc
b/mysql-test/suite/rpl/include/rpl_charset.inc
--- a/mysql-test/suite/rpl/include/rpl_charset.inc +++ b/mysql-test/suite/rpl/include/rpl_charset.inc @@ -145,4 +145,11 @@ sync_slave_with_master; --echo #
+connection master; +SET GLOBAL collation_server=utf8mb4_uca1400_ai_ci; +SET SESSION collation_server=utf8mb4_uca1400_ai_ci; +connection slave; +SET GLOBAL collation_server=utf8mb4_uca1400_ai_ci; +SET SESSION collation_server=utf8mb4_uca1400_ai_ci;
why?
This tests changes both global and session @@collation_server. These commands are needed to restore them back, to avoid internal check failure.
+ --source include/rpl_end.inc diff --git
a/mysql-test/suite/sys_vars/inc/secure_timestamp_func.inc b/mysql-test/suite/sys_vars/inc/secure_timestamp_func.inc
--- a/mysql-test/suite/sys_vars/inc/secure_timestamp_func.inc +++ b/mysql-test/suite/sys_vars/inc/secure_timestamp_func.inc @@ -58,7 +58,7 @@ set time_zone='+00:00'; set timestamp=1234567890.101112; select @@timestamp, now(6);
-create table t1 (b varchar(20), a timestamp(6) default current_timestamp(6)); +create table t1 (b varchar(20), a timestamp(6) default current_timestamp(6)) charset=latin1;
why? charset shouldn't affect secure_timestamp functionality (same in all other secure_timestamp tests)
It uses a BINLOG statement a few lines below this CREATE TABLE statement. The BINLOG statement fails without the charset=latin1 clause in CREATE TABLE.
insert t1 (b) values ('replicated'); sync_slave_with_master; create trigger t1rbr before insert on t1 for each row set
new.a=now(6);
diff --git a/storage/connect/mysql-test/connect/r/bson.result b/storage/connect/mysql-test/connect/r/bson.result --- a/storage/connect/mysql-test/connect/r/bson.result +++ b/storage/connect/mysql-test/connect/r/bson.result @@ -321,7 +321,7 @@ WHO CHAR(12), WEEK INT(2) JPATH='$.WEEK[2].NUMBER', WHAT CHAR(32) JPATH='$.WEEK[2].EXPENSE[*].WHAT', AMOUNT DOUBLE(8,2) JPATH='$.WEEK[2].EXPENSE[*].AMOUNT') -ENGINE=CONNECT TABLE_TYPE=BSON FILE_NAME='expense.json'; +ENGINE=CONNECT CHARSET=latin1 TABLE_TYPE=BSON FILE_NAME='expense.json';
is it necessary? are there any connect/bson tests for utf8mb4?
TABLE_TYPE=BSON did not work with CHARSET=utf8mb4 :(
SELECT * FROM t4; WHO WEEK WHAT AMOUNT Joe 5 Beer 14.00 diff --git a/storage/connect/mysql-test/connect/r/endian.result
b/storage/connect/mysql-test/connect/r/endian.result
--- a/storage/connect/mysql-test/connect/r/endian.result +++ b/storage/connect/mysql-test/connect/r/endian.result @@ -10,7 +10,7 @@ birth DATE NOT NULL FIELD_FORMAT='L', id CHAR(5) NOT NULL FIELD_FORMAT='L2', salary DOUBLE(9,2) NOT NULL DEFAULT 0.00 FIELD_FORMAT='LF', dept INT(4) NOT NULL FIELD_FORMAT='L2' -) ENGINE=CONNECT TABLE_TYPE=BIN BLOCK_SIZE=5 FILE_NAME='Testbal.dat'; +) ENGINE=CONNECT CHARSET=latin1 TABLE_TYPE=BIN BLOCK_SIZE=5 FILE_NAME='Testbal.dat';
or any connect utf8mb4 tests whatsoever?
I had to add CHARSET=latin1 into many tests. I just removed a few CHARSET=latin1 from dbf.test, but not all of them. Unfortunately, connect did not work with utf8mb4 in many cases. We'll need to ask Oliver to check. But this should not be a showstopper for MDEV-19123, I think.
SELECT * FROM t1; fig name birth id salary dept 5500 ARCHIBALD 1980-01-25 3789 4380.50 318 diff --git
a/storage/rocksdb/mysql-test/rocksdb/r/checksum_table_live.result b/storage/rocksdb/mysql-test/rocksdb/r/checksum_table_live.result
--- a/storage/rocksdb/mysql-test/rocksdb/r/checksum_table_live.result +++ b/storage/rocksdb/mysql-test/rocksdb/r/checksum_table_live.result @@ -1,7 +1,7 @@ DROP TABLE IF EXISTS t1,t2; -CREATE TABLE t1 (a INT PRIMARY KEY, b CHAR(8)) ENGINE=rocksdb CHECKSUM=1; +CREATE TABLE t1 (a INT PRIMARY KEY, b CHAR(8)) ENGINE=rocksdb CHECKSUM=1 CHARSET=latin1;
again. was it necessary?
If I don't change to latin1, the checksum changes. Should I record with utf8mb4 and a different checksum? I don't know how to make sure the new checksum value is correct. Or perhaps, we should have two tests: with latin1 and with utf8mb4?
are there any rocksdb tests with utf8mb4?
Yes, I modified only 16 *.test files in storage/rocksdb/mysql-test/rocksdb/, while recorded 46 files. Just rewised, adding latin1 into these files were not really necessary: storage/rocksdb/mysql-test/rocksdb/t/col_not_null.inc storage/rocksdb/mysql-test/rocksdb/t/col_null.inc Removed latin1.
INSERT INTO t1 (a,b) VALUES (1,'a'),(2,'b'); -CREATE TABLE t2 (a INT PRIMARY KEY, b CHAR(8)) ENGINE=rocksdb
CHECKSUM=1;
+CREATE TABLE t2 (a INT PRIMARY KEY, b CHAR(8)) ENGINE=rocksdb CHECKSUM=1 CHARSET=latin1; CHECKSUM TABLE t1; Table Checksum test.t1 4259194219
Hi, Alexander, On Jun 14, Alexander Barkov wrote:
Hi Sergei,
Thanks for the review. Please see comments inline.
I commited a patch with a comment "Addressing MDEV-19123 review suggestions" on top of the main patch. Please find it in the same branch:
https://github.com/MariaDB/server/commits/bb-11.6-bar-MDEV-19123/
In the end, we can either squash it, or leave as an incremental patch. I'm fine with both ways. Please suggest.
Commit bff6cb7 is basically ok. See few comments/question below. Squash it please, but not now. After yor branch will pass testing (you might need to do more fixes before that) and before pushing into the main 11.6 - that will be a good time to clean up the history and squash commits as needed. Thanks.
Hi, Alexander,
No comments about actual code changes, only about tests. Please, see below
On Jun 12, Alexander Barkov wrote:
revision-id: 98ebe0a3afc (mariadb-11.5.1-12-g98ebe0a3afc) parent(s): 186a30de58b author: Alexander Barkov committer: Alexander Barkov timestamp: 2024-06-11 14:17:11 +0400 message:
MDEV-19123 Change default charset from latin1 to utf8mb4
Changing the default server character set from latin1 to utf8mb4.
--- a/mysql-test/main/long_unique.result +++ b/mysql-test/main/long_unique.result @@ -41,7 +41,7 @@ Ignored NO
MyISAM file: DATADIR/test/t1 Record format: Packed -Character set: latin1_swedish_ci (8) +Character set: ? (0)
huh?
MyISAM uses only one byte to store the ID of the table-level default collation:
myisamdef.h: uchar language; /* Language for indexes */
It seems this member is used in myisamchk.c only.
Any ideas what to do with that?
please, create a bug report that myisamchk is broken for collations > 255. It seems that it can corrupt the table that way. Or at least it can keep repairing it, thinking that collation changes every time.
diff --git a/mysql-test/main/mysqlbinlog_row_compressed.result b/mysql-test/main/mysqlbinlog_row_compressed.result --- a/mysql-test/main/mysqlbinlog_row_compressed.result +++ b/mysql-test/main/mysqlbinlog_row_compressed.result @@ -40,9 +40,10 @@ SET @@session.sql_mode=#/*!*/; SET @@session.auto_increment_increment=1, @@session.auto_increment_offset=1/*!*/; /*!\C latin1 *//*!*/; SET @@session.character_set_client=latin1,@@session.collation_connection=8,@@session.collation_server=#/*!*/; +SET @@session.character_set_collations='utf8mb3=utf8mb3_uca1400_ai_ci,ucs2=ucs2_uca1400_ai_ci,utf8mb4=utf8mb4_uca1400_ai_ci,utf16=utf16_uca1400_ai_ci,utf32=utf32_uca1400_ai_ci'/*!*/;
why did this appear?
MTR now starts mariadbd with these lines in my.cnf:
character-set-server=utf8mb4 collation-server=utf8mb4_uca1400_ai_ci
When mariadbd processes the first line, it finds the default collation for utf8mb4, which adds the CHARACTER_SET_COLLATIONS_USED flag. This forces the server to write @@character_set_collations to the binary log.
However, when it processes the second line, the flag becomes not needed.
The code could be rewritten to avoid adding CHARACTER_SET_COLLATIONS_USED if both character-set-server and collation-server are specified. The flag is needed when only character-set-server is specified without collation-server.
Do you need to set character-set-server at all? I mean, isn't the second line alone enough? Or may be both can be removed? utf8mb4 is the default now, isn't it?
diff --git a/mysql-test/suite/innodb/t/alter_primary_key.test b/mysql-test/suite/innodb/t/alter_primary_key.test --- a/mysql-test/suite/innodb/t/alter_primary_key.test +++ b/mysql-test/suite/innodb/t/alter_primary_key.test @@ -1,6 +1,9 @@ --source innodb_default_row_format.inc --source include/have_debug.inc --source include/have_debug_sync.inc +--disable_query_log +--source include/test_db_charset_latin1.inc +--enable_query_log
you don't disable query log in other tests, why here?
Removed the disabling.
in all tests, I presume
--echo # --echo # MDEV-23244 ALTER TABLE…ADD PRIMARY KEY fails to flag diff --git a/mysql-test/suite/perfschema/t/short_option_1-master.opt b/mysql-test/suite/perfschema/t/short_option_1-master.opt --- a/mysql-test/suite/perfschema/t/short_option_1-master.opt +++ b/mysql-test/suite/perfschema/t/short_option_1-master.opt @@ -1 +1 @@ --a -Cutf8 -W1 +-a -Cutf8 -W1 --collation-server=utf8_general_ci
was it necessary?
Without it, the server fails to start with this output in mysql-test/var/log/mysqld.1.err:
COLLATION 'utf8mb4_uca1400_ai_ci' is not valid for CHARACTER SET 'utf8mb3'
Can you do -Cutf8mb4 instead? It's short_option_1.test, better to use short option names there :)
create trigger t1rbr before insert on t1 for each row set new.a=now(6); diff --git a/storage/connect/mysql-test/connect/r/endian.result b/storage/connect/mysql-test/connect/r/endian.result --- a/storage/connect/mysql-test/connect/r/endian.result +++ b/storage/connect/mysql-test/connect/r/endian.result @@ -10,7 +10,7 @@ birth DATE NOT NULL FIELD_FORMAT='L', id CHAR(5) NOT NULL FIELD_FORMAT='L2', salary DOUBLE(9,2) NOT NULL DEFAULT 0.00 FIELD_FORMAT='LF', dept INT(4) NOT NULL FIELD_FORMAT='L2' -) ENGINE=CONNECT TABLE_TYPE=BIN BLOCK_SIZE=5 FILE_NAME='Testbal.dat'; +) ENGINE=CONNECT CHARSET=latin1 TABLE_TYPE=BIN BLOCK_SIZE=5 FILE_NAME='Testbal.dat';
or any connect utf8mb4 tests whatsoever?
I had to add CHARSET=latin1 into many tests.
I just removed a few CHARSET=latin1 from dbf.test, but not all of them.
Unfortunately, connect did not work with utf8mb4 in many cases. We'll need to ask Oliver to check. But this should not be a showstopper for MDEV-19123, I think.
Of course, not. But please, create a bug report like "CONNECT doesn't work with utf8mb4", explain that commit for MDEV-19123 had to add CHARSET=latin1 to many tests because many table types (see commit) don't work with utf8mb4. And assign to Andrew. Regards, Sergei Chief Architect, MariaDB Server and security@mariadb.org
Hi Sergei, Thanks for the review. The new patch "Addressing MDEV-19123 review suggestions" is here: https://github.com/MariaDB/server/commit/b3c3adbeeb08625e8816524510d5918cdb1... Please see comments inline: On 6/14/24 22:12, Sergei Golubchik wrote:
Hi, Alexander,
On Jun 14, Alexander Barkov wrote:
Hi Sergei,
Thanks for the review. Please see comments inline.
I commited a patch with a comment "Addressing MDEV-19123 review suggestions" on top of the main patch. Please find it in the same branch:
https://github.com/MariaDB/server/commits/bb-11.6-bar-MDEV-19123/
In the end, we can either squash it, or leave as an incremental patch. I'm fine with both ways. Please suggest.
Commit bff6cb7 is basically ok. See few comments/question below.
Squash it please, but not now. After yor branch will pass testing (you might need to do more fixes before that) and before pushing into the main 11.6 - that will be a good time to clean up the history and squash commits as needed.
Okey.
Thanks.
Hi, Alexander,
No comments about actual code changes, only about tests. Please, see below
On Jun 12, Alexander Barkov wrote:
revision-id: 98ebe0a3afc (mariadb-11.5.1-12-g98ebe0a3afc) parent(s): 186a30de58b author: Alexander Barkov committer: Alexander Barkov timestamp: 2024-06-11 14:17:11 +0400 message:
MDEV-19123 Change default charset from latin1 to utf8mb4
Changing the default server character set from latin1 to utf8mb4.
--- a/mysql-test/main/long_unique.result +++ b/mysql-test/main/long_unique.result @@ -41,7 +41,7 @@ Ignored NO
MyISAM file: DATADIR/test/t1 Record format: Packed -Character set: latin1_swedish_ci (8) +Character set: ? (0)
huh?
MyISAM uses only one byte to store the ID of the table-level default collation:
myisamdef.h: uchar language; /* Language for indexes */
It seems this member is used in myisamchk.c only.
Any ideas what to do with that?
please, create a bug report that myisamchk is broken for collations > 255. It seems that it can corrupt the table that way.
Or at least it can keep repairing it, thinking that collation changes every time.
MDEV-34409 myisamchk is broken for collation IDs >255
diff --git a/mysql-test/main/mysqlbinlog_row_compressed.result b/mysql-test/main/mysqlbinlog_row_compressed.result --- a/mysql-test/main/mysqlbinlog_row_compressed.result +++ b/mysql-test/main/mysqlbinlog_row_compressed.result @@ -40,9 +40,10 @@ SET @@session.sql_mode=#/*!*/; SET @@session.auto_increment_increment=1, @@session.auto_increment_offset=1/*!*/; /*!\C latin1 *//*!*/; SET @@session.character_set_client=latin1,@@session.collation_connection=8,@@session.collation_server=#/*!*/; +SET @@session.character_set_collations='utf8mb3=utf8mb3_uca1400_ai_ci,ucs2=ucs2_uca1400_ai_ci,utf8mb4=utf8mb4_uca1400_ai_ci,utf16=utf16_uca1400_ai_ci,utf32=utf32_uca1400_ai_ci'/*!*/;
why did this appear?
MTR now starts mariadbd with these lines in my.cnf:
character-set-server=utf8mb4 collation-server=utf8mb4_uca1400_ai_ci
When mariadbd processes the first line, it finds the default collation for utf8mb4, which adds the CHARACTER_SET_COLLATIONS_USED flag. This forces the server to write @@character_set_collations to the binary log.
However, when it processes the second line, the flag becomes not needed.
The code could be rewritten to avoid adding CHARACTER_SET_COLLATIONS_USED if both character-set-server and collation-server are specified. The flag is needed when only character-set-server is specified without collation-server.
Do you need to set character-set-server at all? I mean, isn't the second line alone enough?
Or may be both can be removed? utf8mb4 is the default now, isn't it?
Sorry, I misunderstand the reason for the result change when traced it in gdb for the first time. The real reason was that I added the "CHARACTER SET latin1" clause into three CREATE TABLE statements in this test. I just fixed CREATE TABLEs to use: CHARACTER SET latin1 COLLATE latin1_swedish_ci The "SET @@session.character_set_collations=..." command is now gone in the mysqlbinlog output.
diff --git a/mysql-test/suite/innodb/t/alter_primary_key.test b/mysql-test/suite/innodb/t/alter_primary_key.test --- a/mysql-test/suite/innodb/t/alter_primary_key.test +++ b/mysql-test/suite/innodb/t/alter_primary_key.test @@ -1,6 +1,9 @@ --source innodb_default_row_format.inc --source include/have_debug.inc --source include/have_debug_sync.inc +--disable_query_log +--source include/test_db_charset_latin1.inc +--enable_query_log
you don't disable query log in other tests, why here?
Removed the disabling.
in all tests, I presume
There were some more tests. Fixed all of them now.
--echo # --echo # MDEV-23244 ALTER TABLE…ADD PRIMARY KEY fails to flag diff --git a/mysql-test/suite/perfschema/t/short_option_1-master.opt b/mysql-test/suite/perfschema/t/short_option_1-master.opt --- a/mysql-test/suite/perfschema/t/short_option_1-master.opt +++ b/mysql-test/suite/perfschema/t/short_option_1-master.opt @@ -1 +1 @@ --a -Cutf8 -W1 +-a -Cutf8 -W1 --collation-server=utf8_general_ci
was it necessary?
Without it, the server fails to start with this output in mysql-test/var/log/mysqld.1.err:
COLLATION 'utf8mb4_uca1400_ai_ci' is not valid for CHARACTER SET 'utf8mb3'
Can you do -Cutf8mb4 instead?
It's short_option_1.test, better to use short option names there :)
Right. Done.
create trigger t1rbr before insert on t1 for each row set new.a=now(6); diff --git a/storage/connect/mysql-test/connect/r/endian.result b/storage/connect/mysql-test/connect/r/endian.result --- a/storage/connect/mysql-test/connect/r/endian.result +++ b/storage/connect/mysql-test/connect/r/endian.result @@ -10,7 +10,7 @@ birth DATE NOT NULL FIELD_FORMAT='L', id CHAR(5) NOT NULL FIELD_FORMAT='L2', salary DOUBLE(9,2) NOT NULL DEFAULT 0.00 FIELD_FORMAT='LF', dept INT(4) NOT NULL FIELD_FORMAT='L2' -) ENGINE=CONNECT TABLE_TYPE=BIN BLOCK_SIZE=5 FILE_NAME='Testbal.dat'; +) ENGINE=CONNECT CHARSET=latin1 TABLE_TYPE=BIN BLOCK_SIZE=5 FILE_NAME='Testbal.dat';
or any connect utf8mb4 tests whatsoever?
I had to add CHARSET=latin1 into many tests.
I just removed a few CHARSET=latin1 from dbf.test, but not all of them.
Unfortunately, connect did not work with utf8mb4 in many cases. We'll need to ask Oliver to check. But this should not be a showstopper for MDEV-19123, I think.
Of course, not. But please, create a bug report like "CONNECT doesn't work with utf8mb4", explain that commit for MDEV-19123 had to add CHARSET=latin1 to many tests because many table types (see commit) don't work with utf8mb4. And assign to Andrew.
MDEV-34410 CONNECT doesn't work well with utf8mb4
Regards, Sergei Chief Architect, MariaDB Server and security@mariadb.org
participants (2)
-
Alexander Barkov -
Sergei Golubchik