|
I imported the test innodb.alter_crash.
|
|
I imported the changes to innodb.innodb-index and innodb.innodb-index-debug.
|
|
commit a6a4c25bf7d943ff5a429fbf32fe04fdaca25b16
Author: Jan Lindström <jan.lindstrom@mariadb.com>
Date: Thu Oct 12 17:21:04 2017 +0300
MariaDB adjustments to innodb_gis tests.
Some test are still disabled as they provide different results,
fail on different error or crash.
Fixes also MDEV-14060 InnoDB: Foreign key constraint to
geometry type does not work where test intentionally tried
incorrect cases.
commit 42e3e574260dc6ccfe161f8c0d8e8af41e599353
Author: Jan Lindström <jan.lindstrom@mariadb.com>
Date: Thu Oct 12 16:21:21 2017 +0300
MDEV-13626: Merge InnoDB test cases from MySQL 5.7
Add missing innodb_gis tests.
|
|
Jan Lindström, please take a look on this test, what is the correct result her/
https://github.com/MariaDB/server/blob/mdev-13626/mysql-test/suite/innodb_gis/t/rtree_split.test
innodb_gis.rtree_split 'innodb' [ fail ]
|
|
@@ -80,6 +80,5 @@
|
set @g1 = ST_GeomFromText('Polygon((2 2,2 800,800 800,800 2,2 2))');
|
select count(*) from t1 where MBRWithin(t1.c2, @g1);
|
count(*)
|
-57344
|
+65536
|
drop table t1;
|
mysqltest: Result length mismatch
|
|
|
innodb_gis.gis_split_nan needed 4k size page on options, as --source include/have_innodb_4k.inc just checks if it is not 4k, then skip test. Now it fails with error 1416: Cannot get geometry object from data you send to the GEOMETRY field
https://github.com/MariaDB/server/blob/mdev-13626/mysql-test/suite/innodb_gis/t/gis_split_nan.test
innodb_gis.gis_split_nan 'innodb' w1 [ fail ]
|
Test ended at 2017-10-30 14:03:37
|
|
CURRENT_TEST: innodb_gis.gis_split_nan
|
mysqltest: At line 285: query 'INSERT INTO geom_index SELECT * FROM gis_geometrycollection' failed: 1416: Cannot get geometry object from data you send to the GEOMETRY field
|
|
The result from queries just before the failure was:
|
< snip >
|
'LINESTRING(0 0,10 10),'
|
'GEOMETRYCOLLECTION('
|
'GEOMETRYCOLLECTION())),'
|
'GEOMETRYCOLLECTION(),'
|
'GEOMETRYCOLLECTION('
|
'GEOMETRYCOLLECTION()),'
|
'POLYGON((0 0,0 10,10 10,10 0,0 0)),'
|
'MULTIPOINT(0 0,2 2,4 4,6 6,8 8,10 10),'
|
'MULTILINESTRING((0 0,10 10),(0 10,10 0)),'
|
'MULTIPOLYGON(((0 0,0 5,5 5,5 0,0 0)),((5 5,5 10,10 10,10 5,5 5))))'))));
|
# Creating a table with spatial index column
|
CREATE TABLE geom_index (fid INTEGER NOT NULL PRIMARY KEY, g GEOMETRY NOT NULL, SPATIAL INDEX(g));
|
# Inserting geometry values
|
INSERT INTO geom_index SELECT * FROM gis_point;
|
INSERT INTO geom_index SELECT * FROM gis_linestring;
|
INSERT INTO geom_index SELECT * FROM gis_polygon;
|
INSERT INTO geom_index SELECT * FROM gis_multi_point;
|
INSERT INTO geom_index SELECT * FROM gis_multi_linestring;
|
INSERT INTO geom_index SELECT * FROM gis_multi_polygon;
|
INSERT INTO geom_index SELECT * FROM gis_geometrycollection;
|
|
|
please check if result is correct/ test is here https://github.com/MariaDB/server/blob/mdev-13626/mysql-test/suite/innodb_fts/t/sync.test
CURRENT_TEST: innodb_fts.sync
|
--- /home/alice/aliska/10.2/mysql-test/suite/innodb_fts/r/sync.result 2017-11-02 14:13:42.975845201 +0100
|
+++ /home/alice/aliska/10.2/mysql-test/suite/innodb_fts/r/sync.reject 2017-11-02 14:14:01.639849689 +0100
|
@@ -133,9 +133,9 @@
|
SET GLOBAL innodb_ft_aux_table="test/t1";
|
SELECT * FROM INFORMATION_SCHEMA.INNODB_FT_INDEX_TABLE;
|
WORD FIRST_DOC_ID LAST_DOC_ID DOC_COUNT DOC_ID POSITION
|
-database 4 4 1 4 0
|
-good 5 5 1 5 0
|
-mysql 2 2 1 2 0
|
+database 2 2 1 2 0
|
+good 3 3 1 3 0
|
+mysql 1 1 1 1 0
|
SELECT * FROM INFORMATION_SCHEMA.INNODB_FT_INDEX_CACHE;
|
WORD FIRST_DOC_ID LAST_DOC_ID DOC_COUNT DOC_ID POSITION
|
SET GLOBAL innodb_ft_aux_table=default;
|
|
mysqltest: Result content mismatch
|
|
|
|
also problem in MATCH AGAINST IN BOOLEAN MODE --https://github.com/MariaDB/server/blob/mdev-13626/mysql-test/suite/innodb_fts/t/opt.test
CURRENT_TEST: innodb_fts.opt
|
mysqltest: At line 977: query 'select * from t1 where MATCH a,b AGAINST('"Now sUPPort"' IN BOOLEAN MODE)' succeeded - should have failed with errno 1191...
|
|
The result from queries just before the failure was:
|
< snip >
|
DROP TABLE t1;
|
#
|
# Bug#20261601 ASSERTION FAILED: !FIRST_QEP_TAB->TABLE()->NO_KEYREAD
|
#
|
CREATE TABLE t1(a INT PRIMARY KEY);
|
INSERT INTO t1 VALUES(1),(2);
|
SELECT (SELECT MATCH(`a`)AGAINST('1') FROM t1) FROM t1;
|
ERROR HY000: Can't find FULLTEXT index matching the column list
|
SELECT 1, a IN (SELECT a FROM t1) FROM t1;
|
1 a IN (SELECT a FROM t1)
|
1 1
|
1 1
|
DROP TABLE t1;
|
#
|
# Bug#20442572 ASSERTION `!FIRST_QEP_TAB->TABLE()->NO_KEYREAD' FAILED.
|
# Bug#75688 Assertion `!first_qep_tab->table()->no_keyread' failed.
|
#
|
CREATE TABLE t1(a INT,b POINT NOT NULL,KEY(a));
|
HANDLER t1 OPEN;
|
select * from t1 where MATCH a,b AGAINST('"Now sUPPort"' IN BOOLEAN MODE);
|
|
|
|
please see attached file force-restart.task, it lists tests, which had option --force-restart, that our mtr does not support.
I deleted it, but tests might fail because of it.
|
|
The file mysql-test/suite/innodb/t/cascade_lock_wait.test became useless in 5.7 after the needed DBUG_EXECUTE_IF was removed.
For any debug tests that we import from MySQL, we should check that the debug instrumentation is still available.
|
|
Resuming work on this. innodb.innodb_buffer_pool_dump_pct needed adjustment for MDEV-11454.
|
|
I will not import two tests for temporary tables:
- innodb_zip.wl6915_1 does not seem to add much value to MariaDB. In MariaDB, the limits for rollback segments were preserved (
MDEV-12289) for compatibility with older versions.
- innodb_zip.wl6470_2 is only testing basic DML functionality with temporary tables. We can get better test coverage with RQG runs.
Both tests expect ROW_FORMAT=COMPRESSED to be available for temporary tables. This is intentionally not supported by MariaDB due to the increased memory and processing overhead.
|
|
The test innodb.blob-crash needs some code changes, to replace DBUG_SUICIDE with DEBUG_SYNC and external killing.
|
|
The test innodb.iodku was added in a fix of Bug #25966845 INSERT ON DUPLICATE KEY GENERATE A DEADLOCK. As I had noted in MDEV-20850, MariaDB fixed the problem differently:
the original problem was partly fixed in MDEV-17073 (and later fully fixed in MDEV-17614).
The MySQL fix ensures that statement-level replication for INSERT…ON DUPLICATE KEY UPDATE works without causing an increased amount of deadlocks. The MariaDB fix will simply mark such statements unsafe for statement-level replication, so that they will be deterministically replicated using row-level events. The MySQL fix avoids deadlocks by registering next-key (gap) locks on a partial rollback, that is, by escalating to coarser locking. That is why the "Test 6" and "Test 7" sections in the test result in a lock wait timeout in MySQL 5.7 but not MariaDB.
While I was discussing this with vlad.lesin, I pointed out that by design, InnoDB allows a transaction to modify a record that was committed by a later-started transaction. I verified that a lock wait timeout would be reported by MariaDB if both transactions were using the SERIALIZABLE isolation level.
|
|
I would also add to the comment that MySQL fix sets implicit lock on just-inserted clustered index record during IODKU execution. The lock is converted to explicit one during partial rollback. Then the lock is inherited to a gap lock when just inserted record is deleted on partial rollback. Such a way the gap should be locked in clustered index, and, as any IODKU statement inserts a record to clustered index before secondary indexes duplicates check, any other transaction which executes IODKU will be blocked on "INSERT" stage.
I think it explains the reason of implicit to explicit lock conversion in the fix.
|
force_restart.task