Hi Sergei,

Here is the patch: http://bazaar.launchpad.net/~elenst/maria/5.5-engines-tests/revision/3605

No harm is done, it's only tests and results in suites which are not a part of the standard set. Is it okay to push into 5.5?

Hi, Elena!

The only questionable change is in the file
mysql-test/suite/storage_engine/autoinc_vars.result

monty, is that change ok?

Regards,
Sergei

For Monty's reference, I'll put the change and the context here.

=====================================
The change:

=== modified file 'mysql-test/suite/storage_engine/autoinc_vars.result'
— mysql-test/suite/storage_engine/autoinc_vars.result 2012-07-16 02:17:56 +0000
+++ mysql-test/suite/storage_engine/autoinc_vars.result 2012-12-28 13:02:33 +0000
@@ -44,12 +44,10 @@
SET auto_increment_offset = 300;
CREATE TABLE t1 (a TINYINT <CUSTOM_COL_OPTIONS> AUTO_INCREMENT, <CUSTOM_INDEX>(a)) ENGINE=<STORAGE_ENGINE> <CUSTOM_TABLE_OPTIONS>;
INSERT INTO t1 (a) VALUES (NULL);
-Warnings:
-Warning 1264 Out of range value for column 'a' at row 1
+ERROR 22003: Out of range value for column 'a' at row 1
SELECT LAST_INSERT_ID();
LAST_INSERT_ID()
-127
+850
SELECT * FROM t1;
a
-127
DROP TABLE t1;

=====================================
Context:

This part of the test does the following:

SET auto_increment_increment = 500;
SET auto_increment_offset = 300;
CREATE TABLE t1 (a TINYINT AUTO_INCREMENT PRIMARY KEY);
INSERT INTO t1 (a) VALUES (NULL);

Thus, the inserted value should have been 300, but TINYINT doesn't allow for that. So, earlier the value was truncated with the warning and the max possible value was inserted (and hence LAST_INSERT_ID was updated), while now it throws an error instead, nothing is inserted (and LAST_INSERT_ID remains old, 850 is the value from the previous part of the test)

Yes, this is ok. There is also a test case for this behavior change in t/auto_increment.test

Elena, please feel free to push

pushed to maria/5.5.
Not closing because 10.0 might require more changes, still need to check it.

Hi Sergei,

Here is the 10.0 part (on top of the previous change):
http://bazaar.launchpad.net/~elenst/maria/10.0-base-mdev3990/revision/3468

I listed sources of the differences, as much as I could identify them, in the commit comment.

Everything, but the insert_time.result is certainly ok.
changes in the insert_time.result look plausible too, but I didn't track every 00:00:00 change back to its origin, incorrectly formatted time.

What I did was take this test:

create table t1 (c1 time);
insert into t1 values ('00:00:00');
select * from t1 where c1 >='12colon34colon56';
c1
00:00:00
Warnings:
Warning 1292 Truncated incorrect time value: '12colon34colon56'

And ran it before and after MDEV-457 fix. Before it was returning the value, after it wasn't – so I figured it's related to that change. MySQL doesn't return a value either, so at least it's consistent.

My understanding is this:
'12colon34colon56' is truncated to '12' and considered to be '00:00:12';
so, everything less than that (up to and including '00:00:11' in that test) is filtered out, and the rest is included in the result set.

ok to push in 10.0-base

pushed into 10.0-base