[MDEV-24457] main.sp-big result mismatch on windows Created: 2020-12-21  Updated: 2022-01-18

Status: Open
Project: MariaDB Server
Component/s: Tests
Affects Version/s: 10.5.8
Fix Version/s: 10.5

Type: Bug Priority: Major
Reporter: Otto Kekäläinen Assignee: VicenČ›iu Ciorbaru
Resolution: Unresolved Votes: 0
Labels: None

Attachments: PNG File screenshot-1.png    

 Description   

While working on https://github.com/MariaDB/server/pull/1717 I noticed the following test failure:

main.sp-big                              w5 [ fail ]
 
        Test ended at 2020-12-21 01:58:29
 
 
 
CURRENT_TEST: main.sp-big
 
--- /<<PKGBUILDDIR>>/mysql-test/main/sp-big.result	2019-05-21 13:27:19.000000000 +0000
 
+++ /<<PKGBUILDDIR>>/mysql-test/main/sp-big.reject	2020-12-21 01:58:26.820594746 +0000
 
@@ -84,6 +84,6 @@
 
 call select_test();
 
 select unix_timestamp() - @before < @time;
 
 unix_timestamp() - @before < @time
 
-1
 
+0
 
 drop procedure select_test;
 
 drop table t1;
 
 
 
mysqltest: Result content mismatch

Full log at https://launchpadlibrarian.net/512326996/buildlog_ubuntu-focal-riscv64.mariadb-10.5_1%3A10.5.0~ubuntu20.04.1~1608492466.1e061559eae+ok.10.5.riscv64_BUILDING.txt.gz

and amd64-windows
https://buildbot.mariadb.org/#/builders/234/builds/777/steps/7/logs/stdio

This should be fixed as it can cause a problem for protected branches

 Comments   
Comment by Oleksandr Byelkin [ 2021-04-21 ]

I changed title because we even do not have riscv64, which we can take care about is only windows

Comment by Vladislav Vaintroub [ 2021-04-23 ]

Last time it failed on classic buildbot, was in 10.2, in 2017

Comment by Vladislav Vaintroub [ 2021-04-23 ]

The new buildbot on Windows does not work

Comment by Oleksandr Byelkin [ 2021-04-23 ]

1) it was not failing before my experiments since 2017 on old buildbot
2) the new buildbot has misconfigured windows (12 threads on 2 cores), so time based can fail there
3) the new buildbot has no cross reference to really check there results

Comment by Vladislav Vaintroub [ 2021-04-23 ]

the test is not quite predicable , simplifying it should prove that

call select_test()

takes less than 1 minute (10 with valgrind), which, on an absoutely overloaded, or slow box, might not hold
In case of experimental foundation-run buildbot on Windows, this was overload due to misconfiguration . We have no information on lauchpadlibrarian and RISC, but perhaps, that thing was overloaded, or slow as well.

Generated at Thu Feb 08 09:30:08 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.