Bug #78975 SYSSCHEMA.PR_STATEMENT_PERFORMANCE_ANALYZER TEST still FAILING ON PB2 ON WINDOWS
Submitted: 27 Oct 2015 8:02 Modified: 13 May 2016 14:51
Reporter: Horst Hunger Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Server: Performance Schema Severity:S3 (Non-critical)
Version:8.0.0-m17 OS:Any
Assigned to: CPU Architecture:Any

[27 Oct 2015 8:02] Horst Hunger
Description:
The test running into a race condition, so that results are not always deterministic. 
A similar bug has been solved in this test in bug#21845866, but not for all occurring cases. 
More wait conditions are necessary to make the test reliable.
The log shows that the synchronization of the parallel processes stops before source code line 148, so that the later cases can still fail on slow machines:

At line 148: Can't initialize replace from 'replace_result $query_insert QUERY_INSERT $query_select QUERY_SELECT $query_update QUERY_UPDATE $digest_insert DIGEST_INSERT $digest_update DIGEST_UPDATE $digest_select DIGEST_SELECT $d_upd_total_latency LATENCY $o_upd_max_latency LATENCY $o_upd_first_seen FIRST_SEEN $o_upd_last_seen LAST

How to repeat:
The test fails regularly in daily-trunk on windows:

see e.g. http://pb2.no.oracle.com/?template=mysql_show_test_failure&search=yes&push_id=8615836&test...

Suggested fix:
Insert wait conditions.
[13 May 2016 14:51] Paul Dubois
Posted by developer:
 
Fixed in 5.7.14.

Work was done for test suite. No changelog entry needed.
[18 Jun 2016 21:28] Omer Barnir
Posted by developer:
 
Reported version value updated to reflect release name change from 5.8 to 8.0