Bug #49167 No longer possible to specify that no validators should be used
Submitted: 27 Nov 2009 15:38 Modified: 5 Oct 2010 13:19
Reporter: John Embretsen Email Updates:
Status: Closed Impact on me:
Category:Tools: Random Query Generator Severity:S3 (Non-critical)
Version: OS:Any
Assigned to: Bernt Marius Johnsen CPU Architecture:Any

[27 Nov 2009 15:38] John Embretsen
Since the introduction of new option handling in the RQG (revno 160, 2009-11-24 10:05:52), RQG behavior has changed (regression).

In the PB2 test rqg_rpl_semisync, the runall.pl option --validator was set to nothing (--validator=) to make sure no validators would be used.

Now, this no longer works. Setting --validator= results in the default validators being used.

This affects this particular test, since one of the default validators may report test failure when the slave I/O thread is stopped, which is OK in this context. See http://forge.mysql.com/wiki/RandomQueryGeneratorTests#Semisynchronous_replication

How to repeat:
See description

Suggested fix:
We need a way to specify "use no validators" via runall.pl. One solution may be to introduce an empty validator called "None" or some such.
[27 Nov 2009 15:42] Philip Stoev
I think we need --no-default-validators or a similar option. The "--validators=" syntax was not very descriptive.
[30 Nov 2009 9:00] Bernt Marius Johnsen
What about --validators=None as a syntax for specifiying no validators. Alternatively, let the default be no validators?
[30 Nov 2009 12:53] Philip Stoev
--validators=None is fine with me.

However, I would like us to retain the option to have default validators and reporters, since there are a lot of useful bug-finding checks that we can perform on each query, and the user may not know how to enable all of them.
[1 Dec 2009 7:41] Bernt Marius Johnsen
[1 Dec 2009 7:41] Bernt Marius Johnsen
Both --validators= and --validators='' should now work
[5 Oct 2010 12:35] Bernt Marius Johnsen
is --validators=None an ok solution, and the problem considered sovled?
[5 Oct 2010 13:19] John Embretsen
Yes, --validators=None seems to work fine for me as well. I'm closing the bug.