Bug #49526 Continuing issues with RQG's --seed option
Submitted: 8 Dec 2009 8:05 Modified: 9 Dec 2009 14:44
Reporter: Philip Stoev Email Updates:
Status: Closed Impact on me:
Category:Tools: Random Query Generator Severity:S1 (Critical)
Version:2 OS:Any
Assigned to: Bernt Marius Johnsen CPU Architecture:Any

[8 Dec 2009 8:05] Philip Stoev
We continue to experience issues with --seed, as follows:

* The actual seed in effect is not printed when --seed=time, so that such tests can not be replayed
* --seed=time does not appear to work, 1 is used instead
* if --seed=X is used, some PRNGs are still initialized with seed of 1 as evident when the PRNG constructor is made to print the seed for each new PRNG.

How to repeat:
Enable debugging printout in Random.pm new() and then run RQG with --seed=time

Suggested fix:
* Make sure --seed is handled properly, no matter which script is calling which script (gendata, gensql, etc.)
* Print the seed that is currently in effect for the user to see
* Add a test for --seed=time to the test suite that ensures that two runs with --seed=time do not generate the same queries.
[9 Dec 2009 14:44] Bernt Marius Johnsen
Fixed (revision 216)