Hi everyone, I've added our randgen transaction log tests to Hudson. We run these tests for every push to staging and a param-build job is also available.
These tests do the following: 1) run various DML against the test server 2) create SQL from the contents of the transaction log 3) populate a validation server with the SQL 4) perform a comparison of drizzledump files from master and slave. We run several grammars, which vary in 'fuzz' levels - they range from very messy SQL to tighter, more valid queries. The SQL includes transactions, rollbacks, savepoints and individual queries. We also make use of the --seed=time option for the randgen - this means that we run a standard set of test + randomize data and queries for some smaller runs. Each run gives us a potentially new set of test queries, but the randgen is quite informative and we can always repeat a failing run if needed. So far, this approach has proven quite effective and we've found lots of bugs through the magic of random brute-force testing : ) If you happen to be working on that code and want to know how things will run, try pushing your branch through param-build - the tests are tuned to take ~20 minutes. Thanks, Patrick
_______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
