Code review comment for lp:~laurynas-biveinis/percona-server/merge-5.6.12
Revision history for this message
| Laurynas Biveinis (laurynas-biveinis) wrote | # |
© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
67d34a1
(Get the code!)
The tests still run on large ulimit -n machines, which both we and upstream have in our CI farms (and it is apparent that there is not a single small ulimit -n machine in both of them) and the tested functionality is mostly orthogonal to the ulimit value. Thus IMHO don't we don't lose much by skipping testing on small ulimit machines. While the defaults have some calculating algorithm behind them, which is possible to break and not know about it, since it would not be checked anywhere at all.
Maybe technically better would be splitting the default checks to their own testcases, that would be skipped on low-ulimit machines and the rest that is runnable everywhere? But then again I wouldn't see much return on the time spent here.