> > Also, this will only work if you run it from command-line (an ant task) >
That's fine. The purpose (for me) is to ensure the robustness of a test / code change. Once a seed is found which fails the test, we should be able to debug the test in eclipse using -Dtests.seed, at least for single-threaded tests. Apologies for the inconvenience it causes. > No apologies needed ! Your work is a blessing. This runner is really great. Shai On Wed, Dec 19, 2012 at 9:40 PM, Dawid Weiss <dawid.we...@cs.put.poznan.pl>wrote: > > BTW, I don't know if it should be like that or not, but I ran this test > with > > -Dtests.iters=1000 and printed at the end of each iteration > > Codec.getDefault(). > > For all iterations it printed the same Codec, but if I ran the test many > > times (no iteration), it printed different Codecs in each run. > > I thought that tests.iters should simulate the behavior of running the > test > > many times? And therefore pick a new seed + Codec (among other things) > for > > each iteration? > > Sorry, I missed this. What you're observing is correct -- the Codec is > picked at the class level (in TestRuleSetupAndRestoreClassEnv.java). > -Dtests.iters duplicates each test much like if you put two or more > test methods in a suite class. Any static rules and static hooks > execute *once* for a class, regardless of the number of individual > tests, so the codec is picked once, depending on the master seed (not > the per-method derivative seed). > > This would be of course solved if tests.dups could run each class with > a different seed. I wish I had more time to look into this; I kind of > know how to solve this I think but it'll require a significant effort > put into rewriting core parts of the runner. Also, this will only work > if you run it from command-line (an ant task); there is no way to > duplicate an entire test suite in Eclipse or other IDEs. > > I'll try to look into this, be patient. Apologies for the > inconvenience it causes. > > Dawid > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > For additional commands, e-mail: dev-h...@lucene.apache.org > >
