Jonathan Costers wrote:
Actually, not *all* tests, just all the categories I configured to run .. But nearly all tho.
Certainly a lot more tests than we had running before.
Categories that are still missing (not configured to run by Hudson for now): - javaspace: found quite some issues running this category (see previous report) - txnmanager: one of these tests hangs on me .. finding out why, also some other issues to work out - scalability: not sure about the purpose of these tests
Maybe I should take a look at that category. Most of the tests I've looked at so far use very small numbers of services etc. There is a need to test both for functional correctness and for performance under higher load. Maybe some of those tests are in the "scalability" category.
In general, I suspended work on tests that seemed to be missing while you are working on getting the existing tests working. As your work stabilizes, I want to go back and repeat some tests to see whether some dangerous situations are now being tested.
- end2end: one test needs Kerberos, for the others I get some other errors And of course there's the regression tests we have in the jtreg directory. I am experimenting to port these to either JUnit or QA tests, eventually removing the dependency on the jtreg tool.
I dropped trying to get the jtreg tests running in my preferred environment, MS WindowsXP+cygwin. If you can get rid of the jtreg dependency I won't have to pick it up again.
Any chance of integrating the QA source code tree into a single source code tree with the actual code? Under Eclipse, I've so far only been able to work with River by creating a pair of matched projects, one for src and the other for qa/src, and making the QA project depend on the main source project.
Patricia
