On Sep 24, 2007, at 1:40 PM, kate rhodes wrote:
On 9/24/07, Nigel Daley <[EMAIL PROTECTED]> wrote:
On Sep 22, 2007, at 4:35 PM, kate rhodes wrote:
https://issues.apache.org/jira/browse/HADOOP-1936
contrib tests -1. The patch failed contrib unit tests.
There are know stability issues with the contrib unit tests on the
Solaris build machine. Committers can make a judgment call when they
commit.
Sorry, I'm not following... do you mean that the thing that decides if
a unit test is present has stability issues or that the unit tests
that are typically found in contrib have stability issues?
There is no analysis of the patch done. The automated patch process
runs the test-contrib and test-core ant targets regardless of what is
in the patch. Some of the unit tests in the contrib directory (run
by the test-contrib target) are unstable (ie. give false negatives).
The patch WAS unit tests. Actually it was one unit test, one new
class, and one test for the new class.
Your test was never run. I suggest you move it to
src/test/org/apache/hadoop/examples
both because I think it makes more sense there
and it will get picked up by the existing
ant test target.
Ok, I can do that tonight, but I'm still in the dark as to what I
could have done to not get the penalty if that was the correct place
for it.
It sounds like you're taking the -1 comment too personally ;-) I
realize that http://wiki.apache.org/lucene-hadoop/HowToContribute
needs to be updated with some of the latest patch process details.
Also If it's not going to bother to test in there then it shouldn't
penalize patches for files in that/those directories.
Patches for the patch process are welcome:
http://svn.apache.org/viewvc/lucene/hadoop/nightly/
hudsonBuildHadoopPatch.sh?view=markup
Cheers,
Nige
