hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jakob Homan (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HDFS-669) Add unit tests
Date Thu, 12 Nov 2009 22:36:39 GMT

    [ https://issues.apache.org/jira/browse/HDFS-669?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12777241#action_12777241

Jakob Homan commented on HDFS-669:

OK, latest patch looks fine once the script has been run (or the unit directory manually created).

Of note, I tried adding a unit test into the commit-test.txt and running ant run-commit-test,
but the unit test was not executed.  This should be fixed so that unit tests included in the
file are run, since hopefully that will be the source of most of them.

Also, another JIRA should be opened to change the directory from hdfs to functional, so that
we can start separating out the tests and not have a confusing directory name.

As part of test-patch it may be worth enforcing that all tests included in /unit don't run
more than a few seconds.  This will provide a bit of a stick for enforcing that we end up
with only real unit tests.

> Add unit tests 
> ---------------
>                 Key: HDFS-669
>                 URL: https://issues.apache.org/jira/browse/HDFS-669
>             Project: Hadoop HDFS
>          Issue Type: Test
>          Components: test
>    Affects Versions: 0.21.0, 0.22.0
>            Reporter: Eli Collins
>            Assignee: Konstantin Boudnik
>         Attachments: HDFS-669.patch, HDFS-669.patch, HDFS-669.patch, HDFS-669.patch,
HDFS-669.patch, HDFS-669.patch, HDFS-669.patch, HDFS-669.sh, HDFS669.patch
> Most HDFS tests are functional tests that test a feature end to end by running a mini
cluster. We should add more tests like TestReplication that attempt to stress individual classes
in isolation, ie by stubbing out dependencies without running a mini cluster. This allows
for more fine-grain testing and making tests run much more quickly because they avoid the
cost of cluster setup and teardown. If it makes sense to use another framework besides junit
we should standardize with MAPREDUCE-1050. 

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message