[
https://issues.apache.org/jira/browse/HCATALOG-67?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13085930#comment-13085930
]
Alan Gates commented on HCATALOG-67:
------------------------------------
I don't see the generic harness code here anywhere. I also don't see any test
drivers. The conf files reference a TestDriverHCat.pm.
A lot of this is the pig specific portions of the code. This is fine, since
hcat will share a lot of that. But rather than checking that in as code,
would it be better to check it in as a tarball? Or at very least in a
separate directory.
HCatTypeCheck throws an exception if the schema does not meet its
expectations. This will cause the test to abort. You want it to
fail. Aborts mean the test failed to run, not that it returned wrong
results. I suspect the same applies to HCatTypeCheckHive.
Nice Javadoc throughout. It makes reviewing the patch much easier.
TypeCheckData claims to produce output to be checked against golden files.
We want to minimize the number of golden files we put in SVN, or at least keep
them quite small.
Is the purpose of bootstrap_hcat.conf to setup for future tests or is it
actually testing things? If it's setting up it should be implemented as a
deployer instead. The harness makes no guarantees about what order conf files
are run in, so we can't guarantee this will get run first. And a failure in
one of these "bootstrap" tests will cause a cascade of other failures.
> Utilities for system testing
> ----------------------------
>
> Key: HCATALOG-67
> URL: https://issues.apache.org/jira/browse/HCATALOG-67
> Project: HCatalog
> Issue Type: Test
> Reporter: Ashutosh Chauhan
> Assignee: Ashutosh Chauhan
> Fix For: 0.2
>
> Attachments: hcat-67.patch
>
>
> Various utilities (like hive udf, pig udf, map-reduce program) to do system
> testing for hcatalog.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
