[
https://issues.apache.org/jira/browse/HAWQ-1558?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16274275#comment-16274275
]
Kuien Liu commented on HAWQ-1558:
---------------------------------
Yes, but it is not a good behavior that we split one utility tool into two
parts and force user to use them in highly coupled way. 'feature-test' is
straightforward to run a seperate test, I like it, and I guess it is desigen
for it :)
> feature-test result is not environment independent
> --------------------------------------------------
>
> Key: HAWQ-1558
> URL: https://issues.apache.org/jira/browse/HAWQ-1558
> Project: Apache HAWQ
> Issue Type: Bug
> Components: Tests
> Reporter: Kuien Liu
> Assignee: Jiali Yao
>
> When I run ./feature-test with seperate test, for example,
> 1. {{./feature-test --gtest_filter=TestType.*}}
> Some test sets, e.g., 'date.sql', 'int8.sql', 'time.sql' etc will fail. BUT
> if I run it with
> 2. {{./parallel-run-feature-test.sh 4 ./feature-test
> --gtest_filter=TestType.*}}
> 3. Or, {{PGTZ="PST8PDT" PGDATESTYLE="Postgres, MDY" ./feature-test
> --gtest_filter=TestType.*}}
> At least the test 'date.sql' will pass! That means, feature-test result is
> not environment independent, some environment variables are missed to be set
> for some reason.
> I think we can borrow some idea from pg_regress.c to 'putenv()' those
> variables (LC_*, PG*), what do you think?
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
