[ https://issues.apache.org/jira/browse/HIVE-13686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15270752#comment-15270752 ]
Mohit Sabharwal commented on HIVE-13686: ---------------------------------------- I think this will affect only the subset of tests that (a) get distributed to the same ptest batch as TestRecordReaderImpl (b) get run after TestRecordReaderImpl. My guess is that either this subset doesn't rely on hive-site.xml for some reason or is size zero (i.e. TestRecordReaderImpl runs last). > TestRecordReaderImpl is deleting target/tmp causing all the tests after it to > fail > ---------------------------------------------------------------------------------- > > Key: HIVE-13686 > URL: https://issues.apache.org/jira/browse/HIVE-13686 > Project: Hive > Issue Type: Bug > Reporter: Rajat Khandelwal > > The issue was introduced in HIVE-12159 > (https://github.com/apache/hive/blame/master/ql/src/test/org/apache/hadoop/hive/ql/io/orc/TestRecordReaderImpl.java). > This test deletes target/tmp directory. Because of this, the subsequent > tests don't get to read target/tmp/conf/hive-site.xml, which contains > test-specific configurations. Also, target/tmp has metastore db directory, > which also gets deleted causing subsequent tests that use metastore db to > fail too. > I'm surprised this issue wasn't caught in pre-commit builds. [~spena] I see > that even the latest pre-commit jobs aren't reporting any errors, while > building on local and running hive-exec test cases causes a bunch of tests to > fail. -- This message was sent by Atlassian JIRA (v6.3.4#6332)