[
https://issues.apache.org/jira/browse/MAHOUT-1302?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13751794#comment-13751794
]
Stevo Slavic commented on MAHOUT-1302:
--------------------------------------
I've just committed a change to fix the issue by making order of processing
mail archives and (sub)directories deterministic and non-OS specific - first
processing files then nested directories, just as expected by
{{SequenceFilesFromMailArchivesTest.testSequential}} unit test.
I don't like the design of {{SequenceFilesFromMailArchives}} - it's using
{{PrefixAdditionFilter}} which is a {{FileFilter}} to traverse the FS tree.
That felt unnatural before my change, and feels even more unnatural now after
the change.
> SequenceFilesFromMailArchivesTest.testSequential failing
> --------------------------------------------------------
>
> Key: MAHOUT-1302
> URL: https://issues.apache.org/jira/browse/MAHOUT-1302
> Project: Mahout
> Issue Type: Bug
> Components: Integration
> Affects Versions: 0.8
> Environment: ubuntu-3 and ubuntu-6 Apache Jenkins nodes
> Reporter: Stevo Slavic
> Assignee: Suneel Marthi
> Priority: Minor
> Labels: test
> Fix For: 0.9
>
>
> SequenceFilesFromMailArchivesTest.testSequential is failing only on ubuntu3
> and ubuntu6 Jenkins nodes. Because of that, MahoutQuality and integration job
> builds either fail or are successful depending on where they get run.
> Test fails because it expects entries in chunk-0 SequenceFile to be in
> specific order, but that order is not guaranteed because of the way the
> chunk-0 is created/filled - SequenceFilesFromMailArchives traverses input
> using Java's
> File[] java.io.File.listFiles(FileFilter filter)
> which does not guarantee order of files/directories.
> Unless we want in SequenceFileIterator to guarantee order by sorting, test
> needs to be changed to verify presence of given files and their content, but
> not their exact order.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
