[ 
https://issues.apache.org/jira/browse/LUCENE-2980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13009869#comment-13009869
 ] 

Doron Cohen commented on LUCENE-2980:
-------------------------------------

Thanks Shai!

I fixed the super class and the assert as suggested.

For those nocommits, they stand for a larger problem - I was ready for a 
trivial fix for this bug - just lower case the extension in ContentSource 
before consulting with the map. However the test failed, and I found out that 
this is because the input stream returned by 
CompressorStreamFactory.createCompressorInputStream() does not close its 
underlying stream when it is exhausted or when its close method is called. 

I opened COMPRESS-127 for this.

As a workaround to this bug, ContentSource now returns a wrapper on the input 
stream created by the CsFactory, delegates all methods to it, except for 
close() which is also delegated to the underlying stream. This fix is required 
for the extension letter cases tests to pass, but it fixes a more serious 
problem, - leaking file handles in ContentSource.

As Solr also makes use of CommonCompress I searched in it for references to 
CompressorStreamFactory.createCompressorInputStream() but found none, so it 
seems Solr is not affected by COMPRESS-127.

> Benchmark's ContentSource should not rely on file suffixes to be lower cased 
> when detecting file type (gzip/bzip2/text)
> -----------------------------------------------------------------------------------------------------------------------
>
>                 Key: LUCENE-2980
>                 URL: https://issues.apache.org/jira/browse/LUCENE-2980
>             Project: Lucene - Java
>          Issue Type: Bug
>          Components: contrib/benchmark
>            Reporter: Doron Cohen
>            Assignee: Doron Cohen
>            Priority: Minor
>             Fix For: 3.2, 4.0
>
>         Attachments: LUCENE-2980.patch
>
>
> file.gz is correctly handled as gzip, but file.GZ handled as text which is 
> wrong.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

Reply via email to