[
https://issues.apache.org/jira/browse/SOLR-8778?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15177978#comment-15177978
]
Uwe Schindler commented on SOLR-8778:
-------------------------------------
+1 to fix this!
Ideally we should remove our CSV fork from Solr's repository! I tried to do
that and cut over to the official release, but I gave up; see SOLR-3213. The
commons-csv release is a total rewrite not having any overlap with our early
fork in the internal package. The official release does not support most
"hackish" stuff, introduced to spare allocations (which is no longer an issue
in Java 8, we can do {{new CSV*()}} as often as we like, Hotspot optimizes that
away). I think this stuff also is the reason for this issue, [~steve_rowe]?
So we should really upgrade to a better CSV parser (official commons-csv) and
remove the "reuse" of instances in thread unsafe ways.
> Deprecate CSVStrategy's setters, and make its pre-configured strategies
> immutable
> ---------------------------------------------------------------------------------
>
> Key: SOLR-8778
> URL: https://issues.apache.org/jira/browse/SOLR-8778
> Project: Solr
> Issue Type: Bug
> Reporter: Steve Rowe
> Attachments: SOLR-8778.patch
>
>
> Removing some deprecated things in CSVStrategy (SOLR-8764) exposed a bug:
> it's possible to redefine the public static
> {{CSVStrategy.\{DEFAULT,EXCEL,TDF}_STRATEGY}} strategies, simply by calling
> their setters.
> Right now that's happening in {{CSVParserTest.testUnicodeEscape()}}, where
> the default unicode escape interpretation is changed from false to true. And
> then if that test happens to run before
> {{CSVStrategyTest.testSetCSVStrategy()}}, which tests that the unicode escape
> interpretation on the default strategy is set to false, then the latter will
> fail.
> Example failures:
> http://jenkins.thetaphi.de/job/Lucene-Solr-trunk-Linux/16079/ and
> http://jenkins.thetaphi.de/job/Lucene-Solr-trunk-MacOSX/3126/
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
