[ https://issues.apache.org/jira/browse/HADOOP-15465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16483700#comment-16483700 ]
Steve Loughran commented on HADOOP-15465: ----------------------------------------- bq. My only concerns are about FileAlreadyExists and UnsupportedOperation. We should catch and return 1. this is an interesting problem. We get it with mkdirs and delete in the FS APIs: nobody knows what "failure" means. In delete(), false => dir isn't there, but for mkdirs it means "something is there...without any check for the destination being a file or not". If you look at how code uses this often they don't check results (mkdirs) or overreact delete) I think throwing a FileAlreadyExists exception removes ambiguity: callers can choose their own policy. We can then update those three uses to do the right thing, which, for DefaultContainerExecutor , may mean: throw something where today it doesn't > Deprecate WinUtils#Symlinks by using native java code > ----------------------------------------------------- > > Key: HADOOP-15465 > URL: https://issues.apache.org/jira/browse/HADOOP-15465 > Project: Hadoop Common > Issue Type: Sub-task > Reporter: Íñigo Goiri > Assignee: Giovanni Matteo Fumarola > Priority: Major > Attachments: HADOOP-15465.v0.patch, HADOOP-15465.v0.proto.patch, > HADOOP-15465.v1.patch, HADOOP-15465.v2.patch, HADOOP-15465.v3.patch > > > Hadoop uses the shell to create symbolic links. Now that Hadoop relies on > Java 7+, we can deprecate all the shell code and rely on the Java APIs. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org