[ 
https://issues.apache.org/jira/browse/HADOOP-14843?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16157146#comment-16157146
 ] 

Jason Lowe commented on HADOOP-14843:
-------------------------------------

Thanks for the patch!

I believe the proposed change will break backwards compatibility.  For example, 
"+r" is a valid input that should return mode 0444 and "+rwx" should return 
mode 0777.  The patch makes those invalid.

After thinking about this more, I'm leaning towards closing this JIRA as 
Invalid.  "-rwr" unfortunately appears to be a valid symbolic argument even for 
POSIX {{chmod}}.  It means "remove read and write for all" because "-rwr" is 
semantically equivalent to "-rw" because it ignores duplicate flags.  The only 
thing I can think of to catch this case of misunderstanding by the user would 
be to treat duplicate flags as errors, but since {{chmod}} allows them I'm 
reluctant to diverge from that behavior.


> FsPermission symbolic parsing failed to detect invalid argument
> ---------------------------------------------------------------
>
>                 Key: HADOOP-14843
>                 URL: https://issues.apache.org/jira/browse/HADOOP-14843
>             Project: Hadoop Common
>          Issue Type: Bug
>          Components: fs
>    Affects Versions: 2.7.4, 2.8.1
>            Reporter: Jason Lowe
>            Assignee: Bharat Viswanadham
>         Attachments: HADOOP-14843.patch
>
>
> A user misunderstood the syntax format for the FsPermission symbolic 
> constructor and passed the argument "-rwr" instead of "u=rw,g=r".  In 2.7 and 
> earlier this was silently misinterpreted as mode 0777 and in 2.8 it oddly 
> became mode 0000.  In either case FsPermission should have flagged "-rwr" as 
> an invalid argument rather than silently misinterpreting it.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to