[
https://issues.apache.org/jira/browse/HDFS-126?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jakob Homan updated HDFS-126:
-----------------------------
Attachment: COMMON-126.patch
Done with most of the work. I very much would like to use HADOOP-6105, but
it's not yet committed. I'll be out of town until next Wednesday, so attaching
prelim patch for review and hopefully 6105 will be ready to go when I get back.
Patch:
* Creates a new config option, dfs.umaskmode, which will take either an octal
or symbolic umask value.
* Separates out the permission parsing from the shell code. This allows us to
use the exact same code for parsing in both places.
* Does not change external view of any of the permission processing
* Creates a new PermissionParser and a package-private UmaskParser to handle
the subtle differences between what regular permissions and umask permissions
allow
* Creates new unit tests to verify we're getting the right value from the conf
file. Most of the parsing code testing is free, since we're now using common
code in all the parsing.
Still need to:
* Use 6105 to handle old key
* Test against hdfs
> Permission configuration files should use octal and symbolic
> ------------------------------------------------------------
>
> Key: HDFS-126
> URL: https://issues.apache.org/jira/browse/HDFS-126
> Project: Hadoop HDFS
> Issue Type: Bug
> Reporter: Allen Wittenauer
> Assignee: Jakob Homan
> Attachments: COMMON-126.patch
>
>
> Currently, the settings for the default umask in Hadoop configuration files
> require the input format be in decimal. Considering that every admin in the
> world thinks of permissions in octal and/or symbolic format, the config files
> should really use those two formats and drop decimal.
> [... and, yes, I'm aware this breaks backwards compatibility. But in this
> case, I think that is perfectly acceptable.]
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
