On Fri, Mar 15, 2013 at 10:05 PM, Greg Martin <[email protected]> wrote:
> In all Unices I've used the behavior for unlink is to mark a file for
> deletion when no file descriptors reference it. However, in using a UNIX
> socket with nodejs I have found that calling unlink after the socket is
> created causes it to be deleted immediately. This means to delete the socket
> the process on exit handler must be used. Should it fail to get called the
> socket won't be deleted. IT seems to me this is a bug in fs.unlink's
> behavior.
>
> -Greg.

Your description of how unlink() works is incorrect.  You're confusing
directory entries with inodes.

UNIX sockets are file system entities.  When you unlink() a file
system entity, its directory entry is removed (subject to permission
checks, of course), making it inaccessible to subsequent file system
operations.

A directory entry points to an inode.  That inode has a reference
count that roughly corresponds with the number of open file
descriptors.  As long as the reference count > 0, the resources
associated with the inode are not reclaimed.

That's why you can keep reading from or writing to a file that's been
removed by another process, provided you had it open at the time of
deletion. But you cannot open it again - its directory entry is gone.

-- 
-- 
Job Board: http://jobs.nodejs.org/
Posting guidelines: 
https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines
You received this message because you are subscribed to the Google
Groups "nodejs" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/nodejs?hl=en?hl=en

--- 
You received this message because you are subscribed to the Google Groups 
"nodejs" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.


Reply via email to