Re: touch

[anon]

This function may not be that generally useful.  I write security software 
and it helps give the software hints if I can adjust the dates on the 
symbolic links.  That way I don't have 2 minute cron jobs running every 
hour.

If I wrote an ltouch would it be useful enough to include in the source or 
should I just drop it only sourceforge?  I guess I'd have to convince ppl 
to accept the kernel patches as well, but that's probably do-able.

Jason
address@hidden

On Fri, 1 Mar 2002, Bob Proulx wrote:

> 
> > This really isn't a bug, but it looks like there isn't a way to touch a 
> > symbolic link. 
> 
> That is generally true of BSD like systems which implement symbolic
> links.  Actions upon symlinks pass through the symlink and act upon
> the target of the symlink.
> 
> > If you do a 
> > 
> > ln -s foo.tgz foo
> > touch foo
> > 
> > it update the date on foo.tgz.
> 
> Not of which I am aware.  File times are changed by the utimes(2)
> system call.  I know of no lutimes(2) call.  Perhaps others will have
> different information.
> 
> In order to act upon the symlink special kernel routines need to be
> added such as lstat(2) [as opposed to stat(2)] which act upon the
> symlink itself.  Fortunately the ower, group, and mode of a symlink
> are completely irrelevant to a symlink.
> 
> For what purpose would one desire to change the time of a symink?
> 
> > Is there a work around?
> 
> The only way I know to do this on BSD like systems is to remove and
> recreate the symlink.
> 
> [I am hoping and assuming that you won't but if you actually did that
> I would create one off to the side as a temporarily named object and
> then 'mv' it into place since rename(2) is an atomic operation.  That
> way the symlink will always exist.  Otherwise there will be a time,
> although small, where the symlink will appear not to exist to other
> programs and a race condition will be created.  If you cared.]
> 
> Bob
>