--- On Tue, 9/16/08, Denys Vlasenko <[EMAIL PROTECTED]> wrote:
> From: Denys Vlasenko <[EMAIL PROTECTED]>
> Subject: Re: busybox cp vs gnu cp
> To: [email protected], [EMAIL PROTECTED]
> Date: Tuesday, September 16, 2008, 9:58 PM
> On Tuesday 16 September 2008 03:49, Ming-Ching Tiew wrote:
> >
> > Tried this with busybox :-
> >
> > # cp /etc/hosts /tmp/a
> > # ln -s /tmp/a /tmp/b
> > # cp /etc/passwd /tmp/b
> >
> > After this, /tmp/b which was a symbolic link has
> become a normal
> > text file which content is /etc/passwd. File /tmp/a
> > remaind unchanged.
> >
> > When I tried this with gnu 'cp' the end result
> > is /tmp/b remain as a symbolic link, and /tmp/a was
> modified
> > accordingly.
>
> This is intended. Coreutils cp behavior is dangerous.
> Just imagine that the user running this command wants this:
>
> "copy /etc/hosts to /tmp/a, creating /tmp/a if it
> doesn't exist,
> or overwrite if it exists. In other words, I don't care
> whether it exists or not, I simpy want to make an exact
> duplicate of /etc/hosts."
>
> which is what English word "copy" implies, right?
>
Dangerous or not it all depends on your intent.
I think the current busybox behavior is more dangerous,
because it wipes out the symbolic link and created
a separate copy of the file without the user
realizing it.
Coreutils cp existed first, and it's behaviour has
been accepted as defacto. Busybox cp comes into
the picture to "resemble" coreutils cp, and yet
there are differences in behaviour. That will break
their way of thinking or if they already have code,
that will break their code.
> If you really want "overwrite symlink
> destination" semantic,
> why won't you do "cat /etc/hosts >/tmp/b"?
> In such command, your intention would be much clearer.
>
It's all in the mind about what's considered
the accepted "intention" or "behaviour".
Cheers
_______________________________________________
busybox mailing list
[email protected]
http://busybox.net/cgi-bin/mailman/listinfo/busybox
