Re: Fix for checking out inside of repository with nfs and automo unt

[Donald Sharp]

On Thu, May 24, 2001 at 08:25:00AM -0500, Cameron, Steve wrote:
> > On Wed, May 23, 2001 at 05:11:11PM -0500, Cameron, Steve wrote:
>       I could do something dumb like this:
> 
>       [me@somewhere ] $ rsh cvsserver
>       [me@cvsserver me] $ echo $CVSROOT
>       :pserver:me@cvsserver:/cvsroot
>       [me@cvsserver me[ $ cd /cvsroot
>       [me@cvsserver cvsroot] $  # At this point, maybe I went out to
> lunch...
>       [me@cvsserver cvsroot] $ cvs co whatever  # I forgot what preceded,
> thought I was in a different window, etc....
> 
>       I'm not sure what CVS did before in this situation, so your patch
> may not make this any worse than it was before, and no doubt improves some
> situations.  All I'm saying is, "look at this, is this a hole somebody could
> step into?"  It may be the case that the assumption is deemed "safe
> enough"...and if you shoot yourself in the foot, you will be hurting for
> awhile, so "don't do that".
> >  

Original behaviour would have refused to allow you to checkout in
this scenario.

Sure that's a scenario.  But the second that cvs goes into remote mode
we have no way of knowing how things are mounted.

Suppose we have this scenario

machine_a - Holds the repository in /usr/cvs

machine_b - I do work in /usr/cvs.

With the current setup If i attempt to use remote cvs in this mode
CVSROOT=:pserver:sharpd@machine_a:/usr/cvs

on machine_b:

cd /usr/cvs
cvs co <foo>

I would get a refusal to work checkout inside of repository message.

This is equally likely to happen as well.

The reason I changed the code( and listed the assumption ) is that
it is neigh unto impossible for cvs to guess correctly, when it 
is used in a remote manner, what the correct thing to do is.

Therefore it shouldn't guess...

donald