On Saturday 21 May 2005 12:43 pm, Bernd Warken wrote, quoting me:
> >�You *don't* need this export of CVSROOT -- your previously checked out
> > copy of the repository already "knows" where its CVSROOT is. �You *may*
> > need the CVSROOT definition when you do the initial checkout, but even
> > then it's more convenient to use the "-d" option of cvs, IMHO. �When you
> > do an update, it's best to leave cvs to determine the CVSROOT from the
> > definition stored within the working copy itself.
>
> For me, the "knowing" of cvs does not work. �So it is necessary to add some
> more information.
Strange. On my box, I have a directory "~/sandbox/cvs", which is the parent
of all CVS working copies I currently work on. In this directory, I did
export CVS_RSH=ssh
cvs -d :ext:[EMAIL PROTECTED]:/cvsroot/groff -z5 co groff
creating "~/sandbox/cvs/groff", which becomes the "top" of my working copy.
(Actually, I have the "export CVS_RSH=ssh" in my ".profile", so that's set up
for me automatically, at login).
Now, when I come to start a new session, working on groff, and want to get my
working copy up to date, I do
cd ~/sandbox/cvs/groff
cvs -z5 up -dP
No need to have CVSROOT defined at all -- the important thing is to make the
top directory of the working copy the current directory, so cvs knows which
repository it has to work with. If this doesn't work for you, then your
setup would appear to be broken, in some way. Where does your counterpart of
my "~/sandbox/cvs/groff/CVS/Root" say its repository is?
>�If "cvs -d" is better please provide an update with this.
I don't say it's better, just more convenient, IMHO. The command format is
as I've indicated above. (Of course, if you find yourself needing to specify
the repository location with every cvs command you invoke, then setting
CVSROOT would be the more convenient option, but if you've set it up
correctly, you shouldn't need to do that).
> >�Also, if you work on more than one project, having an exported CVSROOT
> > could lead to strange behaviour, if you forget to redefine it, as you
> > switch between projects.
>
> If you write a shell script for each project the call of them does not
> disturb except when you call them by `. script.sh'.
But why bother with shell scripts at all? Isn't it much more convenient to
just "cd" to the appropriate working directory for each project, as you
commence work on it?
Best regards,
Keith.
_______________________________________________
Groff mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/groff
