RE: cvs-1.11.1p1 fails if CVSROOT is symbolic link

[Bill Biessman]

You have summed up a concern I have running CVSNT.  I have several
repositories set up for the main projects, documents, and other.  They are
segregated based upons size, who needs visibility, and life cycle of the
contents.

Other organizations have asked me to add different types of data
repositories that may become arbitrarily large and so at some point I will
be forced to move things around.

Since the developers do not have direct access to the files anyway, they do
not need to know that their work is under a directory on the server named
e:/cvs_roots/proj as is specified in their CVSROOT variable.  If I move
something I have to socialize the new directory name and get them to change
any scripts they have written.  I would prefer if the CVSROOT variable only
had a symbolic name for the repository.  The server could reference
configuration file that would tell it what physical directory the repository
is in.  The server already looks into the registry for a list of
repositories, so the registry could just as easily contain a symbolic name
as well.  The name could be treated as an alias so that the server would be
backward compatible.

bill

-----Original Message-----
From: bug-cvs-admin@gnu.org [mailto:bug-cvs-admin@gnu.org]On Behalf Of
Cameron, Steve
Sent: Tuesday, August 14, 2001 3:45 PM
To: bug-cvs@gnu.org
Subject: RE: cvs-1.11.1p1 fails if CVSROOT is symbolic link



I think a lot of system administrators will use symbolic
links as a matter of course for things like this (CVSROOT),
because they _know_ from experience that at some point,
_something_  will change (disk full/failed/obsolete, etc.)
that causes them to have to move the repository (or whatever),
and they don't want to break everything that users invariably
build which depend on locations.  That's why things like
symlinks and automounters were invented, of course.
(I'm preaching to the choir, no doubt.)

Might be nice if CVS could detect a symlink CVSROOT
and complain about it (and possibly refuse to continue?)
rather than getting an assertion failure only in some
specific cases.

Nicer still of course would be to make it tolerant of symlinked
CVSROOTs somehow, (though, I imagine if that were possible
it would have already been done)

-- steve

> -----Original Message-----
> From: Gianni Mariani [mailto:gianni@mariani.ws]
> Sent: Tuesday, August 14, 2001 2:20 PM
> To: Larry Jones; James Youngman
> Cc: bug-cvs@gnu.org
> Subject: RE: cvs-1.11.1p1 fails if CVSROOT is symbolic link
>
>
>
> oops, I have been using cvsroot as a symbolic link for months
> and I have no
> problems ... yet.
>
> What kind of problems should I be seeing ?
>
> My CMSROOT is /cvssrc/main/cvsroot and /cvssrc links to /d0/cvssrc
>
> Am I missing somthing ?
>
> G
>
> -----Original Message-----
> From: bug-cvs-admin@gnu.org [mailto:bug-cvs-admin@gnu.org]On Behalf Of
> Larry Jones
> Sent: Tuesday, August 14, 2001 12:19 PM
> To: James Youngman
> Cc: bug-cvs@gnu.org
> Subject: Re: cvs-1.11.1p1 fails if CVSROOT is symbolic link
>
>
> James Youngman writes:
> >
> > If you use LockDir and CVSROOT is a symbolic link,
> cvs-1.11.1p1 falls
> > over with an assertion failure :-
>
> If CVSROOT is a symbolic link, many things will fail in "interesting"
> and non-obvious ways.  Don't do that.
>
> -Larry Jones
>
> Any game without push-ups, hits, burns or noogies is a sissy
> game. -- Calvin
>
> _______________________________________________
> Bug-cvs mailing list
> Bug-cvs@gnu.org
> http://mail.gnu.org/mailman/listinfo/bug-cvs
>
>
> _______________________________________________
> Bug-cvs mailing list
> Bug-cvs@gnu.org
> http://mail.gnu.org/mailman/listinfo/bug-cvs
>

_______________________________________________
Bug-cvs mailing list
Bug-cvs@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-cvs