Re: PATCH: backport of symlink fix for issue 142 to cvs 1.11.9.1

[Mark D. Baushke]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Paul Edwards <kerravon@nosppaam.w3.to> writes:

> "Derek Robert Price" <derek@ximbiot.com> wrote:
> > | If there is an expectation that CVS works with repositories
> > |
> > |that have symbolic links, and it doesn't currently actually
> > |work, then it should be fixed.  You can issue a travel advisory
> >
> > There was no previous expectation or documentation that CVS would work
> > with a symlinked root.
> 
> We're talking about an ordinary softlink in Unix, right?
> 
> I don't expect ANY application to fail just because I happen
> to softlink to the repository rather than hardlink.
> 
> BTW, I have softlinked to the repository many times, and it
> has worked every time.
> 
> I don't know which particular operation makes it fail when
> using symbolic links, but whoever added that feature, made
> it fail IMO, and is thus a bug that can have <xyz> effects.

The bug is documented in 

    http://ccvs.cvshome.org/issues/show_bug.cgi?id=142

In brief, the use of the CVSROOT/config LockDir=/some/directory option
has problems if the path in the $CVSROOT contains symbolic links.

Another issue is the interaction of LockDir=/some/directory with
symbolic links in the $CVSROOT/. tree pointing to various directories.
To be honest, I do not have such a thing in my repository, so I don't
know if it is a problem or not in either the feature or the stable
branches of cvs.

> > Cool as it might be that feature does work with
> > one now, I am against including the backport on 1.11.x for this reason.
> >
> > If you guys can raise enough relative clamor and few objections here and
> > on info-cvs, I will consider changing my vote.
> 
> There is no clamour calling for it to NOT being implemented,
> either here or info-cvs.
> 
> There is 2 votes for, one against.  The relative clamour is on
> the side of bugs being fixed in cvs 1.11.x, not throwing bug
> fixes away.  There was a time, only a few months ago, when
> there was only one version of CVS, and you could add new
> development willy-nilly and break anything you wanted.
> 
> I can only raise as much "clamour" as one person can raise,
> but throwing away bug fixes to achieve a goal (infinite
> stability, unlike cvs 1.11.7) that is impossible to achieve,
> in lieu of a goal (empty BUGS) that is possible to achieve,
> is not what I had in mind.
> 
> BTW, can we have this latest bug, dealing with symlinks,
> added to BUGS?

I'll try to get to that this week after I verify the state of the second
condition of symbolic links in the repository itself rather than just in
the path to the repository... after I figure out an easy to to test it. :-)

        -- Mark
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (FreeBSD)

iD8DBQE/nder3x41pRYZE/gRAr0zAKC8SIkztWkLs3joslddq0DxXtwsYgCgqk84
qhgd3pIWIihVoM3Wt5z0Gts=
=mB+e
-----END PGP SIGNATURE-----