[cvsnt] cvs rename bugs

[David Somers]

Mark Levedahl wrote:
[snip]
> The bugs I noted were inadvertently discovered by a user of a repository
> I maintain, by using DOCUMENTED features of cvsnt. The manual certainly
> does not indicate this is a dangerous operation, much less that you risk
> corrupting a repository.

Running the command yields:
$ cvs rename fred.c fred.cpp
cvs rename: Warning: rename is still experimental and may not behave as you
would expect

RTFM yields:
http://www.cvsnt.org/manual/html/rename.html
"Rename is currently experimental."

http://march-hare.com/cvspro/faq/faq1.asp
March Hare Software are currently planning to deliver the following features
in CVSNT 2.6.01: 
[snip]
Rename / Move / Directory Rename 
[snip]

What could be more explicit than that?

Oh, from your OP, "Also, there is an annoyance: the documentation
indicates  after a move, cvs commit should be run from the parent
directory. I find cvs commit must be run from within each directory where a
rename action has occurred to actually do anything. When run from a parent
directory, cvs commit does nothing."

Again, RTFM: "Rename information is held at the directory level, so the
rename/move is not committed to the repository until cvs commit is called
on the directory containing the file." so don't be surprised that when
doing a move you need to commit in both the *source* and *destination*
directories.

> And, as I noted, the STABLE cvsnt release has bugs in this
> implementation that can corrupt a repository.
> I would think the last statement would cause pause: my belief is that
> repository corrupting bugs should be fixed. Apparently, that is a very
> strange viewpoint.

No really, since rename is *experimental*.

> As Charlie Brown would say, "Good grief."

Yes, but then again Snoopy would fly over Charlie, tell him not to play with
experimental commands, then fly off to fight the Red Barron.

-- 
David Somers
typographer/programmer/whatever