[cvsnt] cvs update overwrites changes

[Bo Berglund]

Please do not reply privately to list issues, you are much more likely
to get help if you address the list than a single participant like me.

An example is this what you describe below about using edit etc.
We *never ever* use cvs edit or any locking or reserved checkouts or
the like in our development. We all work concurrently and we never
have any problems like you describe.

As far as I know the *only* action performed when you use cvs edit
on a file is that the readonly flag on the local file is reset so
it becomes writeable. This can equally well be performed using Windows
Explorer. The only other thing that happens is that anyone having
a watch on this file will be notified that you have done cvs edit on it.
But this only happens if you have scripted the notify action in the
CVSROOT/notify script file.

I have a hard time believing that doing an unedit on an edited file
with changes will magically return the original file contents into
it...

But then again, we never ever use cvs edit ourselves.

/Bo


-----Original Message-----
From: David Vo [mailto:vo.david at gmail.com]
Sent: Tuesday, March 01, 2005 8:59 PM
To: bo.berglund at telia.com
Subject: Re: [cvsnt] Re: cvs update overwrites changes


Thanks, Bo for the replies.

I'm looking into user error as this is very strange.

I have another question, however.

Do cvs edit commands provide the pivot point for changes? For example,
if you run a cvs checkout (read/write) and modify a file before
running a cvs edit. Then, remembering to do so, run the cvs edit after
the changes. If you do not re-save the file, commit won't know that it
has been modified and no response is given. And if you modify the file
further after the edit, then perform an unedit and choose to revert
the file, it's back to the same state when you initially performed the
cvs edit.

Thanks,
David

On Fri, 18 Feb 2005 23:42:06 +0100, Bo Berglund <bo.berglund at telia.com>
wrote:
> On Fri, 18 Feb 2005 13:36:48 -0800, David Vo <vo.david at gmail.com>
> wrote:
>
> >Thanks for your reply.
> >
> >They are having problems with VB text files. I understand the binary
> >limitation. The messages just show a <M filename> message. When the
> >developers open the updated file, they have the newest revision
> >checked out but all of their local changes will be lost. Are there any
> >reasons why this is happening?
> >
> >Thanks, D
> >
> >
> If they see a message with M <filename> that simply means that the
> file in their sandbox has been detected as locally modified. It means
> that they do have edits in the file that have not been committed yet.
>
> During the update, if the base revision they are working on is no
> longer the server HEAD revision, the changes on the server between the
> base revision and the HEAD revision will be mereged into the local
> file leading to a message similar to this:
> RCS file: <path>/<to>/<filename>,v
> retrieving revision 1.14
> retrieving revision 1.15
> Merging differences between 1.14 and 1.15 into <filename>
> M <filename>
> P <filename>
>
> So in this case you will see all of these messages and two lines (M
> and P) for the file. It all means that the server data have been
> applied to your local file *while retaining* the local edits as well.
> This renders the file still modified (the M signal) so it can be
> committed to the server.
>
> I don't think that I ever heard of local edits being lost like you
> describe on a text file. Please repeat the operation and cut the full
> message interchange from cvs and post it here for analysis.
>
> Also: Please make 100% sure that the developers have closed their
> development system editors or IDE:s so they are not referring to edits
> that they have in memory on their development systems instead of on
> file. This has happened before: File is edited in the IDE but not
> saved, then the cvs update is done and the IDE discovers that the file
> has changed on disk and loads it on top of the memory file, thus
> losing the edits!
>
>
> /Bo
> (Bo Berglund, developer in Sweden)
> _______________________________________________
> cvsnt mailing list
> cvsnt at cvsnt.org
> http://www.cvsnt.org/cgi-bin/mailman/listinfo/cvsnt
>