[cvsnt] Possible bug in outdating a removed (dead) revision

Kim Hansen kha at mita-teknik.com
Thu Jun 10 10:26:04 BST 2004


Hi

We are using

Client: Concurrent Versions System (CVSNT) 1.11.1.3 (Build 72) (client/server) on Win2K
Server: Concurrent Versions System (CVSNT) 2.0.9 (client/server)

Recently, a cvs remove operation was performed (in error) on the main branch, whereafter the new revison
was marked dead.

Since, this was clearly an error (and I knew nobody else had updated their sandbox, created branches, etc.) I decided right await
to outdate the latest revision of the affected files using cvs admin -o cmd.

I know that you could just resurrect the file by adding it again and committing, but in my experience this surrection process
causes big problems when merging in developement branches afterwards. Therefore I chose the cvs admin -o solution.

Apparently, the operation was succesfull. When looking at the revision graphs in WinCVS, I verified that the 'dead' revisions had gone,
such that the main branch had the 'resurrected' in head.

However, after doing the outdate operation, I have had some very strange behavior concerning these files:

1. When branch rtagging a module which contains the affected files, the affected files are *not* given a branch tag.
I can tag them manually thereafter using tag on the specific files.

2. When updating a module to a branch, the affected files are removed from the sandbox (as if they did not have the branch tag).
However, inspection in the graph view shows, that they do have the branch.

I suspect, that some the cvs admin -o cmd on a dead revision has left the corresponding RCS files in an inconsistent state.

Is this a known bug?

I guess it would be good help for debugging to get an example of an affected RCS file. Where can I upload one?

Is the problem fixed in newer versions of cvsnt?

Cheers,

Kim



More information about the cvsnt mailing list