3 messages in com.perforce.perforce-user[p4] CVS to P4 Conversion - Label Pro...| From | Sent On | Attachments |
|---|---|---|
| Venters, Cheryl | 04 Mar 2005 09:33 | |
| Peter Stephenson | 04 Mar 2005 09:59 | |
| Matthew Janulewicz | 04 Mar 2005 11:15 |
| Subject: | [p4] CVS to P4 Conversion - Label Problem |
|---|---|
| From: | Matthew Janulewicz (Matt...@cardinal.com) |
| Date: | 03/04/2005 11:15:13 AM |
| List: | com.perforce.perforce-user |
I played around with this script at one time and have used it on small projects. My personal conclusion/opinion was that the way you are forced to use PVCS vs Perforce is so fundamentally different, that importing labels and branches from PVCS to Perforce was almost useless. The engineer that requested I import his project ended up scrapping most of it and just importing a baseline.
Not to get too philosophical on anyone, but we upgraded from PVCS to Perforce for a reason, and we decided early on that importing the code as-is was only going to replicate the same problems we had in a new tool (except faster ;) ) The arrangement of code in our PVCS databases was very much affected by limitations of the tool. We didn't want to impose those limitations on Perforce.
We opted to let projects migrate over when it became a good time to do so. Finish release -> import single baseline. It's taken a while and we haven't completely killed off PVCS yet, but our many Perforce depots are much prettier.
-Matt
On Fri, 04 Mar 2005 17:59:58 +0000, Peter Stephenson <pws at csr.com> wrote:
"Venters, Cheryl" wrote:
I'm converting many CVS repositories to P4 and am having trouble with the labels. In one of my imports, the files that are in a label do not even slightly match the files that are in that same label in CVS. Has anyone else had this problem and how did you fix it?
Yes, our problem was with branches; the models in CVS and Perforce are so different that the conversion script has problems. Therefore you end up with labels on huge numbers of branches unnecessarily. If you're getting files with the label that are unrelated even by a branch, it's a different problem.
What I did was hack the script to force the label only to take place on a likely looking branch. We force certain files (with version numbers) onto the main line to ensure consistent numbering, so I had to add a fix-up for files which were mostly in one branch but with a few files on the main line.
I can dig the hacked version of the script out if you're interested.