CVS on Windows

May 4, 2004 3:26 PM

CVS on Windows help. I've made a complete mess of things trying to get a local copy of CVS running on Windows 2000. [more, in time perhaps much more, inside]

I've come to rely on the versioning control software (StarTeam) we use at work and would like to have a similar safety net here at home. Here is my idealized setup:

1. Install CVS on under-used Windows 2000 machine upstairs


2. Setup CVS to use the external hard drive (e:) for storage


3. Make CVS accessible on my internal home network


4. Provide limited CVS access from outside

I would settle for just point the first. All else is gravy.

posted by yerfatma to computers & internet (9 comments total)

What I've done so far: installed CVSNT and a front-end. Each works, but I don't think it's setup correctly. I followed the install tips, but that just got the thing up and running.
posted by yerfatma at 3:40 PM on May 4, 2004

I know its now what you're after but after going through the same experience as you I gave up and went with Subversion. I had a heck of time getting sspi (or pserver for that matter) working on my XP Home box. CVS support for windows seems like a bit of a hack, whereas the Subversion setup was very simple.
posted by trillion at 4:18 PM on May 4, 2004

I'm not committed to CVS. I had seen Subversion last time it was posted to /. -- I'll check it out. Thanks.
posted by yerfatma at 5:23 PM on May 4, 2004

The wimpy way out, I've taken: install tortoise cvs instead of wincvs as the shell (you may occasionally need wincvs, but tortoise is way nicer) and opt to create a local repository (i.e. on a local drive on the same machine). You can always mount the drive from the other machine to make it local to your development machine.

This will probably be slower than a real cvs client-server setup but not by much (AFAIK, cvs doesn't do deltas both ways, only one, and most home networks are plenty fast enough anyway).
posted by costas at 10:02 PM on May 4, 2004

i read this last night before going home and thought you were going to post more details. what is the problem?

in the past i've installed and used cvs on nt with no problems. i'd recommend installing just cvs until you know it's working, and then adding front ends later (but i'm a command line kind of guy ;o)

cvs for nt comes packaged as an windows installer fo some kind, iirc - you run it and choose the option "install as service" on the machine you want to run as server (you don't have to install as a service, but it makes life easier because it restarts when you reboot) (if we're talking about the same thing it has a fish as logo). one installed, it adds something to the directory where all the settings are, and you can click on that and alter various things (like where the repository is), iirc.

then, on the same machine, from the msdos command line, try getting a cvs client working. do cvs login or whatever. you may need to update some paths and environment variables - see the instructions.

once that works, go to a different machine and install cvs for nt again (you need the client part - i can't remember if there's an option to install just the client part, it doesn't hurt to install everything, i guess). then try using cvs from across the network (don't enable the cvs server on that client machine!). this is a bit tricky because you need to look at different verification methods. i can't remember if we used some kind of windows specific verification, or whether we used pserver. but again, this works following the instructions, as far as i remember.

once you have things working from the command line, fancy guis should work straight away once installed.

i'm on-shift now, but if you are stuck and can wait a week, i'll be back home next tuesday, where i have a windows machine that we used in this way and can look to see what the settings are for connecting across a network (it's turned off at the moment, so i can't get connected remotely).

i've used cvs on many projects over many years. it rocks. it's nasty to learn, and it's nothing fancy, but it does what it does (for small to medium scale projects) absolutely solidly - i've never had any problems that couldn't be traced back to user errors. i'd be very wary about committing my precious source code to newer software unless i had pressing needs that cvs can't solve (maybe it has to be used by a bunch of non-technical people and there's no nice enough interface, or you need a more non-localised system, or to merge corrections across many different branches non-sequentially).

if you're going to use it across the 'net, rather than a private network, i'd suggest going through ssh, but that's a whole other problem...
posted by andrew cooke at 6:23 AM on May 5, 2004

what is the problem?

I've never used CVS/ am completely ignorant of how to set it up. I did get the little fishy installed as a service, but I got stuck at the command line. I'll give it another shot tonight with a second machine as client (and stick with command line until I get things going). Thanks for the advice. I'm going to give CVS and Subversion both a shot and see which one is easier to get running.
posted by yerfatma at 8:10 AM on May 5, 2004

Just to clarify the problem a little further (reading back over the thread I realize I did a poor job of it): I setup CVS per the instructions and then added the WinCVS front-end, but I couldn't get any projects added to CVS. Every directory was listed as non-CVS and the one time I did get a folder added (with recursion turned on), it didn't add anything inside it.

Also, setting up the root directory for CVS seemed screwy. I created the folders suggested in the instructions (on my E drive though) and set one as the root, which added a folder under it called CVS_ROOT or something. But the root shows up as "\" in CVS and WinCVS can't seem to resolve it (due to something I've screwed up).
posted by yerfatma at 8:15 AM on May 5, 2004

bleagh. i don't think i can give you more specific advice because i don't have a windows machine to look at here. i guess i'm repeating myself, but i would clear things out, install again, and try to get "cvs login" and "cvs checkout CVSROOT" working (which might mean defining the CVSROOT environment variable and updating PATH - both in system properties). good luck. i don't know what "recursion turned on" means (i guess it's something in the GUI), but from the command line you don't need any fancy options for this to work - tweak the environment variables above, not the command line arguments.

once cvs login and cvs checkout work, look at the docs for "cvs import", which will get a pre-existing directory into cvs (you probably need "cvs commit" after import has worked correctly).

it's a horrible interface, so trying another system makes sense if you're not happy with command line stuff (what i tried to say above is that when it works, it works well, but it *is* painful, and that might be sufficient reason to try something else).
posted by andrew cooke at 9:33 AM on May 5, 2004

Subversion seemed to work well (I installed it last night but didn't have time to play with it). The only problem I had (and the reason I'm posting in case someone else has the same problem and doesn't get as Google-lucky as I did) was integrating Subversion with Apache. 'Twould appear Apache couldn't see the Berkley db .dll, so I just copied it from Subversion's /bin folder (which was in my system's %PATH%, so I'm not sure what the problem was) to Apache's modules directory. And that was that (after an hour or so).
posted by yerfatma at 10:13 AM on May 6, 2004