My head is exploding. Bad CVS! Bad CVS!

Well, okay, what's bugging me is methodology, not the application.

We need:

• code control
• test environment

Now, CVS does the first, and we already have the second. I'm not having any joy with how to combine them both in a safe and lovely manner.

Ideas from those with expertise?

Because just having folks FTP stuff up isn't working, with the stepping on, and vanilla CVS isn't allowing for testing.

If the problem is people overwriting changes on the test server, you could check a file into CVS that contains a list of people in line to use the test server. The top person on the list can do what they want with the server. When they're done, they take themselves off the list.

That, at least, will keep accidental stepping on.

I don't imagine two people can share the test server at the same time, at least not reliably.

I don't imagine two people can share the test server at the same time, at least not reliably.

There must be a way. I don't say that because I can even halfway think of one, but because someone must be doing, it, right? A test site per coder is way too many.

I'm hoping I can get a fake phoenix running on my Mac. That might help.

The other way would be to break up the Phoenix source into modules that we're convinced are independant. Then folks can test the modules simultaneously.

But otherwise, no, I don't think two people can be modifying and testing the same program at the same time on the same machine and not go insane.

Right now, apart from the pages you see in the URL, there are class files -- user, quotes, post, thread, search, e-mail.

In theory, many changes will be happening in those, and need not overlap.

In theory.

Bleargh.

Yeah, it sounds like reserving the test site will be required.

I'm guessing that since coding Phoenix will be a spare-time only activity for all of us, we won't tend to collide on the test server.

How CVS performance look?

I was futzing with it on Buffistas.org, but I'm stuck at authentication methods. And I don't think there's any advantage to doing it there, if Karl doesn't mind hosting -- at least not to wait on me to work stuff out.

One advantage would be being able to check out directly into the test area, as opposed to having to FTP stuff to the server.

Can you run pserver on the test server?

Hmm. Don't know. How does that work client-wise?

cvs pserver access lets clients have their own accounts and passwords for CVS without having a regular account on the box. All the CVS clients I've seen support it.

You'd still need to log into the test server to install the sources for testing, though, but it should be faster.

For example, you could create a branch tag, make your changes, check in the change on the branch, then check out that branch on the server for testing. We could even write a script to automate the checking out of the branch on the server for testing.