Re: What is the true use ?

[Michael Sims]

At 10:07 AM 11/22/2001 +0000, Michael L. Hostbaek wrote:
> Hey,
>
> A buddy of mine, and I are having a somewhat loud discussion, as to how
> we are using cvs on our server.
>
> We want to put our www docroot in cvs.
[...]
> How is this done in the real world ? I think, my way of doing is the
> correct way, but I do still have some issues, that are hard to explain..
> fx...
> When comitting, you have to export and copy every time - which can be a
> pain..
> And if you are working on a local copy in your homedir, how do you test
> to see if it works, before comitting ? (Let's say you setup and test
> env, but some dynamic pages depends on other things with absolute paths.
> ??)

Ok, I'm far from the expert on CVS, having only been using it for a few 
weeks, but I'm going to take a crack at this one. :)

I am a beginning PHP/MySQL developer.  I have a production site that has 
about 4-5 web "applications" running on it.  I have an inventory/asset 
management system, a documentation archive, and a few other things.  I used 
to do all of my development in the production area, which was rather silly, 
I know, but workable at first because not many people actually used my 
sites for anything.  Once they started becoming more popular I realized 
that something was going to have to change and that I couldn't confine my 
development time to 10 PM or later to make sure no one was accessing my site.

By this time I had already heard of CVS so I decided to investigate it.  I 
spent some time studying and reading the manual, and this is the setup I 
ended up with:

I actually have TWO development servers that I work off of.  At work, I 
have the main production server and one development server.  At home, on my 
home network, I have another development server (of course, one of the 
great advantages of Linux/PHP/MySQL is that there are no licensing issues 
to worry about so you can have as many development servers as you like).

I created a CVS repository on my production machine (just because this 
machine is backed up to tape every night...I could have easily put the CVS 
repository on any computer I wished) and checked in my 4-5 different web 
apps as separate modules.  I backed up my current ./htdocs directory, then 
dumped it, then did a CVS checkout from the repository directly into the 
production ./htdocs (or wwwroot, as the case may be).  Next I held my 
breath and tested everything to make sure it still worked, and it did.

Next, I set up my two development servers to mimic the configuration of the 
production server as closely as possible.  I created development branches 
for each of my modules (with a DEVEL tag) and checked out the DEVEL 
branches into my development servers.  Of course I also had to export the 
MySQL databases from my production machine and import them into the 
development servers.

I do not version control the databases, BUT, if I ever need to make a 
change to the database structure (add tables, drop tables, add keys, etc.) 
then I go ahead and make the change to development, then create a SQL patch 
file that will perform these same changes.  I put that SQL patch file into 
my current working directory and "cvs add" it.  As work continues and I 
make more changes, I update this patch file and it is included on each 
commit back to the repository.

As I work on the development server at work, I'm making changes to the 
checked out copy of the DEVEL branch.  When I'm finished, I commit my 
changes back into the repository.  Now, when I get home if I decide I'm not 
sick of working yet, I'll go to my home development server and "cvs update" 
to get my latest changes.

When I'm ready to merge these changes back into production, I will merge 
the DEVEL changes back into the main trunk, then go to my production server 
and "cvs update" to get the new changes.  Next I will have to run the SQL 
patches to bring the databases up to date (part of the "build" process) and 
then theoretically if everything goes well, I should be good. :)

But, to answer your question, your method is much preferred over what your 
friend is suggesting, IMHO.  Now, the best approach IMHO is to have a 
completely separate machine so you can mimic your production environment, 
absolute paths and all.  If you can't afford to do that, you will have a 
little extra work on your hands.  I'm sure I don't have to tell you this, 
but absolute hard coded path's should be avoided wherever possible because 
of issues like these.  For example, in PHP I always use the built in server 
vars such as $HTTP_SERVER_VARS["document_root"] and relative paths on all 
my links and references, that way I can move things around fairly easily 
without it breaking.  Of course, sometimes this is easier said than done.

Maybe someone else will have some other ideas, but it seems to me that you 
have two options:

1. Setup a seperate devel server and mimic the configuration of your 
production server.
2. Edit/update your code to remove references to absolute paths wherever 
possible.

Here's a good link I found explaining how to use CVS in a web development 
environment.  The author uses Oracle as the database server, but the 
article is general and easily applies to almost any development environment:

http://www.arsdigita.com/asj/cvs

Hope this was of some help. I know some of this might not have been very 
clear, so if you have questions, feel free to ask. :)