info-cvs
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Using branches for safely updating a website


From: Matthew Nuzum
Subject: Using branches for safely updating a website
Date: Mon, 1 Jul 2002 10:54:54 -0400

I've been using CVS successfully for some time now to manage a web-based
application.  I'm to the point where I think I understand (conceptually)
CVS.  However this means I know what I can do, but not exactly how to do
it or IF I should do it.

We now commit our changes using CVS and run a cgi script on the server
that updates the checked out copy in the website.  However our needs are
growing and I'm not sure what the best solution is.

Because our application is very complex in nature, we can't simply open
the pages in a web browser.  They need to be uploaded to a specially
configured webserver.  Because of this, all of our testing has been done
on a live system.  This is proving to be a problem as our user base is
growing.  It is now almost impossible to do any testing without
generating calls to customer support.

The answer is to install a test server and test there.  Here is my
proposed solution, please give me your feedback so that I know if I'm on
the right track.

I imagine all files will continue to reside in the same CVS server.
Developers  will have the application checked out.  When we make a
change, we have to note that this change is for our own testing
purposes.  We may make several commits before we get it right.  Once
everything is just the way we want it, we have to mark our changes as
"need testing".  Once testing has been completed, our changed files will
be marked as clean and good.  Only these clean and good files can be
updated on the main webserver.  Therefore there are three states to a
file, "clean", "need testing", and "working".

Changes will happen very frequently.  As a matter of fact, it's about
all we do during the day.  There are over 1,500 files (not counting
binary files) in the website spread over a very intricate folder
structure.  It's very common for a specific task to involve 1-5 files.
Sometimes this constitutes all of the files in a folder, other times it
doesn't.

Here's the idea I have so far.
====TRUNK==CLEAN==================+=====
  \----TESTING BRANCH------------/

The main trunk will be considered clean. Everything here should be
copied over to the webserver.

A branch will be created off of this called testing.  Once an item has
been tested and is known to work, it is merged back into the clean
trunk.  A test webserver will be set up to keep an entire copy of
testing branch available for browsing.

What I don't understand is exactly how this is going to work.  There is
always going to be a TESTING BRANCH that will for the most part, be
identical to the TRUNK.  All of the FAQ's I've read discuss the
difficulties that are involved with repeated merges from a branch and
etc.  This scares me a little as some of our developers are not very
technology savvy.  As a matter of fact, I don't think they can even
spell CVS.

An alternative that I haven't seen documented too much would be to do
the opposite as above.  The trunk exists as NEEDS TESTING.  Once a file
has been tested it is moved into a CLEAN branch.  The live server could
then just keep a checked out copy of the branch.

====TRUNK==TESTING==============+=====
  \----CLEAN BRANCH------------/


Using your experiences with CVS, can you recommend one of the two
solutions or suggest another?  Manageability is high on my list of
priorities.

Also would someone be willing to demonstrate the commands I would use to
perform some common tasks?  I've got some books and command references
that I can use to understand the commands better, but it helps to see
what order to do them in and what results to expect.

I greatly appreciate any help you can give and the fact that you're
taking time to read my e-mail.

Matthew Nuzum
www.bearfruit.org
address@hidden





reply via email to

[Prev in Thread] Current Thread [Next in Thread]