We do it in one huge tree. Currently it is about 1300 .java files. We have
one package for each ejb deployment, and keep the META-INF and orion files
in separate directories. Ant then combines these files with the packages
upon deployment.
We find this a simple solution, since we have lots of common packages and
interfaces which are used by many parts of the system. At the current
project size, it takes about five to ten seconds to scan the source tree
during compilation (with javac, on a PentiumIII/600).
CT
-----Original Message-----
From: Jeff Schnitzer [mailto:[EMAIL PROTECTED]]
Sent: 23. januar 2001 02:21
To: Orion-Interest
Subject: Compiling, packaging, and directory structure
Here's an open question: How do you have your source tree set up?
Do you have a separate source tree for each component, or do you compile
a "single big tree"?
I currently have a separate directory for each of my components, and
it's getting out of hand. I have a couple layers of RMI components, so
the interfaces are shared among several components, and the dependencies
are ugly.
I'm starting to think that what I really want is a single big java
source tree (in a package hierarchy, of course) which is compiled in one
big step, then each component build step would simply copy the
appropriate class files and deployment descriptors into the staging area
and package them up.
Am I asking for trouble by doing this? Or is this the natural way to
build large interdependent java projects?
Sorry if this is a silly question, but I've never had to deal with with
a dependency graph this incestuous before. My build process has always
looked like the way you would set up a C++ project, mostly out of habit.
I'd like to know if I'm pointed in the right direction before I wreck my
CVS tree :-)
Thanks,
Jeff Schnitzer
[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>
