We have a growing number of modules. Each one is a java package, and
each module may depend on another modules. We are using server
cvs-1.10.8 running on solaris, and wincvs 1.1b13 on clients (nt4.0).
I am looking for an automated way of handling dependencies, because if a
module is changed all users needs to update & recompile it, and keeping
track of changed modules is rather painful. I was thinking of doing the
following:
1) A script to "cvs -n update" every module -- this would generate a
"list of modified module names".
2) A script to query a module for locally modified files (in WinCVS
there is a macro that does this) -- this would generate a "list of
locally modified files".
3) If a module is modified in the repository and has not been locally
modified, then a cvs update should be done automatically. After the
update the module would be compiled by invoking make-like tools.
4) If a module is modified both locally and in the repository, then the
user should do a manual update, this to ensure no conflicts are resolved
without warning the user.
So I am thinking of this four-step process to be a single script or
application, invoked manually (or perhaps invoked periodically as some
kind of daemon).
This is a lot of work! I will have to do it if I have no better choice,
but I am sure a lot of people have/had trouble with this kind of
dependencies, what do you think of this particular solution? any other
ideas? Is there a cvs-way of checking module dependencies ?
Some details: User number = 5-10, Module number = 100+, Module size =
Small, about 50k each.
Thanks.
