SVN Reorg & Automated Build

[Bill Dudney <bd...@mac.com>]

Hi All,

After the reorg is complete I was thinking that moving to something  
like CruiseControl for our automated builds would be good.

I propose;

1) Using CruiseControl for the automated builds
     a) continuous builds - every check-in would do a build and  
report to the dev list any error discovered
     b) historic reports from the JUnit and cactus tests
     c) automate the tagunit tests so we get consistent deployment on  
WLS 8.1.x
2) Using Cargo (http://cargo.codehaus.org) to start and stop the  
various app server instances (tomcat, orion etc)
     a) We could use this to make sure that all the 'supported' app  
servers actually work with the examples
3) Use Cannoo webtest to preform sanity tests on all app server  
instances
     a) along with 2 would provide a means to make sure that at least  
our examples will work with all the 'supported' web containers

I would be glad to host this continuous build process on dudney.net.

Thoughts?

TTFN,

-bd-

Re: SVN Reorg & Automated Build

[John Fallows <jo...@gmail.com>]

I have been recently been investigating Maven Continuum, which looks
very promising.

http://maven.apache.org/continuum/index.html

Although it is only currently in Technology Preview, I believe it
already has the features we would need (and then some!)

    * Support for Any, Maven, Maven2
    * An easy to use web interface
    * A very easy way to add new projects
    * A XML-RPC interface for automation and remoting
    * Mail notification

The Maven2 project is already using it for their nightly builds. 
Perhaps Maven Continuum will be production by the time we are ready to
evaluate the strategy for MyFaces nightly builds.

Kind Regards,
John Fallows.

Re: SVN Reorg & Automated Build

[Sean Schofield <se...@gmail.com>]

Bill,

We currently have an automated build that uses ant that runs nightly
on a unix cron.  I like the idea of more automated tests but those
could be run via a variety of mechanisms.  I looked into cruise
control a while ago and I seem to recall I wasn't too impressed.  We
use Anthill at work but we refer to that as "antpile" for what I think
are obvious reasons.

I've never heard of Cargo but that sounds interesting.  I've only
dabbled in Cactus and other web based tests so I'm not really an
expert on that.  I've always got tired of trying to get it to work
when I had code to write.

What does Maven have to offer in this area?  We've had offers from
James Mitchell to help us "mavenize" the project.

I'm willing to take a look at changes to the build/testing process for
sure.  I would say that its a very massive effort we are talking about
so it will take a lot of committment from several people.

Here is my thinking of the priorities related to the SVN reorg:

1.) get agreement on the jars (we basically have that)
2.) get agreement on the svn regor (close)
3.) test the reorg
4.) perform the reorg
5.) tweak ant scripts to build according to the new reorg
6.) get nightlies back up and running (using current approache for now)

Step 6 might be unecessary if replaced with cruise control or maven
but it will be relatively quick because I've already done most of the
work already.  Step 5 will be the real pain but that is necessary no
matter what we do.  IMO we need to have ant scripts for our users even
if we use soemthing like maven for the builds.

So I'm ok with looking at what happens beyond steps 1-6 but I feel
like steps 1-6 are needed first.  Then we can look at adding on test
cases, etc.

sean

ps. I will write up a MyFaces reorg checklist wiki (outlining these
steps) for us shortly


On 5/31/05, Bill Dudney <bd...@mac.com> wrote:
> Hi All,
> 
> After the reorg is complete I was thinking that moving to something
> like CruiseControl for our automated builds would be good.
> 
> I propose;
> 
> 1) Using CruiseControl for the automated builds
>      a) continuous builds - every check-in would do a build and
> report to the dev list any error discovered
>      b) historic reports from the JUnit and cactus tests
>      c) automate the tagunit tests so we get consistent deployment on
> WLS 8.1.x
> 2) Using Cargo (http://cargo.codehaus.org) to start and stop the
> various app server instances (tomcat, orion etc)
>      a) We could use this to make sure that all the 'supported' app
> servers actually work with the examples
> 3) Use Cannoo webtest to preform sanity tests on all app server
> instances
>      a) along with 2 would provide a means to make sure that at least
> our examples will work with all the 'supported' web containers
> 
> I would be glad to host this continuous build process on dudney.net.
> 
> Thoughts?
> 
> TTFN,
> 
> -bd-
>