Re: SCA Java samples for M2

[Jim Marino]

On Oct 10, 2006, at 12:47 PM, ant elder wrote:
Also, as I think there has been some consensus on large parts of  
this how
about making some incremental changes? For example, I think there's
agreement that the vast majority of the existing top level  
'samples' folder
would move to 'sca/samples', and that the top level folder which  
contains
bigbank will be renamed from 'sampleapps' to 'samples'. So how  
about I go
ahead and make those changes now?
I think we should be clear on what the incremental changes are...Will  
the existing samples specific to extensions then be refactored into  
their respective extension projects, e.g. JavaScript samples under  
the JavaScript extension? Also, I'm not sure we are clear on what the  
move entails. Specifically, the samples should *not* be built as part  
of the standard checkin build. What I propose is:

1. There is a samples directory under root that contains a version of  
BigBank which spans multiple Tuscany sub-projects (SDO/DAS/SCA)

2. There is a samples directory under /sca for "baseline" samples.  
These baseline samples would comprise a samples distribution  
consisting of source. The source would be buildable by end-users from  
the distribution and we could add a variety of artifacts to assist in  
that, including a build.xml, .project (Eclipse), and .iml (IDEA)  
files. This distribution would not be built by the standard Tuscany  
check-in build. Since it is important for samples to *work* (not just  
build), there would be a series of integration tests run periodically  
somewhere. Samples would only be guaranteed to work on a stable  
snapshot of core.

3. An extension could also choose to include samples with it. These  
samples would also be distributed as source and would *not* be built  
as part of the main checkin process. The difference with #2 is that  
these samples would be non-baseline (e.g. "contrib" samples) and  
could release independently from the main samples distribution. In  
terms of the svn repo format, the sample artifacts could either go  
under a sibling folder of java/main (e.g. java/samples) or be  
contained within a Maven sub-module of the extension.

I think this would address issues we are having around modularity as  
well as provide a way for us to verify samples are working through  
integration testing (which cannot be done by seeing if they compile).

I'd like to have a plan agreed on before we make repo changes given  
we are trying to also get a release out.

Jim


I'll go ahead and do this unless I hear
any objections (but it would be nice to hear at least couple of  
okays).

  ...ant

On 10/10/06, ant elder <[EMAIL PROTECTED]> wrote:

Putting aside the when they get built question for now, what i was  
really
asking (in a very unclear way) a few emails ago was if I've really  
got the
directory structure quite right yet?

For example, there's the existing JavaScript container maven  
project at:

   sca/services/containers/container.javascript

I'd suggested samples would go at:

   sca/services/containers/container.javascript/src/samples/ 
calculator

But can maven projects really be nested like this? If there's a  
pom.xml in
that calculator folder does that work nested in the JavaScript  
project with
a pom.xml in the container.javascript folder? What references the  
sample
pom.xml and how would it ever get built? I've not tried this but  
it seems
odd, could someone who understands maven projects comment on if  
this ok?

   ...ant

On 10/9/06, Jim Marino <[EMAIL PROTECTED]> wrote:
>
>
> On Oct 9, 2006, at 8:37 AM, Simon Nash wrote:
>
> > I agree with Andy.  An integration test framework ro build and
> > run the samples is the right long term solution, but until we  
have
> > that framework, we should not remove building the samples from
> > the regular build.
> The samples are not part of the "regular" build, they are part  
of the
> root one. The root build should never have to be run for a  
checkin of
> Java SCA as it includes SDO and DAS, which we agreed should not be
> required by the former for modularity.
>
> I'm not going to rehash old ground but I will state that we need an
> integration test framework run as part of a continuous integration
> process that will include samples testing but goes well beyond  
that.
> In addition, the check-in build should contain test cases that  
verify
> basic things like bootstrapping (standalone, webapp); this can  
all be
> done using EasyMock without the need for code to execute in-
> container. The reason the samples are breaking is not because we
> don't run them as part of the build, it's because our basic test
> coverage is lacking.
>
> I propose we work to resolve this by the following:
>
> 1. Put testcases that verify bootstrapping in the checkin build.
> these should only use EasyMock and not run any containers (e.g.  
Jetty)
> 2. Move to a continuous integration framework, perhaps Continuum.
> These tests should
>
>         - Run the samples in a number of container hosts we support
> (Tomcat,
> Jetty, Standalone, etc.)
>         - Perform a deeper level of integration tests between  
various
> subsystems, such as wiring between component implementation types,
> interop testing, etc.
>         - Be run periodically. We may eventually want to break  
them into
>
> parts. Some which run every 15 minutes or so, others (more resource
> intensive) which run around twice a day
>
> I'm willing to help out if there are some volunteers. Any  
volunteers?
>
> Jim
>
>
>
>  
---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>



---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]