Maven NBM plugin CCE best practice

[Martin Scholl]

Hi all,


I'm an employee of a small company that develops applications on the NB RCP, amongst others. Right now I'm moving one of our RCP applications to maven and experience the well-known CCE exception at runtime. I'm aware of the problem because I read the notice on the NBM plugin site as well as the link NB prints on the console about arbitrary class loading. What I did not get is how do I address that issue. I'll explain my situation.


I've got:


A - a normal netbeans independent jar-packaged maven project
B - a netbeans module (nbm-packaged) maven project, that depends on A
C - a netbeans module (nbm-packaged) maven project, that depends on B and also (transitively) depends on A


So C uses classes of both A and B. NBM plugin tells me that I'll have to explicitly depend on A if I want to use its classes at runtime. Understood. But now I'm running into CCE at runtime since classes of A are loaded by B and C. That's clear to me, too. But I do not understand how to declare my dependencies and not run into CCE. Anyone with suggestions/best practices?


Thanks in advance.


Cheers,


Martin



-----------------------------------------------------
cismet GmbH
Martin Scholl, Dipl.Inf. (FH)
IT Park Saar - Geb

[Milos Kleint]

On Wed, Mar 10, 2010 at 9:31 AM, Martin Scholl <address-removed ([email]address-removed[/email])> wrote:

[Martin Scholl]

Hm, you're right, I forgot about the public packages. This seems to work for me. However, "B" offers a hell of a lot of packages now since it is a utilities module but I think that's ok.

Thanks for your help!


Cheers,


Martin



-----------------------------------------------------
cismet GmbH
Martin Scholl, Dipl.Inf. (FH)
IT Park Saar - Geb

[Milos Kleint]

if both B and C need the jar A, either B has to depend on C or vice versa or you create a simple wrapper module (D)  for A, that contains no code on it's own but only wraps around the jar and defines public packages. Then both B and C depend on D only.

Milos

On Wed, Mar 10, 2010 at 10:33 AM, Martin Scholl <address-removed ([email]address-removed[/email])> wrote: