Re: [egit-dev] jgit and eclipse meta-data.

[Thomas Hallgren <thomas@xxxxxxx>]

Jason van Zyl wrote:
> It's a Maven project I don't think it's unreasonable to expect that.
>
>   
It's an Eclipse project at Eclipse.org. I don't think its unreasonable 
to request that it should build in an Eclipse workspace using Eclipse 
tooling.

> > I also have to install m2eclipse into our own automated build system.
> >     
>
> No you wouldn't. I can help with that but all you would need is Tycho there. Why would you need to install M2Eclipse, it's simply the UI aspect of the Maven integration. Tycho is all you need in an automated environment.
>
>   
I would appreciate your help very much. What we need is a way to call 
out to maven from our workflow.  Buckminster has 'actors' which are 
extendable (through eclipse extension points). A Maven actor that would 
enable us to generate a manifest from the POM would be a great 
enhancement and a very useful improvement to the Buckminster Maven 
provider. Would you be willing to provide such a bundle from Sonatype? 
At present, we cannot do that from Eclipse.org due to the IP restrictions.

> > I don't think a build system should be that intrusive and I actually don't think that m2eclipse/maven *is* that intrusive. That's why I'm advocating an alternative approach where the manifest can be checked in.
> >     
>
> In the native project there is no manifest. So you're asking the project to work in an artificial way.
But what harm will a manifest do? The bundle jar can be used as a common 
jar too right? If OSGi isn't present, the OSGi specific entries of the 
manifest will not matter. Or am I missing the point here?

Regards,
Thomas Hallgren