[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cross-project-issues-dev] More on greedy install attribute ...

A clarification.

It's easy to forget that the Eclipse IDE isn't the only environment where OSGi bundles can be used and that the Eclipse release train isn't the only distribution channel. In many cases it makes a lot of sense to declare a wish to "install X if it is available and doesn't cause conflict" and that's actually what the optional attribute was designed for. I can understand why this in some cases could be considered bad practice in the limited scope the release train (after all, everything is indeed available so it will be installed), but it's often necessary to think beyond that scope when packaging general purpose bundles.

For bundles commonly installed in the IDE, an optional + greedy will cause no harm. As Laurent points out, they will be present anyway. In other environments this combination is very useful. I must confess that I don't understand why projects that exploit this feature are hunted down and blamed. It seems to me that everyone suspects that they don't know what they are doing. Obviously they do.

Just my 2c,
- thomas


On 05/29/2012 10:06 AM, Thomas Hallgren wrote:
On 05/29/2012 09:35 AM, Laurent Goubet wrote:
Hi,

Some bundles from Acceleo (org.eclipse.acceleo.*) appear in this list. For example :

  1. org.eclipse.core.filesystem
  2. Number of IUs using optional, but greedy for this case: 2

    1. org.eclipse.acceleo.common


Acceleo is developped as an Eclipse plugin, yet it is also meant to be useable in a standalone environment. When installed in Eclipse, we depend on org.eclipse.core.* bundles to provide additional functionality and integration. However, none of these dependencies are mandatory when using it as a standalone generation tool. So yes, we are indeed greedy (but in fact, it should be quite difficult to _not_ have o.e.c.filesystem installed before us right? org.eclipse.core.runtime and org.eclipse.core.resources should probably be ignored too by this report), _and_ optional since the dependency is not really needed in other environments.

I believe that these cases should be exceptions to the "greedy" report. How could we document/track these? Or is my understanding wrong here?

No, you're not wrong. This is exactly the kind of scenario that the greedy attribute was supposed to cover.

Kind Regards,
Thomas Hallgren



_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev