Hi Martin, Yes, there is a cyclic dependency between the SLF4J API and impls which is frowned upon by the OSGi community. Proposals for better OSGi integration are most welcome subject to backward compatibility restrictions. (Given the wide use of SLF4J we have to be extra-careful not to break compatibility). Thus, modifying the manifest files should probably be OK while changes in package structure or in classes not OK. Have you looked at previous discussions of the subject? See for example: http://bugzilla.slf4j.org/show_bug.cgi?id=75 http://bugzilla.slf4j.org/show_bug.cgi?id=283 Thank you in advance for your suggestions, -- Ceki 65% of statistics are made up on the spot On 23.01.2013 12:07, Martin Ellis wrote:
Hi Ceki and others,
I thought you might be interested to know about "Metadata Advice": it's an outreach project by folk involved in the OSGi community, who are willing to offer advice to people trying to package their software so it behaves well in an OSGi environment.
http://blog.osgi.org/2013/01/get-help-adding-osgi-metadata-to-your.html
Since there's been talk on this list of problems due to cyclic dependencies between slf4j-api and the logging implementations, I thought I'd point it out. Although it's called "Metadata Advice", I think they're willing to offer some suggestions for other aspects of using OSGi, beyond simply adding manifest headers (which slf4j already has).
Perhaps it's worth creating an issue on their bugzilla inviting suggestions for improving OSGi support in slf4j?
Regards, Martin