Equinox Minutes - 20080205
Revision as of 16:09, 5 February 2008 by Dj houghton.ca.ibm.com (Talk | contribs)
Contents
Call-in Information
- Tuesdays from 2:05pm to 3:05pm (Eastern)
- 613.287.8000 or 866.362.7064, passcode 892048#
Attendees
- John Arthorne
- Oleg Besedin
- Matt Flaherty
- DJ Houghton
- Simon Kaegi
- Tim Mok
- Andrew Niefer
- Pascal Rapicault
- Tom Watson
- Mike Wilson
Discussion Topics
Bug about adaptor manager
- get adaptor doesn't return anything if the bundle hasn't been activated yet
- one recommendation is to have a flag for people who provide factories to force activation
- if true, then getAdaptor will activate. (behave the same as loadAdaptor
- bug number is 82973
- similar to one in the extension registry about activation
Substitutable Exports
- some people recommend exporting and importing all packages
- Apache is doing it and it is listed as a best practice
- if you specify both, then the framework chooses you as only one... either importer or exporter
- breaks required bundle clients
- tom has opened a bug against OSGi
- equinox bug https://bugs.eclipse.org/bugs/show_bug.cgi?id=217724
Install Location
- Install area has always been where we find Eclipse, not where the framework is
- used to default to where the exe is
- assumptions were made about the shape of things on disk
- if launcher jar is in a folder called plug-ins, then install area is the parents of the plug-ins folder
- this works ok if you are self-contained
- p2 allows users to have a bundle pool which is separate from your install
- path magic doesn't work anymore
- people have made assumptions
- we need a new property which describes where the executable is located
- how useful is the osgi.install property?
- should we deprecate this and create a new one which represents the location of the launcher?
- Tom to work with Pascal on a proposal to describe these locations
- 2 problem areas are: osgi.install.area and Platform#getInstallLocation