Community
Participate
Working Groups
After installing a patch/efix, attempting to start against a different previously existing workspace shows orphaned features with warnings after applying pending changes. Consider this very simple scenario: Feature 'master' includes Feature 'ftrA'. Both are at version 1.0.0. An e-fix called 'fix1' is created to patch ftrA. It has the following feature.xml: <?xml version="1.0" encoding="UTF-8"?> <feature id="fix1" image="wsFeatUpdate.jpg" label="%featureName" provider-name="abc" version="1.0.0" colocation-affinity="RootFeature"> <description>fix1 will install FeatureA version 1.0.0.1</description> <license url="%licenseURL">This is for test purposes</license> <includes id="ftrA" match="equivalent" version="1.0.0.1"/> <requires> <import feature="master" version="1.0.0" patch="true"/> </requires> <plugin id="fix1" version="1.0.0"/> </feature> To reproduce the problem: 1) Install Eclipse/Workbench 2.0.3 2) Add master and ftrA along with their attribution plugins 3) Start the workbench againt a fresh workspace eclipse -data ws1 4) Exit and start again against a 2nd fresh workspace eclipse -data ws2 5) Install the "fix1" efix while attached to ws2, and restart when prompted. Notice that it installed cleanly. No warnings in your update manager congifuation. It shows the fix1 feature and master as the top level features, as expected. 6) Exit and start against the first workspace again: eclipse -data ws1 7) It detects the pending changes. Say Yes to apply them, and allow it to restart when prompted. The congifuration now shows an orphaned ftrA at version 1.0.0 with a warning. The warning complains that ftrA version 1.0.0 expected to find plugin 'ftrA' version 1.0.0, but version 1.0.0.1 was found instead. Our biggest concern is that this state could leave the workspace in a state where future fixes couldn't be applied safely.
Action Taken: Reproduced with Eclipse 203 Reproduced regression test: Install Root 1.0.0, start, install e334 natively, start, A Feature is orphaned Reproduced regression test on the same machine we ran our tests from. If after the test you laydown Root 1.0.0, the orphaned feature disappears. Action Plan: Debug
Action Taken: Not a regression. It didn't work in 2.0.2 nor in any of the 2.0.3 build. When processing a delta, we do not diable patched features. Action Plan: Create fix
Action Taken: Implemented fix Tested: 1) start workspace, shutdown, install e334 natively, start -> no warning 2) start workspace, shutdown, install all efixes natively, start -> no warning 3) start workspace, shutdown, install all efixes natively + root 1.0.1, start -> no warning 4) start workspace, shutdown, install e334 natively + XYZ1.0.1, start -> no warning (efix 344 enabled, child is XYZ 1.0.1) 4bis) start workspace, shutdown, install e334 natively + XYZBogus1.0.1, start -> warning (XYZ feature e334 is enabled : see bug 32108) 5) start workspace, shutdown, install e452+e388 natively + XYZbogus 1.0.1, start -> no warning (e388 disabled) 6bis) start workspace, shutdown, install all efixes natively + XYZ 1.0.1, start -> no warning 6bis) start workspace, shutdown, install all efixes natively + XYZbogus 1.0.1, start -> warning (XYZ feature e335 is enabled : see bug 32108) Action Plan: Investigate 4bis and 6bis, Pass new core to cisco for test
Action Taken: released in 2.0.3 and 2.1. Fix issue for 4bis and 6bis in 2.1 Action Plan: Close
close
