Community
Participate
Working Groups
Medic reads its configuration via ConfigAdmin in MedicActivator. The problem is that the configuration is published by kernel.core bundle that is started much later. The scenario is as follows: 1. medic's configuration listener is registered 2. medic tries to read configuration, but the configuration is not present 3. medic uses default configuration with only 3 entries 4. kernel.core publishes the medic configuration 5. configuration is updated via the registered in step 1 listener 6. the code used by the activator to init logging, system.out/err is not invoked
Only log.wrapSys* entries cannot be changed. This is due to the fact that they are read in MedicActivator and used to configure System.out/err stream. The rest of the medic configuration entries are simply queried when needed.
Commit 3f8af6254ee96790897b4f04844d7045d204a491 adds unregister method to ServiceRegistrationTracker. Tests are updated with commit 44863e15158509b22ad1f83c0fd01e0d4acbf21a Medic is refactored with 5090b48a3a6de4c2edee220677871a41d495b3c8 and d7945dc186bbf7b010f1c3865d59f7b7574297f5. Tests updated with commit 76d0d345b3cc28324920c0649da1d616d1dff307
Cherry picked in 3.0.2 with: * medic: 08e4cbf2bde2ecda2ddbac4376f816bdce4b1fa7 , eae42117b2d1f42773610e46eca5acf92828245b and 5dca4896669c3302db05a5bcd2943268ed975218 * util: b286d2b6206c615ea06b948a2bc55f7bccaa3cc4 and 6997a3699b6aaf18ecafc07d95023f01a8b83a26
