This change causes regression in bundle lifecycle handling. I added test coverage for this in the framework's BundleLifecycleTestCase
The issue is that a bundle may automatically progress to state RESOLVED after it gets installed . The expected behaviour is that a bundle stays in state INSTALLED until the first class load triggers a state change to RESOLVED. It is important that bundles do not get resolved one-by-one as they get installed because this may lead to undesired wirings. i.e. the resolver might no be able to correctly wire a set of bundles together
Consider, bundle B imports a package from bundle X. B gets installed. Starting B fails because of the the unresolved constraint. Later X gets installed and is expected to be in state INSTALLED. Currently, X progesses unexpectedly to state RESOLVED.
Bundle bundleB = installBundle(assembleArchive("lifecycle-bundleB", "/bundles/lifecycle/bundleB", ActivatorB.class, ServiceB.class));
assertBundleState(Bundle.INSTALLED, bundleB.getState());
try
{
bundleB.start();
fail("Unresolved constraint expected");
}
catch (BundleException ex)
{
// expected
}
Bundle bundleX = installBundle(assembleArchive("lifecycle-bundleX", "/bundles/lifecycle/bundleX", X.class));
System.out.println("FIXME [JBDEPLOY-245] Unexpected dependee state changes");
//assertBundleState(Bundle.INSTALLED, bundleX.getState());
http://fisheye.jboss.org/browse/JBossOSGi/projects/runtime/framework/trunk/bundle/src/test/java/org/jboss/test/osgi/bundle/BundleLifecycleTestCase.java?r=103300
I still need to look into the details of why this would happen.