Some more thoughts: Even if the original org.apache.jackrabbit.webdav.lock.Scope did not have a hashCode() it implicitly inherited the one from Object. Therefore overwriting the hashCode() with a better implementation should not change anything about compatibility and therefore not require any version change. I am pretty convinced that bnd should in fact just ignore the @HotSpotIntrinsicCandidate annotation.
> On 3. Oct 2017, at 16:41, Konrad Windszus <[email protected]> wrote: > > If I understand the output of https://issues.apache.org/jira/browse/JCR-4191 > correctly it means, it detected that hashCode() has modified after version > 1.0.0 (not added). > In this case the addition of the annotations has been the modification (which > is actually an annotation being added by javac and only with Java9, > https://bugs.openjdk.java.net/browse/JDK-8076112). > > I guess this is worth a bug report at bnd, because those annotations should > IMHO clearly be ignored by the baselining since it depends on the compile > time options, whether the annotation is there or not. > > I am still wondering why this cannot be reproduced with bnd alone... > Are you sure that both the maven-bundle-plugin and bnd are looking at exactly > the same classes (being produced by the same compiler run?) > Also to me it is not clear whether only the Java compiler can add that > annotation or whether it could also be added at runtime by the JIT... > > Konrad > >> On 3. Oct 2017, at 15:53, Julian Reschke <[email protected]> wrote: >> >> Hi there, >> >> I'm looking for help in understanding the above issue. >> >> Short summary: the baseline check asks for a version increase for a class >> where "hashCode()" was added, but only when run with jdk9. >> >> So should I trust the diagnostics, or is this jdk9-related issue? >> >> Best regards, Julian >
