What steps will reproduce the problem? 1. create Topic assocType with History off 2. createAssociation(assocType) 3. call revisionIndex.firstRevision.changeSet

What is the expected output? What do you see instead? A changeSet with at least the change of type "ASSOCIATION_CREATED". Instead: empty changeSet and the next revision holds the information I expected in the first (empty) one.

createAssociation(assocType) with History on, of course.

Hi,

I write the following short test which works, as expected.

IRevisionIndex index = topicMap.getIndex(IRevisionIndex.class); index.open(); assertNull(index.getFirstRevision());

topicMap.getStore().enableRevisionManagement(false); ITopic type = createTopic(); assertNull(index.getFirstRevision());

topicMap.getStore().enableRevisionManagement(true); assertNull(index.getFirstRevision());

topicMap.createAssociation(type); assertNotNull(index.getFirstRevision()); IRevision r = index.getFirstRevision(); assertEquals(3, r.getChangeset().size());

assertNull(r.getFuture());

Can you post your code snippet which does not works as expected?

Sorry for the Ruby Spec code.

@tm.topics.should be_empty @tm.revision_index.last_revision.should be_nil @tm.disable_revision_management @tm.get!("assoc_type") @tm.enable_revision_management @tm.revision_index.last_revision.should be_nil @tm.create_association("assoc_type") @tm.all_revisions.each do |revision| changeset = revision.changeset changeset.should_not be_empty end

all_revisions is an Array of revisions, starting with last_revision and calling past until no revision is left.

If I call

first_revision = @tm.revision_index.first_revision first_revision.should_not be_nil first_revision.changeset.should_not be_empty

then the changeset of the first revision is also empty.

I see the problem, the bug only exists in the in-memory topic map store.

I fixed it and deploy it to maven dependency management