[
https://issues.apache.org/jira/browse/CAY-2851?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17841298#comment-17841298
]
Matt Watson commented on CAY-2851:
----------------------------------
[~ntimofeev] I had only been running the tests on cayenne-server, so I missed
testing that one. However, looking at that test, I think it is a bad test. It
should have been written with {{assertEquals(5,
changeMap.getValue().getUniqueChanges().size());}} instead of {{4}} and then it
would have caught this bug that I found. The extra change that shows up with my
fix is {{a2}} {{toManyRelationshipChanges}} for {{{}children1{}}}. It used to
have a child (\{{ac2}}), but when {{ac2}} was added to {{{}a1{}}}'s children,
then it should have been removed from {{{}a2{}}}'s children. Now with my fix
that is happening.
Let me know what you think about that?
> Replace Existing OneToOne From New Object
> -----------------------------------------
>
> Key: CAY-2851
> URL: https://issues.apache.org/jira/browse/CAY-2851
> Project: Cayenne
> Issue Type: Bug
> Components: Core Library
> Affects Versions: 4.0.3, 4.1.1, 4.2, 5.0-M1
> Reporter: Matt Watson
> Assignee: Nikita Timofeev
> Priority: Major
> Fix For: 4.2.1, 5.0-M1
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> We have a One to One (belongsTo on both sides) relationships between 2 Models
> (A & B), very similar to the Painting < - > PaintingInfo in the test
> environment.
> When we have an existing {{a1}} related to {{b1}} (and vice-versa b1 to a1),
> and we create a new b2. Then set {{{}b2.toA = a1{}}}. We find out that
> {{b1.toA}} is still related to {{a1}} instead of {{null}}
> The closest test that I found is in
> {code:java}
> CDOOneDep2OneIT.testReplace{code}
> but just need to set from the other side of the relationship instead.
> I can push up a breaking test soon. Not sure if you prefer it to be on
> {{master}} or {{{}STABLE-4.2{}}}.
> It seems that during {{b2.setToA(a1)}} the call to {{setToOneTarget("toA",
> a1, true)}} triggers a call to {{setReverseRelationship("toA", a1)}} , which
> triggers a call to {{{}setToOneTarget("toB", b2, false){}}}. In here it
> accesses the {{oldTarget}} which is {{{}b1{}}}. This is where I feel it
> should nullify {{b1.toA}} using {{{}unsetReverseRelationship("toB", b1){}}}.
> However, it does NOT because {{setReverse}} was passed in as {{{}false{}}}.
> Since it does not, at this point if you check the relationships:
> {{b2.toA == a1}} _GOOD_
> {{a1.toB == b2}} _GOOD_
> {{b1.toA == a1}} _BAD_
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
