[
https://issues.apache.org/jira/browse/IGNITE-10396?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16698863#comment-16698863
]
PetrovMikhail commented on IGNITE-10396:
----------------------------------------
The problem can be in HashMap field initialization via
{code:java}
HashMap(Map<? extends K, ? extends V> m) {code}
constructor. The way how initial map size is setted in HashMap#putMapEntries is
different from the approach in IgniteUtils#capacity method,which is used during
deserialization. It can cause the differences in keys' hashcode calculation and
leads to different entries order.
> Serialization issue of class with HashMap field
> -----------------------------------------------
>
> Key: IGNITE-10396
> URL: https://issues.apache.org/jira/browse/IGNITE-10396
> Project: Ignite
> Issue Type: Bug
> Reporter: PetrovMikhail
> Assignee: Dmitriy Govorukhin
> Priority: Major
> Fix For: 2.8
>
> Attachments: SerializationTest.java
>
>
> If class which cointains HashMap field be serialized to BinaryObject,
> deserialized and serialized again, hash codes of first and second
> BinaryObjects will be different.
> That can case the following situation.
> IgniteCache#containsKey or IgniteCache#remove methods return false when in
> their parameters passed instance of key class produced via cache iterator.
> Even in so situation when IgniteCache.Entry with such key is existing.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
