You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@geode.apache.org by "ASF GitHub Bot (Jira)" <ji...@apache.org> on 2022/05/10 14:06:00 UTC
[jira] [Commented] (GEODE-10276) Refactor PDX (de)serialziation code to align it with Java client
[ https://issues.apache.org/jira/browse/GEODE-10276?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17534346#comment-17534346 ]
ASF GitHub Bot commented on GEODE-10276:
----------------------------------------
gaussianrecurrence opened a new pull request, #969:
URL: https://github.com/apache/geode-native/pull/969
- Code related to PDX de(serialization) has been refactored and aligned
(in all possible means) to the Java client, solving the following
issues:
* GEODE-9968 - Fix deserialization for new fields in PdxSerializable
class
* GEODE-9753 - Coredump during PdxSerializable object serialization
* GEODE-10220 - Coredump while initializing PdxType remoteToLocal
* GEODE-10255 - PdxSerializable not working correctly for multiple
versions of the same class
- Reverted GEODE-8212, to solve potential issues during (de)serialization.
This is due to the fact that the server orders PdxType fields
sequentially, rather than alphabetically.
- In order to achieve what was originally intended with GEODE-8212,
now PdxInstances fields are written in alphabetical order, rather
than in the order writeXX is called.
- Also, code now should be easier to understand
- Several tests modified to adapt to code changes
- Added several UTs to further incrase coverage on PDX (de)serialization code.
- Added several ITs to verify PDX (de)serialization scenarios.
> Refactor PDX (de)serialziation code to align it with Java client
> ----------------------------------------------------------------
>
> Key: GEODE-10276
> URL: https://issues.apache.org/jira/browse/GEODE-10276
> Project: Geode
> Issue Type: Improvement
> Components: native client
> Reporter: Mario Salazar de Torres
> Assignee: Mario Salazar de Torres
> Priority: Major
>
> Currently there are the following open issues regarding PDX (de)serialization:
> * [GEODE-9968 - Fix deserialization for new fields in PdxSerializable class|https://issues.apache.org/jira/browse/GEODE-9968]
> * [GEODE-9753 - Coredump during PdxSerializable object serialization|https://issues.apache.org/jira/browse/GEODE-9753]
> * [GEODE-10220 - Coredump while initializing PdxType remoteToLocal|https://issues.apache.org/jira/browse/GEODE-10220]
> * [GEODE-10255 - PdxSerializable not working correctly for multiple versions of the same class|https://issues.apache.org/jira/browse/GEODE-10255]
> Also, the implementation on this ticket ([GEODE-8212: Reduce connections to server to get type id|https://issues.apache.org/jira/browse/GEODE-8212]) poses some issues with PDX entries which fields are a permutation. Thing is that PdxTypes which fields are a permutation might use the wrong offsets, leading to a corrupt serialization. This is something that was not taken into account at the time of getting this PR merged.
> So this ticket should be reverted and possibly an alternative solution proposed.
> In order to tackle these issues, a code refactoring is needed to introduce the following implementations:
> * Single type of PdxWriter
> * An implementation PdxReader that tracks unread data, and other that don't.
> * An implementation for PdxInstances that guarantees that fields are actually written in alphabetical order, independently of the writeFields call order. This should tackle the issue described above regarding GEODE-8212.
> * Also, it'd be ideal to make it so PDX code is cleaner and easier to understand, though that's a complex matter, and also, subjective.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)