[
https://issues.apache.org/jira/browse/FLINK-21385?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Flink Jira Bot updated FLINK-21385:
-----------------------------------
Labels: auto-deprioritized-major stale-minor (was:
auto-deprioritized-major)
I am the [Flink Jira Bot|https://github.com/apache/flink-jira-bot/] and I help
the community manage its development. I see this issues has been marked as
Minor but is unassigned and neither itself nor its Sub-Tasks have been updated
for 180 days. I have gone ahead and marked it "stale-minor". If this ticket is
still Minor, please either assign yourself or give an update. Afterwards,
please remove the label or in 7 days the issue will be deprioritized.
> AvroDeserializationSchema corrupted after getting invalid data
> --------------------------------------------------------------
>
> Key: FLINK-21385
> URL: https://issues.apache.org/jira/browse/FLINK-21385
> Project: Flink
> Issue Type: Bug
> Components: API / Type Serialization System
> Affects Versions: 1.12.1
> Reporter: Leonid Ilyevsky
> Priority: Minor
> Labels: auto-deprioritized-major, stale-minor
> Attachments: Test.out, Test.scala, TestRec.avsc
>
>
> After getting the data which cannot be deserialized,
> AvroDeserializationSchema goes into some corrupted state, which prevents it
> from properly deserializing good data that comes later.
> Looks like some buffer inside it gets messed up; the "bad" data is not
> properly skipped.
> Please see the attached files that help to reproduce the issue:
> TestRec.avsc with Avro schema, Test.scala with test code, and Test.out with
> the output of my test run.
> In my test, I generate 10 testing records, serialize them, and then
> deserialize using two methods: "decode2" using AvroDeserializationSchema and,
> for comparison, "decode" using SpecificDatumReader.
> Every other record is intentionally broken by adding extra 5 bytes in the
> front (I simulated the Confluent schema ID, because this is how I discovered
> this problem).
> Ideally, the "bad" records should just result in the exception, and the
> "good" records should be properly decoded. This is true for the simple
> "decode" method.
> However, "decode2" using AvroDeserializationSchema, after the first "bad"
> record cannot decode the good one. Further down, it does decode something,
> but not the data just passed in; it returns some previous record instead.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
