[ 
https://issues.apache.org/jira/browse/STORM-2937?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16354673#comment-16354673
 ] 

Stig Rohde Døssing commented on STORM-2937:
-------------------------------------------

Okay. I thought the whole reason we were doing this issue was to avoid 
releasing 1.0.6 with a broken storm-kafka-client?

Regarding setting the deserializer, for the 1.1.1 release it's done via 
KafkaSpoutConfig.Builder.setKey/setValue, or one of the Builder constructors. 
The configure method won't be called on the deserializer because we're using 
this constructor internally 
https://kafka.apache.org/0100/javadoc/org/apache/kafka/clients/consumer/KafkaConsumer.html#KafkaConsumer(java.util.Map,%20org.apache.kafka.common.serialization.Deserializer,%20org.apache.kafka.common.serialization.Deserializer).
 We realized this was a bad way to do it (and it was making the 
KafkaSpoutConfig.Builder interface a mess), so in 1.2.0 and the overwritten 
1.1.x-branch, the deserializers are set with e.g. 
KafkaSpoutConfig.Builder.setProp(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, 
your-deserializer-here.class). The configure method should be called when the 
consumer is instantiated.

Regarding commit squashing, I agree while reviewing the PR that it's easier if 
the squashing happens later. In this case it was only two files outside 
storm-kafka-client that changed though. We might consider doing merges the same 
way they do on the Kafka project. They squash PRs when merging, where we squash 
during the PR. It leaves the history intact in the PR. 

> Overwrite storm-kafka-client 1.x-branch into 1.0.x-branch
> ---------------------------------------------------------
>
>                 Key: STORM-2937
>                 URL: https://issues.apache.org/jira/browse/STORM-2937
>             Project: Apache Storm
>          Issue Type: Task
>          Components: storm-kafka-client
>    Affects Versions: 1.0.6
>            Reporter: Jungtaek Lim
>            Assignee: Erik Weathers
>            Priority: Blocker
>
> This is to track the effort of syncing up divergence between 
> storm-kafka-client 1.x-branch and 1.0.x-branch so that critical fixes can be 
> go in 1.0.x-branch as well.
> Note that it can modify storm-core as well (unlikely in a 
> backwards-incompatible way but not 100% sure), so we should make a decision 
> whether we allow the change in bugfix version line.
> Linking discussion thread:
> [https://lists.apache.org/thread.html/0451fed132bb982b618d9e0780282a87554f1bc5747827599f276944@%3Cdev.storm.apache.org%3E]



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to