[
https://issues.apache.org/jira/browse/SLIDER-1005?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15028534#comment-15028534
]
Steve Loughran commented on SLIDER-1005:
----------------------------------------
I'd be up for the generated-sources path, though it should still be in SCM.
Otherwise everyone needs a copy of protoc to hand *of the right version*. Which
is surprisingly hard to do, especially on windows. That's why its a
{{-Pprotoc}} profile: so most people don't have to do it. I think HBase does
the same
IF we do this, the avro code should go the same way.
(FWIW, I often use the command line protoc while adding new messages/fields, as
its the only way to see the error messages). Once working I use the protocl
> The java code generated by protobuf should not be placed in src directory
> -------------------------------------------------------------------------
>
> Key: SLIDER-1005
> URL: https://issues.apache.org/jira/browse/SLIDER-1005
> Project: Slider
> Issue Type: Improvement
> Components: build, core
> Affects Versions: Slider 0.90
> Environment: Centos 6.5, Maven 3.2.5
> Reporter: Pan Yuxuan
> Priority: Minor
>
> Now slider using protobuf for RPC, but the output directory is
> <output>${basedir}/src/main/java</output> in slider-core/pom.xml. That means
> the generated java code was been placed in the src directory.
> I have seen the generated java code for protobuf in Hadoop, they put the code
> in build directory
> <output>${project.build.directory}/generated-sources/java</output>. And they
> put the compile-protoc in <plugin> not <profile> in the pom.xml.
> I think the way Hadoop used is more reasonable. The code is generated by
> protobuf, not by ourselves.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
