[
https://issues.apache.org/jira/browse/HADOOP-8924?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13510565#comment-13510565
]
Alejandro Abdelnur commented on HADOOP-8924:
--------------------------------------------
If I'm not mistaken the votes for using Maven plugins didn't have a single -1s,
including a +1 from you Randim.
Writing Maven plugins is more complex that writing scripts, I don't dispute
this. The main motivations for using Maven plugins is to keep things in the POM
declarative and to hide (if necessary) different handlings for different
platforms.
IMO, having a bunch of stand alone java programs that do this will have the
same complexity than a Maven plugin, plus we'll lose all the benefits from
Maven, version management, distribution/download via maven repo, declarative
builds.
Regarding HADOOP-8887, it has been sitting around for a while, not because it
is incomplete or complex but because we are not solving how the plugin gets
into a Maven repo so it is available for the Hadoop build just to download it.
I thinking we should have a discussion in common-dev@ on how to address this.
Going back to this particular JIRA, the other approach is to do it in the way
Oozie does it:
For building a release tarball, there is a script 'mkdistro' which computes the
SCM (SVN or GIT) info (uri, branch, commit), the build timestamp and passes it
the Maven invocation as -D options, then they are avail as properties in the
POM and the version-info.properties file gets filtered with those.
Essentially is this patch minus the plugin plus a script to call Maven.
The drawback of this approach is that the script has to do very much what
saveVersion.sh does, meaning we are at square one.
We never tackled this in Oozie as a plugin because, so far, nobody as attempted
to build/run Oozie server in Windows. I assume in Oozie we'd go the plugin way
when such need arises.
I'm OK in updating this patch to do things how Oozie does it today, using bash.
For how to do that in Windows command without using cygwin, I don't know.
Again, I think we are back to square one with this approach, we are just the
script logic of saveVersion.sh around.
> Hadoop Common creating package-info.java must not depend on sh, at least for
> Windows
> ------------------------------------------------------------------------------------
>
> Key: HADOOP-8924
> URL: https://issues.apache.org/jira/browse/HADOOP-8924
> Project: Hadoop Common
> Issue Type: Improvement
> Components: build
> Affects Versions: trunk-win
> Reporter: Chris Nauroth
> Assignee: Alejandro Abdelnur
> Fix For: trunk-win
>
> Attachments: HADOOP-8924-branch-trunk-win.patch, HADOOP-8924.patch
>
>
> Currently, the build process relies on saveVersion.sh to generate
> package-info.java with a version annotation. The sh binary may not be
> available on all developers' machines (e.g. Windows without Cygwin). This
> issue tracks removal of that dependency in Hadoop Common.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
