[
https://issues.apache.org/jira/browse/HADOOP-2207?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
stack updated HADOOP-2207:
--------------------------
Attachment: 2207-v3.patch
v3 includes improvements suggested up on IRC (plus the one added just now by
Doug that src/contrb/build.xml not have a jar target). Transcript below:
{code}
[09:25] * st^ack wonders if any chance of hadoop-2207 going in...
[09:33] * cutting wishes he had another coffee
[09:35] <cutting> st^ack: i looked over 2207 quickly last friday and it
looked good, but i wanted to give it one more glance before committing or voting
[09:35] <st^ack> cutting: no problem chief
[09:36] <cutting> is the jar-contrib target used by anyone?
[09:36] <cutting> or just there for symmetry?
[09:37] <st^ack> no. its the deploy target recast. thought might be of
utility having such a thing available in the top-level build
[09:39] <cutting> in build-contrib.xml, should "test" require "jar", or
just "compile". in the top-level, we don't jar for testing.
[09:39] <cutting> s/"./"?/
[09:41] <st^ack> test used depend on deploy which used depend on jar.
[09:41] <st^ack> let me try depending on compile only
[09:41] <cutting> st^ack: okay. so that's probably the safest, but, yes,
it'd be good to see if it'
[09:41] <cutting> s required
[09:42] * st^ack trying it...
[[09:44] <st^ack> cutting: Its looking like it works. Let me be
sure. Will upload new patch since this would be an improvement. Will ping you
when done. Thanks for review
[09:46] <cutting> & you can remove the jar-contrib target from top-level
& src/contrib/build.xml's jar target too, i think, in the interests of
minimalism
[09:48] <st^ack> cutting: will do
{code}
> Add 'package' target to contrib/build-contrib.xml
> -------------------------------------------------
>
> Key: HADOOP-2207
> URL: https://issues.apache.org/jira/browse/HADOOP-2207
> Project: Hadoop
> Issue Type: Improvement
> Reporter: stack
> Assignee: stack
> Priority: Minor
> Fix For: 0.16.0
>
> Attachments: 2207-2.patch, 2207-v3.patch, 2207.patch
>
>
> Currently, contribs each build a jar file that is deployed to the contrib
> directory when hadoop is packaged. Here is a listing:
> {code}
> ls build/hadoop-0.16.0-dev/contrib/
> hadoop-0.16.0-dev-abacus-examples.jar
> hadoop-0.16.0-dev-abacus.jar
> hadoop-0.16.0-dev-datajoin-examples.jar hadoop-0.16.0-dev-datajoin.jar
> hadoop-0.16.0-dev-hbase.jar
> hadoop-0.16.0-dev-streaming.jar
> {code}
> This works when the contrib product is a jar only.
> The hbase contrib needs to add scripts, configurations and its own jar
> dependencies. HADOOP-1648 proposed amending the general cluster CLASSPATH to
> add contrib jars (contrib dependencies would be added to lib/contrib).
> HADOOP-2196 was proposing that hbase scripts be installed into the bin
> directory beside the hadoop core scripts.
> Discussion up on #hadoop IRC channel this morning came up w/ a solution that
> better keeps the separation between whats core and whats contrib. In
> synopsis, responsibility for the contrib packaging is passed down to contribs
> as a new target in build-contrib.xml. The default would build the jar as is
> done now only rather than package the jar under the contrib directory,
> instead, the contrib jar would be put into a subdirectory named for the
> contrib. So for hbase, instead of packaging the hbase jar at
> contrib/hadoop-X.X.X-hbase.jar, instead it would sit at
> contrib/hbase/hadoop-X.X.X-hbase.jar . Those contribs that comprise of more
> than just a jar product -- e.g. hbase with its start/stop scripts, dependent
> libraries and configurations -- on invocation of the 'package' target, would
> create a bin, lib, and conf directory inside contrib/hbase/.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
