You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by Alan Burlison <Al...@oracle.com> on 2015/05/13 13:14:13 UTC
HADOOP-9279 still not fixed, needs reopening
https://issues.apache.org/jira/browse/HADOOP-9279
Document the need to build hadoop-maven-plugins for eclipse and separate
project builds
Although the BUILDING.txt mentions the need to to build
hadoop-maven-plugins for Eclipse, it's actually needed for *all* builds,
even from the top level, otherwise a top-level build will fail with:
[ERROR] Failed to parse plugin descriptor for
org.apache.hadoop:hadoop-maven-plugins:3.0.0-SNAPSHOT
(/pool1/home/alanbur/bleaklow/bigdata/hadoop/hadoop-maven-plugins/target/classes):
No plugin descriptor found at META-INF/maven/plugin.xml -> [Help 1]
Manually building hadoop-maven-plugins fixes the problem, but if I'm
reading HADOOP-9279 right it shouldn't be required. What's the correct
course of action here, reopen HADOOP-9279 or log a new bug? If it's the
former I can't see any way of reopening it in JIRA, presumably because
I'm not blessed enough to do so.
Thanks,
--
Alan Burlison
--
Re: HADOOP-9279 still not fixed, needs reopening
Posted by Alan Burlison <Al...@oracle.com>.
On 13/05/2015 17:38, Chris Nauroth wrote:
> I usually tell people starting fresh to do "mvn clean install -Pnative
> -DskipTests" as a first step. That's a good validation of your build
> environment, and it will put a SNAPSHOT version of each sub-module into
> your local Maven repository. After that, you can do smaller incremental
> builds by cd'ing into hadoop-hdfs or hadoop-yarn-server-nodemanager or
> anything else and running things like "mvn test" for just that sub-module.
Thanks, I've logged
https://issues.apache.org/jira/browse/HADOOP-11968
BUILDING.txt is still unclear about the need to build
hadoop-maven-plugins first
And assigned it to myself.
--
Alan Burlison
--
Re: HADOOP-9279 still not fixed, needs reopening
Posted by Chris Nauroth <cn...@hortonworks.com>.
I usually tell people starting fresh to do "mvn clean install -Pnative
-DskipTests" as a first step. That's a good validation of your build
environment, and it will put a SNAPSHOT version of each sub-module into
your local Maven repository. After that, you can do smaller incremental
builds by cd'ing into hadoop-hdfs or hadoop-yarn-server-nodemanager or
anything else and running things like "mvn test" for just that sub-module.
--Chris Nauroth
On 5/13/15, 9:28 AM, "Alan Burlison" <Al...@oracle.com> wrote:
>On 13/05/2015 16:53, Chris Nauroth wrote:
>
>> Last time I checked, this was particular to "mvn compile" and would not
>> occur for other kinds of builds from the root of the source tree like
>>"mvn
>> test" or "mvn package". This is a Maven behavior that is unavoidable
>> unless we pull hadoop-maven-plugins out into its own project separate
>>from
>> the Hadoop source tree, which is something the community already decided
>> against. We can only document the behavior.
>
>Thanks for the explanation, I've added a comment to HADOOP-9279 to that
>effect. What's the best recommended target for someone to use when
>starting from scratch, test?
>
>> Please feel free to file a new issue if you'd like to see this clarified
>> in BUILDING.txt. We'd welcome a patch if you want to provide one.
>>Thanks!
>
>Yes, I'll do that.
>
>--
>Alan Burlison
>--
Re: HADOOP-9279 still not fixed, needs reopening
Posted by Alan Burlison <Al...@oracle.com>.
On 13/05/2015 16:53, Chris Nauroth wrote:
> Last time I checked, this was particular to "mvn compile" and would not
> occur for other kinds of builds from the root of the source tree like "mvn
> test" or "mvn package". This is a Maven behavior that is unavoidable
> unless we pull hadoop-maven-plugins out into its own project separate from
> the Hadoop source tree, which is something the community already decided
> against. We can only document the behavior.
Thanks for the explanation, I've added a comment to HADOOP-9279 to that
effect. What's the best recommended target for someone to use when
starting from scratch, test?
> Please feel free to file a new issue if you'd like to see this clarified
> in BUILDING.txt. We'd welcome a patch if you want to provide one. Thanks!
Yes, I'll do that.
--
Alan Burlison
--
Re: HADOOP-9279 still not fixed, needs reopening
Posted by Chris Nauroth <cn...@hortonworks.com>.
Last time I checked, this was particular to "mvn compile" and would not
occur for other kinds of builds from the root of the source tree like "mvn
test" or "mvn package". This is a Maven behavior that is unavoidable
unless we pull hadoop-maven-plugins out into its own project separate from
the Hadoop source tree, which is something the community already decided
against. We can only document the behavior.
Please feel free to file a new issue if you'd like to see this clarified
in BUILDING.txt. We'd welcome a patch if you want to provide one. Thanks!
--Chris Nauroth
On 5/13/15, 4:14 AM, "Alan Burlison" <Al...@oracle.com> wrote:
>https://issues.apache.org/jira/browse/HADOOP-9279
>
>Document the need to build hadoop-maven-plugins for eclipse and separate
>project builds
>
>Although the BUILDING.txt mentions the need to to build
>hadoop-maven-plugins for Eclipse, it's actually needed for *all* builds,
>even from the top level, otherwise a top-level build will fail with:
>
>[ERROR] Failed to parse plugin descriptor for
>org.apache.hadoop:hadoop-maven-plugins:3.0.0-SNAPSHOT
>(/pool1/home/alanbur/bleaklow/bigdata/hadoop/hadoop-maven-plugins/target/c
>lasses):
>No plugin descriptor found at META-INF/maven/plugin.xml -> [Help 1]
>
>Manually building hadoop-maven-plugins fixes the problem, but if I'm
>reading HADOOP-9279 right it shouldn't be required. What's the correct
>course of action here, reopen HADOOP-9279 or log a new bug? If it's the
>former I can't see any way of reopening it in JIRA, presumably because
>I'm not blessed enough to do so.
>
>Thanks,
>
>--
>Alan Burlison
>--