[
https://issues.apache.org/jira/browse/HADOOP-10530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14239250#comment-14239250
]
Steve Loughran commented on HADOOP-10530:
-----------------------------------------
colin, sorry must have missed this.
bq. In the future, we should make sure that all the precommit jobs work before
committing stuff like this, not just the hadoop-common one.
I actually spent the whole weekend spinning up jenkins builds and seeing the
results, which included the precommit builds. Somehow that bit escaped me.
I do note that if I browse to
[[https://builds.apache.org/view/H-L/view/Hadoop/job/PreCommit-HDFS-Build/]] I
can see a large number of builds over the weekend; pretty much everything from
#8949 to #8966. The fact that this build took two hours meant that it was one
of the
Even identifying the fact that not only was java/latest the wrong value, so was
java/latest7; I ended up having add ls commands to the build to identify which
JDKs were present, then experimenting to see how much worked. Then there were,
as I noted, race conditions. Things get confused if you are running enough
builds that >1 precommit is on the same node.
There's also
# the fact that the output handling of the different precommits vary. Why do
some builds end up with a {{../patchprocess}} dir, when others do not -and
pasting in the {{mv}} command fails?
# the need to go through every build and enter the memory requirements
# fixing the XML test capture so they pick up stuff like the contrib/ section
in the hbase project.
A lot of the work I ended up doing was not actually the java7 migration, it was
fixing a set of jenkins builds that were in a strongly inconsistent state.
I did not expect so much of my weekend to be frittered away with returning to
my keyboard every hour —if you look at the build logs for all the projects you
can see how many times I did end up triggering builds, from 10 am to midnight
both days. I therefore consider the implication that I didn't look at the HDFS
build invalid and unappreciative of the effort I put in.
I shall give you the honour of repeating this process when we move to java 8.
> Make hadoop trunk build on Java7+ only
> --------------------------------------
>
> Key: HADOOP-10530
> URL: https://issues.apache.org/jira/browse/HADOOP-10530
> Project: Hadoop Common
> Issue Type: Improvement
> Components: build
> Affects Versions: 3.0.0, 2.6.0
> Environment: Java 1.7+
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Blocker
> Fix For: 2.7.0
>
> Attachments: HADOOP-10530-001.patch, HADOOP-10530-002.patch,
> HADOOP-10530-003.patch, HADOOP-10530-004.patch, HADOOP-10530-005.patch,
> HADOOP-10530-debug.000.patch, Screen Shot 2014-09-20 at 18.09.05.png
>
>
> As discussed on hadoop-common, hadoop 3 is envisaged to be Java7+ *only*
> -this JIRA covers switching the build for this
> # maven enforcer plugin to set Java version = {{[1.7)}}
> # compiler to set language to java 1.7
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
