[ https://issues.apache.org/jira/browse/HADOOP-11937?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14538141#comment-14538141 ]
Allen Wittenauer commented on HADOOP-11937: ------------------------------------------- [~busbey] and I had a long talk about HADOOP-11929, which effectively covers this and a few other non-filed issues (e.g., a pom change in something like dist should really trigger a full build). I'll likely have a sample solution in the next week or two, as it requires some invasive surgery. But I think I know how to do it without increasing the build times for every patch and may actually shrink the build times for other types of patches. Of course, in the case of the native code, the docker bits in HADOOP-11933 will still likely need to get used. [I need to do one or two things to clean that up, then I'll start running some tests with it on Jenkins later this week.] > Guarantee a full build of all native code during pre-commit. > ------------------------------------------------------------ > > Key: HADOOP-11937 > URL: https://issues.apache.org/jira/browse/HADOOP-11937 > Project: Hadoop Common > Issue Type: Improvement > Components: scripts > Reporter: Chris Nauroth > Fix For: 2.8.0 > > > Some of the native components of the build are considered optional and either > will not build at all without passing special flags to Maven or will allow a > build to proceed if dependencies are missing from the build machine. If > these components do not get built, then pre-commit isn't really providing > full coverage of the build. This issue proposes to update test-patch.sh so > that it does a full build of all native components. -- This message was sent by Atlassian JIRA (v6.3.4#6332)