[
https://issues.apache.org/jira/browse/YETUS-559?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16722202#comment-16722202
]
Allen Wittenauer commented on YETUS-559:
----------------------------------------
A year later and several projects later, I have a suspicion that it isn't the
raw ordering that's the problem, but the very first 'mvn install'. In other
words, I find myself writing code similar to this :
{code}
mvninstall)
if [[ "${repostatus}" = branch || "${BUILDMODE}" = full ]]; then
clear_personality_queue
personality_enqueue_module .
{code}
quite often. In other words, in qbt and in patch-testing mode, the VERY FIRST
mvn install should probably be done at the root level. After that, the maven
ordering should be absolutely fine. I also suspect this situation is true for
other build systems.
If that's the case, should instead of changing the ordering the fix be that the
mvn install (and other builds?) just always install root for the branch-part of
the build?
> mvn reordering should put root first
> ------------------------------------
>
> Key: YETUS-559
> URL: https://issues.apache.org/jira/browse/YETUS-559
> Project: Yetus
> Issue Type: Improvement
> Components: Test Patch
> Affects Versions: 0.4.0
> Reporter: Mike Drob
> Assignee: Mike Drob
> Priority: Major
> Attachments: YETUS-559.patch
>
>
> when doing the maven module reordering, we currently put root module at the
> end. however, if child modules depend on changes in the root module, then
> they will fail (at mvn install) since they will be looking at the old root
> pom.
> See comment at
> https://issues.apache.org/jira/browse/HBASE-16338?focusedCommentId=16208146&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16208146
> for an example of how this can fail.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)