[
https://issues.apache.org/jira/browse/HBASE-24319?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17099585#comment-17099585
]
Istvan Toth commented on HBASE-24319:
-------------------------------------
Both of the above suggestions would in effect make any -Dhadoop.profile setting
ineffective for master, so we'd still need to distinguish 2.x from 3.x, to
avoid running the same test twice for 3.x.
> Clearly document how profiles for the sake of Hadoop compatibility work
> across all branches
> -------------------------------------------------------------------------------------------
>
> Key: HBASE-24319
> URL: https://issues.apache.org/jira/browse/HBASE-24319
> Project: HBase
> Issue Type: Task
> Components: build, hadoop2, hadoop3
> Reporter: Josh Elser
> Priority: Major
>
> In HBASE-24280, we investigated a test failure which was ultimately caused by
> the simultaneous activation of the (intended mutually exclusive) hadoop-2 and
> hadoop-3 profiles.
> After master has moved to only supporting profile activation via the profile
> itself (rather than a system property) with the removal of the hadoop-2
> profile, the build was inadvertently broken as all branches (or is it just
> 2.x branches and master?) use the one build/yetus scripts in dev-support.
> To make sure that these scripts continue to work against all branches, we
> need to have a clear decision on how profile activation is expected to work
> in our HBase build. Otherwise, we'll come back to this problem where each
> branch does things ever-so-slightly different, requiring a bunch of {{if
> branch-2; else if branch-2.2; else if branch-2.3}} type changes to our yetus
> scripts.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
