[
https://issues.apache.org/jira/browse/HBASE-18782?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16160761#comment-16160761
]
stack commented on HBASE-18782:
-------------------------------
They are in place thoughout, yes. They were added once and then have not been
maintained since. Original idea was that we could feed hbase piecemeal to
external CI sites like Travis -- Travis doesn't like projects of our size and
test-duration -- first passing it the i/o tests, then the region tests, then
the mapreduce tests, and so on. It cross-cuts our modularization. Not sure it
the way to go. Either we double-down on the "test-type" categorization -- we
already have another categorization dimension of small/medium/large which has
decent definition up in refguide -- and exploit it in someway (at mo. it is
just unused noise) or we purge it as a failed experiment. I'm for the latter.
> Module untangling work
> ----------------------
>
> Key: HBASE-18782
> URL: https://issues.apache.org/jira/browse/HBASE-18782
> Project: HBase
> Issue Type: Task
> Reporter: stack
> Assignee: stack
> Fix For: 2.0.0-alpha-3
>
> Attachments: HBASE-18782.master.001.patch,
> HBASE-18782.master.002.patch, HBASE-18782.master.002.patch,
> HBASE-18782.master.003.patch
>
>
> Here is a bit of module untangling work done using structure 101. In a few
> places we depend on modules but are not making use of the dependency. In
> other places we depended on a module just because we had one references only,
> one we could undo.
> After this little exercise, we have nicer layering and stuff like
> hbase-protocol and hbase-protocol-shaded no longer depend on other hbase
> components. Not everything depends on annotations anymore.
> There is way more work to do here but this will do for now.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
