> On Mar 8, 2017, at 10:55 AM, Marton Elek <me...@hortonworks.com> wrote:
>
> I think the main point here is the testing of the release script, not the
> creation of the official release.
… except the Hadoop PMC was doing exactly this from 2.3.0 up until
recently. Which means we have a few years worth of releases that are
effectively untrustworthy despite being signed. One of the (many) reasons I
rewrote the release process was to get Hadoop back in line with ASF policy.
Given the massive turn over in committers, I don’t want us to repeat the same
mistakes (like we usually do).
> I think there should be an option to configure the release tool to use a
> forked github repo and/or a private playground nexus instead of official
> apache repos. In this case it would be easy to test regularly the tool, even
> by a non-committer (or even from Jenkins). But it would be just a smoketest
> of the release script…
This is already possible:
* don’t use —asfrelease
* use —sign, —native, and, if appropriate for your platform,
—docker and —dockercache
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-dev-h...@hadoop.apache.org
