So just out of interest: Is there ASF regulation that totally forces releases to be built on personal machines or could this in theory be done in an CICD pipeline (if the user identity is correctly propagated to fulfil "by definition releases are individual acts in the Apache Software Foundation")?

On 2019-03-29 22:40, Georg Henzler wrote:
Sorry for the late feedback, in general I really like the idea to
improve the release process!

The wiki states:

Many of these must be run by the release manager, since by definition releases are individual acts in the Apache Software Foundation.

Isn't pushing a button in Jenkins when you are logged in an individual
act? IMHO it would be best to just have one pipeline that you start in
Jenkins on master, builds (it's easy with git to have Jenkins commit
as a certain user) and stages the release (the private GPG key might
be problem, but I'm sure that can be solved somehow) and then sends
the [VOTE] email to the mailing list. The votes can be done by simple
input elements (that send an email to the mailing list if +1 is voted,
-1 can be another button that would also sent that email and abort the
pipeline).

Having it all in Jenkins would be simpler than a CLI tool (that again
needs documentation). Also Jenkins can automatically acknowledge user
rights (release start only for committers), it can also count the
votes (binding/non-binding) after timeout and promote (or not promote)
the release automatically.

Just imagine a workflow where
- one committer starts the release by just a click in Jenkins on Thursday night
- three PMC members vote +1 within an hour (because things had been
discussed already)
- all these people move on to other tasks
- on Sunday, the release pipeline automatically acknowledges the
sufficient votes and does everything (including updating the site
etc.) that is needed to finish the release

I think this would greatly improve productivity :)

There are probably reasons why this is not allowed (e.g. the release
has to be built on a computer of an individual), but personally (if it
is a problem) I think it could even be worthwhile to question these
requirements with the ASF board.

-Georg


On 2019-03-06 14:28, Robert Munteanu wrote:
Hi,

I have been thinking lately about how we can improve our development
process for contributors/developers/release managers.

I have listed some issues that I perceived as important with proposed
solutions at

https://cwiki.apache.org/confluence/display/SLING/Scaling+Sling+Development

All comments are welcome:

- there are probably pain points I have missed
- maybe there are solutions I have overlooked
- maybe some pain points are my own and others have solved them :-)

Thanks!

Robert

Reply via email to