[Libreoffice-qa] Lifecycle of builds?

[Petr Mladek]

Robinson Tryon píše v St 10. 04. 2013 v 10:12 -0400:
> On Wed, Apr 10, 2013 at 8:02 AM, Rainer Bielefeld
> Rainer wrote:
> "So I believe we should mark the not-release-3.3 as inactive, what means
> they will stay for the bugs where they are used, but can not be used for
> new bug reports (except by the LibO Bugzilla Administrators) because
> they will no longer be shown in the Versions selector."

Nice feature

> To riff off of your suggestion, perhaps we can do our cleanup in stages:

Makes sense.

> 1) When we EOL a stable release such as 3.5.5, we remove (i.e. "mark
> as inactive") all of the not-release-3.5.5.x versions from the
> bugtracker.

I think that it is hard to define life time for a bugfix release. It is
basically obsoleted by the next bugfix release for the given minor
version.

If we accept this, it would mean to remove betas and rcs for "X.Y.Z"
when "X.Y.Z+1" bugfix release is released. For example, remove 4.0.2 RCs
when 4.0.3 is released.

It could work. The logic here is that people, who install RCs from
prerelease, usually install newer RCs when they are available. The RCs
granularity gets outdated with next release because most people do not
have them installed. If anyone wants to find when it exactly got broken,
they  probably use bibisect which uses another identification.

Maybe we could wait one more bugfix release, just for sure. See below.


> 2) When we EOL a release series such as 3.5, we *could* do more
> cleanup, but there still may be end users running those stable builds.
>
> As Michael pointed out, we still have companies interested in
> maintenance on those older versions. Ubuntu 12.04 still has a 3.5
> build, and I'm not sure if/when that will be upgraded. Realistically,
> I think we can punt on this piece of the puzzle until another day.

I would probably leave it a bit longer. It seems that people still use
different 3.4.x bugfix releases. I did a query for bugs reported this
year:

      + 64 bugs are marked against the version "3.4.*"
      + 1 is against "3.4.*RC.*"  (fdo#53725)
      + 2 are against "3.4.6 release" (last one)
      + 61 are against other "3.4.x" releases

Best Regards,
Petr