Am 2015-10-28 13:34, schrieb Maximiliano Curia:
On 27/10/15 13:51, Martin Graesslin wrote:
I was thinking about the problem of how we can get bug fixes quicker to our user. With a three month release cycle a one-month bug fix cycle sounds too
long to me.

Fair enough.

So I thought we should make bug fix releases faster and more often. In 5.4 we already went for this partially by having the first bug fix earlier. I wanted to know how much work this would mean for our distributions. If we ship out way more bug fix releases, would you be able to work with it? Would it block you? Would you have to skip releases? Or is it just pressing a button to run
automatic scripts which upload your packages?

We can keep up with bug fixes, in most cases the bugs will affect our users anyway and having the bug fixes released would reduce the time spent dealing with bugs, but it would be a waste of time and resources to package things
that only contain a version bump.

understood. Maybe we can do some scripting to only create tarballs if there were changes. Jonathan?


So my suggestion would be to publish a 5.5 version and use the dot releases to
publish bug fixes without a fixed schedule.

Sorry I don't understand what you mean with that. Do you want a new bug fix release for each commit to stable branch? That certainly wouldn't scale.


The next discussion would be what should be considered a bug fix to trigger a release, as much as I like to have l10n support, the automatic: l10n daemon
script commit hardly qualify.

While l10n might not seem to be that important to you, it might be the difference between an unusable system and a useable system. So let's not start downplaying the importance of translations ;-)


What had I been thinking about? I was thinking about a Fibonacci based release schedule. This gives us quick bug fix releases directly after the release with slowly larger intervals. Of course it would mean tag and release happens on
same day.

Agreed on the release and tag on the same day. Can it be enforced that all the
tests need to pass to consider the part for the release

Please start a dedicated discussion for this point.

Also, if possible, could you use signed tags for the releases? (on the Debian side of things we are considering using the upstream signed git tags as a
replacement of tarballs and signatures/sums)

Please also start a dedicated discussion for this point.

Cheers
Martin
_______________________________________________
release-team mailing list
[email protected]
https://mail.kde.org/mailman/listinfo/release-team

Reply via email to