Re: Killing the gerrit to dev-list spam ...

Hi,

On Fri, May 03, 2013 at 04:58:13PM +0200, Jan Holesovsky wrote:
> Before we go ahead with any actions here, let me explain why I think it
> is good to have some sort of mails going to the ML, and what I think is
> broken at the moments.

No worries, Im drowned with work and will not suddenly change everything
tommorrow. ;)
> The reason why we should have the mails going to the ML is simple - we
> want to integrate the patches as quickly as possible, and for many, the
> mailing list is the primary way of operation - see Lionel's  workflow.
>
> I use something similar too - I read the ML so that first I kill threads
> ending with [PUSHED], and then check the not yet merged patches; and
> when it is something that I know something about, I go to gerrit &
> integrate it.

For those using such a workflow, both a second list of gerrit watches can keep
the flow of information.
 
> What seems to be broken (to me) though are 2 things:
>
> - the mails for the stable branches
>   - no need for announcements there, as people usually know whom to CC
>     to get a review, and it is later to be scanned by the person doing
>     tags before the tagging

That should be easy to fix.

 
> - the mails when "LibreOffice gerrit bot" is explicitly in the CC
>   - for those, not only [PATCH] and [PUSHED] go to the ML, but all the
>     discussion etc.

Hmmm, that _could_ be a feature though to bring a discussion to the dev-list
(see Thorstens mail). Maybe we just need to clarify how to use this.

> So before we do any radical action, what about to fix those 2 above [if
> we can agree that the above is bug, and not a feature], and then see how
> much the situation improves?

The first should be a technical fix, the second is a social one and thus
harder. But overall both will not help anything at all with the huge number of
threads started by automated mails.

> I can imagine 3rd thing to improve later, and that is to mail the
> initial [PATCH] mail to the ML only if it does not get merged in eg. 1/2
> day after submission ;-) - but that would probably need a patch to
> gerrit, or something.

Well, yeah. And when we wait half a day, we could also collect all the unpushed
fixes in _one_ mail and not start one thread for each. We could go fancy and
call this mail "digest". ;)

I dont see the need to track the status of patches via flaky "PUSHED" "PATCH"
subjects of emails -- at least for patches on gerrit, gerrit has a much better
way of keeping track of things as it tracks and watches the repository directly.

Best,

Bjoern