[Libreoffice-qa] New Whiteboard Statuses

[Kohei Yoshida]

On Thu, 2014-06-05 at 15:41 +0200, Joren DC wrote:
> Hi Kohei,
> 
> Kohei Yoshida schreef op 5/06/2014 15:06:
> > So, to me personally, this practice of "witch-hunting" (or
> > finger-pointing) really bogs me down, especially I receive such notice
> > hundreds of times during a typical development cycle.
> Well, that's at least not what I'm talking about right now.
> (1) we (I) are (am) only talking about bugs we can track down to 1 
> single commit or developer. Not a developer in general by component 
> (writer, calc, ...), which we do now to try to avoid as much as 
> possible. I'm not sure that there are that many QA'ers or reporters 
> which can track down to 1 single developer/commit?
> I'm not a developer at all and just to provide me an idea: do you still 
> receive that much CC's on bugs compared to months/a year ago?

No because I turend off the notification.  But occasional query for such
bug reports turns up still quite a bit, and most of them are general
Calc bugs and the reporters just add me, Eike and Markus as part of
their routine bug triaging.

> 
> (2) we discussed this yesterday on the QA-call too. Our conclusion was 
> to just kindly ping a developer in particular on IRC. 
> (https://wiki.documentfoundation.org/QA/Meetings/2014/June_04#Topics_for_ESC.3F) 
> If not I still am 'pro' an active approach and mail the particular 
> developer (in private) or put him/her in CC.
> 
> (3) and as far my experience concerns... I already know which developers 
> are and are not open for a nice (not pointing, just asking) SINGLE 
> message. I think other core QA members do have such experience too.
> >   Since statistically every change one makes can and will cause *some*
> > regressions in some obscure corners, this disadvantages those who make
> > lots of changes, even when those changes are to fix other regressions
> > and bugs.
> True, but luckely not all regression cases are 'obscure' and border cases.

It's pretty subjective what a "border" case is.  For the bug reporter,
the bug he/she reported is not a border case but a serious regression
"that needs immediate fixing!" even though others don't see it that way.

> > And some of these sometimes escalate to a (often repeated) demand of a
> > revert of the commit, which is another blow especially when the change
> > itself took weeks and weeks of careful coding to get conceived. One can
> > be as careful as possible, and still (and almost always) break somethign
> > somewhere.
> Again: true. But we are not talking about: bug the particular developer 
> as much as possible, if he doesn't react/revert/fix spam his email and 
> IRC with threats to revert that commit ...

I'm just throwing that in because it happens quite often, and is a
growing concern for me.  It's not targeted toward you personally.

> I think we have to find the most constructive approach to get a 
> regression bug fixed, with respect to the situation (developer, commit 
> message, ...) and severity.

Sure, but please keep in mind that we currently don't have enough
developers to fix all regressions, and I dare say with the current
development resources (and people's (un)willingness to fix bugs), we
could probably only fix 5% or less of all bugs tagged regressions.  And
each regression fix will (yes I'm using the word "will" here on purpose)
create at least 2 or 3 new ones, and the cycle only continues.

Maybe I'm being pessimistic, but that's how I see the situation.

Kohei