[Mageia-dev] [Freeze] please let in xonotic

[Michael Scherer]

Le dimanche 11 mars 2012 à 17:58 +0100, Samuel Verschelde a écrit :
> Le dimanche 11 mars 2012 13:21:57, Michael Scherer a écrit :
> > Le samedi 10 mars 2012 à 10:26 +0100, Thierry Vignaud a écrit :
> > > Hi
> > > 
> > > please let in xonotic-0.6.0
> > > It's just a game that impacts nothing else.
> > 
> > Niet.
> > 
> > That's bugfix or security fixes.
> 
> Communication tip for the future : when such a policy change happens (this is 
> a policy change, there were more exceptions in the past before package freeze, 
> for leaf packages without weird deps and post/pre scripts and little impact), 
> let's try to highlight it *before*.

We asked if people understood what version freeze mean, and people all
said yes. 

http://meetbot.mageia.org/mageia-dev/2012/mageia-dev.2012-02-15-20.13.log.html#l-75

So I assumed that people were aware of what was posted last time :

http://permalink.gmane.org/gmane.linux.mageia.devel/4015

and that "very good reasons" where basically the same for everybody. 

It seems there is some misunderstanding, likely coming from another
discussion that the one I found. Can people just tell me where they have
seen their interpretation of freeze, cause I have read the one I posted
twice, I still think it correspond that what i just said.

> There was a communication effort to announce the freeze: I read freeze 
> announce, was present at the last packager meeting. However it didn't struck 
> me, it all looked like the usual end of release cycle, with their common-sense 
> exceptions.
>
> A message such as "contrarily as what happened in the past" (this is the 
> important part, highlight a *change*), "there will be no exception even for 
> leaf non-critical packages, so that you all focus on bugfixes, as we know that 
> otherwise you will not fix bugs and continue updating your packages". 

I took a look at the archives for the last round ( ie mageia 1 ), thanks
to evolution. Since people were upset because this exception should have
been granted, I assumed that the same type of exception would have been
granted in the past numerous times ( I mean, if people didn't remember
the mail saying "this would not be granted", that's likely because we
granted it several time, enough time to have people totally forget what
was posted in the first place ).

I found exactly 1 package that not at least saying "that's a bug fix
release", "this fix a CVE", "it fix upgrade"., ie all reasons that were
announced "we would likely let this pass".

It was alienarena on 06/05/2011, and it was pushed by boklm around 2h
after the mail was sent.

That's the only one I can find that was not corresponding at the
criteria we laid out for being a regular exception ( cf url given sooner
), but it was still granted. We trusted boklm as well as others to
choose, so he did. But still, if we look at the list given before ( and
since no one answered at all and no one complained later, I assumed that
everybody agreed, especially since this was based on common sense ), it
was said "it would likely not pass".


So if there wasn't others ( and I will assume that's the case unless
someone show others examples that I could have missed ), we can see that
the whole "but it was granted last time" assumption is either based
on : 
- 1 single rpm being a exception that would have been refused and that
wasn't ( ie, a weak example )
or :
- based on a different "last time" than last year ( and then why did no
one complained last year is left as a exercise )
or :
- based on different perception of what type of exception were granted
last time, perception that doesn't align with the reality of the
archives I checked.

But so far, I think that we are basically and roughly coherent when
compared to the last freeze period. Therefore, I do not see the need to
communicate a change when there was in fact no changes on the side of
the policy. And sorry, I cannot communicate to say that what people
remember do not correspond to what was done. 

And we can see that as tmb said, the problem is not that granting would
disturb the distribution, but that granting once for a minor package
would lead to the same type of lengthy and heated discussion as we are
having now, but for each packages. The more complex the rules are, the
more discussion and the more problem we will have to solve. 

-- 
Michael Scherer