[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Uploading during freeze time



On 11 October 2010 19:02, Russ Allbery <rra@debian.org> wrote:
> Jordi Gutiérrez Hermoso <jordigh@gmail.com> writes:
>

>> The *real* problem is that labelling Firefox 3.6 as "experimental" is
>> downright silly.
>
> Hm, okay.  I guess I'm not feeling particularly inspired to do any work
> based on that reaction.

Sorry if I'm frustrated, but there's a real issue that you seem to
insist as dismissing as a non-issue. Perhaps you're happy mixing three
different distribution lines in your sources.list, but I'm not.

> I can see the argument of people who'd like to have new software in
> unstable during the freeze period since they're not very interested in the
> stable release, and that's come up before (such as in the CUT
> discussions), but as with your note above, this seems most often to be an
> argument made theoretically by people who aren't personally having
> significant issues.

That's not what I noted above. Or maybe it is. But it is a real
problem. Let me try to convince you again that this is an actual
problem. I was having issues during the last freeze time with squeeze,
with KDE, where software that would normally have gone into unstable
went into experimental instead. Admittedly, KDE 4 was probably buggy
enough to merit the experimental label, but as it stabilised and went
into the 4.1 and 4.2 releases, it still had the experimental label,
still due to the necessity of being able to ship 3.6 in squeeze.

To elaborate, during freeze time, experimental *still* has packages of
actual experimental quality, and it also has packages that just can't
make it into unstable due to the freeze. Firefox is of the latter
category, binutils of the former. So you get a mix, and because
experimental is not a full distribution (afaik, it doesn't do some of
the automatic testing that unstable does), packages that should
receive the attention of unstable don't get it, nor do they get the
infrastructure of unstable (e.g. p.d.o treats experimental
differently).

So it's freeze time, you want to try out software that is new but
mostly ok, and you realise it's not in unstable. After looking around
for a while, someone explains that the DDs are busy using unstable to
upload changes to testing, so if you want to get the newer packages,
you have to go to experimental, which is labelled all over the place
with warnings about how it will eat your firstborn. You decide you
will never have a firstborn anyways, so you update your sources.list
and you get the package you wanted to try out. It is, perhaps, a
little buggy, or perhaps it is stabler than the proposed stable
package, but there you have it.

Now, experimental by default is pinned to lowest priority, and in the
meantime, during the freeze, your friendly packager is still waiting
on the testing package, so updates the experimental package. You may
or may not notice, but because of the low pin priority you won't
automatically upgrade the experimental package that is in effect being
treated as an unstable package by the packager. So perhaps you go and
you change the low pin priority, but you can't really do that easily
because then apt will try to pull in all sorts of assorted crap from
experimental that you really don't want and is genuinely experimental.
Perhaps you're smart enough to deal with this problem automatically,
perhaps you're not and you're stuck closely following the development
of a package that you would be tracking automatically without the
freeze.

Then testing is finally released, champagne for everyone, all the
distributions roll over, and if you're tracking testing, you get a
deluge of packages cascading down from unstable, but that's expected
and normal. Some of the unstable-experimental packages might get
pushed down to unstable, and you may or may not remember to remove
your experimental line from sources.list and/or change the pin
priority of experimental back to its lowest. Some of them don't,
because the packager doesn't get around to it, so you may still be
stuck tracking experimental even though you don't really want
experimental packages, and you still have to deal with the problems
with the freeze of the mix of packages in experimental.

It takes a couple of months, usually things settle back to the way
they are, unstable again is again used as it should be, experimental
again is full of packages that really eat firstborns as announced all
over the Debian pages, you finally purge the experimental line from
your sources.list because you changed your mind about having
firstborns, and things go on their merry way...

Until next freeze time, that is.


Reply to: