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

Re: A new address of an APT source registered



* "Achim" == Achim Bohnet <ach@mpe.mpg.de> wrote:

Achim> It's not the packaging or debian stuff that is buggy.  The
Achim> gnome apps had/have bugs.  So packaging a new version fixes a
Achim> lot of bugs (check http://bugs.gnome.org/)

After the slink update, the core GNOME packages have been rearranged
somewhat. This is a point where one has to wait if problems arrise for
this, before one should even think about throwing them at
stable. There is always also the Debian part that can cause trouble.

Why do we demand of packages for stable-updates to backport the fix
instead of just using the new release? Because it almost always also
contains new code. New code has the tendency to have new bugs.

Achim> Check the gnome-announce list.  Almost very announcement
Achim> mentions fixes of memory leaks and bugs.  The upstream
Achim> maintainers are fixing the bugs so all that the debian package
Achim> maintainers have to do is to package it.  Isn't this an ideal
Achim> situation :)

For unstable, this is OK, and this is how it works in the ideal
case. But we are talking about a set of packages, aimed as an update
for the stable distribution. 

It is not that easy. There are 160 binary packageds in the GNOME
update. Compiling every new version available in unstable for stable
is calling for trouble. Orbit is buggy. I just saw a message in
debian-user about the today-installed-in-unstable new version of gmc
dying with a segfault.

Users using stable and an updated targeted at stable want the thing
running. This is what they expect, as this is what Debian is
promissing to them.

Unstable packages can break at times. This is the ways it works, but
on a stable system, this should not happen.

Well, as I said, we can handle this the following way:

If a maintainer has a new package version and thinks it should go into
the stable update, he can upload to the staging area. If he doesn't
have a stable box, he can ask here for someone to compile the package.
This also shifts the decision to the ones with the knowledge about the
status of a package. I can't and don't track the upstream changes and
uploads of all 160 packages.

Ciao,
	Martin


Reply to: