Re: Be careful of library updates once freeze happens.
>>>>> "Ingo" == Ingo Saitz <Ingo.Saitz@stud.uni-hannover.de> writes:
Ingo> MoiN
Ingo> On Mon, Jun 11, 2001 at 12:43:56PM -0400, Sam Hartman wrote:
>> A few days ago I mentioned that I would probably end up
>> uploading a new version of PAM to experimental rather than
>> unstable. Julian Gilbey asked why and during that discussion
>> it became obvious that you want to be fairly careful how you
>> treat libraries and other depended-on packages once the freeze
>> starts.
Ingo> Yes, and those problems may be even bigger because whole
Ingo> woody is not frozen at the same time. Instead base is frozen
Ingo> first.
This is rather the entire point.
Ingo> I thinbk we even need new pool sections for each frozen part
Ingo> of debian (e.g. frozen-base, frozen-standard,
Ingo> frozen-leftovers), so developers can still track
Ingo> testing/unstable for the non-froozen parts of debian (see
Ingo> apt_preferences(5) on how to do this with apt-get).
Somehow I suspect this won't actually happen.
>> The builder might be able to be clever on their own arch by
>> building against a frozen libpam, but we have autobuilders and
>> such tricks tend not to work.
Ingo> Since new versions of libraries are guaranteed to be
Ingo> available within the freeze an probably will be included in
Ingo> unstable I think autobuilders _have_ to build against
Ingo> frozen/testing. And developers have to make sure they use
As I understand it, this simply does not work with current
autobuilders. The autobuilder will simply not build the package if
the build dependency is not satisfied, rather than pulling a newer
version out of unstable. So if you use autobuilders against testing,
you will fail to build any package that specifies a build depend
against something only in unstable. It's not clear to me this is even
a bug.
Reply to: