considerations about queues

[Andreas Barth <aba@not.so.argh.org>]

Hi,

we're getting slowly more and more queues. What could (theoretically)
happen today is that a package is pointed at proposed-updates, has a new
package in it and has something that needs to be installed by hand.

Currently, the question in which queue a package goes is totally
hardcoded into both jennifer and lisa. This also means that any
additions need to be done in both of them.

My idea would be to make a generic "needstoqueue" variable that could
contain e.g. {'new'} if it only goes to the new queue or {'new',
'byhand'} or {'p-u-new'} - or whatever is appropriate. If a tool is done
with handling a package, it just looks up what is first in that
variable, pushes the package there (also by a generic "what directory
exists, and what permissions should files there have" - nice for places
like p-u-new :) and removes that location from the variable. If there is
no directory left anymore, the next destination is the accepted queue.

On writing this down, a second idea came to my mind - currently, if a
source package exists in new or byhand (or accepted), this is enough for
jennifer - but not if the source exists in p-u-new (and I don't know how
it behaves with the different queues on security.d.o). One could replace
that with a loop that goes through some list of directories in
katie.conf.


Comments on these ideas? Of course, the final ok can only come after a
patch exists, ... :)


Cheers,
Andi
-- 
  http://home.arcor.de/andreas-barth/