On 07.08.2013 21:39, Julian Gilbey wrote: > On Sun, Jul 14, 2013 at 02:39:41PM +0200, Arno Töll wrote: >>> * The only way to ship a package named apache2.2-common is to add a >>> Breaks header listing every single reverse dependency with correct >>> version information. >> >> That would be 100+ Breaks. I do not think that is feasible but that may >> need a wider discussion. > > How did you reach that conclusion? I looked at the current testing > distribution, and the only direct dependency on apache2.2-common is > libapache2-svn, which may go away when subversion is able to > transition to testing.
It's one now. :) It was the state as of Squeeze at the time I wrote this as we were in the middle of the transition. Now we can seriously consider doing the transition package approach. > Also, it is important to realise that without a dependency from > apache2.4 on apache2.2-common, apache2.2-common could be purged by > apt(itude) before the first apache2.4 package is even unpacked: > looking at my dpkg log, this is exactly what happened. So the > mechanism in apache2.2-common.postrm of checking for > /etc/apache2/upgrade-to-2.4-in-progress doesn't provide any benefit in > this case :-( Right. That's also why we do not use this trapdoor in maintainer scripts. > So it seems like having a dependency on a dummy apache2.2-common would > be the sensible (if annoying) thing to do. Thanks for this list. I'm short of time for the next 2-4 weeks, and unless sf beats me with it I will address all the outstanding Apache packaging issues then (or try to find a feasible solution at least). -- with kind regards, Arno Töll IRC: daemonkeeper on Freenode/OFTC GnuPG Key-ID: 0x9D80F36D
signature.asc
Description: OpenPGP digital signature