Hi, Thanks for the quick feedback.
Wookey wrote: > Having this package M-A: same is genuinely useful because to > cross-build cups, for example, which is quite low in the bootstrapping > tree, it is needed in both host and build arch versions. So I'd like > to avoid reverting this is possible. Would it be feasible for people bootstrapping to use the package from experimental in the meantime? > +++ Jonathan Nieder [2012-07-22 10:54 -0500]: >> b) Manage /etc/papersize in a separate libpaper-common package. The >> configuration file's name doesn't vary with the soname so the >> current state seems like a policy ยง10.7.4 violation waiting to >> happen anyway. > > Do we need to do this for every M-A: same lib that has a config file? If I remember correctly, refcounting works fine for conffiles. That would mean that it's only explicitly managed configuration files that would be affected. But keep in mind that this was already a policy violation. What happens when libpaper1 and libpaper2 are installed at the same time? The usual case where the supporting file's name incorporates the soname version is easier and can be taken care of by adding the triplet to the path. [...] >> c) Guard the code that purges /etc/papersize with a test that >> libpaper1 is already purged for all other architectures. > > That seems the most attractive suggestion to me. Perhaps it should be > standard practice? For the reasons described above, I hope not. I really think that (a) is the right fix for wheezy and (b) for wheezy+1. Compare http://bugs.debian.org/682425 where the ABI is unlikely to change (the only other case of this bug that looked severe enough to file right away). Context: https://lists.debian.org/debian-dpkg/2012/07/msg00098.html Thanks, Jonathan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
