On Fri, 9 May 2014 22:54:08 +0200 Christian Hofstaedtler wrote: [...] > * Francesco Poli <invernom...@paranoici.org> [140509 19:20]: > > On Fri, 9 May 2014 11:39:05 +0200 Christian Hofstaedtler wrote: > > > please add a failure catcher to apt-listbugs, so that in common > > > configurations apt doesn't break during an upgrade of ruby and > > > related packages. > > > > Could you elaborate a bit, please? > > Yeah, sorry for writing this in a not really clear way. > > > Do you think that apt-listbugs should realize that it's failing > > to load the debian_version (binary) module and just exit with > > zero status without doing anything? > > I /think/ in cases where it's hopeless for apt-listbugs to start > working again, it should just exit with zero.
But, as I said, this would mean that, whenever apt-listbugs cannot run, the users are left without any protection against unsafe upgrades that might introduce RC bugs into their systems! What is worse, the users could even fail to notice that apt-listbugs is not running at all, and go on upgrading and installing packages with a false sense of safety granted by the fact that apt-listbugs is installed on their systems... I am thinking about a possible alternative way to handle this situation: maybe apt-listbugs could catch the LoadError exception and write some suggestion for the user on how to fix the situation, and *then exit with non-zero status*... Nonetheless, I would strongly prefer seeing the ruby upgrade path become more robust, as I have already told you. [...] > > I want to stop the upgrade immediately, examine the situation, > > fix the minimum needed to let apt-listbugs run correctly, > > and *only after that* proceed with the upgrade! > > > > I am of course open to any suggestion to improve apt-listbugs, > > as long as it takes this need into account. > > I do agree with you on basically everything you said above; altough > I'm concerned about users that somehow use apt-listbugs because it's > there, not because it was their concious choice. This looks really weird to me. AFAICT, apt-listbugs is useful on Debian testing and unstable systems, above all. It's definitely less useful on stable or oldstable systems, if useful at all. I would expect that a user who runs testing or unstable has at least a little knowledge about which leaf packages are installed on his/her system... Maybe I am being naive here! > Now I don't know how they end up with that -- I haven't installed > Debian using d-i in a long time, so I might have missed any defaults > set by d-i -- I am not aware of any tasksel choice that installs apt-listbugs. But I usually only install the base system with d-i, without selecting anything from tasksel, and then I go on installing the packages I need with aptitude, so I might be wrong... > but apparently there are users that can't fix > temporary issues such as this one on their own. I agree that it seems so. [...] > > As I have previously said in the past, I think that some Breaks > > would be highly useful to make the upgrade path more robust... > > Please consider addressing the issue from this side. > > Yes, I agree; ruby (from src:ruby-defaults) needs to Break > ruby-debian, but I didn't think of that -- will fix that (hopefully > tomorrow). That's great, thanks for addressing this issue! > > > I hope this line of reasoning makes sense to you. > > Thanks for the great work on Ruby packaging! > > Yep. > > Thanks for your insight. Most of the Ruby packaging is really > Antonio's work though :-) It goes without saying that I am very thankful to Antonio, as well. -- http://www.inventati.org/frx/frx-gpg-key-transition-2010.txt New GnuPG key, see the transition document! ..................................................... Francesco Poli . GnuPG key fpr == CA01 1147 9CD2 EFDF FB82 3925 3E1C 27E1 1F69 BFFE
pgp6j9OE1fd26.pgp
Description: PGP signature
