Apparently, though unproven, at 13:50 on Saturday 23 October 2010, daid kahl did opine thusly:
> > Don't worry about it. I'm not sure if portage-2.1.9.20 will deal with > > this automagically (I *think* it does these days and 2.2 definitely > > does) but if not just > > > > emerge -C shadow ; emerge -1 shadow > > > > then emerge -avuND world. > > > > No good technical reason for doing shadow first apart from getting it > > over and done with while you watch and confirm it works fine. Then do > > world and wander over to the kettle letting portage go on with doing > > it's thing unattended > > For my own comfort, on a case like this, if I didn't have the portage > FEATURE buildpkg or buildsyspkg turned on, I'd make sure that was on > and that I had a functional backup of shadow to install from binary, > in case something went very wrong. But I tend to be extremely > cautious in terms of how I maintain my system, and a lot of that > caution is just paranoia. Just because you're paranoid doesn't mean they're NOT out to get you :-) Or in this case, it doesn't mean it's not justified. I now have buildpkg enabled for @system - everything else I can re-run emerge to fix. After watching portage break python *twice* exactly a year apart, watching the exciting developments in python-3, after some horrendous shadow breakage 3 years ago and the convoluted upgrade path for bash 2 years ago, and someone's b0rked commit of glibc-2.12 to the tree quite recently, I feel entirely justified in keeping binary copies of @system around. It long ago stopped being paranoia and started being good old common sense (right up there with backups). -- alan dot mckinnon at gmail dot com
