Paul Mackinney wrote: > Thanks for posting this! It kept my down-time < 5 minutes. > > The moral: when something big goes wrong after an apt-get upgrade, try > again, then try deb-user.
If you can still get to it with a b0rked system. Just imagine if your only mail and web clients had been written in C++... or if it had been the C, rather than C++, library that was unusable... The real moral: if you don't have a good chance of figuring out what's wrong on your own, and fixing, backing out of, or jury-rigging around it without outside help... don't run unstable. It's unfortunate that this happened, but this is unstable, and developers aren't perfect. I'm actually less disturbed that this happened than that so many people have posted desperate "please help" messages about it, apparently unable to figure out on their own either to downgrade libstdc++2.10-glibc2.2 or create a symlink. I mean, seriously, let's look at the process in detail... First, you run a package manager to upgrade Sid. Hopefully you do this fairly often (I try to do it every day) to minimize the number of packages affected by each update. (It might also be considered smart to do this daily update several hours after the mirrors are updated, so that if there are any catastrophic problems, you will be forewarned about them from debian-user.) While the package manager is installing the new packages, suddenly you start getting errors (from update-menu, IIRC?) about a missing library, /usr/lib/libstdc++-libc6.2-2.so.c (IIRC). So most likely something has just now happened to cause this problem -- probably a bad package. And indeed, one of the packages you have just updated is libstdc++2.10-glibc2.2. Seems like a likely candidate, no? So you look in /usr/lib, and the missing library is indeed missing. Not good. Looks like the file's missing in the new package. Now, since all the other /usr/lib/libstdc++-libc6* files are just symlinks, you could try making a similar symlink to see if that fixes the problem. There's even one symlink that looks like it has a typo -- libstdc++libc6.2-2.so.3 (missing the - before libc6). So you could try fixing that (or, to be safe, creating another one with what you think would be the right name). Or you could try to just back out of the problem like this: # dpkg -i --force-downgrade /var/cache/apt/archives/libstdc++2.10-glibc2.2_1%3a2.95.4-12_i386.deb because OF COURSE, since you run unstable, you keep at least one previous version of every critical package locally, since you can't be sure you'll be able to use the network in the event of a catastrophic problem. IMHO, the above process is not terribly complex, does not require kernel-hacker skills, and solves a problem that is minor compared to some of the catastrophes that have befallen Sid in the past. If someone can't fix this one, then they'll really be in trouble if something _really_ bad happens -- in which case maybe they should consider running testing instead of unstable, for their own sake. I'm really not trying to be 'leet here or sneer at people who can't match my skills. Seriously, for your own sake (and this is the generic "you", I'm not addressing Paul specifically, as I don't know what he can or can't do for himself), why run unstable, which is _intended_ as a place for leading-edge testing -- "catch it here before it breaks something really important" -- if you aren't able to deal with the problems that sooner or later _will_ arise? Craig -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
