Hi, Sven Joachim wrote: On 2009-04-27 17:41 +0200, Daniel Burrows wrote: > On Mon, Apr 27, 2009 at 04:23:50PM +0200, Javier Barroso > <[email protected]> was heard to say: >> Daniel Burrows wrote: >> > Could you please send the output of: >> > >> > $ /usr/sbin/update-alternatives --display aptitude >> > >> With aptitude 4.11 >> # LANG=C /usr/sbin/update-alternatives --display aptitude >> aptitude - status is manual. >> link currently points to /usr/bin/aptitude-curses >> No versions available. > > Would I be guessing right if I said that you downgraded to aptitude > 0.4.11.1? Did you have aptitude 0.5.1 installed at some point, then > downgrade, then upgrade again?
I would think so; I once upgraded to 0.5.1, found it unusable due to a broken apt-xapian-index and downgraded to 0.4.11.11. Now I have this left-over alternative: ,---- | % LANG=C /usr/sbin/update-alternatives --display aptitude | aptitude - status is auto. | link currently points to /usr/bin/aptitude-curses | No versions available. `---- > It sounds like something in the upgrade path is broken, but I'm not > sure what. Maybe it's just not possible to safely replace a file with > an alternative :-/; I thought it might work since the symlink isn't > known to dpkg (so that whole bag of issues shouldn't come up), but it > sounds like update-alternatives is getting confused. It seems to be http://bugs.debian.org/87677, claimed to be fixed in dpkg 1.15.0. Javier, what dpkg version do you use? dpkg from unstable (~ 1.14.26) wou ! that bug 87677 was reported in 2001, and is now fixed, and only has one comment (I see it curious) > Hrm: actually, I think a downgrade would explain this. aptitude > removes the alternative in prerm when the package is removed, but not > when it's upgraded. Downgrades invoke prerm with the arguments > "upgrade new-version". So probably aptitude should be checking for > a new version earlier than 0.5.1-1 and removing the alternatives link > if that's the case. Otherwise the alternatives symlink will be > clobbered but the alternatives system won't know that, and I bet that > leads to what you saw. That makes sense to me. To me too Sorry for the mail (replying from webmail compose sucks!) -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
