On Mon, Dec 18, 2006 at 08:03:37PM +0100, Holger Levsen wrote: > Hi Kurt, > > On Monday 18 December 2006 18:35, Kurt Roeckx wrote: > > I've recently seen a bug where a package is unpacked but not setup, so > > postinst didn't get run. > > Isn't that the case of a local administrator willingly (or due to a bug in > apt/dpkg) choose to change default behaviour of the packaging system? Are > there any normal circumstances when the following can happen?
The bug that made me post this is #403641, it took some time to get this. So, the problem is that because of a failure in some other package, dpkg didn't completly install everything. When then trying to remove package packages that weren't in completly installed, it failed. > > When trying to remove it, it failed in the postrm, because a file > > created in the postinst didn't exist. > > See above :) So, the bug above is actually not what I initially thought. But I think that is one way things can break too. > > It might also find other type of bugs, not sure, but it atleast looks > > like a useful test. > > What test do you propose exactly / what other bugs do you have in mind? Some things I have in mind are: - postinst gave an error, so it failed to install, then you try to remove the package. - preinst does something, postinst also does something, pre/post-rm doesn't clean things up if postinst didn't run. There might be other various that I can't think of right now. Kurt -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
