On Fri, Nov 14, 2003 at 09:32:48AM -0600, Manoj Srivastava wrote: > On Fri, 14 Nov 2003 10:46:35 +1100, Matthew Palmer <[EMAIL PROTECTED]> said: > > the user hasn't used debconf to make the change. To cover that > > circumstance, I've started to make a file containing an md5sum of > > the config file, and before I change it I make sure the user hasn't. > > If they've changed it at all (so the md5sum has changed), I don't > > make any modifications and it's then the user's problem to keep it > > up to date. It's by no means perfect, but it follows Policy's "must > > keep local config changes", and doesn't require pre- and > > post-knowledge of what the options were in Debconf. > > This is suboptimal. If there are changes that would be made > to a file if the user has not modified the file, then the user should > be offered the choice to accept these changes even if they have > modified the configuration file on their own. The "my way or the high > way" approach is a dissservice to our users.
OK, so I'll add a question asking part to my script. It doesn't mean the entire concept is flawed. > Check out ucf; it has been specifically written to allow you > to offer the user a choice, exactly like dpkg does for conffiles. But it's still "all my changes or all your changes", right? Which is "my way or the highway" asked on every upgrade... <g> Anyway, last time I looked at ucf it's documentation spent a lot of time discussing esoteric examples and absolutely no time showing me how to apply it to my packages. On a hunch, I just went and installed the latest version of ucf, and it appears that the documentation has morphed into something useful, so my homegrown hack's days may be numbered. - Matt -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
