On Tue, 2005-08-30 at 16:46 +0200, Holly Bostick wrote: > Jerry Turba schreef: > > As I understand the process etc-update lists new configuration files > > provided by the program authors. I have tried to define some rules for > > myself to determine how to handle these new files. > > > > 1. If I made a change to a file I will never allow the new config file > > to overwrite the old file. > > I disagree. Certainly there are some 'new' config files that you should > never, ever allow etc-update to overwrite, such as /etc/fstab. However, > if the format of the config file has been changed in the meantime, some > of the settings in the old config file may be invalid, and new, valid > default settings (for areas that you have not changed) will not be added. > > This is what the '3' option is for, after the changes have been > displayed: 'Interactively merge update with original'. > > I use this in those cases to preserve those settings that I want to > keep, while upgrading the config header, comments, and other settings to > the new defaults. > > In those very rare cases where the line ordering has changed so much > that the diff utility would overwrite one or more settings, I accept the > new file, and immediately edit it with nano to change the (usually) one > or two lines that were 'wrongly' diff-ed. > > > > > 2. If the new config file is a new default file I will accept the new file. > > Agreed. > > > > > 3. I will never change a file that is program code, (I am not a > > programmer). > > Agreed. > > > > I have tried dispatch-conf but I > > still have to make the same decisions. Am I missing something? > > Not really; that would be Gentoo. Decision is not meant to be taken out > of your hands. But the power to choose how your system is configured > carries the responsibility to pay attention to the offered changes and > think about their effects (which means you have to know what their > effects are going to be, which means you have to learn wtf is going on > on your system in the first place). > > Holly
While I agree that etc-update is a vast improvement over other package systems, it would be nice to have a CVS type merge where I only have to make choices when the "system can't figure it out". It seems like etc-update (and friends) should be able to take advantage of mtime metadata and md5 checksums to determine if I've made any modifications to the default config file. That way an unmodified default config from version N can just safely be replaced with the new default for version N +1. Does this functionality already exist with the current etc-update? Eric C. -- gentoo-user@gentoo.org mailing list
