On Fri, Aug 12, 2011 at 07:53:47PM +0200, Corinna Vinschen wrote: >On Aug 12 12:56, Christopher Faylor wrote: >> On Fri, Aug 12, 2011 at 05:13:22PM +0100, Dave Korn wrote: >>>I'd just suggest on by default and remembers its state between runs. >> >>Ditto. > >That's fine, but let's remember the original questions. > >- *Shall* setup call rebaseall at all? > >- If yes, hardcoded or as postinstall script? > >As a sidenote, if by postinstall script, there's no way to influence it >with a checkbox. In theory I'd rather see a postinstall script since >that's a more flexible way to do the right thing.
Don't we still have sporadic reports of rebaseall "breaking" things? I think it was libncurses* that had problems. If so, I'd hate to deliver a broken system to people with no easy way to fix it. I can see why a postinstall script would be much easier but I think we want to have an straightforward way out if something doesn't work. I know that the checkbox is problematic though. One problem with checkbox/postinstall is that the checkbox shows up after the postinstall scripts have been run, right? As a potential workaround we could have a completely separate "runrebaseall" package which consisted of nothing but a postinstall script that ran rebaseall and peflagsall. It could go in the base category. Then the workaround if there was a problem with *baseall would be to uncheck that package when reinstalling. It would be a pain to have to remember to do that but, in theory, it shouldn't come up too often. OTOH, we could also have some sort of non-setup.exe configuration file in /etc which controlled what you want *baseall to do when setup.exe was run, i.e., what starting base address do you want or do you want it to run at all. Btw, is there an "undo" flag in rebaseall now that it stores things in a database? Does it store the old settings somewhere so that you can revert to the previous base address settings? cg