On Tue, 13 Apr 2010 10:38:07 -0400 Jim Heck <jsurf@heckheck.com> wrote: > On Tue, Apr 13, 2010 at 9:32 AM, Neil Williams <codehelp@debian.org> wrote: > > > > Does this mean that there won't be an opportunity to keep these scripts > > > granular per package? > > > > Better to have scripts that are granular per device variant. > > > I was suggesting that instead of having a single monolithic script per > variant, to have an aggregated set of script fragments each specific to a > package that get gathered and run en masse. You are correct that the > overall combined contents of the scripts should be tightly tied to > processing a given variant, but the flexibility I refer to is to have a > fragment stored such that it can be reused across variants (possibly by > simply copying it to a different variants script directory), to comprise the > full scripted set of actions working on the root filesystem. That is up to each developer to decide. Multistrap is gaining support for "cascading configuration" which basically means that configurations can share basic config settings. These settings include which scripts to use. There's no reason why your scripts cannot be configured to call different sub routines depending on which frontend is called. How the scripts are arranged is independent of the rest of the rootfs build. All that's needed is that multistrap can be given a single filename to run for this particular run. Aggregating on a per package basis doesn't make sense to me. Makes more sense if the variants that use a particular package all use the same scripts - no point writing the same script multiple times to do the same thing for different packages. I think we're talking about different levels of granularity - package maintainer scripts are NOT that disparate. Indeed, maintainer scripts prepared by debconf are likely to be identical for lots of packages. It makes no sense to have dozens of package-specific scripts that all call ldconfig, for example, yet there are hundreds (possibly thousands) of maintainer scripts that call ldconfig and many do nothing else. > I meant a fragment script (off box) that was dealing with the steps > necessary to integrate busybox into the root filesystem. Since it is > conceivable that busybox might be a package used in every system, I was > suggesting that other generalized stuff (the boilerplate that might exist in > a monolithic script) needed to handle system setup could be conveniently > piggybacked into this packages script fragment. Perhaps that is gross, and > having a generalized variant master script would be much more elegant. Such > a master script could invoke the execution of the individual fragment > scripts that are specific to each package. No, it's not gross, it makes sense because busybox is quite unusual in the amount of configuration and the differences in configuration commands compared to any other package. Just don't expect other maintainer scripts to be as complex. :-) -- Neil Williams ============= http://www.data-freedom.org/ http://www.linux.codehelp.co.uk/ http://e-mail.is-not-s.ms/
Attachment:
pgptWeaTMmO3S.pgp
Description: PGP signature