* Steve Langasek <vor...@debian.org> [110405 20:29]:
> > I think it might be nice if those two aspects could be isolated somehow.
> > This could also reduce the size of some build chroots and the set of
> > packages
> > any boot-strap code has to handle specially[1]. With all the essential
> > stuff only needed for a full system to boot, those are larger than they
> > needed to be.
> >[...]
> > and their dependencies (passwd, initscripts, the whole pam stack)
> > are mostly not needed in that set[2].
> > (Util-linux might have one or two programs one might want to move
> > to another package then, and something for update-rc.d needs to be
> > done).
>
> I think this is a false optimization. How does reducing the set of packages
> in a buildd chroot help anything? A typical package has build-dependencies
> many times the size of the Essential set.
It might not be so big a saving for build chroots sizes (though I guess
it will in the mayority of cases be more than 5%). For package install
tests I guess it would be much more significant.
More importantly is the boot-strap code. Currently everything in the
required set gets a special handling, must be unpackaged twice and is
supposed to work in quite a special situation (all files existing in the
file system but no pre/postinst yet run, ...).
Such a step would almost reduce the packages to a half (something like 61 to
33).
Bernhard R. Link
--
To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive:
http://lists.debian.org/20110406072426.ga31...@pcpool00.mathematik.uni-freiburg.de
