Re: patch to split fsck
On Wed, May 12, 2004 at 03:34:52PM +0200, Robert Millan wrote:
> tags 111651 patch
> thanks
>
> I'm attaching a patch that splits out fsck and adds the proper relationships
> in debian/control.
>
> debian-boot people: Please drop any comments concerning the udeb, if
> applicable; and take the appropiate measures in d-i, if any.
The dependency structure of the base system has been frozen since 29
March so that d-i images stop getting repeatedly broken by
rearrangements like this one. It's too late to do this for sarge.
> diff -Nur e2fsprogs-1.35.old/debian/control e2fsprogs-1.35/debian/control
> --- e2fsprogs-1.35.old/debian/control 2004-05-12 13:50:03.000000000 +0200
> +++ e2fsprogs-1.35/debian/control 2004-05-12 14:22:38.000000000 +0200
> @@ -5,6 +5,16 @@
> Build-Depends: texi2html, gettext, texinfo, dc, debhelper (>= 4)
> Standards-Version: 3.6.1
>
> +Package: fsck
> +Essential: yes
> +Depends: ${shlibs:Depends}, libblkid1 (= ${Source-Version}),
> + libuuid1 (= ${Source-Version}), fsck-backend
Essential packages generally need to use Pre-Depends, since they need to
work even when unconfigured.
> Package: e2fsprogs
> -Essential: yes
> Pre-Depends: ${shlibs:Depends}
> -Depends:
> +Depends: fsck
> +Provides: fsck-backend
This allows the new e2fsprogs to be unpacked before fsck, at which point
/sbin/fsck will disappear until fsck is unpacked, which is a
showstopper. It's very hard to solve this and avoid breaking fresh
installs at the same time.
I think we'll need something like:
Package: fsck
Essential: yes
Pre-Depends: ${shlibs:Depends}, libblkid1 (= ${Source-Version}), libuuid1 (= ${Source-Version})
Depends: e2fsprogs | fsck-backend
Replaces: e2fsprogs (<< first-split-version)
Package: e2fsprogs
Pre-Depends: fsck
Depends: ${shlibs:Depends}
This guarantees that fsck remains available during upgrades no matter
what, and ensures that fresh installs are at least possible. This isn't
optimal, but, given that you have to install one of fsck and a
fsck-backend after the other, I don't see a better possibility.
Aside from the base dependency freeze, I think the complexity of this
change alone marks it as post-sarge.
Cheers,
--
Colin Watson [cjwatson@flatline.org.uk]
Reply to: