[Pkg-uml-devel] Re: Really reviving pbuilder-uml!
Junichi Uekawa
dancer at netfort.gr.jp
Sat Sep 16 07:54:17 UTC 2006
Hi,
I'm a bit worried about these output:
? debian/control-nouml
? debian/rules-nouml
> here we go!
> I cooked a patch that makes pbuilder-u-m-l and pdebuild-u-m-l run
> Successfully(TM) here.
>
> The main issue was the absence of devfs that needs to be dealt with.
> We need to mount /dev, create the neccessary device nodes to mount the
> FS image and immediately after bind mount ${UML_CHROOT_MOUNTPOINT}/dev
> to /dev in order to have the full /dev in place (see hunk #3).
>
> To avoid trashing the FS image IGNORE_UMOUNT="no" is needed as it
> umounts /dev/pts that otherwise keeps ${UML_CHROOT_MOUNTPOINT} busy (see
> hunk #2).
>
> Pdebuild-user-mode-linux was running pbuilder-u-m-l with a spurious
> --debbuildopts flag that seems to be meaningful for it.
>
> The rest is just a copy of {rules,control}-uml on {rules,control}.
>
> Oh, and I added ${HOME}/.pbuilder-umlrc in the search path for config
> files. Yes, a very very ugly name :)
>
> One thing I left behind (and that is worth discussing) is rootstrap
> configuration in the image creation. Rootstrap now has some more options
> that can cause small issues when running pbuilder-uml expecially:
> - the "inlcude" and "install" options that can cause additional stuff
> being installed (defeating the "clean" build environment)
> - the "kernel_modules" option that deals with additional modules could
> cause warnings on boot about modules.dep not being found.
>
> Moreover, pbuilder-uml is currently overwriting /etc/network/interfaces
> which is probably unnecessary with the proper rootstrap configuration at
> creation time.
>
> That's all!
>
> PS: I hope I'll be able to ship UML for amd64 soon.
> --
Thanks!
I'll merge them to my tree.
regards,
junichi
--
dancer@{debian.org,netfort.gr.jp} Debian Project
More information about the Pkg-uml-devel
mailing list