On Mon, Apr 8, 2013 at 3:43 AM, Zeeshan Ali (Khattak) <zeeshanak@xxxxxxxxx> wrote:
On Sat, Apr 6, 2013 at 9:32 AM, Fabiano Fidêncio <fabiano@xxxxxxxxxxxx> wrote:
> Howdy!
Hi Fabiano,
Hi Zeeshan,
One important thing to note is that according to the documentation,
> Since we start using libosinfo to provide the install-scripts for
> gnome-boxes we have a pending task related to the command-line passed to
> load the install-script.
> Nowadays, the command-line hardcoded for Fedora format and it is okay, once
> we only support unattended-installations for Fedora's flavors of Linux. But
> the family is growing fast, fortunately.
>
> A few days ago Zeeshan found out how the unattended-installations work on
> Debian/Ubuntu. A few hours ago I found out how it works for OpenSuSE.
>
> So, our current cases are:
> - Fedora/RHEL/CentOS: ks=hd:<device>:/<filename>
> - OpenSuSE: autoyast=device://<device>/autoinst.xml
> - Debian/Ubuntu: decompress the initrd, put the file into it, compress again
> and boot with this fresh compressed initrd.
>
> The 3rd method is also supported by Fedora/RHEL/CentOS, but is not by
> OpenSuSE).
> Ah, so bad, looks like we don't have a standard way to follow.
the same method we use for Fedora should also work for Debian/Ubuntu
so this initrd repacking method is a work around to a bug. We should
file a bug on Debian/Ubuntu about this and use this workaround in Apps
for now. Libosinfo should simply provide working installer scripts for
Debian/Ubuntu.
I don't know if it's supposed to work in the same way that Fedora does.
After your successfully with Debian I re-read the docs and they are not exactly clear about that. I could understand that the expected file is the new initrd, with the preseed file.
Yup. Sounds right.
> I'm going to add OpenSuSE install-scripts for Libosinfo and I think we will
> want to use these scripts on Boxes.
What would '!' say to app? Do they substitute it with 'ks' or
> Thinking a bit here, at least for Fedora* and OpenSuSE, we could add a
> property in the install-scripts, "command-line-format" to specify the
> format. Something contenting:
> - installation method (ks, autoyast) -- could be represented by "!"
'autoyast'? Why not just give them 'ks' or 'autoyast' itself?
Answered above, my bad.
Again why use a variable and how does app know what it means for each script/os?
> - installation method option (hd, device) - could be represented by "@"
Same here.
Now is this something that really should come from app but thing is
> - device where the script will be put in (sda, sdb) - could be represented
> by "#"
that app is already providing this in installer config.
What I suggest is that we also use XSL template (just like the script
template) for generating this commandline for the app. App will pass
the same install config as it will for generation of scripts.
I will try to draft something in this way in the next days and then I will return to this discussion with a patch. I think the things are clear enough to start a draft
Thank you.
Best Regards,
--
Fabiano Fidêncio
_______________________________________________ Libosinfo mailing list Libosinfo@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libosinfo
