> But not those that you would have as a part of the "installer OS"
But we do not want to have an "installer OS". We want to use the exact
same tools as the installed system. We've spent years getting away from
having our own code with its own bugs. We much prefer using the same
system tools and sharing bugs.
> Somehow I doubt that the command used by those utility change that
> often so is it really that frequent?
>
> Can you give me any examples when that has happened.
The most recent example I can think of was from October 30:
commit 3f0a1f9668f42dabe5ec728a76347e25cdc551d7
Author: David Lehman <dlehman@xxxxxxxxxx>
Date: Tue Oct 30 12:35:00 2012 -0500
Apparently necessary kpartx changes (#867593)
Don't always force kpartx to use a delimiter. The problem is that there
are all these places that could be triggering creation of the partitions
and they need to all use the same rules for when to use a delimiter
between disk and partition number. Otherwise, how can parted know when
to use a delimiter when adding a partition to a dm disk? You'd think the
answer would be "always use a delimiter" or "never use a delimiter" but
that's not how the world works.
Also, pass -s to kpartx to wait for devices to be created before
returning since udev should be doing all the work nowadays.
Just search through git history for udev, udev, systemd, lvm,
NetworkManager, and so on. This kind of thing happens all of the time.
> Afaik the method to do that is unchanged since the get go ( setting
> the default runlevel in systemd ).
But we didn't always have systemd.
> Anyway what steps remain to isolate the installer OS from the
> "installed OS " so we can have better contingency plan in the future
> for the installer ?
No, this is absolutely not the direction we will be moving in.
- Chris
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
