Hello,

divoplade writes:

> Hello,
>
> I tried to add my comment to issue 31365, but it has been archived and
> made read-only.
>
> This issue is not fully fixed. I’m using gnome-boxes to create virtual
> machines, and I noticed something similar to what is experienced in
> this issue.
>
> The hard-coded path is in /gnu/store/xxx-profile/, so I think it has
> been found in /run/current-system/profile/bin, but the symlink has been
> resolved too early.
>
> I have recently gc’ed my system, and that’s when I noticed I could not
> run my VMs anymore.
>
> My workaround is to replace the hard-coded paths in the XML files.

Ouch :-/.

Do you know at which guix commit this happened?  I'm wondering which
version of libvirt triggered this.  AFAICT, right now, if you create a
fresh VM using gnome-boxes, it initializes the xml config with the qemu
in /run/current-system/profile, which looks like the right thing to do.
I've just tested this downloading a fedora image via gnome-boxes, using
either libvirt 5.8 or 7.2 under the hood (updated just now).

Could it be that you created a VM before the bug was fixed? Which would
have left a xml config that refers to GC-able qemu in .config/libvirt,
which a guix update cannot automatically migrate (although one day it might!).

Let me know if I've got this wrong.

Thanks,
Pierre