On Fri, Aug 26, 2022 at 07:39:31AM -0500, Eric Blake wrote:
On Fri, Aug 26, 2022 at 01:39:05PM +0200, Laszlo Ersek wrote:
> Extract and somewhat generalize the recipe for the $(PHYSICAL_MACHINE)
> target to a separate shell script. In preparation for the multiple steps
> we're going to introduce later, redirect virt-builder to a temp file at
> first (placed in the same directory as the finally expected disk image),
> and rename that file upon success.
>
> Signed-off-by: Laszlo Ersek <lersek(a)redhat.com>
> ---
> +++ b/make-physical-machine.sh
> @@ -0,0 +1,37 @@
> +#!/bin/bash -
See the response in the other thread about not needing the - here.
Are we sure that /bin/bash is on all systems where this script will be
run, or is it better as '#!/bin/env bash'?
Realisticly virt-p2v only makes sense on Linux, both when used to
prepare the P2V ISO, as well as the environment in which the ISO
boots.
eg: Using virt-p2v on (say) FreeBSD to prepare a Linux ISO? The
FreeBSD folk would prefer that the built ISO contained FreeBSD, but
that would require a small mountain of work.
> +
> +set -e -u -C
My personal opinion is that 'set -e' is a crutch that should be
avoided because of its unintended interaction with functions; but I'm
not adamant enough about it to tell people to rip it out of scripts.
For short scripts, like this one, it's easy enough to check that we
aren't tripping over any of -e's surprises.
I'm confused too: what's the problem with set -e? Compared to _not_
having set -e which allows scripts to continue running after errors.
Rich.
--
Richard Jones, Virtualization Group, Red Hat
http://people.redhat.com/~rjones
Read my programming and virtualization blog:
http://rwmj.wordpress.com
nbdkit - Flexible, fast NBD server with plugins
https://gitlab.com/nbdkit/nbdkit