On Mon, Apr 19, 2021 at 05:42:01PM +0200, Andrea Bolognani wrote:
> On Mon, 2021-04-19 at 11:45 +0200, Michal Privoznik wrote:
> > On 4/15/21 2:52 PM, Andrea Bolognani wrote:
> > > the latter is also very small and, just like virsh,
> > > is something that you likely want to have available on any
> > > virtualization host to help with management and debugging
> > > tasks.
> >
> > Exactly! Host, not client. Not every virt client is virt host too. Also,
> > what's stopping management from installing libvirt-admin among with
> > other RPMs? It's tiny, as you say. But I think I've expressed my concern
> > enough.
>
> Okay, let's not have virt-admin as part of -client then. Can we at
> least have it in -daemon as opposed to its own package? And while
> we're at it, we can move there also the other commands we agreed are
> probably not a good fit for -client.
>
> virt-xml-validate just performs validation on the input XML, so
> having it in -client seems reasonable; on slighly closer inspection,
> virt-pki-validate can validate not only the server-side PKI situation
> but also the client-side one, so again -client is probably the most
> sensible home for it.
Yep, the RNGs are shipped with the library, so server can use tem
already.
> virt-host-validate is intended for the hypervisor host, so we can
> move it to -daemon; libvirt-guests should go there too, because it
> should live where the daemon does.
virt-host-validate clearly is daemon only. libvirt-guests is a little
odd because in theory it doesn't depend on the daemon - it can work
with stateless drivers. IIRC thats why it is in the clients package.
I'm sceptical that anyone will ever even think about using the
libvirt-guests script to automate startup/shutdown of VMware/Hyperv/Etc
guests though.
> What about the systemtap probes? Do they belong in -client, or maybe
> they should go into -libs? I'm not familiar with their use.
Probes can be compiled into pretty much any binary artifact - daemon,
virsh, libvirt.so, driver modules. -libs is the common place to all
of those.
> On Mon, 2021-04-19 at 11:45 +0200, Michal Privoznik wrote:
> > On 4/15/21 2:52 PM, Andrea Bolognani wrote:
> > > The former is ridiculously tiny and doesn't have any use on
> > > its own, so it hardly warrants the overhead of an additional
> > > package;
> >
> > The idea was, that if you'd install virt-admin only, you would have bash
> > completion too, hence a separate package.
>
> On Mon, 2021-04-19 at 10:57 +0100, Daniel P. Berrangé wrote:
> > Per discussion on the previous version though, we better to have
> > installed any time the daemon is installed and so if we do that
> > it means we also want the bash completion separate from the client.
>
> I would really like to get rid of the -bash-completion package.
>
> $ rpm -ql libvirt-bash-completion
> /usr/share/bash-completion/completions/vsh
> $ du -sh /usr/share/bash-completion/completions/vsh
> 4.0K /usr/share/bash-completion/completions/vsh
> $ du -sh .../libvirt-bash-completion-7.2.0-1.fc32.x86_64.rpm
> 8.0K .../libvirt-bash-completion-7.2.0-1.fc32.x86_64.rpm
>
> The completion file itself uses as much disk space as the RPM
> metadata for it. With that in mind, can we just install the file
> twice instead of having a single file with two additional symlinks
> pointing to it?
>
> We could also turn it into a .in file so that instead of
>
> #
> # virsh & virt-admin completion command
> #
>
> _vsh_complete()
> {
> # ...
> } &&
> complete -o default -o filenames -F _vsh_complete virsh &&
> complete -o default -o filenames -F _vsh_complete virt-admin
>
> the installed file for each command looks like
>
> #
> # @cmd@ completion command
> #
>
> _vsh_complete()
> {
> # ...
> } &&
> complete -o default -o filenames -F _vsh_complete @cmd@
>
> Basically, if we can't avoid the overhead of duplication altogether,
> I'd rather pay it on the filesystem than in the RPM database, where
> it's felt over and over again when dowloading repodata and when
> performing package management.
I think this worrying about micro-optimizing a problem that doesn't
really have any real world impact. 8k is lost in the noise of an
RPM transaction.
Regards,
Daniel
--
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org -o- https://fstop138.berrange.com :|
|: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
