On Mon, Oct 08, 2018 at 13:04:17 +0200, Andrea Bolognani wrote:
On Mon, 2018-10-08 at 12:59 +0200, Peter Krempa wrote:
> On Fri, Oct 05, 2018 at 17:18:44 +0200, Andrea Bolognani wrote:
> > On Fri, 2018-10-05 at 16:43 +0200, Ján Tomko wrote:
> > > On Thu, Sep 20, 2018 at 05:25:26PM +0200, Andrea Bolognani wrote:
> > > > @@ -756,6 +755,13 @@ virQEMUCapsInitGuest(virCapsPtr caps,
> > > > */
> > > > binary = virQEMUCapsFindBinaryForArch(hostarch, guestarch);
> > > >
> > > > + /* RHEL doesn't follow the usual naming for QEMU binaries
and ships
> > > > + * a single binary named qemu-kvm outside of $PATH instead */
> > >
> > > This does not look like something upstream libvirt should worry about.
> >
> > I would be simple enough to turn this into a downstream-only patch;
> > however, by doing so we would lose the ability to run unchanged
> > upstream libvirt on top of RHEL, so I think it's preferable to
> > leave it alone.
>
> Well, if we are going to bend the rules and make a hack for the mess
> some downstream distro made, then we should keep the original code which
> allows to add mess for other distros as well.
The downstream path names I dropped are no longer necessary because
Debian and friends are using the standard binary names in all
releases we support; the same is not true of RHEL, which is why that
one path has not been removed along with the rest.
That is perfectly fine with me. I'm arguing that we should keep the
whole function that iterates the array of override paths even if it
contains the exception only for RHEL.
Doing a tailored special case for RHEL seems like we are favoriting it
somehow which we should not. So either the loop is deleted without
replacement and RHEL maintainers fix their mess or we keep it as-is so
that other distros can add their mess as well.