On Mon, 2020-06-29 at 12:01 +0200, Michal Privoznik wrote:
On 6/29/20 11:58 AM, Andrea Bolognani wrote:
> On Mon, 2020-06-29 at 09:43 +0200, Michal Privoznik wrote:
> > +++ b/src/qemu/qemu_validate.c
> > @@ -1028,6 +1028,12 @@
qemuValidateDomainDeviceDefZPCIAddress(virDomainDeviceInfoPtr info,
> > return -1;
> > }
> >
> > + if (virDeviceInfoPCIAddressExtensionIsWanted(info)) {
> > + virReportError(VIR_ERR_XML_ERROR, "%s",
> > + _("Missing uid or fid attribute of zPCI
address"));
> > + return -1;
> > + }
>
> Throwing an XML_ERROR here doesn't seem right, because from the
> user's point of view it's perfectly fine to only provide a partial
> zPCI address: by the point we get here, we've already gone through
> the PostParse callback, so if either attribute is missing that's
> because of a developer mistake.
>
> Do we even need this check at all? I'm the one who asked for it to
> be introduced, but now I'm questioning that :)
>
Maybe we don't need it then. Regardless, it doesn't belong into XML
formatter. Should I post a v2 where the check is removed?
I think so, yes. If someone finds a flaw in my reasoning, we still
have a week to change it back.
--
Andrea Bolognani / Red Hat / Virtualization