On Thu, Jul 14, 2016 at 08:23:18AM -0600, Eric Blake wrote:
> On 07/14/2016 08:16 AM, Daniel P. Berrange wrote:
> > Add a qmp_mixed_input_visitor_new() method which returns
> > a QMP input visitor that accepts either strings or the
> > native data types.
> >
> > Signed-off-by: Daniel P. Berrange <berra...@redhat.com>
> > ---
> >
> > NB, just a demo - this should have tests added before submitting
> > for real.
> >
>
> >
> > +static void qmp_input_type_int64_mixed(Visitor *v, const char *name,
> > int64_t *obj,
> > + Error **errp)
> > +{
> > + QmpInputVisitor *qiv = to_qiv(v);
> > + QObject *qobj = qmp_input_get_object(qiv, name, true);
>
> This consumes the key out of the QDict input stream...
I did say it was not tested :-)
> > +
> > + if (qobj && qobj->type == QTYPE_QSTRING) {
> > + qmp_input_type_int64_str(v, name, obj, errp);
>
> then calls a helper function that also wants to consume the key. You'll
> have to use qmp_input_get_object(,false) for this to work (in other
> words, do a peek instead of consume when deciding which other helper to
> use).
>
>
> > +
> > +Visitor *qmp_mixed_input_visitor_new(QObject *obj, bool strict)
> > +{
>
> I guess it's not too hard to generate 'qmp_mixed_input_visitor_new'
> instead of 'qmp_input_visitor_new' when 'autocast':true is set in QAPI,
> so that's a minor tweak to my series. Your version also has the benefit
> of not changing all existing callers to add another parameter.
Yep, that's the main reason I went for separate constructor. It'd be
equally fine to have a enum param in the existing constructor which
just picked different callbacks.
Regards,
Daniel
--
|: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org -o- http://virt-manager.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
