Eric Blake <ebl...@redhat.com> writes: > The testsuite was the only client that attempted to reuse a > QmpOutputVisitor for a second visit after encountering an > error and/or calling qmp_output_get_qobject() on a first > visit. The next patch is about to tighten the semantics to > be one-shot usage of the visitor, like all other visitors > (which will enable further simplifications down the road). > > Signed-off-by: Eric Blake <ebl...@redhat.com> > --- > > This commit is designed to completely replace commit 51750617 > on Markus' qapi-next branch; all other patches in that series > remain unchanged and the testsuite still passes. Doing this > now will make it easier for my next series to introduce a > generalized visit_free(Visitor*) instead of our current > duplication of a per-subtype foo_visit_cleanup(FooVisitor*), > without having to revert the addition of qmp_output_visitor_reset().
Patch looks good to me. I massaged qapi-next as instructed. Thanks!