Re: [Libvirt-cim] [PATCH 2 of 3] SDC: returned instances are not of type specific subclass

HE> + if (rasd_classname_from_type(rasd->resource_type, &base) != CMPI_RC_OK) { HE> + cu_statusf(broker, s, HE> + CMPI_RC_ERR_FAILED, HE> + "Resource type not known"); HE> + goto err; HE> + } HE> + HE> inst = get_typed_instance(broker, HE> CLASSNAME(ref), HE> - "ResourceAllocationSettingData", HE> + base, HE> NAMESPACE(ref)); HE> CMSetProperty(inst, "InstanceID", inst_id, CMPI_chars); HE> @@ -723,6 +732,7 @@ static CMPIInstance *sdc_rasd_inst(const HE> prop_list[i].value, prop_list[i].type); HE> } HE> + err: HE> CU_DEBUG("freeing prop_list"); HE> free_rasd_prop_list(prop_list); HE> out: I didn't notice this before, but shouldn't we initialize prop_list to NULL, make sure free_rasd_prop_list() can handle NULL, and then put the free call below the out: target? There is another place in the function that looks like it jumps to out after prop_list could have been initialized. Further, it would be better to have it below out, so that additions like this one don't have to add or utilize an err: target. Heidi, I know this isn't really a part of your patch, but I think it's worth looking at while we're here. Jay, can you comment? Thanks! -- Dan Smith IBM Linux Technology Center Open Hypervisor Team email: danms(a)us.ibm.com