[Libvirt-cim] [PATCH] [RFC] [CU] Turn std_indication's awesome knob to eleven

[Jay Gagnon]

Dan Smith wrote:
> JG> +char *classname_from_inst(const CMPIBroker *broker,
> JG> +                          CMPIInstance *inst,
> JG> +                          CMPIStatus *s)
> JG> +{
> JG> +        char *ret = NULL;
> JG> +
> JG> +        CMPIObjectPath *ref;
> JG> +        ref = CMGetObjectPath(inst, s);
> JG> +        if ((s->rc != CMPI_RC_OK) || CMIsNullObject(ref)) {
> JG> +                cu_statusf(broker, s,
> JG> +                           CMPI_RC_ERR_FAILED,
> JG> +                           "Could not get objectpath from instance");
> JG> +                goto out;
> JG> +        }
> JG> +        
> JG> +        ret = strdup(CLASSNAME(ref));
> JG> +
> JG> + out:
> JG> +        return ret;
> JG>  }
>
> This could probably be a distinct patch as well.
>
> Also, why strdup the result of CLASSNAME()?  It's CIMOM-managed
> memory, and most of the other libcu functions have been modified to
> avoid dynamic memory where possible.  The return value should be const
> char * after this change.
>   
As this illustrates, I'm never really sure whether or not strdup is the 
right thing to do.  It would appear that it's unnecessary here, so I'll 
take it out.
> Also, why not just signal your single error case by returning NULL?
> That way you can avoid having to pass in the broker and the status
> pointers and the function becomes a bit cleaner.
>   
Probably just my natural tendency to over-complicate.  I'll fix that as 
well.

-- 

-Jay