On 04/09/2015 05:26 AM, Peter Krempa wrote: >> Because you touched it - Coverity whines that 'error' is not checked for >> NULL: >> >> (8) Event returned_null: "virJSONValueObjectGet" returns null (checked >> 96 out of 99 times). [details] >> (16) Event var_assigned: Assigning: "error" = null return value from >> "virJSONValueObjectGet". >> Also see events: >> (17) Event dereference: Dereferencing a pointer that might be null >> "error" when calling "virJSONValueObjectGetString". [details] > > I hate coverity for this. That's a false positive due to the fact that > we check right away that the reply object contains the error subobject. > How about a "sa_assert(event);" Coverity is only complaining because so many places check "if (error) ? xxx : yyy" And yes it does work, I just checked John >> >> >> There's many examples (96) of checking... > > Many examples, but here it's useless. The check will make it ugly. > >> >> ACK with the check > > I'll send a new version that will have different control flow. If we are > going to uglyfy this function for coverity, we might as well as fix it > so that the error object is not looked up several times which will then > make the coverity error disappear. > > The downside will be that it will not be straight move of code. > > Peter > -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list