Re: Why check NPP instance != NULL?

On 2011-04-15 23:52, Benjamin Smedberg wrote:
> On 4/15/11 5:32 PM, Oleksandr Gavenko wrote:
>>
>> They use such pattern:
>>
>> if (instance==NULL)
>> return NPERR_INVALID_INSTANCE_ERROR;
>>
>> for functions that pass 'NPP instance' arg (like NPP_New, NPP_Destroy,
>> NPP_GetValue).
>>
>> Why this need?
> We are practicing defensive programming. The null check is cheap, and
> since we really don't know whether plugin code might have bugs, we just
> do it for sanity's sake.
>
 From 
http://devedge-temp.mozilla.org/library/manuals/2002/plugin/1.0/plugin.pdf

NPError NPP_GetValue(NPP instance, NPPVariable variable, void *value)
{
     if(instance == NULL)
         return NPERR_INVALID_INSTANCE_ERROR;
     ...
     else if (variable == NPPVpluginScriptableIID) {
         nsIID* ptr = (nsIID *)NPN_MemAlloc(sizeof(nsIID));
         *ptr = scriptableIID;
         *(nsIID **)value = ptr;
     ...
}

Here check for 'instance' but no check for 'value'!

I like defensive programming. But if you call 'memcpy'
you SHOULD NOT pass NULL because this call have no sense.

So question: all input pointer MUST be checked for NULL?
Or 'NPP instance' is special case?

Seems that docs say that 'instance' always non NULL:

   https://developer.mozilla.org/En/NPP_GetValue
   https://developer.mozilla.org/en/NPP_New

-- 
Best regards!