Re: mshtml: Add IHTMLXMLHttpRequest::responseXML() method implementation. (try 2)

[Nikolay Sivov]

I don't know if it's terribly important in practice or not, but it could 
be better to always initialize out pointer to NULL.
> - hres = IXMLDOMDocument_QueryInterface(xmldoc, &IID_IObjectSafety, (void**)&safety);
> - if(SUCCEEDED(hres)) {
> -
>    hres = IObjectSafety_SetInterfaceSafetyOptions(safety, NULL,
> -
>        INTERFACESAFE_FOR_UNTRUSTED_CALLER | INTERFACESAFE_FOR_UNTRUSTED_DATA | INTERFACE_USES_SECURITY_MANAGER,
> -
>        INTERFACESAFE_FOR_UNTRUSTED_CALLER | INTERFACESAFE_FOR_UNTRUSTED_DATA | INTERFACE_USES_SECURITY_MANAGER);
> -
>    IObjectSafety_Release(safety);
> -
>    if(FAILED(hres)) {
> -
>        ERR("SetInterfaceSafetyOptions(%p) failed: %08x\n", safety, hres);
> -
>        IXMLDOMDocument_Release(xmldoc);
> -
>        return hres;
> -
>    }
> - } else {
> -
>    ERR("QueryInterface(IID_IObjectSafety) failed: %08x\n", hres);
> -
>    IXMLDOMDocument_Release(xmldoc);
> -
>    return hres;
> - }

This looks too complicated to me, I don't think any failure here is 
critical.

mshtml: Add IHTMLXMLHttpRequest::responseXML() method implementation. (try 2)

[Zhenbo Li]

Hi Nikolay,

2015-11-04 22:44 GMT+08:00 Nikolay Sivov :
> I don't know if it's terribly important in practice or not, but it could be
> better to always initialize out pointer to NULL.

So how about adding such code at the beginning of this function

if(!p)
        return E_POINTER;
    *p = NULL;
> > - hres = IXMLDOMDocument_QueryInterface(xmldoc, &IID_IObjectSafety,
> >
> > (void**)&safety);
> > - if(SUCCEEDED(hres)) {
> > -
> >    hres = IObjectSafety_SetInterfaceSafetyOptions(safety, NULL,
> > -
> >        INTERFACESAFE_FOR_UNTRUSTED_CALLER |
> >
> > INTERFACESAFE_FOR_UNTRUSTED_DATA | INTERFACE_USES_SECURITY_MANAGER,
> > -
> >        INTERFACESAFE_FOR_UNTRUSTED_CALLER |
> >
> > INTERFACESAFE_FOR_UNTRUSTED_DATA | INTERFACE_USES_SECURITY_MANAGER);
> > -
> >    IObjectSafety_Release(safety);
> > -
> >    if(FAILED(hres)) {
> > -
> >        ERR("SetInterfaceSafetyOptions(%p) failed: %08x\n", safety,
> >
> > hres);
> > -
> >        IXMLDOMDocument_Release(xmldoc);
> > -
> >        return hres;
> > -
> >    }
> > - } else {
> > -
> >    ERR("QueryInterface(IID_IObjectSafety) failed: %08x\n", hres);
> > -
> >    IXMLDOMDocument_Release(xmldoc);
> > -
> >    return hres;
> > - }
>
> This looks too complicated to me, I don't think any failure here is
> critical.

The code is too complicated to me, either. But I haven't find a better
solution for that.

mshtml: Add IHTMLXMLHttpRequest::responseXML() method implementation. (try 2)

[Nikolay Sivov]

On 04.11.2015 18:11, Zhenbo Li wrote:
> Hi Nikolay,
>
> 2015-11-04 22:44 GMT+08:00 Nikolay Sivov :
> > I don't know if it's terribly important in practice or not, but it could be
> > better to always initialize out pointer to NULL.
>
> So how about adding such code at the beginning of this function
>  if(!p)
>      return E_POINTER;
>  *p = NULL;

Sure, if test proves that it handles NULL arguments.
> > > - hres = IXMLDOMDocument_QueryInterface(xmldoc, &IID_IObjectSafety,
> > >
> > > (void**)&safety);
> > > - if(SUCCEEDED(hres)) {
> > > -
> > >    hres = IObjectSafety_SetInterfaceSafetyOptions(safety, NULL,
> > > -
> > >        INTERFACESAFE_FOR_UNTRUSTED_CALLER |
> > >
> > > INTERFACESAFE_FOR_UNTRUSTED_DATA | INTERFACE_USES_SECURITY_MANAGER,
> > > -
> > >        INTERFACESAFE_FOR_UNTRUSTED_CALLER |
> > >
> > > INTERFACESAFE_FOR_UNTRUSTED_DATA | INTERFACE_USES_SECURITY_MANAGER);
> > > -
> > >    IObjectSafety_Release(safety);
> > > -
> > >    if(FAILED(hres)) {
> > > -
> > >        ERR("SetInterfaceSafetyOptions(%p) failed: %08x\n", safety,
> > >
> > > hres);
> > > -
> > >        IXMLDOMDocument_Release(xmldoc);
> > > -
> > >        return hres;
> > > -
> > >    }
> > > - } else {
> > > -
> > >    ERR("QueryInterface(IID_IObjectSafety) failed: %08x\n", hres);
> > > -
> > >    IXMLDOMDocument_Release(xmldoc);
> > > -
> > >    return hres;
> > > - }
> >
> > This looks too complicated to me, I don't think any failure here is
> > critical.
>
> The code is too complicated to me, either. But I haven't find a better
> solution for that.

If I was doing it, I'd simply QI for IObjectSafety, if SUCCEEDED() -> 
SetInterfaceSafetyOptions(), and Release(), no additional error 
checking. Jacek, what do you think?