Chump Is Stumped By VARIANT Problem

[Eurosid]

So I have this COM object that's packing up an array floats for me. Here's the loop where the VARIANT* array is getting stuffed: for (i = 0; i < cElements; i++) {      VariantInit(&vDataValues[i]);      V_VT(&vDataValues[i]) = VT_R4;            //Works ok UNLESS float_data[i] == 0.0;      V_R4(&vDataValues[i]) = float_data[i];      if (!V_R4(&vDataValues[i]))      {           *pErrorCode = 1;           m_strMessage.Format("V_R4 failed!");           hr = E_FAIL;           goto Error;      } } The VT_R4() operation works unless the element in float_data happens to be 0.0. When that happens, VT_R4 fails. I know it fails only with 0.0, because I tried this: if( float_data[i] != 0.0 )      V_R4(&vDataValues[i]) = float_data[i]; else      V_R4(&vDataValues[i]) = float_data[i] + 0.1; Then, it never fails. I can't figure out why.

[led mike]

Well I have no idea what your float_data is or if it has been properly initialized but this works without error

VARIANT DataValues[4];
    V_R4( &DataValues[0]) = 0.0;

led mike

[Eurosid]

I'm new to using COM, so maybe I'm missing something obvious here. The data I can verify is OK.

led mike wrote:

VARIANT DataValues[4];
V_R4( &DataValues[0]) = 0.0;

Those lines work for me, too. The examples I'm looking show that you verify it's OK by doing this sort of thing: if (!V_R4(&vDataValues[i])) { //Do some error stuff } And !V_R4(&vDataValues[i]) evaluates to false if the data is 0.0. Why is that?:confused: If I comment out that error check, everything seems to work fine. Is it not needed?

[led mike]

Eurosid wrote:

The examples I'm looking show that you verify it's OK by doing this sort of thing:

I don't think that's correct. Where are you getting these examples? V_R4 is just a macro that evaluates to the specified VARIANT member during pre-compile. It does not return an status indication value.

led mike

[Iain Clarke Warrior Programmer]

I'm slightly curious what you mean by "fails". Do you put the variant in a watch window, and see wrong numbers? From your later post, I think you think that V_R4 is a function which passes success / failure. Looking in OLEAUTO.H,

#define V_UNION(X, Y) ((X)->Y)
#define V_VT(X) ((X)->vt)
#define V_R4(X) V_UNION(X, fltVal)

So...

VARIANT var;
VariantInit (&var);
V_VT(&var) = VT_R4;
V_R4(&var) = 0.0;

becomes

V_UNION(&var,ftlVal) = 0.0;

becomes

(&var)->ftlVal = 0.0;

which looks fine to me. BUT: You are using this in an if statement. Imagine the following:

if (1)
{
// pretty guaranteed to get here
} else {
// not likely...
}

and

if (a=0)
{
// pretty unlikely to get here
} else {
//
}

The second if evaluates to 0, which will fail the if. Cutting to the chase:

V_R4(&blah) = a number

evaluates to whatever the number is - and 0 is FALSE. Here endeth the lesson. Iain.

Iain Clarke appears because CPallini still cares.

[Eurosid]

:doh: :doh: :doh: You are exactly correct. I misunderstood what I was looking at. The example was for packing up a BSTR, so they did such a check to make sure the pointer wasn't NULL. Doing the same thing on a float isn't needed, but will look OK unless the value happens to be 0, which will look like FALSE. I told you I was a Chump! Thank you for help!

[Eurosid]

Iain Clarke wrote:

Here endeth the lesson. Iain.

Yes. That was it. I was being a Chump. I knew it, but I just couldn't see why. :-O Thanks!