Question

I have the following function:

void Register(Data* _pData, uint32 _Line, const char* _pFile, ...)
{
   va_list Args;
   va_start(Args, _pFile);
   for(uint i = 0;i m_NumFloats; ++i)
   {
      _pData->m_Floats[i] = va_arg(Args, fp32);
   }
   va_end(Args);
}

Which is called by the macro:

#define REG(_Name, ...)\
{\
   if(s_##_Name##_Data.m_Enabled)
      Register(&s_##_Name##_Data, __LINE__, __FILE__, ##__VA_ARGS__);\
}\

With the usage:

REG(Test, (fp32)0.42f);

The Data-struct looks like:

struct Data
{
   int m_NumFloats;
   fp32 m_Floats[4];
}

The creation-macro of Data creates the static Data g_YourName_Data and initializes it correctly with a maximum of 4 m_NumFloats.

The va_arg call resolves to 0.0. s_Test_Data exists and the Register-function is called appropriate. va-list just simply won’t let me resolve the first argument into the float that I passed it into. Is there anything specific that I’m missing?

Answer 1

Try:

#define REG(_Name, ...)\
{\
   if(s_##_Name_Data.m_Enabled)\
      Register(&s_##_Name_Data, __LINE__, __FILE__, __VA_ARGS__);\
}

Get rid of the token-pasting operator. You we’re also missing a ‘\’ in your macro (maybe a copy-n-paste error?).

Also, use va_arg(), not va_args(). And I’m not sure if you meant for _Name to be _Name_Data or the other way around.

Finally, I assumed that fp32 was an alias for float; GCC had this to say to me:

C:\TEMP\test.c:22: warning: `fp32' is promoted to `double' when passed through `...'
C:\TEMP\test.c:22: warning: (so you should pass `double' not `fp32' to `va_arg')
C:\TEMP\test.c:22: note: if this code is reached, the program will abort

You should heed that warning. The program does crash for me if I do not.