Re: Compatibility: g++-2.7.2 and g++-2.95.2

["Martin v. Loewis" <martin at loewis dot home dot cs dot tu-berlin dot de>]

> thanks for the quick and helpful answer. However, I thing this was a 
> very unfortunate decision: New semantics should have used a new name.
> I have some real trouble converting thinks like
> ``char foo[] = __FILE__ "/" __FUNCTION__;''
> It seems I must write a C++ class with a constructor now, because 
> string concatenation in initialization will not work any longer.

Maybe, but that's a decision that dates a while back, so it is
unlikely to change, again. Interestingly enough, the C99 standard
follows your view: It uses __func__ in the new gcc way (as an array
variable), but with a name not used in existing compilers before.

Regardless, I guess you don't have to *write* such a class - just use
the existing string class:

#include <string>
#include <cstdio>

int main()
{
  string s = __FILE__ "/" + string(__FUNCTION__);
  puts(s.c_str());
}


> As a side-note: What is the correct preprocessor GNUC minor version 
> to detect "egcs 1.0"? Maybe I want to use the older compiler...

It is 91. However, I believe this change is shared with 2.8, so you
may also check for 8 instead.

Regards,
Martin