[v3] Possible fix for PR 5730

[Gabriel Dos Reis]

Phil Edwards <phil@jaj.com> writes:

| On Thu, Mar 20, 2003 at 11:36:51PM +0100, Gabriel Dos Reis wrote:
| > Phil Edwards <phil@jaj.com> writes:
| > 
| > | > | > Now, I realize the macro __FAST_MATH__ may not be defined with
| > | > | > definite value -- that is really sad.  
| > | 
| > | I believe you are mistaken here.  It's defined exactly as if you had written
| > | 
| > |     #define __FAST_MATH__
| > | 
| > | in normal code, i.e., the value is 1 by default.
| > 
| > But that is insufficiant for use in the code I posted.  The reason
| > being that when it is not defined you get a hard error.
| 
| Sorry, hadn't found that message yet.
| 
| Certainly, then,
| 
| 	#ifdef __FAST_MATH__
| 	# define _GLIBCPP_FAST_MATH 1
| 	#else
| 	# define _GLIBCPP_FAST_MATH 0
| 	#endif
| 
| makes the most sense.

That is what I suggested in the message you didn't get :-)

| For now I'd suggest putting it in std_complex.h
| unless we also find a use for it elsewhere.

The right place for these sorts of compiler level flags is c++config.
Nowehere else.

-- Gaby