<bits/features.h> setting of __STDC_IEC_559__ not universally accurate

[Mark Mitchell]

Ulrich Drepper wrote:
> Mark Mitchell wrote:
> 
>> At present, <bits/features.h> unconditionally sets __STDC_IEC_559__ to
>> 1.  This is not accurate on (at least) ARM platforms using software
>> floating-point.  On those platforms, <fenv.h> does not define
>> FE_INEXACT, and friends, because the software floating-point emulation
>> provided by libgcc does not honor those rounding modes.  According to
>> the C99 standard, that would seem to preclude setting __STDC_IEC_559__.
> 
> 
> The interface glibc provides requires this.  If the broken arm runtime
> cannot handle it, fix it.

I don't fully understand.

C99 certainly allows implementations not to define this macro, and 
(AFAICT) GLIBC is still C99-conformant for ARM, with respect to these 
sorts of issues, modulo the setting of __STDC_IEC_559__.

I do agree that it would be best to fix the runtime to support the full 
IEC 559 functionality, but I'm concerned about the fact that, until that 
happens, GLIBC is defining this macro incorrectly.  It seems like we're 
making it unnecessarily hard for people to write portable programs that 
conditionalize their use of IEC 559 functionality on whether or not this 
macro is defined.

-- 
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304