generic fenv wrapup

generic fenv wrapup

[Joel Sherrill]

Hi

We discussed a lot for this patch. I see FE_NOMASK_ENV isn't included.
When looking at that, the comment in /usr/include/bits/fenv.h on CentOS 7
has a subject verb agreement issue that leads me to question its intent.
 I would like someone to confirm that the glibc manual and this comment
mean the same thing:

ifd#f __USE_GNU
/* Floating-point environment where none of the exception is masked.  */
# define FE_NOMASK_ENV  ((const fenv_t *) -2)
#endif

And from the manual:

    If possible, the GNU C Library defines a macro FE_NOMASK_ENV which rep‐
       resents  an  environment  where every exception raised causes a trap
to
       occur.  You can test for this macro using #ifdef.  It is  defined
 only
       if  _GNU_SOURCE  is defined.

I think the manual version is clearer on intent.

Also.. there are three other GNU specific methods:

      #define _GNU_SOURCE         /* See feature_test_macros(7) */
       #include <fenv.h>

       int feenableexcept(int excepts);
       int fedisableexcept(int excepts);
       int fegetexcept(void);

Should all those be added conditionally?

What else did we think needed to be done with the fenv generic code before
moving on to focusing on implementations?

--joel

Re: generic fenv wrapup

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 1867 bytes --]

On Aug 12 08:51, Joel Sherrill wrote:
> Hi
> 
> We discussed a lot for this patch. I see FE_NOMASK_ENV isn't included.
> When looking at that, the comment in /usr/include/bits/fenv.h on CentOS 7
> has a subject verb agreement issue that leads me to question its intent.
>  I would like someone to confirm that the glibc manual and this comment
> mean the same thing:
> 
> ifd#f __USE_GNU
> /* Floating-point environment where none of the exception is masked.  */
> # define FE_NOMASK_ENV  ((const fenv_t *) -2)
> #endif
> 
> And from the manual:
> 
>     If possible, the GNU C Library defines a macro FE_NOMASK_ENV which rep‐
>        resents  an  environment  where every exception raised causes a trap
> to
>        occur.  You can test for this macro using #ifdef.  It is  defined
>  only
>        if  _GNU_SOURCE  is defined.
> 
> I think the manual version is clearer on intent.
> 
> Also.. there are three other GNU specific methods:
> 
>       #define _GNU_SOURCE         /* See feature_test_macros(7) */
>        #include <fenv.h>
> 
>        int feenableexcept(int excepts);
>        int fedisableexcept(int excepts);
>        int fegetexcept(void);
> 
> Should all those be added conditionally?

_GNU_SOURCE definitely.  Conditionally on what else?  

> What else did we think needed to be done with the fenv generic code before
> moving on to focusing on implementations?

Just have a look into Cygwin's fenv.h.  There's also fegetprec/fesetprec
which have been defined with SunOS and made it into a standarization
proposal http://www.open-std.org/jtc1/sc22//WG14/www/docs/n752.htm They
have not been standarized, though, so this is kind of optional.  They
should probably get defined for x86{_64} but not necessarily for any
other target.


Corinna

-- 
Corinna Vinschen
Cygwin Maintainer
Red Hat

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]