Re: [PATCH 2/2] Remove broken ieeefp.h macros

[Corinna Vinschen <vinschen at redhat dot com>]

On Feb 12 10:24, Yaakov Selkowitz wrote:
> On 2016-02-12 07:14, Corinna Vinschen wrote:
> >On Feb 11 20:27, Yaakov Selkowitz wrote:
> >>Any attempt to use isnanf, isinff, or finitef from <ieeefp.h> on
> >>platforms other than SPU result in a "lvalue required as unary '&'
> >>operand" error.
> >>
> >>	newlib/libc/
> >>	* include/ieeefp.h (__ieeefp_isnanf): Remove broken macro.
> >>	(__ieeefp_isinff, __ieeefp_finitef): Ditto.
> >
> >This one puzzles me.  How is just removing the macros going to help,
> >other than avoiding a compiler warning?  Wouldn't the right thing to
> >do to fix the macros for GCC-5 or provide the functionality otherwise?
> 
> The macros are (supposed to be) just optimized replacements of the
> respective function calls.  They don't compile, however, and probably didn't
> even before GCC 5.

They do compile under GCC 4, albeit producing a warning when using -Wall:

ie.c:10:3: warning: dereferencing type-punned pointer will break strict-aliasing rules [-Wstrict-aliasing]
   if (isnanf (f))
   ^

The expression in the macro was valid code once, but apart from GCC and
the C standard having moved on, on second glance it suffers a big problem.
It assumes sizeof(float)==sizeof(long).  It can be made valid again, but
how useful is that?

So the patch is ok to apply.  If somebody complains, we can discuss
reinstantiating the macros in a valid way.


Thanks,
Corinna

-- 
Corinna Vinschen
Cygwin Maintainer
Red Hat

Attachment: signature.asc
Description: PGP signature