fenv.h

[Joel Sherrill]

On 10/20/2015 10:43 AM, Corinna Vinschen wrote:
> On Oct 20 17:21, Corinna Vinschen wrote:
>> On Oct 20 08:53, Joel Sherrill wrote:
>>>
>>>
>>> On 10/19/2015 8:34 AM, Corinna Vinschen wrote:
>>>> On Oct  7 18:49, Stefan Heinzmann wrote:
>>>>> Hi all,
>>>>>
>>>>> I have seen a couple of discussions about fenv.h in the past on this list,
>>>>> and there seems to be support for fenv.h in newlib for the spu target, but
>>>>> for no other. As fenv.h belongs to C99, I wonder why that is so. Is fenv.h
>>>>> support meant to be the responsibility of the C library or of the compiler
>>>>> support library? What about implementations for the more mainstream
>>>>> architectures, such as X86 or ARM?
>>>>
>>>> I guess the answer is simply that nobody provided them yet, so SPU
>>>> and Cygwin are the only ones so far.
>>>
>>> Is the Cygwin implementation general enough for x86?
>>
>> I think so.  It includes two Cygwin-specific headers winsup.h and
>> wincap.h but both includes are apparently unnecessary.
>>
>> Would you like to give it a whirl?  Just take the files
>>
>>    winsup/cygwin/fenv.cc
>>    winsup/cygwin/include/fenv.h
>>
>> and in fenv.cc change
>>
>>    #include "winsup.h"
>>    #include "fenv.h"
>>    #include "errno.h"
>>    #include "wincap.h"
>>    #include <string.h>
>>
>> to
>>
>>    #include "fenv.h"
>>    #include <errno.h>
>>    #include <string.h>
>
> There's no reason to stick to C++ either, so renaming to fenv.c should
> be fine with minor style tweaks at worst.
>
> If you think this is fine for x86 and x86_64 in general, there's no
> problem in moving the file to newlib and relicensing to BSD 2-clause.

I can try to build it for an RTEMS x86 target. It is in the queue now.

Stefan.. how does it look to you for x86? This could address one of the
targets for you.

--joel