Missing implementation ?

Jeff Johnston jjohnstn@redhat.com
Thu Dec 16 22:45:00 GMT 2010

On 12/08/2010 05:08 AM, Corinna Vinschen wrote:
> On Dec  8 11:14, Roi Illouz wrote:
>> Hi
>> I'm no expert with newlib so please bare with me :)
>> I've built a powerpc-eabi tool chain gcc 4.4.5 including newlib 1.18.
>> When compiling my program I got a linker error: in the file
>> sf_logarithm.c the call to the function isfinitef could not be resolved.
>> I looked around in libm there were implementation of isfinite but no
>> implementation of isfinitef the FPU version was missing (or that I
>> didn't have the wits to find it). Is this a bug ? should I fix the code
>> ? Am I missing something ?
> I'm rather fuzzy on math stuff, but AFAICS isfinite() is supposed to be
> a float-type agnostic macro.  Something like isfinitef doesn't exist.
> Jeff, that's your code.  Is that just a typo?
> Corinna

No, its just old code that was using the former isfinitef macro which no 
longer exists.  I do not advise using the mathfp code which you can only 
use by specifying a configuration option.  It has been proven to be 
inaccurate compared to the regular math versions due to the series 
algorithms used.

I have fixed the reference nonetheless.

-- Jeff J.

More information about the Newlib mailing list