Re: math_error.7 draft 4, for review

[Andreas Jaeger <aj-Et1tbQHTxzrQT0dZR+AlfA@public.gmane.org>]

[-- Attachment #1: Type: text/plain, Size: 1740 bytes --]

"Michael Kerrisk" <mtk.manpages-gM/Ye1E23mwN+BqQ9rBEUg@public.gmane.org> writes:

> On Fri, Jul 25, 2008 at 12:36 PM, Andreas Jaeger <aj-Et1tbQHTxzrQT0dZR+AlfA@public.gmane.org> wrote:
>> "Michael Kerrisk" <mtk.manpages-gM/Ye1E23mwN+BqQ9rBEUg@public.gmane.org> writes:
>>
>>> Well, for example, according to my tests, cos(3) does not set errno.
>>> Is that what you would have expected?
>>
>> No, I wouldn't.
>
> (Slightly confused here, since you seem to contradict my result, but
> then the text below seems to agree with my result.  But maybe I'm just
> misreading your text.)

I was confused as well ;)

>> And double checking, it's the other way round :-(.  We
>> test the exception handling, not the error handling, so tests like
>> (libc/math/libm-test.inc) these
>>
>>  TEST_f_f (cos, plus_infty, nan_value, INVALID_EXCEPTION);
>>  TEST_f_f (cos, minus_infty, nan_value, INVALID_EXCEPTION);
>>
>> test that cos with input plus/minus infinity returns a NaN and raises
>> and invalid exception which is tested with fetestexcept.  And if you
>> have different results, please tell me,
>
> When I run the program belwo, this is what I see for a domain error from cos():
>
> $ ./a.out inf
> errno == 0
> fetestexcept() says: 1 FE_INVALID
> cos(inf)=nan

Which is the correct behaviour for the exception AFAIK.  Setting of
errno is indeed missing - if math_errhandling has the right value which
is unfortunately missing,

Andreas
-- 
 Andreas Jaeger, Director Platform/openSUSE, aj-l3A5Bk7waGM@public.gmane.org
  SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
   Maxfeldstr. 5, 90409 Nürnberg, Germany
    GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]