Re: In libm, sin(qNaN) doesn't expect FE_INVALID ?

[Adhemerval Zanella <adhemerval.zanella@linaro.org>]

On 03/09/2020 09:34, Ruinland ChuanTzu Tsai wrote:
> Hi all,
> sorry for the bothering.
> 
> Recently, as I'm testing some modification on libm, I happen to realize
> the fact that glibc's testsutie doesn't expect sin(+-qNaN) to trigger
> FE_INVALID, which is designed in `math/libm-test-sin.inc` :

AFAIK this is the expected behavior for *quiet* NaN for most symbols, it should
not signal invalid exception where applicable.

> 
> TEST_f_f (sin, qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED)
> TEST_f_f (sin, -qnan_value, qnan_value, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED)
> 
> Yet I think that if the program issued sin(qNaN), it should be okay to
> be given an invalid operation excpetion. 
> 
> If that's the case, then appending `INVALID_EXCEPTION_OK` to the
> expected exception list of sin(qNaN) and sin(-qNaN) should be benign.

I think we should honor the quiet nan specification where applicable. Why
exactly are you trying to do that is requiring return a signaling NaN
for quiet NaN input?

> 
> Though I'm neither an expert on libm nor IEEE standards, I'm wondering
> will there be any concern of such behavior ( raising FE_INVALID on
> sin(+-qNaN) ) ?
> 
> Sincerely,
> Ruinland
>