[PATCH] Provide a C++ version of iseqsig

[Joseph Myers]

On Fri, 3 Nov 2017, Gabriel F. T. Gomes wrote:

> +__iseqsig_type (float fmt, __T1 x, __T2 y)

As noted, namespace requirements mean it should be __fmt, __x, __y.

> +  __iseqsig_type (__MATH_EVAL_FMT2 (x, y), x, y)

As far as I can tell, __MATH_EVAL_FMT2 (x, y) *does* get evaluated (with 
consequent floating-point exceptions), just the value of that function 
argument isn't used.  I'd think something like

(__typeof (__MATH_EVAL_FMT2 (x, y)) 0)

(adjust to use C++ features as appropriate) would be better to ensure no 
evaluation.

-- 
Joseph S. Myers
joseph@codesourcery.com