spurious warning for exp() underflow

[jghub]

I have reported this issue a few days ago to the gawk maintainer Arnold Robbins where it is described in detail:
https://lists.gnu.org/archive/html/bug-gawk/2026-05/msg00014.html

bottom line is that depending on system/OS/ibrary used/linked exp() underflow might or might not trigger 'out of range' warning. this is bad isince it get's massively in the way when some script uses a substantial loop where the underflow just is basically fine (e.g. exp(-900) → 0 within floating point precission) but hundreds of warnings might flood the screen, forcing to use categorical 2> /dev/null redirection to silence those warnings (which in turn also will hide all true errors and is therefore not a real solution).

Arnold has confirmed that this will be fixed in future gawk like so:

--------------------------------------
diff --git a/builtin.c b/builtin.c
index f33525d5..a797d7fc 100644
--- a/builtin.c
+++ b/builtin.c
@@ -207,7 +207,7 @@ do_exp(int nargs)
 	DEREF(tmp);
 	errno = 0;
 	res = exp(d);
-	if (errno == ERANGE)
+	if (errno == ERANGE && res != 0.0)
 		warning(_("exp: argument %g is out of range"), d);
 	return make_number((AWKNUM) res);
 }
--------------------------------------

I believe the corresponding place in the nawk code is here:

awk/run.c

Lines 2085 to 2087 in 5739fd7

	case FEXP:
	errno = 0;
	u = errcheck(exp(getfval(x)), "exp");

and my request/suggestion would be to fix it accordingly: issue ERANGE warning for overflow, but not underflow, of exp() (provided this is considered acceptable to do for the one true awk at all ...)