fix excess precision in return value of i386 log-family functions
[musl] / src / math / i386 / log1p.s
index 354f391..f3c95f8 100644 (file)
@@ -10,10 +10,14 @@ log1p:
        cmp $0x00100000,%eax
        jb 2f
        fyl2xp1
+       fstpl 4(%esp)
+       fldl 4(%esp)
        ret
 1:     fld1
        faddp
        fyl2x
+       fstpl 4(%esp)
+       fldl 4(%esp)
        ret
                # subnormal x, return x with underflow
 2:     fsts 4(%esp)