sgnGamma(-0.5) should be -1

[tedgin]

When evaluating the expression std.mathspecial.sgnGamma(-0.5) recently, I got 1. This surprised me, because Γ(x) < 0 when -1 < x < 0. Looking at the source code, I see that rndtol is used to convert x to an integer for binning. Depending on the rounding mode, -0.5 will be rounded to either 0 or -1. I think this is the source of my error. Since the domain boundaries are the no positive integers, it might be better to use a function like trunc that will map negative numbers in the same domain to the same integer, since it doesn't rely on the rounding mode.

[tedgin]

I observed the reported failure on Ubuntu 24.04 using ldc2.

? ldc2 --version
LDC - the LLVM D compiler (1.36.0):
  based on DMD v2.106.1 and LLVM 17.0.6
  built with LDC - the LLVM D compiler (1.36.0)
  Default target: x86_64-pc-linux-gnu
  Host CPU: alderlake
  http://dlang.org - http://wiki.dlang.org/LDC

When I use dmd, I don't see the error for -0.5, i.e. sgnGamma(-0.5) == -1. However, I do see the error for -0.6, sgnGamma(-0.6) == 1.

? dmd --version
DMD64 D Compiler v2.098.0
Copyright (C) 1999-2021 by The D Language Foundation, All Rights Reserved written by Walter Bright

I think the error has the same cause, rndtol being dependent on rounding mode. I suspect ldc2 has a different default rounding mode than dmd.