Re: [MIT-Scheme-devel] Fixing non-standard floenv failures

[Chris Hanson]

Test failures due to unimplemented primitives, and places where values were returned rather than traps thrown or vice versa.

On Fri, Nov 3, 2017 at 10:36 PM, Taylor R Campbell <address@hidden> wrote:

> Date: Fri, 3 Nov 2017 22:09:27 -0700
> From: Chris Hanson <address@hidden>

>
> I added some support for testing the existence of feenableexcept and
> fedisableexcept, which are not part of C99 and not present in macOS.
>
> I used this to disable tests that depend on their existence, which
> otherwise failed.
>
> Could you please check what I did and confirm that it is correct? I'm not
> sure that I got it right, other than that the tests now pass on macOS.

Unless you're on some exotic hardware that doesn't actually support
the traps, there shouldn't be any need to test this at runtime.
There's three possible ways to get them:

1. from GNUish libm, with feenableexcept/fedisableexcept
2. from BSDish libm, with fpgetmask/fpsetmask
3. from x86 mxcsr/x87 cw

All this worked when I last used Apple stuff (which admittedly was
some years ago now).  What motivated this -- what's the symptom you're
seeing?