Re: [Sbcl-help] floating point traps silently unset

[Nikodemus S. <nik...@ra...>]

On 26 January 2010 19:20, Tamas K Papp <tk...@gm...> wrote:
> In my code, I was relying on (log 0) signalling an error.  It worked
> fine.  After a bit of development (basically a few hours), I noticed
> my code misbehaving, and found that the cause is:
>
> CL-USER> (sb-int:get-floating-point-modes)
> (:TRAPS NIL :ROUNDING-MODE :NEAREST :CURRENT-EXCEPTIONS (:INEXACT)
>  :ACCRUED-EXCEPTIONS (:INEXACT) :FAST-MODE NIL :PRECISION :53-BIT)
>
> I never explicitly unset the traps.  Is there anything else in SBCL
> that could unset them without my knowledge?

One of the reason setting traps isn't supported is just this. :/ SBCL
should not be unsetting them behind your back, but calls to libc may
do so -- and SBCL doesn't notice if that happens.

Cheers,

 -- Nikodemus