uniform vector byte signed or unsigned

uniform vector byte signed or unsigned

[Kevin Ryde]

In scm_cvref it looks like a byvect byte is fetched as a "char", which
is of course normally signed, but on some systems can be unsigned.

Is a byte meant to be treated as signed?  It's probably worth making
that explicit so scheme level stuff doesn't depend on the vagaries of
C type conventions.


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: uniform vector byte signed or unsigned

[Neil Jerram]

Kevin Ryde wrote:

> In scm_cvref it looks like a byvect byte is fetched as a "char", which
> is of course normally signed, but on some systems can be unsigned.
> 
> Is a byte meant to be treated as signed?  It's probably worth making
> that explicit so scheme level stuff doesn't depend on the vagaries of
> C type conventions.

Good spot.  I've hit this on Windows, where it comes out as signed, and 
so I have to do a trivial conversion to an unsigned value.

My preference would be for the value to be unsigned in the first place, 
though.

Regards,
       Neil



_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel

Re: uniform vector byte signed or unsigned

[Kevin Ryde]

Neil Jerram <neil@ossau.uklinux.net> writes:
>
> My preference would be for the value to be unsigned in the first
> place, though.

Me too, but it's probably too late to change.  I'd imagine everyone
who used it would have quickly found its signed, and written code for
that.

Actually, I see there's always a -128 to +127 check in array-fill!, so
I guess that answers the question what it ought to be.


_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel