Re: Bug in scm_getaffinity

[Ludovic Courtès]

Eli Zaretskii <address@hidden> skribis:

> While working on the MS-Windows emulation of this function, I bumped
> into something that looks like a bug in its subroutine:
>
>   static SCM
>   cpu_set_to_bitvector (const cpu_set_t *cs)
>   {
>     SCM bv;
>     size_t cpu;
>
>     bv = scm_c_make_bitvector (sizeof (*cs), SCM_BOOL_F);
>
>     for (cpu = 0; cpu < sizeof (*cs); cpu++)
>       {
>       if (CPU_ISSET (cpu, cs))
>         /* XXX: This is inefficient but avoids code duplication.  */
>         scm_c_bitvector_set_x (bv, cpu, SCM_BOOL_T);
>       }
>
> I think using 'sizeof (*cs)' is incorrect here, we need to use
> CPU_SETSIZE instead.  The cpu_set_t data type could be an array of bit
> masks, in which case counting only bytes in it is wrong: the result is
> too small.

Indeed, good catch.  Could you commit the obvious fix?

Thanks,
Ludo’.