Re: results of gnulib tests with -fcheck-pointer-bounds

[Bruno Haible]

> > +      int user_key =
> > +        ((opt & (1 << (USER_BITS - 1))) ? ~USER_MASK : 0) | (opt & 
> > USER_MASK);
> 
> Yes, thanks, that's simpler than what is there now, and (in theory at least) 
> is 
> more portable.

OK, pushed.