[plt-scheme] ports, make-srcloc and recursive reads

At Mon, 23 Apr 2007 22:20:49 -0400 (EDT), Dimitris Vyzovitis wrote:
> On Mon, 23 Apr 2007, Dimitris Vyzovitis wrote:
> 
> > On Tue, 24 Apr 2007, Matthew Flatt wrote:
> >
> > > Is there some other way that this shows up, though?
> >
> > Not that I have seen.
> 
> port-next-location is defined like this [portfun.c]:
> static Scheme_Object *port_next_location(int argc, Scheme_Object *argv[])
> {
>   Scheme_Object *a[3];
>   long line, col, pos;
> 
>   if (!SCHEME_INPUT_PORTP(argv[0]) && !SCHEME_OUTPUT_PORTP(argv[0]))
>     scheme_wrong_type("port-next-location", "port", 0, argc, argv);
> 
>   scheme_tell_all(argv[0], &line, &col, &pos);
> 
>   a[0] = ((line < 0) ? scheme_false : scheme_make_integer_value(line));
>   a[1] = ((col < 0) ? scheme_false : scheme_make_integer_value(col));
>   a[2] = ((pos < 0) ? scheme_false : scheme_make_integer_value(pos+1));
> 
>   return scheme_values(3, a);
> }
> 
> So it can conceivably be triggered by ports that have 0s at line/col/pos
> in plain reads.

The `line' result reported by scheme_tell_all() should never be 0; the
line counter is initialized to 1 when line counting is on.

The `col' result can be 0, but column counting starts from 0. 

The `pos' result can be 0 (because it's easier to count from 0
internally), but 1 is added for the (external) result.

So, while it's a bit tricky, I think it works out at this level.

Thanks,
Matthew