var
   x:array[1..20] of integer;
   y:array[1..20] of longint;
   min,s:longint;
   n,i,v:longint;
procedure outit;
var i:integer;
    s,c:longint;
begin
   s:=0;
   for i:=1 to n do
    inc(s,x[i]*y[i]);
   c:=v-s;
   if c>s then  s:=c-s else s:=s-c;
   if s<min then min:=s;
end;
procedure search(k:longint);
var i:integer;
begin
   if k>n then begin outit;exit;end;
   for i:=0 to 1 do
         begin x[k]:=i;search(k+1);end;
end;
begin
   readln(n);
   min:=0;
   v:=0;
   for i:=1 to n do read(y[i]);
   for i:=1 to n do begin min:=min+y[i]; end;
   v:=min;
   search(1);
   writeln(min);
end.

I think I'm right.

program ural1005;
var
    a:array[1..20]of boolean;
    w:array[1..20]of longint;
    i,j,k,n,m:longint;
function dif:longint;
var diff:longint;
begin
    diff:=0;
    for i:=1 to n do
        if a[i] then inc(diff,w[i]) else dec(diff,w[i]);
    exit(diff);
end;
procedure dfs(i:longint);
begin
    if i>n then
        begin
            if abs(dif)<m then m:=abs(dif);
            exit;
        end;
    a[i]:=true;
    dfs(i+1);
    a[i]:=false;
    dfs(i+1);
end;
begin
    readln(n);
    for i:=1 to n do
        readln(w[i]);
    m:=maxlongint;
    a[1]:=true;
    dfs(2);
    writeln(m);
end.
Me too

not readln(w[i]); but read(w[i]);
I had this mistake at the first time (because of spaces between numbers)