Re: Possible code defects: macros and precedence

[Julia Lawall <julia.lawall@lip6.fr>]

The semantic patch below finds a binary operator in a macro and a binary
operator in the use of the macro, and checks if the priority of the
operator in the macro is higher (lower number) than the priority of the
operator in the use.  If this is the case, it adds parentheses in the use,
which is not what one wants, but serves to show where the problem is.

It doesn't turn up anything, except an occurrence of (u32)-1, which
Coccinelle parses as a subtraction, due to not having any nearby evidence
that u32 is a type.

I didn't make any special effort on the include files, which means that
only local include files and ones with the same name as the C file are
taken into account.  I can try with more aggressive include options.

This only works with the github version of Coccinelle, as it required
quite a lot of improvement to the treatmern of #define.

julia

@initialize:ocaml@
@@

let binoptbl =
    [("*",3);("/",3);("%",3);
      ("+",4);("-",4);
      ("<<",5);(">>",5);
      ("<",6);(">",6);("<=",6);(">=",6);
      ("==",7);("!=",7);
      ("&",8);
      ("^",9);
      ("|",10);
      ("&&",11);
      ("||",12)]

@r@
identifier i,j;
identifier list[n] is;
binary operator b;
expression e;
@@

#define i(is,j,...) (<+... \(j b e \| e b j\) ...+>)

@s@
identifier r.i;
expression list[r.n] es;
binary operator b1;
expression e1,e2;
position p;
@@

 i@p(es,e1 b1 e2,...)

@script:ocaml@
_p << s.p;
b << r.b;
b1 << s.b1;
@@

try
  let p1 = List.assoc b binoptbl in
  let p2 = List.assoc b1 binoptbl in
  if p1 >= p2 then Coccilib.include_match false
with Not_found -> ()

@@
identifier r.i;
expression list[r.n] es;
expression e;
position s.p;
@@

i@p(es,
+(
e
+)
,...)