Re: [Cocci] A complex set of rules: can it be done ?

[Guenter Roeck <linux@roeck-us.net>]

Hi Julia,

On Sun, Mar 31, 2019 at 09:58:15AM +0200, Julia Lawall wrote:
> 
> 
> On Sat, 30 Mar 2019, Guenter Roeck wrote:
> 
> > Hi Coccinelle experts,
> >
> > is it possible to define a set of rules to automate the conversion done
> > in the attached patch ? I tried doing it myself, but did'nt get anywhere.
> 
> @r@
> initializer list elements;
> identifier i;
> @@
> 
> -static const u32 i[] = {
> -  elements,
> -  0
> -};
> 
> @s@
> identifier r.i,j,ty;
> @@
> 
> -static const struct hwmon_channel_info j = {
> -       .type = ty,
> -       .config = i,
> -};
> 
> @script:ocaml t@
> ty << s.ty;
> elements << r.elements;
> shorter;
> elems;
> @@
> 
> shorter :=
>    make_ident (List.hd(List.rev (Str.split (Str.regexp "_") ty)));
> elems :=
>    make_ident
>     (String.concat ","
>      (List.map (fun x -> Printf.sprintf "\n\t\t\t   %s" x)
>        (Str.split (Str.regexp " , ") elements)))
> 
> @@
> identifier s.j,t.shorter;
> identifier t.elems;
> @@
> 
> - &j
> + HWMON_CHANNEL_INFO(shorter,elems)
> 
> --------------------------------------
> 
> Theer is some ugliness here.  The elements of the array are considered to
> be an initializer list, while the arguments to the macro have to be an
> expression list.  The script code related to elems and elements is to deal
> with that.
> 
> There is also script code to constuct the first argument of
> HWMON_CHANNEL_INFO from gthe name of the hwmon_channel_info structure.
> 
> You should be able to drop the static const stuff on the structures, if it
> is not appropriate in general.  It should still be removed when the whole
> structure is removed.
> 
I dropped the "static const". With that, the above catches all instances
in the upstream kernel.

> If you want to write some other script code that makes metavariables in
> this way and prefer to use python, there are examples in
> 
> demos/pythontococci.cocci (only for making identifiers)
> tests/python_mdecl.cocci  (for making identifiers and others)
> 

And I thought this might be super-complicated or just impossible.
You are just amazing. Thanks a lot for your help!

Guenter
_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci