On 07/10/2012 09:26 AM, Julia Lawall wrote:
> On Tue, 10 Jul 2012, Lars-Peter Clausen wrote:
>
>> On 07/10/2012 02:22 AM, Joe Perches wrote:
>>> I want to elide void * casts in uses of functions with
>>> pointers only when the function prototypes are declared
>>> with the appropriate void * arguments.
>>>
>>> Any idea/suggestion how to do this?
>>>
>>> I tried variants of this but had no success:
>>>
>>> @ func_1 @
>>> type rtn_type;
>>> identifier arg;
>>> @@
>>>
>>> static int func(void *arg)
>>> {
>>> ...
>>> }
>>>
>>> @@
>>> identifier func1.func;
>>> expression e;
>>> @@
>>>
>>> - fn((void *)e)
>>> + fn(e)
>>>
>>
>> It works fine if you get rid of all the typos ;)
>>
>> This one is a bit more generic since it allows the void parameter to be at
>> any position in the argument list:
>>
>> @r1@
>> identifier arg;
>> identifier fn;
>> parameter list[n] P;
>> @@
>> fn(P, void *arg, ...)
>> {
>> ...
>> }
>>
>> @depends on r1@
>> expression e;
>> identifier r1.fn;
>> expression list[r1.n] El;
>> @@
>> - fn(El, (void *)e
>> + fn(El, e
>> , ...)
>
> A further improvement is the following. It checks that the casted argument
> already has pointer type and takes into account the case where there is only
> a signature for the function, but no definition.
Yea, looks much nicer :)
>
> @r@
> parameter list[n] ps;
> identifier fn,arg;
> @@
>
> fn(ps,void *arg,...) { ... }
>
> @@
> identifier r.fn;
> expression *E;
> expression list[r.n] es;
> @@
>
> fn(es,
> - (void *)
> E,...)
>
> @rr@
> parameter list[n] ps;
> identifier fn,arg;
> type T;
> @@
>
> T fn(ps,void *arg,...);
>
> @@
> identifier rr.fn;
> expression *E;
> expression list[r.n] es;
rr.n instead of r.n, I guess
Btw. if a rule uses identifiers from another rule is this like an implicit
depends on?
> @@
>
> fn(es,
> - (void *)
> E,...)
>
> julia
_______________________________________________
Cocci mailing list
[email protected]
http://lists.diku.dk/mailman/listinfo/cocci
(Web access from inside DIKUs LAN only)
