[Cocci] Are types re-evaluated between subsequent rules?

[Cocci] Are types re-evaluated between subsequent rules?

[Christoph Böhmwalder]

Hi,

I'm having trouble understanding coccinelles behaviour here. Consider 
the following C code and cocci rules:


#include <stdio.h>

int x;

int main(int argc, char **argv)
{
     f(x);
}


@@
identifier x;
@@
- int x;
+ int *x;

@@
int *x;
@@
- f(x)
+ g(x)



Since I read on some slides[0] that "Later rules see the results of 
earlier rules", I would assume that this would pick up the type change 
introduced by the first rule and replace f by g because it now has an 
"int *" parameter. However, spatch merely outputs the patch to change 
the type of x. If I change the type of the "expected" x in rule 2 to 
"int", cocci picks it up correctly.

Am I missing something?


[0]: 
http://events17.linuxfoundation.org/sites/events/files/slides/part1.pdf, 
page 43

--
Regards,
Christoph
_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci

Re: [Cocci] Checking change scope for a data type replacement

[Markus Elfring]

> Am I missing something?

It depends on details.

Your initial transformation approach can be written also as follows.

@replacement@
identifier x;
@@
-int
+int*
 x;


In which scopes would you like to add the asterisk for the usage of a pointer
data type?

Do you expect that function parameters should be also modified here?

Regards,
Markus
_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci

Re: [Cocci] Checking change scope for a data type replacement

[Christoph Böhmwalder]

On 27.05.19 16:16, Markus Elfring wrote:
>> Am I missing something?
> 
> It depends on details.
> 
> Your initial transformation approach can be written also as follows.
> 
> @replacement@
> identifier x;
> @@
> -int
> +int*
>   x; >
> 
> In which scopes would you like to add the asterisk for the usage of a pointer
> data type?
I'm trying to replace a particular function call with a one-liner if the 
function is not defined, so I control the scope myself. In other words, 
in my original code "int x" is passed to "void f(int)" as a paramter, 
and I would like to apply the following transformations:
1) "x" has a type of "int *"
2) the new "int *x" gets passed to a function "void g(int *)" instead.

> 
> Do you expect that function parameters should be also modified here?
> 
> Regards,
> Markus
> 

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

Re: [Cocci] Checking change scope for a data type replacement

[Markus Elfring]

>> @replacement@
>> identifier x;
>> @@
>> -int
>> +int*
>>   x; >
>>
>> In which scopes would you like to add the asterisk for the usage of a pointer
>> data type?
…
> 1) "x" has a type of "int *"

The asterisk addition seems to work for (local) variables.


> 2) the new "int *x" gets passed to a function "void g(int *)" instead.

The succinct SmPL change specification looks like it could express
data type adjustments also at other places.

But an additional SmPL rule seems to be needed at the moment.

@replacement2@
identifier f, x;
type t;
@@
 t
 f(...,
-  int
+  int*
   x, ...)
 { ... }


Do you distinguish any references to global variables here?

Regards,
Markus
_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci

Re: [Cocci] Checking change scope for a data type replacement

[Markus Elfring]

> In other words, in my original code "int x" is passed to "void f(int)" as a paramter,
> and I would like to apply the following transformations:

How do you think about to try a SmPL change specification out like the following?

@replacement3@
identifier x;
@@
-int
+int*
 x;
 <+...
-f
+g
  (x);
 ...+>


Regards,
Markus
_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci

Re: [Cocci] Checking change scope for a data type replacement

[Julia Lawall]

On Mon, 27 May 2019, Markus Elfring wrote:

> > In other words, in my original code "int x" is passed to "void f(int)" as a paramter,
> > and I would like to apply the following transformations:
>
> How do you think about to try a SmPL change specification out like the following?
>
> @replacement3@
> identifier x;
> @@
> -int
> +int*
>  x;
>  <+...
> -f
> +g
>   (x);
>  ...+>

His example shows that he wants to change a parameter type, not a local
variable.

julia

>
>
> Regards,
> Markus
> _______________________________________________
> Cocci mailing list
> Cocci@systeme.lip6.fr
> https://systeme.lip6.fr/mailman/listinfo/cocci
>
_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci

Re: [Cocci] Checking change scope for a data type replacement

[Markus Elfring]

>> @replacement3@
>> identifier x;
>> @@
>> -int
>> +int*
>>  x;
>>  <+...
>> -f
>> +g
>>   (x);
>>  ...+>
>
> His example shows that he wants to change a parameter type,

He would like to call a function which gets a single pointer passed
instead of an integer by possibly varying variables.


> not a local variable.

The support for source code transformations together with global variables
might need further software development considerations.

Regards,
Markus
_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci

Re: [Cocci] Are types re-evaluated between subsequent rules?

[Julia Lawall]

[-- Attachment #1: Type: text/plain, Size: 1621 bytes --]



On Mon, 27 May 2019, Christoph Böhmwalder wrote:

> Hi,
>
> I'm having trouble understanding coccinelles behaviour here. Consider the
> following C code and cocci rules:
>
>
> #include <stdio.h>
>
> int x;
>
> int main(int argc, char **argv)
> {
>     f(x);
> }
>
>
> @@
> identifier x;
> @@
> - int x;
> + int *x;
>
> @@
> int *x;
> @@
> - f(x)
> + g(x)
>
>
>
> Since I read on some slides[0] that "Later rules see the results of earlier
> rules", I would assume that this would pick up the type change introduced by
> the first rule and replace f by g because it now has an "int *" parameter.
> However, spatch merely outputs the patch to change the type of x. If I change
> the type of the "expected" x in rule 2 to "int", cocci picks it up correctly.
>
> Am I missing something?

Your first rule changes only variable declarations, which consist of a
type, a variable and a semicolon.  If you want to change all int types to
int *, then you can change the first rule to remove the x and the ;.  If
you want to change int x in the parameter list of main, or in the
parameter list of all functions, then you could do

@@
identifier f; // all functions
identifier x;
@@

f(...,
- int x
+ int *x
  ,...) { ... }

You can factor out the common parts, ie int and x from the transformation,
as Markus suggests, but I doubt this would be useful.

julia

>
>
> [0]: http://events17.linuxfoundation.org/sites/events/files/slides/part1.pdf,
> page 43
>
> --
> Regards,
> Christoph
> _______________________________________________
> Cocci mailing list
> Cocci@systeme.lip6.fr
> https://systeme.lip6.fr/mailman/listinfo/cocci
>

[-- Attachment #2: Type: text/plain, Size: 136 bytes --]

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