[RFC] Sharing data processing resources for selected SmPL scripts?

[RFC] Sharing data processing resources for selected SmPL scripts?

[Markus Elfring]

Hello,

I became curious if further developers would like to try the following script out
for the semantic patch language.


@display@
identifier action;
type t;
@@
*t * action(...)
 {
 ... when any
 }


I interpret such a search approach in the way that it should extract some data
from a lot of source code places.
Thus I imagine that it results also in special software run time characteristics
which matter especially for big codebases like Linux.

How do you think about to clarify corresponding data collections
and computation resources?

Related source code analysis concerns can eventually be also clarified.

* Handling of function declarations in header files

* Relevance of some preprocessor directives


Regards,
Markus

Re: [cocci] [RFC] Sharing data processing resources for selected SmPL scripts?

[Julia Lawall]

On Fri, 5 Sep 2025, Markus Elfring wrote:

> Hello,
>
> I became curious if further developers would like to try the following script out
> for the semantic patch language.
>
>
> @display@
> identifier action;
> type t;
> @@
> *t * action(...)
>  {
>  ... when any

when any is not useful here.

>  }
>
>
> I interpret such a search approach in the way that it should extract some data
> from a lot of source code places.
> Thus I imagine that it results also in special software run time characteristics
> which matter especially for big codebases like Linux.

I guess it would be quite fast, because it will just find the deader of
each function.

julia

> How do you think about to clarify corresponding data collections
> and computation resources?
>
> Related source code analysis concerns can eventually be also clarified.
>
> * Handling of function declarations in header files
>
> * Relevance of some preprocessor directives
>
>
> Regards,
> Markus
>

Re: [cocci] [RFC] Sharing data processing resources for selected SmPL scripts?

[Markus Elfring]

>> @display@
>> identifier action;
>> type t;
>> @@
>> *t * action(...)
>>  {
>>  ... when any
…>>  }
>>
>>
>> I interpret such a search approach in the way that it should extract some data
>> from a lot of source code places.
>> Thus I imagine that it results also in special software run time characteristics
>> which matter especially for big codebases like Linux.
> 
> I guess it would be quite fast,

I am curious how such development views will evolve further
(depending on the data processing environment you may use).


>                                 because it will just find the deader of
> each function.

Header information should be found for several function implementations
from selected directories.
If you would dare to try the shown SmPL script out on the complete Linux codebase,
a huge diff output will probably be generated.
It might be more helpful to try it out on a few header/source files
with a more reasonable size.

Other data structures might become also more attractive for the storage
and documentation of found pointer types in combination with known functions.

Regards,
Markus