[cocci] Does SmPL for functions with prototype disjunctions need to be complete?

[cocci] Does SmPL for functions with prototype disjunctions need to be complete?

[Eric Wheeler]

This seems to work when the body of the funciton is defined in each disjunction:

	@ prefix_nec_arg @
	type T;
	identifier F, j;
	typedef nec_t;
	parameter list PL;
	@@
	( 
	T F(
	-void
	+nec_t *nec
	 )
	{
	... when exists
	    when any
	nec->j
	... when exists
	    when any
	}
	| 
	T F(
	+nec_t *nec
	 )
	{
	... when exists
	    when any
	nec->j
	... when exists
	    when any
	} 
	| 
	T F(
	+nec_t *nec,
	PL
	 )
	{
	... when exists
	    when any
	nec->j
	... when exists
	    when any
	}
	)

But placing the function body after the disjunction block did not, even 
though it is easier to read and does not duplicate content:

	@ prefix_nec_arg @
	type T;
	identifier F, j;
	typedef nec_t;
	parameter list PL;
	@@
	( 
	T F(
	-void
	+nec_t *nec
	 )
	| 
	T F(
	+nec_t *nec
	 )
	| 
	T F(
	+nec_t *nec,
	PL
	 )
	)
	{
	... when exists
	    when any
	nec->j
	... when exists
	    when any
	}

Is that expected?  

It means there must be a longer SmPL, which is ok, but for my own 
curiosity: is there a shorter form?


--
Eric Wheeler

Re: [cocci] Does SmPL for functions with prototype disjunctions need to be complete?

[Julia Lawall]

On Sun, 6 Mar 2022, Eric Wheeler wrote:

> This seems to work when the body of the funciton is defined in each disjunction:
>
> 	@ prefix_nec_arg @
> 	type T;
> 	identifier F, j;
> 	typedef nec_t;
> 	parameter list PL;
> 	@@
> 	(
> 	T F(
> 	-void
> 	+nec_t *nec
> 	 )
> 	{
> 	... when exists
> 	    when any
> 	nec->j
> 	... when exists
> 	    when any
> 	}
> 	|
> 	T F(
> 	+nec_t *nec
> 	 )
> 	{
> 	... when exists
> 	    when any
> 	nec->j
> 	... when exists
> 	    when any
> 	}
> 	|
> 	T F(
> 	+nec_t *nec,
> 	PL
> 	 )
> 	{
> 	... when exists
> 	    when any
> 	nec->j
> 	... when exists
> 	    when any
> 	}
> 	)
>
> But placing the function body after the disjunction block did not, even
> though it is easier to read and does not duplicate content:
>
> 	@ prefix_nec_arg @
> 	type T;
> 	identifier F, j;
> 	typedef nec_t;
> 	parameter list PL;
> 	@@
> 	(
> 	T F(
> 	-void
> 	+nec_t *nec
> 	 )
> 	|
> 	T F(
> 	+nec_t *nec
> 	 )
> 	|
> 	T F(
> 	+nec_t *nec,
> 	PL
> 	 )
> 	)
> 	{
> 	... when exists
> 	    when any
> 	nec->j
> 	... when exists
> 	    when any
> 	}
>
> Is that expected?
>
> It means there must be a longer SmPL, which is ok, but for my own
> curiosity: is there a shorter form?

It might be possible to put a disjunction around only the parameter list,
ie inside the parentheses.

julia

Re: [cocci] Does SmPL for functions with prototype disjunctions need to be complete?

[Markus Elfring]

>> It means there must be a longer SmPL, which is ok, but for my own
>> curiosity: is there a shorter form?
> It might be possible to put a disjunction around only the parameter list,
> ie inside the parentheses.


How useful would such an SmPL code variant be here actually?

Regards,
Markus

Re: [cocci] Does SmPL for functions with prototype disjunctions need to be complete?

[Eric Wheeler]

On Mon, 7 Mar 2022, Markus Elfring wrote:

> 
> >> It means there must be a longer SmPL, which is ok, but for my own
> >> curiosity: is there a shorter form?

> > It might be possible to put a disjunction around only the parameter list,
> > ie inside the parentheses.

Thanks Julia, that is a great hint!

> How useful would such an SmPL code variant be here actually?

It just makes it more readable.  I always try to avoid duplicate code no 
matter what language I'm writing for.

--
Eric Wheeler


> 
> Regards,
> Markus
> 
> 
> 

Re: [cocci] Does SmPL for functions with prototype disjunctions need to be complete?

[Julia Lawall]

On Mon, 7 Mar 2022, Eric Wheeler wrote:

> On Mon, 7 Mar 2022, Markus Elfring wrote:
>
> >
> > >> It means there must be a longer SmPL, which is ok, but for my own
> > >> curiosity: is there a shorter form?
>
> > > It might be possible to put a disjunction around only the parameter list,
> > > ie inside the parentheses.
>
> Thanks Julia, that is a great hint!
>
> > How useful would such an SmPL code variant be here actually?
>
> It just makes it more readable.  I always try to avoid duplicate code no
> matter what language I'm writing for.

There is some value to rules that are readable, rather than minimal.

julia

Re: [cocci] Does SmPL for functions with prototype disjunctions need to be complete?

[Markus Elfring]

> It just makes it more readable.  I always try to avoid duplicate code no
> matter what language I'm writing for.


How do you think about “code” which you would not need to specify at all?

Regards,
Markus

Re: [cocci] Does SmPL for functions with prototype disjunctions need to be complete?

[Eric Wheeler]

[-- Attachment #1: Type: text/plain, Size: 433 bytes --]

On Tue, 8 Mar 2022, Markus Elfring wrote: 
> > It just makes it more readable.  I always try to avoid duplicate code no
> > matter what language I'm writing for.
> 
> 
> How do you think about “code” which you would not need to specify at all?

At the moment my needs are met, but if you think an SmPL notation for this 
would benefit others then of course that would be great.

--
Eric Wheeler



> 
> Regards,
> Markus
> 
> 
>