[cocci] Development challenges for handling of selected code repetitions (with SmPL)?

[cocci] Development challenges for handling of selected code repetitions (with SmPL)?

[Markus Elfring]

Hello,

I hope that some software development concerns can be clarified also together
with a script variant (like the following) for the semantic patch language.


@adjustment@
expression call;
expression * pointer;
statement action;
@@
 if (pointer)
+{
    action
 <+... when != puts("Hello");
 call(pointer);
 ...+>
+}


Questionable test result:
Markus_Elfring@Sonne:…/Projekte/Coccinelle/janitor> spatch --parse-cocci move_call_into_if_branch2.cocci
…
12: no available token to attach to


A) Can any collateral evolution become possible according to the shown error message?

B) The information “Dots may be modified with a when clause,
   indicating a pattern that should not occur anywhere within the matched sequence.”
   is provided in the documentation.
   https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/ff57802f2bd1ab1bca063810dbeacf6fdce08b95/docs/manual/cocci_syntax.tex#L1030

   Such when clauses influence extra code which may eventually appear around
   the searched code pattern.
   But I would appreciate here if only selected repetitions of the specified code
   will be taken into account for this use case.
   https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/ff57802f2bd1ab1bca063810dbeacf6fdce08b95/docs/manual/cocci_syntax.tex#L797


Regards,
Markus

Re: [cocci] Development challenges for handling of selected code repetitions (with SmPL)?

[Julia Lawall]

[-- Attachment #1: Type: text/plain, Size: 1535 bytes --]



On Wed, 11 Dec 2024, Markus Elfring wrote:

> Hello,
>
> I hope that some software development concerns can be clarified also together
> with a script variant (like the following) for the semantic patch language.
>
>
> @adjustment@
> expression call;
> expression * pointer;
> statement action;
> @@
>  if (pointer)
> +{
>     action
>  <+... when != puts("Hello");
>  call(pointer);
>  ...+>
> +}

As noted, there is nothing to attach the } to.  The ... goes to the end of
the function unless there is the when.

julia

>
>
> Questionable test result:
> Markus_Elfring@Sonne:…/Projekte/Coccinelle/janitor> spatch --parse-cocci move_call_into_if_branch2.cocci
> …
> 12: no available token to attach to
>
>
> A) Can any collateral evolution become possible according to the shown error message?
>
> B) The information “Dots may be modified with a when clause,
>    indicating a pattern that should not occur anywhere within the matched sequence.”
>    is provided in the documentation.
>    https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/ff57802f2bd1ab1bca063810dbeacf6fdce08b95/docs/manual/cocci_syntax.tex#L1030
>
>    Such when clauses influence extra code which may eventually appear around
>    the searched code pattern.
>    But I would appreciate here if only selected repetitions of the specified code
>    will be taken into account for this use case.
>    https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/ff57802f2bd1ab1bca063810dbeacf6fdce08b95/docs/manual/cocci_syntax.tex#L797
>
>
> Regards,
> Markus
>

Re: [cocci] Development challenges for handling of selected code repetitions (with SmPL)?

[Markus Elfring]

>> @adjustment@
>> expression call;
>> expression * pointer;
>> statement action;
>> @@
>>  if (pointer)
>> +{
>>     action
>>  <+... when != puts("Hello");
>>  call(pointer);
>>  ...+>
>> +}
>
> As noted, there is nothing to attach the } to.

I suggest to reconsider the software situation a bit more.


>                                                 The ... goes to the end of
> the function unless there is the when.

I might be looking for other variations of indicated system functionality.

* May any statement combinations (also behind an if check) be converted to
  a compound statement (by adding extra curly brackets)?

* Can the possible code repetition ever be expressed in better ways
  besides the usage of the SmPL ellipses so far?


Regards,
Markus

Re: [cocci] Development challenges for handling of selected code repetitions (with SmPL)?

[Markus Elfring]

> As noted, there is nothing to attach the } to.

Would you like to help more to enable such a connection
even if code filter repetitions get involved?


>                                                 The ... goes to the end of
> the function unless there is the when.

Will such information influence any imaginations for possible software documentations?


I see further design options.

1. Additional notations for indicating code repetitions

2. It is demonstrated already how something can be selected by the means of when clauses.
   Further settings can become relevant. They should probably not be confused with when clauses.
   Thus I imagine that additional parameters can be specified behind a key word like “config”
   for SmPL ellipses.


Regards,
Markus