[cocci] Reproducing a known patch with SmPL?

[cocci] Reproducing a known patch with SmPL?

[Markus Elfring]

Hello,

I proposed another source code adjustment for a selected function implementation
according to the development branch “Linux next-20260226”.

[PATCH] i3c: dw: Use more common code in dw_i3c_master_i2c_xfers()
https://lore.kernel.org/lkml/2411a872-5bf5-4353-a754-47fa54d69c46@web.de/
https://lkml.org/lkml/2026/2/27/700


I tried the following SmPL script variant out also together with the software
combination “Coccinelle 1.3.1” accordingly.

@adjustment exists@
expression action, context, last_action;
expression list el;
identifier member, rc;
@@
 {
 ...
 {
 ...
-   action(context);
-   return rc
+   goto target
 ;
 }
 ...
+   last_action(el);
    rc = context->member;
+target:
    action(context);
-   last_action(el);
    return rc;
 }


Questionable test result:
Markus_Elfring@Sonne:…/Projekte/Linux/next-analyses> time /usr/bin/spatch …/Projekte/Coccinelle/janitor/use_more_common_code.cocci drivers/i3c/master/dw-i3c-master.c
…
@@ -1672,9 +1672,7 @@ err_disable_pm:
        pm_runtime_dont_use_autosuspend(&pdev->dev);
 
 err_assert_rst:
-       reset_control_assert(master->core_rst);
-
-       return ret;
+       goto target;
 }
 EXPORT_SYMBOL_GPL(dw_i3c_common_probe);
 

real    0m0,582s
user    0m0,539s
sys     0m0,033s



* Why was a source code place shown here at all which does not contain a variable
  assignment before an interesting function call?

* How should the data processing be improved in more desirable ways?


Regards,
Markus

Re: [cocci] Reproducing a known patch with SmPL?

[Markus Elfring]

> * How should the data processing be improved in more desirable ways?

The following SmPL script variant seems to work as expected.


@movement exists@
expression action, context, last_action;
expression list el;
identifier member, rc;
@@
 {
 ...
+   last_action(el);
    rc = context->member;
+target:
    action(context);
-   last_action(el);
    return rc;
 }


How should related if branches be adjusted on demand here?

Regards,
Markus

Re: [cocci] Reproducing a known patch with SmPL?

[Markus Elfring]

> * Why was a source code place shown here at all which does not contain a variable
>   assignment before an interesting function call?

May I expect that selected source code should be found in nested compound statements
even if it was not restricted as belonging to a known condition check?

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

> * How should the data processing be improved in more desirable ways?

I constructed another source file example together with a corresponding SmPL script.


void my_test(void)
{
 int item;

 if (1)
 {
    item = 1;
 }

 item = 2;
}


@display@
@@
 if (...)
*{ item = 1; }
*item = 2;


Test result:
Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch test-item_assignments.c show_selected_assignments_after_compound_statements3.cocci
…
@@ -3,9 +3,5 @@ void my_test(void)
  int item;
 
  if (1)
- {
-    item = 1;
- }
 
- item = 2;
 }


Can it be achieved that the source code search pattern will work also
after the omission of the filter “if (...)”?

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

> I constructed another source file example together with a corresponding SmPL script.

@display@
statement list sl;
@@
*{ sl }


Compound statements are often used in known programming languages.
https://en.cppreference.com/w/c/language/statements.html#Compound_statements

Another test result:
Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch test-item_assignments.c ../janitor/show_compound_statements.cocci
…
@@ -3,9 +3,6 @@ void my_test(void)
  int item;
 
  if (1)
- {
-    item = 1;
- }
 
  item = 2;
 }


The source code search pattern looks promising (in principle).
But it can be ambiguous in its minimal form.
Thus further case distinctions would occasionally be helpful.

Additional context information can become relevant then.
How would you prefer to handle control flow recursion in safer ways here?

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Julia Lawall]

[-- Attachment #1: Type: text/plain, Size: 1138 bytes --]



On Sun, 8 Mar 2026, Markus Elfring wrote:

> > I constructed another source file example together with a corresponding SmPL script.
>
> @display@
> statement list sl;
> @@
> *{ sl }
>
>
> Compound statements are often used in known programming languages.
> https://en.cppreference.com/w/c/language/statements.html#Compound_statements
>
> Another test result:
> Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch test-item_assignments.c ../janitor/show_compound_statements.cocci
> …
> @@ -3,9 +3,6 @@ void my_test(void)
>   int item;
>
>   if (1)
> - {
> -    item = 1;
> - }
>
>   item = 2;
>  }
>
>
> The source code search pattern looks promising (in principle).
> But it can be ambiguous in its minimal form.
> Thus further case distinctions would occasionally be helpful.
>
> Additional context information can become relevant then.
> How would you prefer to handle control flow recursion in safer ways here?

I have no idea what result you would like to see.  The whole point of
Coccinelle is that if some context information should be taken into
account, then you can provide it.

julia

>
> Regards,
> Markus
>

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

>> Additional context information can become relevant then.
>> How would you prefer to handle control flow recursion in safer ways here?
> I have no idea what result you would like to see.

I would appreciate further data processing refinements.

A) I would find it nice when selected source code transformation patterns
   can trigger usable patches from minimal SmPL script variations.

B) Ambiguity should occasionally be avoided in more succinct ways.


>                                                    The whole point of
> Coccinelle is that if some context information should be taken into
> account, then you can provide it.

I hope that known software limitations can be adjusted accordingly.

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

> I have no idea what result you would like to see.

Under which circumstances would you dare to acknowledge remaining open issues here?

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Julia Lawall]

On Sun, 8 Mar 2026, Markus Elfring wrote:

> > I have no idea what result you would like to see.
>
> Under which circumstances would you dare to acknowledge remaining open issues here?

If you can tell me what result you want maybe I can do something.  It's
not a matter of daring to do something or not.

julia

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

>> Under which circumstances would you dare to acknowledge remaining open issues here?
> 
> If you can tell me what result you want maybe I can do something.

Can you see concrete change opportunities from my recent test results?

1. Inappropriate difference output
   https://lore.kernel.org/cocci/349a12a9-f70d-4c8f-a730-9f8ec90b593f@web.de/
   https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00004.html

2. Questionable hierarchy level
   https://lore.kernel.org/cocci/20a815aa-5fa9-4b34-ab81-d6985ed193c0@web.de/
   https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00015.html


> It's not a matter of daring to do something or not.

Will development interests grow anyhow?

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Julia Lawall]

On Sun, 8 Mar 2026, Markus Elfring wrote:

> >> Under which circumstances would you dare to acknowledge remaining open issues here?
> >
> > If you can tell me what result you want maybe I can do something.
>
> Can you see concrete change opportunities from my recent test results?
>
> 1. Inappropriate difference output
>    https://lore.kernel.org/cocci/349a12a9-f70d-4c8f-a730-9f8ec90b593f@web.de/
>    https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00004.html

There is a problem here.

> 2. Questionable hierarchy level
>    https://lore.kernel.org/cocci/20a815aa-5fa9-4b34-ab81-d6985ed193c0@web.de/
>    https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00015.html

This is the message I was referring to.  I have no idea what you expect as
the output.  I don't know why it is so hard to provide that information.

julia

>
>
> > It's not a matter of daring to do something or not.
>
> Will development interests grow anyhow?
>
> Regards,
> Markus
>

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

>> Can you see concrete change opportunities from my recent test results?
>>
>> 1. Inappropriate difference output
>>    https://lore.kernel.org/cocci/349a12a9-f70d-4c8f-a730-9f8ec90b593f@web.de/
>>    https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00004.html
> There is a problem here.

Thanks for such a feedback finally.


>> 2. Questionable hierarchy level
>>    https://lore.kernel.org/cocci/20a815aa-5fa9-4b34-ab81-d6985ed193c0@web.de/
>>    https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00015.html
> This is the message I was referring to.  I have no idea what you expect as
> the output.

I presented another test result for further development considerations.



>              I don't know why it is so hard to provide that information.

I imagine that corresponding case distinctions can become also more interesting.

A) C++ test variant:
Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch --c++ test-item_assignments.c ../janitor/show_compound_statements.cocci
…
@@ -3,9 +3,6 @@ void my_test(void)
  int item;
 
  if (1)
- {
-    item = 1;
- }
 
  item = 2;
 }


Should declarations belong also to statements then?
https://en.cppreference.com/w/cpp/language/statements.html#Declaration_statements
https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/233b5ceb1d5a701fb270ebdf0a9bd6be0a6ce53b/docs/manual/cocci_syntax.tex#L348-350


B) SmPL script variant:

@display2@
declaration d;
statement list sl;
@@
 {
 d
 *sl
 }

Questionable test result:
Markus_Elfring@Sonne:…/Projekte/Coccinelle/janitor> /usr/bin/spatch --parse-cocci show_compound_statements2.cocci
…
minus: parse error: 
  File "show_compound_statements2.cocci", line 7, column 2, charpos = 56
  around = 'sl',
  whole content =  *sl


How should this implementation detail be improved anyhow?

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Julia Lawall]

[-- Attachment #1: Type: text/plain, Size: 2042 bytes --]



On Sun, 8 Mar 2026, Markus Elfring wrote:

> >> Can you see concrete change opportunities from my recent test results?
> >>
> >> 1. Inappropriate difference output
> >>    https://lore.kernel.org/cocci/349a12a9-f70d-4c8f-a730-9f8ec90b593f@web.de/
> >>    https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00004.html
> > There is a problem here.
>
> Thanks for such a feedback finally.
>
>
> >> 2. Questionable hierarchy level
> >>    https://lore.kernel.org/cocci/20a815aa-5fa9-4b34-ab81-d6985ed193c0@web.de/
> >>    https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00015.html
> > This is the message I was referring to.  I have no idea what you expect as
> > the output.
>
> I presented another test result for further development considerations.
>
>
>
> >              I don't know why it is so hard to provide that information.
>
> I imagine that corresponding case distinctions can become also more interesting.
>
> A) C++ test variant:
> Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch --c++ test-item_assignments.c ../janitor/show_compound_statements.cocci
> …
> @@ -3,9 +3,6 @@ void my_test(void)
>   int item;
>
>   if (1)
> - {
> -    item = 1;
> - }
>
>   item = 2;
>  }
>
>
> Should declarations belong also to statements then?
> https://en.cppreference.com/w/cpp/language/statements.html#Declaration_statements
> https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/233b5ceb1d5a701fb270ebdf0a9bd6be0a6ce53b/docs/manual/cocci_syntax.tex#L348-350
>
>
> B) SmPL script variant:
>
> @display2@
> declaration d;
> statement list sl;
> @@
>  {
>  d
>  *sl
>  }
>
> Questionable test result:
> Markus_Elfring@Sonne:…/Projekte/Coccinelle/janitor> /usr/bin/spatch --parse-cocci show_compound_statements2.cocci
> …
> minus: parse error:
>   File "show_compound_statements2.cocci", line 7, column 2, charpos = 56
>   around = 'sl',
>   whole content =  *sl
>
>
> How should this implementation detail be improved anyhow?

By putting the * in the leftmost column where it belongs.

julia


>
> Regards,
> Markus
>

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

>> How should this implementation detail be improved anyhow?
> By putting the * in the leftmost column where it belongs.

@display2@
declaration d;
statement list sl;
@@
 {
 d
*sl
 }


Another questionable test result:
Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch test-item_assignments.c ../janitor/show_compound_statements2.cocci
…
[no diff output]


Will further software improvements follow?

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Julia Lawall]

[-- Attachment #1: Type: text/plain, Size: 585 bytes --]



On Sun, 8 Mar 2026, Markus Elfring wrote:

> >> How should this implementation detail be improved anyhow?
> > By putting the * in the leftmost column where it belongs.
>
> @display2@
> declaration d;
> statement list sl;
> @@
>  {
>  d
> *sl
>  }
>
>
> Another questionable test result:
> Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch test-item_assignments.c ../janitor/show_compound_statements2.cocci

And I should know what test-item_assignments.c is?

julia

> …
> [no diff output]
>
>
> Will further software improvements follow?
>
> Regards,
> Markus
>
>

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

>> @display2@
>> declaration d;
>> statement list sl;
>> @@
>>  {
>>  d
>> *sl
>>  }
>>
>>
>> Another questionable test result:
>> Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch test-item_assignments.c ../janitor/show_compound_statements2.cocci
> 
> And I should know what test-item_assignments.c is?

Reminder:
This can be possible.
https://lore.kernel.org/cocci/f25025ab-4500-4315-b41c-53cd5775c69f@web.de/
https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00014.html


void my_test(void)
{
 int item;

 if (1)
 {
    item = 1;
 }

 item = 2;
}


Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Julia Lawall]

[-- Attachment #1: Type: text/plain, Size: 851 bytes --]



On Sun, 8 Mar 2026, Markus Elfring wrote:

> >> @display2@
> >> declaration d;
> >> statement list sl;
> >> @@
> >>  {
> >>  d
> >> *sl
> >>  }
> >>
> >>
> >> Another questionable test result:
> >> Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch test-item_assignments.c ../janitor/show_compound_statements2.cocci
> >
> > And I should know what test-item_assignments.c is?
>
> Reminder:
> This can be possible.
> https://lore.kernel.org/cocci/f25025ab-4500-4315-b41c-53cd5775c69f@web.de/
> https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00014.html
>
>
> void my_test(void)
> {
>  int item;
>
>  if (1)
>  {
>     item = 1;
>  }
>
>  item = 2;
> }

A statement list metavariable only matches the code starting right after a
{.  It can't start in random places.  Nothing is going to chenge here.

julia

>
>
> Regards,
> Markus
>

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

>>>> @display2@
>>>> declaration d;
>>>> statement list sl;
>>>> @@
>>>>  {
>>>>  d
>>>> *sl
>>>>  }
>>>>
>>>>
>>>> Another questionable test result:
>>>> Markus_Elfring@Sonne:…/Projekte/Coccinelle/Probe> /usr/bin/spatch test-item_assignments.c ../janitor/show_compound_statements2.cocci
>>>
>>> And I should know what test-item_assignments.c is?
>>
>> Reminder:
>> This can be possible.
>> https://lore.kernel.org/cocci/f25025ab-4500-4315-b41c-53cd5775c69f@web.de/
>> https://sympa.inria.fr/sympa/arc/cocci/2026-03/msg00014.html
>>
>>
>> void my_test(void)
>> {
>>  int item;
>>
>>  if (1)
>>  {
>>     item = 1;
>>  }
>>
>>  item = 2;
>> }
> 
> A statement list metavariable only matches the code starting right after a
> {.  It can't start in random places.

Interesting …


>                                       Nothing is going to chenge here.

Will it matter then to mention for a better software documentation
if declarations would be skipped (or not)?
https://gitlab.inria.fr/coccinelle/coccinelle/-/blob/233b5ceb1d5a701fb270ebdf0a9bd6be0a6ce53b/docs/manual/cocci_syntax.tex#L348-350

Regards,
Markus

Re: [cocci] Reconsidering source code search challenges for nested compound statements

[Markus Elfring]

> A statement list metavariable only matches the code starting right after a {.
> It can't start in random places.  Nothing is going to chenge here.

Will ever any other metavariable type become supported with which a mixture
of items (including statements and variable declarations) could be handled
in a succinct way?

Regards,
Markus

Re: [cocci] Too limited support for SmPL statement lists?

[Markus Elfring]

> A statement list metavariable only matches the code starting right after a {.
> It can't start in random places. …

I got the impression then that the supported functionality is so limited for
the metavariable type “statement list” so far.
It seems that it is unlikely to be usable besides discussed test cases.

Statements are often mixed with variable declarations and definitions.

Regards,
Markus