[Cocci] Match #define in cocci

[Cocci] Match #define in cocci

[Waldemar Rymarkiewicz]

Hi,

I'm new to cocci and try to understand how all this things work. I
have a problem with matching a #define  in a patch.

I wan to match local definition

   #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

and  add #undef  it first. The output should be like this

   #undef  pr_fmt
   #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

I tried many approaches but seems like parser have a problem with
parsing #defines.

Is that possible to achieve  what I expect in cocci ?


Cheers
/Waldek

[Cocci] Match #define in cocci

[Julia Lawall]

On Wed, 14 Sep 2016, Waldemar Rymarkiewicz wrote:

> Hi,
>
> I'm new to cocci and try to understand how all this things work. I
> have a problem with matching a #define  in a patch.
>
> I wan to match local definition
>
>    #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>
> and  add #undef  it first. The output should be like this
>
>    #undef  pr_fmt
>    #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>
> I tried many approaches but seems like parser have a problem with
> parsing #defines.
>
> Is that possible to achieve  what I expect in cocci ?

Could you send the exact semantic patch that you tried?  Coccinelle should
allos parsing #defines.

julia

[Cocci] Match #define in cocci

[Waldemar Rymarkiewicz]

>
> Could you send the exact semantic patch that you tried?  Coccinelle should
> allos parsing #defines.

Simply

@@
@@
-#define pr_fmt(fmt) KBUILD_MODNAME

works, but

@@
@@
-#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

fails  with

Fatal error: exception Failure("minus: parse error: \n = File
\"patches/1021-pr_fmt_undef.cocci\", line 4, column 38,  charpos =
59\n    around = '\"', whole content = #define pr_fmt(fmt)
KBUILD_MODNAME \": \" fmt\n")

but even if I want to match just

@@
@@
-#define pr_fmt(fmt) KBUILD_MODNAME  fmt

it also fails with

Fatal error: exception Failure("minus: parse error: \n = File
\"patches/1021-pr_fmt_undef.cocci\", line 4, column 0,  charpos = 21\n
   around = '#define pr_fmt(', whole content = #define pr_fmt(fmt)
KBUILD_MODNAME fmt\n")


Thanks,
/Waldek

[Cocci] Match #define in cocci

[Julia Lawall]

On Thu, 15 Sep 2016, Waldemar Rymarkiewicz wrote:

> >
> > Could you send the exact semantic patch that you tried?  Coccinelle should
> > allos parsing #defines.
>
> Simply
>
> @@
> @@
> -#define pr_fmt(fmt) KBUILD_MODNAME
>
> works, but
>
> @@
> @@
> -#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>
> fails  with
>
> Fatal error: exception Failure("minus: parse error: \n = File
> \"patches/1021-pr_fmt_undef.cocci\", line 4, column 38,  charpos =
> 59\n    around = '\"', whole content = #define pr_fmt(fmt)
> KBUILD_MODNAME \": \" fmt\n")
>
> but even if I want to match just
>
> @@
> @@
> -#define pr_fmt(fmt) KBUILD_MODNAME  fmt
>
> it also fails with
>
> Fatal error: exception Failure("minus: parse error: \n = File
> \"patches/1021-pr_fmt_undef.cocci\", line 4, column 0,  charpos = 21\n
>    around = '#define pr_fmt(', whole content = #define pr_fmt(fmt)
> KBUILD_MODNAME fmt\n")

OK, thanks I will look into it.  At least in the last case, that is not
the intended behavior.

julia

[Cocci] Match #define in cocci

[Waldemar Rymarkiewicz]

> OK, thanks I will look into it.  At least in the last case, that is not
> the intended behavior.
>

One more, I use 1.0.2 version as I had a problem with  1.0.5
compilation previously.
I will try bump to 1.0. 5 anyway and check the behavior.

/Waldek

[Cocci] Match #define in cocci

[Waldemar Rymarkiewicz]

On 15 September 2016 at 08:41, Waldemar Rymarkiewicz
<waldemar.rymarkiewicz@gmail.com> wrote:
>> OK, thanks I will look into it.  At least in the last case, that is not
>> the intended behavior.
>>
>
> One more, I use 1.0.2 version as I had a problem with  1.0.5
> compilation previously.
> I will try bump to 1.0. 5 anyway and check the behavior.
>

Now I use
  spatch version 1.0.5-00089-g0849d71 compiled with OCaml version 4.02.3
  Flags passed to the configure script: [none]
  Python scripting support: yes
  Syntax of regular expresssions: PCRE

and still cannot mach

@@
@@
-#define pr_fmt(fmt) KBUILD_MODNAME fmt

minus: parse error:
  File "patches/1021-pr_fmt_undef.cocci", line 3, column 1, charpos = 7
  around = '#define pr_fmt(',
  whole content = -#define pr_fmt(fmt) KBUILD_MODNAME fmt


/Waldek

[Cocci] Match #define in cocci

[Julia Lawall]

On Thu, 15 Sep 2016, Waldemar Rymarkiewicz wrote:

> On 15 September 2016 at 08:41, Waldemar Rymarkiewicz
> <waldemar.rymarkiewicz@gmail.com> wrote:
> >> OK, thanks I will look into it.  At least in the last case, that is not
> >> the intended behavior.
> >>
> >
> > One more, I use 1.0.2 version as I had a problem with  1.0.5
> > compilation previously.
> > I will try bump to 1.0. 5 anyway and check the behavior.
> >
>
> Now I use
>   spatch version 1.0.5-00089-g0849d71 compiled with OCaml version 4.02.3
>   Flags passed to the configure script: [none]
>   Python scripting support: yes
>   Syntax of regular expresssions: PCRE
>
> and still cannot mach
>
> @@
> @@
> -#define pr_fmt(fmt) KBUILD_MODNAME fmt
>
> minus: parse error:
>   File "patches/1021-pr_fmt_undef.cocci", line 3, column 1, charpos = 7
>   around = '#define pr_fmt(',
>   whole content = -#define pr_fmt(fmt) KBUILD_MODNAME fmt

Yeah, I didn't expect it to work, but the github version is better in
other ways :)

julia

[Cocci] Match #define with SmPL

[SF Markus Elfring]

> One more, I use 1.0.2 version as I had a problem with  1.0.5
> compilation previously.

Thanks that you added a bit more version information to your bug report.


> I will try bump to 1.0. 5 anyway and check the behavior.

Thanks for your interest to try additional software variants out.

I have tried also the following script out for the semantic patch language
of the software combination "Coccinelle 1.0.5-00089-g0849d71-dirty (OCaml 4.03.0)"
a moment ago.


@macro_deletion@
@@
-#define pr_fmt(fmt) "SVM: " fmt


elfring at Sonne:~/Projekte/Linux/next-patched> spatch.opt ~/Projekte/Coccinelle/Probe/delete_macro1.cocci arch/x86/kvm/svm.c
init_defs_builtins: /usr/local/lib64/coccinelle/standard.h
minus: parse error: 
  File "/home/elfring/Projekte/Coccinelle/Probe/delete_macro1.cocci", line 3, column 29, charpos = 49
  around = 'fmt',
  whole content = -#define pr_fmt(fmt) "SVM: " fmt


Do you eventually know any more software developers who would like to help in the task
"Completion of support for preprocessor statements by the Coccinelle software"?


I would be also looking for an interface to store the text from the macro definition
into a corresponding data structure.

* How do think about to look at another "story" like "Checking printk() calls with SmPL"
  once more?
  https://github.com/coccinelle/coccinelle/issues/54#issuecomment-169988424

* Which metavariable types would be appropriate for the discussed source code transformation?


I would like to point another detail out for further considerations according to
your initial request.

1. I guess that you should use the SmPL asterisk functionality if you are interested
   in only finding (or "matching") a bit of code.

@macro_definition@
@@
*#define pr_fmt(fmt) "SVM: " fmt


init_defs_builtins: /usr/local/lib64/coccinelle/standard.h
minus: parse error: 
  File "/home/elfring/Projekte/Coccinelle/Probe/show_macro_definition1.cocci", line 3, column 29, charpos = 51
  around = 'fmt',
  whole content = *#define pr_fmt(fmt) "SVM: " fmt


2. Do you want to try another source code adjustment out?

@statement_addition@
@@
+#undef  pr_fmt
 #define pr_fmt(fmt) "SVM: " fmt


elfring at Sonne:~/Projekte/Linux/next-patched> spatch.opt ~/Projekte/Coccinelle/Probe/add_undef1.cocci arch/x86/kvm/svm.c
init_defs_builtins: /usr/local/lib64/coccinelle/standard.h
minus: parse error: 
  File "/home/elfring/Projekte/Coccinelle/Probe/add_undef1.cocci", line 4, column 29, charpos = 69
  around = 'fmt',
  whole content =  #define pr_fmt(fmt) "SVM: " fmt


How often would you like to use such search patterns for your source files?

Regards,
Markus

[Cocci] Match #define with SmPL

[Waldemar Rymarkiewicz]

> I would be also looking for an interface to store the text from the macro definition
> into a corresponding data structure.
>
> * How do think about to look at another "story" like "Checking printk() calls with SmPL"
>   once more?
>   https://github.com/coccinelle/coccinelle/issues/54#issuecomment-169988424
>
> * Which metavariable types would be appropriate for the discussed source code transformation?
>
>
> I would like to point another detail out for further considerations according to
> your initial request.
>
> 1. I guess that you should use the SmPL asterisk functionality if you are interested
>    in only finding (or "matching") a bit of code.
>
> @macro_definition@
> @@
> *#define pr_fmt(fmt) "SVM: " fmt
>
>
> init_defs_builtins: /usr/local/lib64/coccinelle/standard.h
> minus: parse error:
>   File "/home/elfring/Projekte/Coccinelle/Probe/show_macro_definition1.cocci", line 3, column 29, charpos = 51
>   around = 'fmt',
>   whole content = *#define pr_fmt(fmt) "SVM: " fmt
>
>
> 2. Do you want to try another source code adjustment out?
>
> @statement_addition@
> @@
> +#undef  pr_fmt
>  #define pr_fmt(fmt) "SVM: " fmt
>
>
> elfring at Sonne:~/Projekte/Linux/next-patched> spatch.opt ~/Projekte/Coccinelle/Probe/add_undef1.cocci arch/x86/kvm/svm.c
> init_defs_builtins: /usr/local/lib64/coccinelle/standard.h
> minus: parse error:
>   File "/home/elfring/Projekte/Coccinelle/Probe/add_undef1.cocci", line 4, column 29, charpos = 69
>   around = 'fmt',
>   whole content =  #define pr_fmt(fmt) "SVM: " fmt
>
>
> How often would you like to use such search patterns for your source files?

I've already tried all above, but always I get parser error.
Apparently parsing #defines is not working as expected specially if
you use argument list.


/Waldek

[Cocci] Match #define with SmPL

[SF Markus Elfring]

>> How often would you like to use such search patterns for your source files?

Would you like to answer this question?


> I've already tried all above,

Your tries were not so clear initially.


> but always I get parser error.

Now we can agree that there open issues for further considerations, can't we?


> Apparently parsing #defines is not working as expected specially if
> you use argument list.

How do you think about to pick this use case up as another software development challenge?

Would you like to extend your development experience in this direction anyhow?

Regards,
Markus