[Cocci] Replacing function definition with a macro

[Cocci] Replacing function definition with a macro

[Petr Malát]

Hi All,
I'm trying to replace a function prototype in the definition with a
macro. The reason is to generate stubs to do some argument mangling.
The following is my cocci file:

@@
type T1;
identifier i1, name;
@@
- long name(T1 i1) {
+ COMPAT_SYSCALL_DEFINE1(name, T1, i1) {
  ...
  }

And the following is my testing code:

#define __SC_DECL1(t1, a1) t1 a1
#define COMPAT_SYSCALL_DEFINE1(name, ...) \
                long compat_sys_##name(__SC_DECL1(__VA_ARGS__))
long test(int first) {
        return 0;
}


And I expect spatch to replace
  long test(int first) {
with
  COMPAT_SYSCALL_DEFINE1(test, int, first) {
but when I try to run it, I get
  init_defs_builtins: /usr/local/lib/coccinelle/standard.h
  82 86
  Fatal error: exception Failure("plus: parse error:
   = File "test.cocci", line 6, column 25,  charpos = 82
      around = 'name', whole content = + COMPAT_SYSCALL_DEFINE1(name, T1, i1) {
  ")
Any idea what am I doing wrong?

Thanks,
  Petr

[Cocci] Replacing function definition with a macro

[Julia Lawall]

On Tue, 4 Aug 2015, Petr Mal?t wrote:

> Hi All,
> I'm trying to replace a function prototype in the definition with a
> macro. The reason is to generate stubs to do some argument mangling.
> The following is my cocci file:
>
> @@
> type T1;
> identifier i1, name;
> @@
> - long name(T1 i1) {
> + COMPAT_SYSCALL_DEFINE1(name, T1, i1) {
>   ...
>   }
>
> And the following is my testing code:
>
> #define __SC_DECL1(t1, a1) t1 a1
> #define COMPAT_SYSCALL_DEFINE1(name, ...) \
>                 long compat_sys_##name(__SC_DECL1(__VA_ARGS__))
> long test(int first) {
>         return 0;
> }
>
>
> And I expect spatch to replace
>   long test(int first) {
> with
>   COMPAT_SYSCALL_DEFINE1(test, int, first) {
> but when I try to run it, I get
>   init_defs_builtins: /usr/local/lib/coccinelle/standard.h
>   82 86
>   Fatal error: exception Failure("plus: parse error:
>    = File "test.cocci", line 6, column 25,  charpos = 82
>       around = 'name', whole content = + COMPAT_SYSCALL_DEFINE1(name, T1, i1) {
>   ")
> Any idea what am I doing wrong?

Coccinelle expects that each element in a parameter list will have the
form type identifier.  In the case of name, you just have an identifier.
I don't really see a solution.

julia

[Cocci] Replacing function definition with a macro

[Petr Malát]

Hi Julia,
thanks for the answer
> Coccinelle expects that each element in a parameter list will have the
> form type identifier.  In the case of name, you just have an identifier.
> I don't really see a solution.
I though similar operations are generally possible, if all required macros
are known. Anyway, I've worked it around by using something like

@@
type T1;
identifier i1, name;
typedef DELME;
@@
- long name(T1 i1) {
+ COMPAT_SYSCALL_DEFINE1(DELME name, T1 DELME_TOO, DELME i1) {
  ...
  }

and then fixing the generated patch with sed.

Cheers,
  Petr

[Cocci] Replacing function definition with a macro

[Julia Lawall]

On Tue, 4 Aug 2015, Petr Mal?t wrote:

> Hi Julia,
> thanks for the answer
> > Coccinelle expects that each element in a parameter list will have the
> > form type identifier.  In the case of name, you just have an identifier.
> > I don't really see a solution.
> I though similar operations are generally possible, if all required macros
> are known.

No.  Macros are somewhat taken into account when parsing the C code.  But
here you are talking about the SmPL (pattern) code, which is different.

> Anyway, I've worked it around by using something like
>
> @@
> type T1;
> identifier i1, name;
> typedef DELME;
> @@
> - long name(T1 i1) {
> + COMPAT_SYSCALL_DEFINE1(DELME name, T1 DELME_TOO, DELME i1) {
>   ...
>   }
>
> and then fixing the generated patch with sed.

Yeah, not ideal, but perhaps easier than making all the changes by hand.

julia

[Cocci] Replacing function definition with a macro

[SF Markus Elfring]

> Coccinelle expects that each element in a parameter list will have the
> form type identifier.

Is the Coccinelle software still limited to some degree for the handling
of function parameters?


> In the case of name, you just have an identifier.
> I don't really see a solution.

How do you think about direct support for macro parameter lists?
Does anybody dare to contribute corresponding software development efforts?

Are there any more limitations to reconsider so that further post-processing
can be avoided for generated patches?

Regards,
Markus