On 02/11/2016 22:40, Julia Lawall wrote:
>
>
> On Wed, 2 Nov 2016, Daniel Lezcano wrote:
>
>> Hi all,
>>
>> I'm very new to coccinelle, so sorry if the question is irrelevant or was
>> already asked.
>>
>> I'm trying to add a new argument to a function which has a couple of
>> callbacks.
>>
>> The original function is:
>>
>> static inline int cpuhp_setup_state(enum cpuhp_state state,
>> int (*startup)(unsigned int cpu),
>> int (*teardown)(unsigned int cpu));
>>
>> I would like to pass a private pointer when setting the state and then pass
>> this pointer to the 'startup' / 'teardown' callbacks where there signature
>> will be
>> different and all call sites must change their callbacks signature also.
>>
>> I tried to first add the new argument to the function, in order to have:
>>
>> static inline int cpuhp_setup_state(enum cpuhp_state state,
>> int (*startup)(unsigned int cpu),
>> int (*teardown)(unsigned int cpu),
>> void *data);
>>
>> ... with the rule:
>>
>> @rule@
>> identifier cpuhp_setup_state;
>> identifier state, name, startup, teardown, data;
>> @@
>>
>> static inline int cpuhp_setup_state(enum cpuhp_state state,
>> const char *name,
>> int (*startup)(unsigned int cpu),
>> int (*teardown)(unsigned int cpu),
>> void *data)
>> {
>> ...
>> }
>>
>> @@
>> identifier rule.cpuhp_setup_state;
>> expression E1, E2, E3, E4;
>> @@
>>
>> cpuhp_setup_state(E1, E2, E3, E4
>> + , NULL
>> )
>>
>> But the parsing gives:
>>
>> ...
>>
>> cpuhp_setup_state(E1, E2, E3, E4
>> >>> , NULL
>
> This is normal. The >>> has to do with whether the NULL is to be put
> after what comes before or before what comes after. <<< would mean the
> opposite. Unfortunately, I don't remember which is which, but it doesn't
> really matter in this case either.
>
>> )
>>
>>
>> grep tokens
>> cpuhp_state || cpu
>> No query
>> warning: line 8: should cpu be a metavariable?
>> warning: line 9: should cpu be a metavariable?
>
> I guess you would want an identifier cpu metavriable declaration in the
> first rule. Otherwise, I don't really see any problems. Do you have a .c
> file on which you expect to get a result?
Yes, that is all around the linux kernel code (eg.
./drivers/clocksource/qcom-timer.c)
If I change the cocci file to:
@@
expression E1, E2, E3, E4;
@@
-cpuhp_setup_state(E1, E2, E3, E4)
+cpuhp_setup_state(E1, E2, E3, E4, NULL)
It seems to work, except the function signature itself is not changed. I
can do add a manual change in the patch but if coccinelle can handle
both that would be nice.
--
<http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs
Follow Linaro: <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog
_______________________________________________
Cocci mailing list
[email protected]
https://systeme.lip6.fr/mailman/listinfo/cocci
