[c++,12/12] ada-lang.h: Add cast in GROW_VECT

  The assignment requires a cast in C++.

gdb/ChangeLog:

	* ada-lang.h (GROW_VECT): Add cast.
---
 gdb/ada-lang.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
  

On 10/26/2015 03:49 AM, Simon Marchi wrote:
> The assignment requires a cast in C++.
> 
> gdb/ChangeLog:
> 
> 	* ada-lang.h (GROW_VECT): Add cast.
> ---
>  gdb/ada-lang.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/gdb/ada-lang.h b/gdb/ada-lang.h
> index 62896f1..32c4b55 100644
> --- a/gdb/ada-lang.h
> +++ b/gdb/ada-lang.h
> @@ -147,7 +147,7 @@ struct ada_task_info
>     least M objects, updating V and S as necessary.  */
>  
>  #define GROW_VECT(v, s, m)                                    \
> -   if ((s) < (m)) (v) = grow_vect (v, &(s), m, sizeof *(v));
> +   if ((s) < (m)) (v) = (typeof (v)) grow_vect (v, &(s), m, sizeof *(v));
>  

typeof in C is a GCC extension.  Probably not all compilers support it.

In my branch I just have:

  #define GROW_VECT(v, s, m)                                    \
 -   if ((s) < (m)) (v) = grow_vect (v, &(s), m, sizeof *(v));
 +   if ((s) < (m)) (v) = (char *) grow_vect (v, &(s), m, sizeof *(v));

Because that works for all current uses of GROW_VECT.

If we wanted to make this work for random types, then we could add
a type parameter to the GROW_VECT macro, like:

- #define GROW_VECT(v, s, m)                                    \
-   if ((s) < (m)) (v) = grow_vect (v, &(s), m, sizeof *(v));
+ #define GROW_VECT(t, v, s, m)                                    \
+   if ((s) < (m)) (v) = (t *) grow_vect (v, &(s), m, sizeof *(v));

Not sure it's worth it though.  It then raises the question of
"why not replace this home grown GROW_VECT stuff with a VEC instead".

Thanks,
Pedro Alves
  

On 26/10/15 12:50 PM, Pedro Alves wrote:
> On 10/26/2015 03:49 AM, Simon Marchi wrote:
>> The assignment requires a cast in C++.
>>
>> gdb/ChangeLog:
>>
>> 	* ada-lang.h (GROW_VECT): Add cast.
>> ---
>>  gdb/ada-lang.h | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/gdb/ada-lang.h b/gdb/ada-lang.h
>> index 62896f1..32c4b55 100644
>> --- a/gdb/ada-lang.h
>> +++ b/gdb/ada-lang.h
>> @@ -147,7 +147,7 @@ struct ada_task_info
>>     least M objects, updating V and S as necessary.  */
>>  
>>  #define GROW_VECT(v, s, m)                                    \
>> -   if ((s) < (m)) (v) = grow_vect (v, &(s), m, sizeof *(v));
>> +   if ((s) < (m)) (v) = (typeof (v)) grow_vect (v, &(s), m, sizeof *(v));
>>  
> 
> typeof in C is a GCC extension.  Probably not all compilers support it.
> 
> In my branch I just have:
> 
>   #define GROW_VECT(v, s, m)                                    \
>  -   if ((s) < (m)) (v) = grow_vect (v, &(s), m, sizeof *(v));
>  +   if ((s) < (m)) (v) = (char *) grow_vect (v, &(s), m, sizeof *(v));
> 
> Because that works for all current uses of GROW_VECT.
> 
> If we wanted to make this work for random types, then we could add
> a type parameter to the GROW_VECT macro, like:
> 
> - #define GROW_VECT(v, s, m)                                    \
> -   if ((s) < (m)) (v) = grow_vect (v, &(s), m, sizeof *(v));
> + #define GROW_VECT(t, v, s, m)                                    \
> +   if ((s) < (m)) (v) = (t *) grow_vect (v, &(s), m, sizeof *(v));
> 
> Not sure it's worth it though.  It then raises the question of
> "why not replace this home grown GROW_VECT stuff with a VEC instead".
> 
> Thanks,
> Pedro Alves

I thought about changing it for a proper VEC, but it's not really in the scope of
the C++ changes.  If I side-track on things like this, it will take an eternity!
And I feel that at this moment, it would be a risk of introducing bugs without
added value.

Is it ok if I simply push your version that adds the (char *) in the macro?  It
seems like the most efficient fix given the situation.
  

On 10/27/2015 02:34 AM, Simon Marchi wrote:
> On 26/10/15 12:50 PM, Pedro Alves wrote:

>> typeof in C is a GCC extension.  Probably not all compilers support it.
>>
>> In my branch I just have:
>>
>>   #define GROW_VECT(v, s, m)                                    \
>>  -   if ((s) < (m)) (v) = grow_vect (v, &(s), m, sizeof *(v));
>>  +   if ((s) < (m)) (v) = (char *) grow_vect (v, &(s), m, sizeof *(v));
>>
>> Because that works for all current uses of GROW_VECT.
>>
>> If we wanted to make this work for random types, then we could add
>> a type parameter to the GROW_VECT macro, like:
>>
>> - #define GROW_VECT(v, s, m)                                    \
>> -   if ((s) < (m)) (v) = grow_vect (v, &(s), m, sizeof *(v));
>> + #define GROW_VECT(t, v, s, m)                                    \
>> +   if ((s) < (m)) (v) = (t *) grow_vect (v, &(s), m, sizeof *(v));
>>
>> Not sure it's worth it though.  It then raises the question of
>> "why not replace this home grown GROW_VECT stuff with a VEC instead".

> I thought about changing it for a proper VEC, but it's not really in the scope of
> the C++ changes.  If I side-track on things like this, it will take an eternity!
> And I feel that at this moment, it would be a risk of introducing bugs without
> added value.
> 
> Is it ok if I simply push your version that adds the (char *) in the macro?  It
> seems like the most efficient fix given the situation.

It's certainly fine with me.

Thanks,
Pedro Alves