On 02/03/2012 08:27 AM, Henrik Rydberg wrote:
> Hi Chase,
>
>>> +#define INPUT_MT_REQUEST(num_slots) \
>>> + { \
>>> + __u32 code; \
>>> + __s32 values[num_slots]; \
>>
>> I think this assumes a userspace C compiler that can handle variable
>> length arrays. This would require only compiling in C source code at the
>> C99 standard or later. It looks like C++ doesn't even allow variable
>> length arrays, though gcc handles it. According to:
>>
>> http://www.cplusplus.com/forum/beginner/1601/
>>
>> it looks like Borland c++ compilers may not be able to compile this :(.
>
> This is resolved on the preprocessor level, so C99 or not does not
> enter the problem. Compile-time constant, as you can see in the code
> example in the patch summary.
You're right, I didn't catch that. It will be compatible with all C
compilers if you use a static number of slots.
However, it will break if you use a non-C99 C compiler and the code
wants to do dynamic number of slots calculations. I imagine most callers
would do:
EVIOCGABS call on ABS_MT_SLOT;
int num_slots = ABS_MT_SLOT.max - ABS_MT_SLOT.min
struct INPUT_MT_REQUEST(num_slots) req;
This will break on non-C99 C compilers and other language compilers. It
also will lead to head-scratcher bugs when someone compiles it just fine
in their C99 project, copies the code to another project with a
different compiler, and is confronted with the issue.
I think this issue should be enough to rethink the interface.
-- Chase
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
