Re: [PATCH net-next] compiler: Document behavior compiling with -O0

[Joe Stringer]

On 25 August 2017 at 04:45, Michal Nazarewicz <mina86@xxxxxxxxxx> wrote:
> On Thu, Aug 24 2017, Joe Stringer wrote:
>> Recent changes[0] to make use of __compiletime_assert() from container_of()
>> increased the scope of this macro, resulting in a wider set of
>> situations where developers cannot compile their code using "-O0". I
>> noticed this when making use of the macro in my own development, and
>> spent more time than I'd like to admit tracking the problem down. This
>> patch documents the behavior in lieu of a compile-time assertion
>> implementation that does not rely on optimizations.
>>
>> Example compilation failure:
>>
>> ./include/linux/compiler.h:547:38: error: call to â__compiletime_assert_94â declared with attribute error: pointer type mismatch in container_of()
>>   _compiletime_assert(condition, msg, __compiletime_assert_, __LINE__)
>>                                       ^
>> ./include/linux/compiler.h:530:4: note: in definition of macro â__compiletime_assertâ
>>     prefix ## suffix();    \
>>     ^~~~~~
>> ./include/linux/compiler.h:547:2: note: in expansion of macro â_compiletime_assertâ
>>   _compiletime_assert(condition, msg, __compiletime_assert_, __LINE__)
>>   ^~~~~~~~~~~~~~~~~~~
>> ./include/linux/build_bug.h:46:37: note: in expansion of macro âcompiletime_assertâ
>>  #define BUILD_BUG_ON_MSG(cond, msg) compiletime_assert(!(cond), msg)
>>                                      ^~~~~~~~~~~~~~~~~~
>> ./include/linux/kernel.h:860:2: note: in expansion of macro âBUILD_BUG_ON_MSGâ
>>   BUILD_BUG_ON_MSG(!__same_type(*(ptr), ((type *)0)->member) && \
>>   ^~~~~~~~~~~~~~~~
>>
>> [0] http://lkml.kernel.org/r/20170525120316.24473-7-abbotti@xxxxxxxxx
>>
>> Signed-off-by: Joe Stringer <joe@xxxxxxx>
>> ---
>>  include/linux/compiler.h | 5 +++++
>>  1 file changed, 5 insertions(+)
>>
>> diff --git a/include/linux/compiler.h b/include/linux/compiler.h
>> index eca8ad75e28b..bb640167fdac 100644
>> --- a/include/linux/compiler.h
>> +++ b/include/linux/compiler.h
>> @@ -517,6 +517,11 @@ static __always_inline void __write_once_size(volatile void *p, void *res, int s
>>  # define __compiletime_error_fallback(condition) do { } while (0)
>>  #endif
>>
>> +/*
>> + * __compiletime_assert() relies on compiler optimizations to remove the check
>> + * against '__cond' if 'condition' is false. As a result, compiling with -O0
>> + * will cause compilation errors here regardless of the value of 'condition'.
>> + */
>>  #define __compiletime_assert(condition, msg, prefix, suffix)         \
>>       do {                                                            \
>>               bool __cond = !(condition);                             \
>
> Could __builtin_choose_expr help here?  Something like:
>
> #define __compiletime_assert(condition, msg, prefix, suffix)            \
>         do {                                                            \
>                 bool __cond = !(condition);                             \
>                 extern int prefix ## suffix(void) __compiletime_error(msg); \
>                 __builting_choose_expr(cond, prefix ## suffix(), 0);    \
>                 __compiletime_error_fallback(__cond);                   \
>         } while (0)
>
> Or better still, _Static_assert?

I tried both of the above, and they both complain that "condition"
isn't a constant.