Re: bk commit into 5.0 tree (thek:1.2489) BUG#28249

[Kristofer Pettersson]

Sergei Golubchik wrote:
>
> No, don't return the table size in engine_data. It creates an impression
> that this is actually needed for something, while, in fact, the value
> will be ignored, so it's completely useless.
>   
Fixed. Engine data will always be returned as 0.

>
> Very confusing comment. Even though I know what happens in the
> bugreport, how insert invalidation works in general and why it fails
> here, and how your fix helps, I still have troubles understanding the
> comment. Try to rewrite it please.
>   
Simplified.
>   
>> +  */
>> +  ulonglong actual_data_file_length;
>> +  ulonglong current_data_file_length;
>> +
>> +  pthread_mutex_lock(&file->lock.lock->mutex);
>>     
>
> As we discussed with Kostja, you don't need a mutex here.
> You only want to see if actual_data_file_length !=
> current_data_file_length, you don't care about the exact value of
> actual_data_file_length. So, it's ok if you'll see garbage there because
> of non-atomic update.
>   
Mutex removed.


/Kristofer