So the questions are
- why doesn't it happen on 5.6?
- apparently because control doesn't reach that codepath.
- So why does it not? And why does it on 5.5? Perhaps the real bug is that this codepath is reached, not that the pointer is not checked for NULL?
In other words, I find it strange that 5.5 needs that check and 5.6 does not.
So the questions are
- why doesn't it happen on 5.6?
- apparently because control doesn't reach that codepath.
- So why does it not? And why does it on 5.5? Perhaps the real bug is that this codepath is reached, not that the pointer is not checked for NULL?
In other words, I find it strange that 5.5 needs that check and 5.6 does not.