Simplification of relational expressions

[John Fine]

Andrew Haley wrote:
> Which it is doing. What, then, is your point?
My point was explaining the aspect of this that the OP probably wouldn't 
have understood after your first reply:  The fact that the optimizer in 
a specific compiler can/should/does base its behavior on the actual non 
optimized code of THAT compiler, not on the universe of possible non 
optimized code specified by the standard.

In other words, I was giving the reason that your first reply is correct 
even with his statement "signed integer overflow is undefined".  I 
expect a beginner reading your reply would conclude you had ignored or 
disagreed with that statement.