[cfe-commits] r76911 - /cfe/trunk/lib/Analysis/CFG.cpp

[Ted Kremenek]

On Jul 24, 2009, at 12:48 AM, Sebastian Redl wrote:

> Ted Kremenek wrote:
>> On Jul 23, 2009, at 9:33 PM, Eli Friedman wrote:
>>
>>> (int)x == false is semantically equivalent to (int)x == 0 due to
>>> integer promotion rules.  That said, it is slightly confusing the  
>>> way
>>> it's written.
>>>
>>
>>
>> The problem isn't with 'x == 0', but with 'x == -1'.  Since (I
>> believe) both '1' and '-1' convert to true when int is casted to a
>> bool, the 'KnownVal == true' would evaluate to true in both cases
>> (which isn't the desired behavior).
>>
> Integer promotion rules are the same no matter the values. In KnownVal
> == true the 'true' literal is converted to the integer 1, and KnownVal
> is not touched.

Ah that's right.  The promotion is from bool to int, not the other way  
around.  Thanks Eli and Sebastian.  I think what my concern was with  
this code is that someone (mistakenly) could write 'KnownVal' instead  
of 'KnownVal == true' or '!KnownVal' instead of 'KnownVal == false'  
and not realize that they were fundamentally changing the logic of the  
code.