[cfe-dev] [PATCH] -Wconversion-null

[Douglas Gregor]

Sent from my iPhone

On Mar 17, 2012, at 11:50 AM, "James K. Lowden" <jklowden at schemamania.org> wrote:

> On Thu, 15 Mar 2012 21:31:39 -0500
> Joshua Cranmer <pidgeot18 at gmail.com> wrote:
> 
>> I should be 
>> able to look at code and not have to reason about the formal
>> underlying algebraic model to figure out what the intent is
> 
> You don't really mean that, do you?  What is programming except the use
> of logical constructs, and how is it possible to program in any
> language without know the meaning expressed through its syntax?  
> 
> Regarding bit-wise operators, I didn't make that case.  They're
> relatively rarely used, and even Dennis Ritchie expressed some regret
> over their precedence.  Logical operators, on the other hand, appear in
> all but the most trivial programs. 
> 
>> most people don't memorize the precedence charts and instantly 
>> plan for minimizing parenthesis usage, especially when you consider
>> that precedence does vary between languages
> 
> I'm not prepared to make any claim about what most programmers do or
> think.  I have a hard time believing, per above, that much useful work
> can be accomplished in C++ without knowing the precedence of C++
> operators.  
> 
> If you know a language that 1) defines operator precedence and 2)
> defines OR before AND, I'd be interested to hear.  I was unable to find
> one.  For that reason I reject the "multi-language confusion"
> hypothesis.  
> 
> AFAIK the question of bugs stemming from logical operator confusion
> hasn't been carefully studied. Rather than assume I'm smarter than the
> poor schlump using my compiler, I prefer to think that whatever I know,
> anyone else can learn, too

Turning these warnings on in our codebase yielded a significant number of real bugs with very few false positives. 

> Note you're positing a very peculiar kind of ignorance: the
> programmer who isn't sure of the precedence *and* decides not to use
> parentheses to make the logic clear to *himself*.  I would not presume
> to try to help such a person, and I don't think Clang should, either.  

They exist, and I've returned their bug reports with an explanation of the precedence rules. Clang helped them already. 

Programmers make mistakes, even when they know the precedence rules perfectly. Code gets refactored, and bugs get introduced. 

You are following a well-trodden path with your line of argument against Wparentheses and similar and I personally am not interested in going down it again. Real data from real codebases supports the hypothesis that these warnings do more good than harm, and no "programmers should know better" argument will change that. 

  - Doug