[cfe-dev] CXXForRangeStmtClass in Static Analyzer

[Jim Goodnow II]

On 9/22/2011 5:22 PM, Ted Kremenek wrote:
>
> On Sep 22, 2011, at 5:03 PM, Jim Goodnow II wrote:
>
>> Just curious. Is there any reason that CXXForRangeStmtClass isn't 
>> treated the same as ForStmtClass in ExprEngine?
>
> They shouldn't be treated the same because they aren't the same 
> semantically.  In the AST their common ancestor is Stmt; they need to 
> be handled separately.
>
>> It shouldn't appear explicitly in the CFG since it is already handled 
>> by the semantic engine.
>
> The CFG is suppose to represent all control-flow, including this 
> statement.  What do you mean by the "semantic engine"?
>
Yes, it is true that they are not the same semantically, but when the 
AST is generated, the semantics of the ForRange and For are replaced 
with the appropriate statements that implement the semantics by 
initializing, incrementing, comparing and branching. So, the CFG does 
model both statement types with other explicit statements which is why 
the For never appears explicitly in the CFG. The same should hold true 
for the ForRange. I'm just suggesting that the CXXForRangeStmtClass case 
be moved from the 'unsupported' part of the switch to the section with 
For, Break, Continue, etc. Now, the ObjectiveC ForCollection statement 
may be handled differently which may require it to be explicitly handled 
in the CFG and in the analyzer.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20110922/844db8cf/attachment.html>