[cfe-commits] r131989 - in /cfe/trunk: include/clang/Basic/DiagnosticGroups.td include/clang/Basic/DiagnosticSemaKinds.td lib/Sema/SemaExprCXX.cpp test/SemaCXX/destructor.cpp

[Anders Carlsson]

On May 30, 2011, at 2:49 PM, Eli Friedman wrote:

> On Mon, May 30, 2011 at 2:40 PM, Argyrios Kyrtzidis <akyrtzi at gmail.com> wrote:
>> On May 30, 2011, at 2:36 PM, Argyrios Kyrtzidis wrote:
>> 
>> Hi Nico,
>> On May 30, 2011, at 12:19 PM, Nico Weber wrote:
>> 
>> Hi Argyrios and Matthieu,
>> this warning found a few problems in chromium – thanks! However, it also
>> finds a few false positives, so I don't think I can turn it on by default,
>> which is a bummer.
>> All of the false positives are of this form:
>> class SomeInterface {
>>  public:
>>   virtual void interfaceMethod() {}  // or = 0;
>>  protected:
>>   ~SomeInterface() {}
>> }
>> class WorkerClass : public SomeInterface {
>>  public:
>>   // many non-virtual functions, but also:
>>   virtual void interfaceMethod() override { /* do actual work */ }
>> };
>> void f() {
>>   scoped_ptr<WorkerClass> c(new WorkerClass);  // simplified example
>> }
>> This is a somewhat standard pattern (see
>> e.g. http://www.gotw.ca/publications/mill18.htm, "Virtual Question #2").
>> Do you have any good suggestions how to deal with this case?
>> 
>> If WorkerClass gets subclassed in the future, deletion in "f()" will be
>> undefined behavior. Ideally WorkerClass should be marked "final" and then
>> there will also be no warning; does this sound reasonable ?
>> 
>> Um, to be exact, undefined behavior if you delete a WorkerClass * pointer.
> 
> Sure, except that there isn't any way to do that outside of C++0x mode. :)
> 

We actually accept final and override as C++0x extensions even outside of C++0x.

- Anders
> -Eli
> 
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits