[cfe-commits] [Patch] Warn about self-initialization of reference variables

[Hans Wennborg]

On Thu, Aug 16, 2012 at 9:24 PM, Richard Smith <richard at metafoo.co.uk> wrote:
> LGTM, modulo a couple of small things:

Thanks for the review!

>> -  if (!VDecl->hasLocalStorage() &&
>> -      (isa<InitListExpr>(Init) || !VDecl->getType()->isRecordType()))
>> -    CheckSelfReference(RealDecl, Init);
>> +  if (!VDecl->hasLocalStorage() || VDecl->getType()->isReferenceType())
>> +    if (isa<InitListExpr>(Init) || !VDecl->getType()->isRecordType())
>> +      CheckSelfReference(RealDecl, Init);
>
> '&&' instead of nested 'if's? It might be clearer to reuse !isTrackedVar()
> (from Analysis/UninitializedValues.cpp) here, but I'm happy for the patch to
> go in either way.

Switched to && instead of nested ifs for now.

!isTrackedVar() doesn't seem to be a one-to-one match here. For
example, in my if statement I'm checking that it's not a record type,
whereas !isTrackedVar() would check that it's not a scalar or vector
type. Happy to work with this post-commit if you think it would be
better, though.

>> +++ b/test/SemaCXX/convert-to-bool.cpp
> Your warning found a bug in the test! I'm pretty sure this intended to say:
> B &b = c;

Fixed.

Committed r162093.

Thanks,
Hans