[cfe-dev] CFRefCount Problem #2: Region Invalidation

[Jordy Rose]

On Aug 27, 2011, at 15:19, Ted Kremenek wrote:

> On Aug 27, 2011, at 3:11 PM, Jordy Rose wrote:
> 
>> 
>> On Aug 27, 2011, at 14:32, Jordy Rose wrote:
>> 
>>> P.S. I just realized that NewExprs are not implemented in terms of ConstructExprs. If they were, it would make things easier. I don't really want to add NewExprs to CallOrObjCMessage too! Guess that's a new patch to work on, in a completely different part of the system.
>> 
>> Uh, combining NewExprs and ConstructExprs doesn't seem to actually make sense outside of the analyzer. Oh well.
>> 
>> OTOH, there are /no/ pre- or post-statement hooks for NewExprs right now, so to even get a stop-tracking result for arguments to NewExprs we either need to add a callback or include an Expr or ProgramPoint in checkRegionChanges.
> 
> Great observation.  I suggest we table this issue for now so we can continue to make progress.  Let's file a PR and tackle it in later in follow-up.  I think we should make progress on the core retain/release checker changes first.  What do you think?

Filed as PR10794.