baloghadamsoftware added a comment.

Thank you for your respone! However, I think you (not you, Artem, but you three 
at Apple) do not really understand that I need to compare A+m to B+n not only 
because of the iterator range checking, but also in later parts. So your 
proposal means that I am not allowed to fix the shortcomings of the constraint 
solver neither in the engine nor in the checker. Maybe part 2 of the iterator 
checkers patch will work, but then we will restart the whole dispute at a later 
part.

As I mentioned, I need these comparisons in a lot of places, not only in range 
checking. This is the core part of my solution. It is primarily needed for 
invalidation, which is the most solid part. The latest version (not part 10, 
but a fixed version of the whole chekcer set) finds only 5 false positives for 
invalidated iterator access, which I could further reduce by adding more 
detailed simulation of insertion and erasement.

So still the options are to fix it in the checker or fix it in the engine (the 
max/4 or the type extension solution), but leaving it unfixed is not an option. 
I am open to any solution, but only full solutions and no partial solutions, 
because they will not help us enough.


https://reviews.llvm.org/D35109



_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to