baloghadamsoftware added a comment.
In D81718#2095043 <https://reviews.llvm.org/D81718#2095043>, @NoQ wrote:
> So it still thinks that the iterator is an object, whereas it should be
> tracking the iterator as symbol instead. As expected, because your code fails
> to discriminate between these two cases. You can't reorder checks in
> `{set,get}IteratorPosition` to fix that (first check for symbol, then check
> for region) because that'd break object iterators that reside in symbolic
> regions; you need to repeat the check on the AST instead. Which makes the
> reuse of that check proposed in this patch worthless.
This one is easy to overcome, because in case of pointers we need the pointer
itself. The checker //knows// in advance, whether it is a pointer or not.
However it //does not know// whether it is a value or a (constant) reference.
However, with such iterator-specific behavior this function must go into the
iterator library. (Pointers are symbols, references are objects.)
Do you have some alternative suggestions?
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D81718/new/
https://reviews.llvm.org/D81718
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
