> > Well, the question is how to identify "operator new and operator delete at the > point of the new-expression and delete-expression". Currently we're > just matching up "is this a new/delete operator" and the dataflow of the > pointer. In the PR it was suggested the actual called methods should have > the same DECL_CONTEXT. Honza suggested the context should have the > "same" ODR type (or be global). I was just arguing that comparing pointers to types does not make much sence in LTO where types can get unshared :) Since the classes have ODR name at least this problem can be solved by using ODR name compare.
However the testcase I sent abuses the fact that if you inherit the class you can rewrite only new operator. In the inerited class DECL_CONTEXT of delete will still point to the oriignal class. This means that you can mix new/delete pair from the original and inherited class. Honza > > You make it sound it's much harder and the parser needs to build the > relation? You also suggest the "validness" is only OK in the context > of std::allocator and based on the unspecifiedness of the number of > allocations from the standard library. That would further suggest that > we need to mark the allocation/deallocation points somehow and _not_ > base the optimization on the called new/delete "function" (maybe with > an exception of the global ::new and ::delete). > > Richard.
